FACULTY OF HEALTH SCIENCES

UNIVERSITY OF COPENHAGEN

Brain microstructural correlates of behavioural and neuroendocrinological phenotypes

DTI studies of response inhibition and choice reaction time in typically-developing children and of negative emotionality and circadian cortisol patterns in healthy adults

PhD thesis
Kathrine S kak M adsen, M Sc

Preface
The work presented in this PhD thesis was carried out between February 1, 2008 and January 31, 2011 at the Danish Research Centre for Magnetic Resonance, MR-department, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark and the Center for Integrated Molecular Brain Imaging, CIMBI, Copenhagen, Denmark.

Supervisors
Terry L. Jernigan, Professor, PhD
Danish Research Centre for Magnetic Resonance, MR-department, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, CIMBI, Copenhagen, Denmark Center for Human Development, University of California, San Diego, La Jolla, California, United states

William F. C. Baar, Senior researcher, PhD

Danish Research Centre for Magnetic Resonance, MR-department, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, CIMBI, Copenhagen, Denmark

Summary
The major aim of the present PhD project was to study relationships between brain microstructure and behavioural and neuroendocrinological phenotypes by employing diffusion- weighted imaging (DWI) in healthy human subjects. The thesis builds on data collected in two different projects; the Hubu (Hjernens Udvilking hos Brn og Unge, in English Brain maturation in children and adolescents) project, and the Cimbi (Center for integrated molecular brain imaging) project. Four manuscripts make up the basis of this thesis, of which two are based on the Hubu cohort, and two are based on the Cimbi cohort. The Hubu studies aimed to examine associations between regional brain microstructure, and response inhibition and visuospatial motor performance, as assessed with the stop-signal task and a basic choice reaction time task in a cohort of typically-developing children aged 7-13 years. Both grey and white matter regions continue to develop during childhood and adolescence, and indices of their maturation can be obtained using DWI. Though response inhibition and visuospatial motor function are known to improve during childhood and adolescence, little is known about how response inhibition and choice reaction time performance is correlated with structural brain development. In the first study, better response inhibition performance was associated with higher fractional anisotropy (FA) in the white matter underlying the right inferior frontal gyrus and presupplementary motor area, adjusted for age effects. In the second study, faster choice reaction times were associated with lower mean diffusivity in the corticospinal tracts and the striatum, independently of age. The observed associations may be related to individual variations in the phase of maturation of children of similar age, to activity-dependent alterations in the networks subserving response inhibition and visuospatial motor function, or to more stable individual differences in underlying neural system architecture. The Cimbi studies aimed to examine associations between microstructural asymmetry of the limbic system, neuroticism a personality trait reflecting the tendency to experience negative emotions, and circadian cortisol measures in a cohort of healthy adults aged 19-86 years. The limbic system plays an important role in regulating the hypothalamic-pituitary-adrenal (HPA) axis as well as aspects of emotions, and both neuroendocrine disturbance and increased negative emotionality are associated with the risk of developing affective disorders. However, the extent to which architectural characteristics of limbic structures may reflect individual differences in HPA-axis activity and negative emotionality is unknown. The first study found that higher neuroticism scores, which were associated with higher cortisol awakening response

ii

(CAR), were also correlated with higher right relative to left cingulum FA. Elevated CAR was associated with the degree of FA asymmetry within both the cingulum and the uncinate fasciculus, but in opposing directions. Specifically, higher CAR was associated with higher right relative to left cingulum FA, and with decreased right relative to left uncinate fasciculus FA. The second study found that higher left relative to right hippocampus mean diffusivity was associated with higher basal cortisol levels. These findings provide a new perspective on the biology of HPA-axis activity and negative emotionality in humans. Overall, the findings suggest that the balance between left- and right-sided limbic circuits may bear an important relationship to HPA-axis activity, and to the tendency to experience negative emotions, raising important questions about the significance of limbic system architecture.

ii

Dansk Resum (Danish summary)

Det primre forml med det aktuelle ph.d.-projekt var at undersge sammenhngen mellem hjernens mikrostruktur, og adfrds- og neuroendokrinologiske fnotyper ved at anvende diffusionsvgtet billeddannelse hos raske forsgspersoner Afhandlingen bygger p data indsamlet i to forskellige projekter; Hubu (Hjernens Udvilking hos Brn og Unge), og Cimbi (Center for integrated molecular brain imaging). Fire manuskripter danner basis for nrvrende afhandling, hvoraf to er baseret p Hubu kohorten, og to er baseret p Cimbi kohorten. Formlet med Hubu studierne var at undersge sammenhngen mellem mikrostruktur i specifikke hjerneregioner, og respons inhibition og visuospatial motor prstation i to forskellige psykologiske tests; stop-signal task (inhibitionstest) og en simpel 5-valgs reaktionstidstest, i en kohorte bestende af raske brn mellem 7 og 13 r. Gennem barndommen og ungdomsrene gennemgr hjernens gr og hvide substans en lbende modning, som kan estimeres ved hjlp af diffusionsvgtet billeddannelse. Selvom bde respons inhibition og visuospatial motor funktion forbedres i lbet af brne- og ungdomsrene, vides der kun lidt om, hvordan disse er korreleret med hjernestrukturel modning. I det frste studie blev bedre respons inhibition associeret med hjere fraktionel anisotropi (FA) i hvid substans underliggende hjre gyrus frontalis inferior og hjre presupplementary motor area justeret for alderseffekter. I det andet studie blev hurtigere 5-valgs reaktionstider associeret med lavere middel-diffusivitet i de kortikospinale baner, samt i striatum, og disse associationer var uafhngig af alder. De observerede sammenhnge kan vre relaterede til individuelle variationer i det opnede modningsstadie hos brn af samme alder, til aktivitetsafhngige ndringer i de netvrk, som medierer respons inhibition og visuospatial motor funktion, eller til stabile individuelle forskelle i arkitekturen af det underliggende neurale system. Formlet med Cimbi studierne var at undersge sammenhngen mellem mikrostrukturel asymmetri i det limbiske system i forhold til neuroticisme - et personlighedstrk, der afspejler en tendens til at opleve negative flelser, og kortisol dgnrytme mlt i en kohorte bestende ad raske voksne mellem 19 og 86 r. Det limbiske system spiller en vigtig rolle i reguleringen af hypothalamus-hypofyse-binyrebark (HPA) aksen s vel som i aspekter af emotioner. Bde neuroendokrine forstyrrelser og forhjet negativ emotionalitet er blevet associeret med hjere risiko for at udvikle affektive lidelser. Det vides dog ikke, hvordan individuelle forskelle i mikrostrukturelle karakteristika af limbiske strukturer og fiberbundter er forbundet med individuelle forskelle i HPA-akse aktivitet og negativ emotionalitet. I det frste studie blev det

iii

fundet, at hjere neuroticisme niveauer ud over at vre associeret med hjere kortisol opvgningsrespons (CAR), ogs var associeret med hjere FA i hjre cingulum i forhold til venstre cingulum. Endvidere blev en sammenhng mellem hjere CAR og graden af FA asymmetri i bde cingulum og fasciculus uncinatus fundet, som var i modsat retning. Specifikt, s blev hjere CAR associeret med hjere FA i hjre relativ til venstre cingulum, og med lavere FA i hjre relativ til venstre fasciculus uncinatus. I det efterflgende studie blev hjere middel- diffusivitet i venstre relativet til hjre hippocampus associeret med hjere basal kortisol niveauer. Disse fund giver et nyt perspektiv p det biologiske grundlag for HPA-akse regulering og negativ emotionalitet i mennesker. Samlet set foreslr resultaterne, at balancen mellem venstre og hjre limbiske strukturer spiller en betydelig rolle for individuelle variationer i funktionen af HPA-aksen og i tendensen til at opleve negative flelser. Disse nye fund rejser vigtige sprgsml om hvordan det limbiske system er relateret til negativ emotionalitet og neuroendokrine funktioner.

iii

Acknowledgements
Special thanks go to my supervisor Terry Jernigan for academic supervision, inspiring ideas and discussions, and for being able to explain even complex things in an understandable way. Special thanks also go to my supervisor William Baar for academic supervision, patient guidance, encouragement, and educative and inspiring discussions. Grateful thanks go to Arnold Skimminge for technical support, and to Sussi Larsen, who did all the MR-scanning. I would also like to thank all my co-authors on the papers included in the thesis for inspiring collaborations, and my colleagues at DRCMR and in Cimbi for a stimulating working environment. I wish to thank all volunteers participating in the studies, and especially the children and their families for their effort in participating in the Hubu project. Finally, I would like to thank family and friends for their support, and in particular Pernille Maj Svendsen and Bettina Hornbll for their support in the final phase of the thesis writing.

iv

List of original papers

Paper 1
Madsen KS, Baar WFC, Vestergaard M, Skimminge A, Ejersboe LR, Ramsy TZ, Gerlach C, keson P, Paulson OB, and Jernigan TL (2009) Response inhibition is associated with white matter microstructure in children. Neuropsychologia 48(4): 854-862.

Paper 2
Madsen KS, Baar WFC, Skimminge A, Vestergaard M, Siebner HR, and Jernigan TL. Brain microstructural correlates of visuospatial choice reaction time in children. Submitted

Paper 3
Madsen KS, Jernigan TL, Iversen P, Frokjaer VG, Mortensen EL, Knudsen GM, and Baar WFC. Cortisol awakening response and negative emotionality linked to asymmetry in major limbic fibre bundle architecture. Submitted

Paper 4

Madsen KS, Jernigan TL, Iversen P, Frokjaer VG, Knudsen GM, Siebner HR and Baar WFC. Hypothalamic-pituitary-adrenal axis tonus is associated with hippocampal microstructural asymmetry. Submitted

Abbreviations
|| CAR Cimbi CST DARTEL DRCMR DTI DWI MD FA HPA Hubu IFG MRI PFC PreSMA ROI(s) RT SD SSRT SST STN TBSS parallel diffusivity perpendicular diffusivity cortisol awakening response center for integrated molecular brain imaging corticospinal tract diffeomorphic anatomical registration through exponentiated lie algebra Danish research centre for magnetic resonance diffusion tensor imaging diffusion-weighted imaging mean diffusivity fractional anisotropy hypothalamic-pituitary-adrenal hjernens udvikling hos brn og unge - brain maturation in children and adolescents inferior frontal gyrus magnetic resonance imaging prefrontal cortex pre-supplementary motor area region(s)-of-interest reaction time standard deviation stop-signal reaction time stop-signal task subthalamic nucleus tract-based spatial statistics

vi

Contents
Preface.......................................................................................................................................................................................i Summary .................................................................................................................................................................................ii Dansk resum...................................................................................................................................................................... iii Acknowledgements .......................................................................................................................................................... iv List of original papers........................................................................................................................................................v Abbreviations ...................................................................................................................................................................... vi

Chapter 1. Diffusion-weighted imaging ...................................................................................3

General introduction to DWI....................................................................................................................................3 Diffusion tensor imaging............................................................................................................................................3 Age-related changes in children and adolescents...........................................................................................4 Biological interpretation of diffusion parameters ..........................................................................................6

Chapter 2. Phenotypes...................................................................................................................9
Response inhibition......................................................................................................................................................9 Visuospatial choice reaction time .......................................................................................................................12 The Hypothalamic-pituitary-adrenal axis .......................................................................................................13 Neuroticism ..................................................................................................................................................................17

Chapter 3. Aims and Hypotheses............................................................................................. 19

Hubu papers: Motivation and hypotheses.......................................................................................................19 Cimbi papers: Motivation and hypotheses......................................................................................................20

Chapter 4. Methods ...................................................................................................................... 21

Projects ...........................................................................................................................................................................21 Subjects...........................................................................................................................................................................22 Behavioural assessments........................................................................................................................................23 Salivary cortisol assessment .................................................................................................................................24 MRI acquisition ...........................................................................................................................................................25 Image evaluation ........................................................................................................................................................25 MRI data processing..................................................................................................................................................25 Regions-of-interest....................................................................................................................................................29 Statistical analyses.....................................................................................................................................................30

Chapter 5. Main results............................................................................................................... 31

Paper 1: Response inhibition and WM microstructure in children......................................................31

Paper 2: Microstructure and visuospatial choice RT in children ..........................................................33 Paper 3: CAR and neuroticism links to limbic fibre bundle asymmetry ............................................34 Paper 4: HPA-axis tonus and hippocampal microstructural asymmetry ..........................................36

Chapter 6. General discussion.................................................................................................. 39

Hubu papers .................................................................................................................................................................39 Cimbi papers.................................................................................................................................................................41

Chapter 7. Future work .............................................................................................................. 45 References ...................................................................................................................................... 47 Papers............................................................................................................................................... 57

Chapter 1
In this chapter, a brief introduction to diffusion-weighted imaging (DWI) is given. The focus will be on diffusion tensor imaging (DTI), and on age-related changes during childhood and adolescence. Finally, the interpretation of diffusion parameters will be discussed.

Diffusion-weighted imaging

General introduction to DWI

DWI is a non-invasive imaging technique that within recent years has become widely used to study the microstructure of tissue, particularly in the brain. DWI is sensitive to the diffusion of water molecules. In white matter, water diffusion is less hindered parallel than perpendicular to fibre bundles, as cellular structures, such as the axonal membranes and surrounding myelin sheaths, hinder the diffusion across the fibre bundle, thereby causing anisotropy (directionality) of the diffusion. In grey matter, diffusion is also hindered by cellular structures, but it is relatively isotropic (equal diffusion in all directions) at clinical voxel sizes (Beaulieu 2009).

Diffusion tensor imaging

A common use of diffusion imaging involves fitting the diffusion measurements of each voxel to a diffusion tensor that estimates the magnitude of water diffusion (diffusivity) along 3 orthogonal axes (Fig. 1). The diffusivities along the 3 principal axes are referred to as eigenvalues (1, 2, 3) and are ordered according to their magnitude. By employing the diffusion tensor model, several useful diffusion parameters can be derived (Basser et al. 1994; Beaulieu 2009). One of the most commonly used diffusion parameters is fractional anisotropy (FA), which estimates the degree of diffusion directionality. FA values range from 0 to 1, where 0 indicate complete isotropic diffusion, and increasing values indicate increasing anisotropy of the diffusion (Fig. 1). FA is generated by using the equation:

Diffusion-weighted imaging

Moreover, useful diffusion parameters regarding the overall magnitude of the diffusion, the mean diffusivity (MD = (1 + 2 + 3 / 3)), as well as diffusivity parallel (|| = 1) and perpendicular ( = (2 + 3) / 2) to the principal diffusion direction can be derived.

Figure 1. Left: Schematic illustration of the diffusion tensor ellipsoid. The three eigenvalues (1, 2, 3) correspond to the diffusivities along the 3 principal axes, and are sorted according to their magnitude (1 2 3). The orientation of the principal axes is given by the 3 orthogonal eigenvectors (1, 2, 3) (Figure from Jones 2009). Right: All 3 diffusion ellipsoids have the same mean diffusivity (0.7 x 10-3 mm2/s), but different fractional anisotropy ranging from almost isotropic to anisotropic. Since the mean diffusivity is fixed, the parallel and perpendicular diffusivity must change in opposite direction to get the change in the shape of the ellipsoid (Beaulieu 2009).

Age-related changes in children and adolescents

Conventional structural imaging studies

Brain maturation is a complex ongoing process during childhood and early adulthood. Conventional structural MRI studies have demonstrated morphological changes in grey and white matter structures during childhood and adolescence consistent with cellular maturational processes, i.e., synaptic pruning and myelination (Giedd et al. 1999; Gogtay et al. 2004; Jernigan et al. 2011; Jernigan et al. 1991; Paus et al. 1999; Sowell et al. 2004); and different grey matter structures exhibit distinct maturational trajectories (Gogtay et al. 2004; Jernigan et al. 1991; Sowell et al. 2004). Furthermore, gradual increases in global white matter volume as well as regional white matter density, have been reported in children and adolescents, possibly reflecting age-related increase in axonal diameter and ongoing myelination across this age- range (Giedd et al. 1999; Paus et al. 2001; Paus et al. 1999). However, conventional structural MRI provides limited information about the underlying white matter microstructural properties.

Diffusion-weighted imaging

DTI studies
In recent years, DTI has been applied in studies of typically-developing children and adolescents to investigate the microstructure of grey and white matter. Age-related increases in FA and/or decreases in MD have been reported in multiple locations with both grey and white matter regions (Barnea-Goraly et al. 2005; Eluvathingal et al. 2007; Lebel et al. 2008; Snook et al. 2005). Typically the increase in FA is reflecting a disproportionate decrease in relative to ||. Although the physiological significance of these changes in diffusion parameters during childhood are still not fully understood, a previous study of infants correlated increased FA (and decreased ) with apparently increased neural conduction speed, as reflected in decreased latency of the first positive wave of the visual evoked potential (Dubois et al. 2008). Moreover, different white matter tracts and grey matter regions have been found to exhibit distinct maturational patterns, with diffusion parameters in some regions approaching adult levels earlier than others (Fig. 2), e.g. the inferior longitudinal fasciculus and the splenium of the corpus callosum reach adult maturity levels early, whereas the major limbic fibre bundles, the cingulum and uncinate fasciculus reach adult maturity levels relatively late in life (Lebel et al. 2008). Finally, it was recently reported that white matter maturation proceed in an inferior-to- superior and posteriorto-anterior manner (Colby et al. 2011), confirming the pattern already assumed by Yakovlev and Lecours (1967).

Figure 2. Left: Magnitude and timing of developmental FA changes, which varies across different brain regions (based on cross-sectional data from 202 subject between 5 to 30 years), The length of the horizontal bars indicate the age in which a given region had reached 90% of its developmental plateau. Right: The maturational curves of 3 white matter tracts that exhibit different maturational trajectories, with early, intermediate and late maturation (Figures from Lebel et al. 2008).

Diffusion-weighted imaging

Biological interpretation of diffusion parameters

In DWI, the displacement of water molecules is in the range of 5-11 m over a typical diffusion time of 40 ms (Beaulieu 2009). For cellular comparison, this is within the range of the size of many cellular structures, such as glial cell bodies (6-10 m) and axonal diameters (0.3-10 m) (Edgar and Griffiths 2009). However, in DWI clinical voxels sizes are in general around 2 x 2 x 2 2.5 x 2.5 x 2.5 mm2 and, thus, DWI gives a very rough estimate of the underlying microstructure within each voxel. This makes the interpretation of differences in diffusion parameters difficult. Even so, DWI has proven to be a sensitive method in detecting variability in microstructure. As reviewed by Beaulieu (Beaulieu 2002; 2009), cellular membranes are one of the key factors in causing anisotropy in white matter. Several studies have found that the degree of myelination also, to some extent, influences anisotropy (for review see Beaulieu 2009; Schwartz et al. 2005; Song et al. 2003; Song et al. 2005). A rodent study, using a model of experimental de- and remyelination of the corpus callosum, found that continuous cuprizone (neurotoxin) treatment caused demyelination of the callosal fibres, reflected by increased . When the treatment was discontinued, the effects were reversed, and the progression of fibre remyelination was consistent with decrease in (Song et al., 2005). Increases in FA is linked to either a decrease in or an increase in ||, or a combination of these. Thus, examining these diffusion parameters may sometimes provide additional information about the underlying white matter microstructure. Using a mouse model of retinal ischemia, a DTI and histology study revealed a gradual decrease in relative anisotropy (as measured with the RA index), caused by, at first, decrease in ||, corresponding with the timing of axonal degeneration, followed by increase in , associated with optic nerve demyelination (Song et al., 2003). Although FA and does not measure myelin directly, these findings suggest that FA and , at least to some extent, reflect the degree of myelination. However, given that the number of axons have been reported to be several 100.000 per mm2 (for review see Edgar and Griffiths 2009) and clinical voxel sizes are 8-15 mm2, a variety of other tissue parameters, such as axonal density, diameter, tortuosity, or number; extracellular volume fraction; or tract geometry may also contribute to changes in diffusion parameters (Beaulieu 2009; Madler et al. 2008; Schwartz et al. 2005) Little is known about the biological bases of variability in grey matter MD, particularly in healthy individuals. Lower MD has been related to higher cell density in tumours (Chenevert et al. 2006; Gibbs et al. 2009). Moreover, a recent study of rats found that over an 18-month period after radiation temporal changes in apparent diffusion coefficient were negatively correlated with changes in glial cell staining in the hippocampus (Huang et al. 2010). Although MD does not measure cell density directly, these findings suggest that differences in cell (neuron and glial)

Diffusion-weighted imaging density may be contributing to observed differences in MD. However, many other tissue parameters may also influence the magnitude of water diffusion, including membrane permeability, cell number and size (Chenevert et al. 2006), myelination and axon diameter of fibres originating or terminating in the hippocampus, as well as number, size and tortuosity of axons, dendrites and astrocytic processes.

Chapter 2
In this chapter, introductions to the behavioural and neuroendocrinological phenotypes or measures examined in the papers are given. The chapter consists of the four sections: Response
inhibition, Visuospatial choice reaction time, Hypothalamic-pituitary-adrenal axis, and Neuroticism.

Phenotypes

Response inhibition
An introduction to response inhibition and the neural network thought to mediate response inhibition. The focus will be on studies using the stop-signal task (SST), which was used in Paper 1. A more detailed description of the SST and the neural network subserving response inhibition is given in Paper 1.

Cognitive control
Cognitive control of behavior and the capacity to inhibit prepotent responses show marked development in the course of childhood and adolescence. Cognitive control of thoughts, actions, and emotions is important for normal adaptive behaviour, and deficits in behavioural control are prominent in a variety of psychiatric disorders, e.g., attention deficit/hyperactive disorder (Pliszka et al. 2006; Tillman et al. 2008) and obsessive-compulsive disorder (Enright and Beech 1993). Cognitive control of behaviour may be defined as the flexible regulation of thoughts and actions in the presence of competing ones (Durston and Casey 2006). Cognitive control is involved in many cognitive functions, including motor response inhibition - the ability to stop/withhold planned or ongoing motor responses. Over the years, investigators have developed experimental paradigms designed to measure motor control, or more specifically the capacity to inhibit primed, or prepotent, motor responses (Chambers et al. 2009). Among these is the SST, which provides a measure of a subjects ability to inhibit a prepotent manual response (Logan and Cowan 1984). Developmental studies have found that response inhibition,

Behavioural and neuroendocrinological phenotypes

as measured with the SST, continues to improve throughout childhood and adolescence (Tillman et al. 2008; Williams et al. 1999).

The stop-signal task

The SST consists of frequent go and infrequent stop trials. In go trials, subjects have to respond quickly and accurately to a go stimulus, whereas in stop trials, an auditory stop-signal is presented at some delay after the go stimulus, and subjects must attempt to withhold their (prepotent) responses. The time between the onset of the go stimulus and the stop signal, i.e., the stop-signal delay (SSD), changes adaptively throughout the test, depending on the subjects past performance, so that responses are inhibited successfully approximately 50% of the time for each subject. The shorter the SSD, the more likely it is that the subject will be able to inhibit his or her response. The primary behavioural outcome measure is the stop-signal reaction time (SSRT), which measures how fast subjects can inhibit a prepotent response (Fig. 3).

Figure 3. Estimation of the stop-signal reaction time (RT) using the race model (Logan and Cowan 1984). The curve illustrates the distribution of RTs on Go trials. On stop trials, the delay between the go signal and stop signal (stop-signal delay - SSD) varies adaptively throughout the task to produce the SSD 50%, where subjects are able to inhibit 50% of their responses. The stop-signal RT is estimated by subtracting the SSD 50% from the median go RT. The race model assumes that the go and stop processes are in a race with each other and are (mainly) independent of each other (Boucher et al. 2007; Logan and Cowan 1984). Shorter stop-signal RTs index better response inhibition performance. (Figure adabted from van den Wildenberg et al. 2006)

Neural network subserving response inhibition

Studies using the SST have implicated several brain structures in a neural network subserving response inhibition. Responding to the go stimulus is thought to be mediated by a premotor-

10

Behavioural and neuroendocrinological phenotypes striatal-pallidal-thalamic-motor cortical network, also referred to as the direct pathway (Fig. 4). A primarily right-lateralized network has been implicated in response inhibition, involving prominently the inferior frontal gyrus (IFG), as well as the presupplementary motor area (preSMA), and the subthalamic nucleus (STN) (Aron et al. 2007a; Aron and Poldrack 2006). Several other regions, such as the neostriatum and the primary motor cortex (Boehler et al. 2010; Zandbelt and Vink 2010) have also been suggested to be involved in response inhibition, although further details on these structures is not part of this thesis. Several functional magnetic resonance imaging (fMRI) studies have shown that inhibiting a prepotent response consistently activates prefrontal regions, particularly the right IFG and preSMA, in adults (Rubia et al., 2003; Aron and Poldrack 2006; Aron et al., 2007; Chevrier et al., 2007) as well as in children (Cohen et al. 2010). There is also converging evidence from human lesion studies suggesting that lesions in the right preSMA (Floden and Stuss 2006; Nachev et al. 2007) and the right IFG (Rieger et al. 2003), particularly in the IFG sub-region pars opercularis (Aron et al. 2003; Aron et al. 2004) impair response inhibition. The latter has been confirmed in a study using transcranial magnetic stimulation, where temporary deactivation of the right pars opercularis selectively impaired the ability to stop an initiated response (Chambers et al. 2006). The role of the STN region in response inhibition is evidenced by fMRI studies showing this region to be activated during stop trials (Aron et al. 2007a; Aron and Poldrack 2006), and by the finding that deep-brain stimulation of the STN improves response inhibition in patients with Parkinsons disease (van den Wildenberg et al. 2006). Consistently, STN lesions impair inhibition in rodents (Eagle et al. 2008).

Finally and importantly, evidence from a tractography study (Aron et al. 2007a) suggests

that the three regions implicated in response inhibition are interconnected. A possible preSMA-

Figure 4. Schematic diagram of the direct pathway (corticostriato-GPi/SNr), hyperdirect pathway (cortico STN-GPi/SNr) and indirect pathway (cortico-striato- GPe-STN-GPi/SNr). Abbreviations: Cx = cerebral cortex; GPe = globus pallidus (external); GPi = globus pallidus (internal); SNr = substantia nigra pars reticulata; STN = subthalamic nucleus; Str = striatum; Th = thalamus (Figure from Nambu et al. 2002). In simplified terms, activation of the direct pathway reduces the inhibition of thalamic neurons allowing a selected motor response to be activated, whereas activation of the hyper-direct and/or indirect pathways increase the inhibition of thalamic neurons allowing motor responses to be inhibited/ cancelled (Haber and Gdowski 2004; Nambu et al. 2002).

11

Behavioural and neuroendocrinological phenotypes

IFG connection is further supported by another tractography study in humans (Johansen-Berg et al. 2004), whereas the existence of direct fibre connections between the preSMA and STN is supported by tract tracing in monkeys (Inase et al. 1999). Given that the IFG and preSMA are connected to the STN, it has been suggested that the cancelling process of prepotent responses may be conveyed to the basal ganglia via the hyperdirect pathway (Fig. 4) (Aron et al. 2007a; Aron et al. 2007b). However, this hypothesis awaits further support.

Visuospatial choice reaction time

In this section, choice reaction time tasks are briefly introduced, as well as the direct pathway of movement.

Choice reaction time tasks and microstructural correlates

Perceptual-motor speed continues to develop throughout childhood and adolescence (Kail 1991). Increasing speed of processing can involve a wide range of sensory, cognitive and motor functions, and has been studied at the behavioural level with many different choice reaction time (RT) tasks (Konrad et al. 2009; Madden et al. 2004; Miller and Low 2001). Choice RT is generally defined as the time elapsed between the presentation of at least two stimuli that each require a different response, and movement initiation. Choice RT tasks differ significantly in their complexity, from very basic tasks, primarily involving motor selection and initiation, to more complex ones that require stimulus detection, response inhibition and/or working memory (Konrad et al. 2009; Madden et al. 2004; Miller and Low 2001). Previous DTI studies of young adults have examined the relationship between choice RT and white matter microstructure. These studies generally employed more complex choice RT tasks that also placed high demands on visual discrimination, response inhibition, and/or working memory (Konrad et al. 2009; Madden et al. 2004; Tuch et al. 2005). Hence, they generally focused their examinations on brain regions outside the motor system. Studies using visual oddball tasks have found that faster choice RTs were correlated with higher FA in the splenium of corpus callosum and the anterior limb of the internal capsule (Madden et al. 2004), as well as with lower MD in many different white matter tracts, such as the optic radiation, inferior longitudinal fasciculus, cingulum and superior longitudinal fasciculus (Konrad et al. 2009). Further, others have found effects in pathways supporting visuospatial attention (Tuch et al. 2005). The task used in Paper 2 was a very basic visuospatial choice RT task that mainly required movement selection and initiation, but little/no visual discrimination, response inhibition, or working memory. Specifically, the task required detection of a spatially-varying

12

Behavioural and neuroendocrinological phenotypes visual target stimulus, and selection, initiation, and execution of an appropriate motor program spatially matched to that target. Thus, we focused on core motor brain structures in this study.

Direct pathway of movement

The direct pathway is a cortico-basal ganglia circuit involved in facilitating and execution of voluntary movements. A schematic diagram of the direct pathway is depicted in Figure 4. Activation of the direct pathway is mediated by cortical inputs from the motor cortices to the neostriatum. When the neostriatum is activated, it sends inhibitory signals to the internal segment of globus pallidus and the substantia nigra pars reticularis. This reduces the inhibitory tone from these regions on the thalamus, which results in thalamic activation of cortical regions (Haber and Gdowski 2004), allowing activation of selected motor responses via the major output pathway, the corticospinal tract (CST). Diffusion parameters within the motor network, including the CST, putamen, and caudate nuclei, exhibit protracted age-related changes continuing into early adulthood (Lebel et al. 2008). Developmental trajectories of the CST, putamen and caudate from the Lebel et al. (2008) study are depicted in Figure 5.

Figure 5. Developmental trajectories of MD and FA in core motor structures (Lebel et al. 2008).

The Hypothalamic-pituitary-adrenal axis

In this section, the diurnal profile of the HPA-axis and the different modelling approaches used to estimate the dynamics of the cortisol awakening response are described. Moreover, studies

13

Behavioural and neuroendocrinological phenotypes

linking the HPA-axis to the limbic system are shortly discussed. A more detailed description of the link between HPA-axis functioning and the limbic system is given in Papers 3 and 4.

Diurnal profile of the HPA-axis

The HPA-axis is a major neuroendocrine system involved in coping with bodily and environmental stressors and challenges (Fries et al. 2009). In most large mammals, the hormonal end product of the HPA-axis is the glucocorticoid cortisol (corticosterone in rodents and birds). Under normal conditions, basal HPA-axis activity follows a distinct diurnal profile. In humans, plasma cortisol levels start to increase during the second half of the night, peaking approximately 30 minutes after morning awakening (Fig. 6). The rise in cortisol levels after morning awakening is commonly referred to as the cortisol awakening response. After peaking levels, cortisol levels gradually decrease during the rest of the day (Fries et al. 2009; Pruessner et al. 1997; Wilhelm et al. 2007). Disruption of the normal diurnal profile of the HPA-axis has been found in several psychiatric disorders. Although not consistently, lower basal cortisol levels and lower CAR have been associated with fatigue, burnout, and post-traumatic stress disorder (Chida and Steptoe 2009; Handwerger 2009), while higher morning cortisol levels have been associated with anxiety disorder (Vreeburg et al. 2010) and major depression (Chida and Steptoe 2009; Handwerger 2009; Vreeburg et al. 2009). Moreover, a prospective study found that individuals with high CAR had an increased risk of developing major depression (Adam et al. 2010).

Figure 6. Curve displaying mean standard error of mean of serum cortisol levels across the night, and within the first hour after awakening. Samples were obtained from 16 young healthy male adults under laboratory conditions. After awakening, a clear increase in cortisol levels that superimpose the rising levels of the night can be observed. This rise in cortisol levels in relation to morning awaking is commonly referred to as the cortisol awakening response (CAR). (Figure from Wilhelm et al. 2007)

Modelling approaches of the cortisol awakening response

Studies of the CAR generally include estimates of both the overall morning cortisol secretion, as well as the dynamic increase of CAR, in order to cover both the phasic and more tonic part of the CAR, which may be mediated by different neural mechanisms (see discussion in Paper 4). The

14

Behavioural and neuroendocrinological phenotypes overall morning cortisol secretion is estimated by either the mean morning cortisol level or as the total area under the curve (ground) of the morning samples (AUCg). These two measures are highly correlated (r=0.99). Note that correlations given in this section are based on the cohort included in Papers 3 and 4. In the literature, several modeling approaches have typically been used to try to capture the dynamics of CAR. Frequently used estimates are the area under the curve increase (AUCi) relative to the awakening sample (Chida and Steptoe 2009; Vreeburg et al. 2010), the absolute difference score between the peak sample and awakening sample (Absdif) (Chida and Steptoe 2009; Knoops et al. 2010), and the mean increase (MnInc: awakening sample subtracted from the mean of the following morning samples) (Chida and Steptoe 2009; Wust et al. 2000). Which CAR estimate one chooses depends on the question one wants to answer. In Papers 3 and 4 we wanted to use an estimate that highly reflected HPA-axis reactivity, but not the overall morning cortisol secretion. However, AUCi, MnInc and Absdif all show substantial correlations with the mean morning cortisol level (r=0.58-0.75), and, thus partly reflect the dynamics of CAR, and partly the overall morning cortisol output. In Papers 3 and 4, we estimated the dynamics of CAR as the difference between the maximum and mean morning cortisol level relative to the mean morning cortisol level (see methods sections in Thesis, or Papers 3 and 4). Importantly, this CAR estimate was not correlated with overall morning cortisol output (p=0.67). Moreover, this estimate of CAR is significantly correlated with Adif (p=0.02), but not with AUCi or MnInc (ps0.29).

Links between HPA-axis activity and limbic system

The limbic system plays a central role in stress regulation. Limbic structures, such as the medial prefrontal cortex (PFC), amygdala, and hippocampus, have been implicated in modulating the response to stress by regulating the hypothalamic-pituitary-adrenal (HPA)-axis (Dedovic et al. 2009; Jankord and Herman 2008; Sullivan and Gratton 1999). The relationship between diurnal cortisol profiles and human limbic brain structures is unclear and available data are sparse. Human studies investigating the relationship between brain structure and cortisol measures have typically focused on the hippocampus (Buchanan et al. 2004; Knoops et al. 2010; Pruessner et al. 2007; Wolf et al. 2002). Bilateral and unilateral hippocampal damage has been reported to abolish the morning response, while preserving overall cortisol levels throughout the day (Buchanan et al. 2004). Volumetric findings relating cortisol measures to the hippocampus have been inconsistent (Knoops et al. 2010; Kremen et al. 2010; Pruessner et al. 2007; Wolf et al. 2002). In a small cohort of 13 young subjects, one study found that higher morning cortisol levels were associated with larger hippocampal volumes (Pruessner et al. 2007). Two other studies in larger, middle-aged cohorts (N>380) did not find

15

Behavioural and neuroendocrinological phenotypes

such a relationship, but one observed that smaller hippocampal volumes were correlated with higher evening cortisol levels, as well as with higher morning cortisol levels after receiving dexamethasone (a synthetic glucocorticoid) the night before (Knoops et al. 2010). Recently,

higher mean cortisol levels were also associated with regionally reduced frontal cortical
thickness (Kremen et al. 2010).

Links between HPA-axis regulation and limbic asymmetry

A growing body of data suggests that HPA-axis functioning may be related to hemispheric lateralization. Indeed, stroke patients with left-sided, but not right-sided, infarctions showed increased morning cortisol levels compared to controls (Lueken et al. 2009). Moreover, healthy subjects with impaired cortisol suppression in response to dexamethasone had smaller left and substantially larger right than left anterior cingulate gyrus volumes as compared to suppressors (MacLullich et al. 2006). Furthermore, presenting an emotionally adverse film to the right hemisphere in healthy controls resulted in a consistently higher increase in phasic cortisol secretion than when presented to the left hemisphere (Wittling and Pfluger 1990). Additionally, findings from animal studies suggest that the left and right hemisphere may be differentially related to HPA-axis functioning (for review, see Cerqueira et al. 2008). In rodents right but not left medial PFC lesions significantly suppressed stress-induced increases in corticosterone levels in repeatedly restrained animals, suggesting that the right medial PFC may be activating physiological stress responses (Sullivan and Gratton 1999). Finally, an EEG study in monkeys found that animals with greater right relative to left frontal activity had higher plasma cortisol levels (Kalin et al. 1998). Overall, the balance between left and right prefrontal structures appears to be important for HPA-axis regulation and function.

Adverse effects of cortisol on limbic structures

Prolonged increases in cortisol level, for instance induced by chronic stress, are known to have adverse effects on hippocampal neurons and glial cells (Jauregui-Huerta et al. 2010; Rajkowska and Miguel-Hidalgo 2007; Sapolsky 2000; Starkman et al. 1992). In patients with elevated cortisol levels due to Cushings syndrome, hippocampal volume was negatively correlated with mean plasma cortisol levels (Starkman et al. 1992), and following treatment to reduce glucocorticoid levels, patients had a significant increase in hippocampal volume, which correlated with a decrease in 24-hour urinary cortisol levels (Starkman et al. 1999). Further, in elderly subjects prolonged higher basal cortisol levels have been associated with increased age- related hippocampal atrophy (Lupien et al. 1998). There are some indications that left limbic structures may be more vulnerable to high cortisol levels than those on the right (Cerqueira et al. 2005; van der Beek et al. 2004). In chronically-restrained (stressed) pigs, higher basal

16

Behavioural and neuroendocrinological phenotypes afternoon salivary cortisol levels were associated with decreased volume and neuron number in the left, but not the right dentate gyrus (van der Beek et al. 2004). Further, in a recent study (Zach et al. 2010) 3 weeks of corticosterone treatment in rats was associated with alterations of hippocampal volume asymmetry as well as measures of the number of neurons in hippocampal subfields. Moreover, the adverse and possible asymmetric effects of cortisol are not only restricted to the hippocampus. A study in rats found that chronic treatment with high doses of dexamethasone significantly reduced the volume of the left anterior cingulate gyrus, and to a lesser extent the right, suggesting that left anterior cingulate may be more vulnerable to prolonged high cortisol levels than the right (Cerqueira et al. 2005). Further, another study in rats found that chronic psychosocial stress suppressed proliferation and survival of newborn cells in to a greater extent in the left medial PFC than in the right (Czeh et al. 2007). Overall, these studies suggest that cortisol have differential effects on left and right limbic structures.

Neuroticism
This section gives a brief introduction to the personality trait neuroticism and to the brain structural correlates of negative emotionality related traits with the focus on limbic structures.

Neuroticism and links to affective disorders

Neuroticism is a fundamental personality trait that reflects an individuals tendency to experience negative emotions. Individuals with high neuroticism scores are more likely to experience emotions, such as anger, anxiety, cheerlessness, guilt, sadness and worries, and show higher susceptibility to stress. High neuroticism scores have been linked to mood and anxiety disorders, such as phobias, panic disorder, and major depression (Bienvenu et al. 2001). Prospective studies indicate that high neuroticism scores increase the risk of developing major depression (Kendler et al. 2006; Kendler et al. 2004). Moreover, twin studies have found that neuroticism show substantial heritability (Jang et al. 1996; Kendler et al. 2006). Additionally, associations between neuroticism and major depression show substantial genetic correlations (0.46-0.47), and, thus, may result from shared genetic factors. Finally, higher neuroticism scores have been associated with higher CAR and mean morning cortisol levels (Portella et al. 2005).

Links between negative emotionality and the limbic system

Several MRI studies have linked negative emotionality related personality traits to limbic structures. Structural MRI studies have reported negative associations between neuroticism and right amygdala grey matter concentration (Omura et al. 2005), and right dorsomedial PFC and left medial temporal lobe volume (DeYoung et al. 2010). The latter study also observed that

17

Behavioural and neuroendocrinological phenotypes

higher neuroticism scores were associated with increased bilateral mid-cingulate cortex volume. Further, higher neuroticism scores have been associated with thinner cortical thickness in the anterior portions of the left orbitofrontal cortex (Wright et al. 2006). Moreover, neuroticism has been positively correlated with frontolimbic serotonin 2A receptor binding (Frokjaer et al. 2008). Other structural MRI studies have investigated the neuroticism related temperamental trait Harm Avoidance (as measured with Temperament and Character Inventory). Higher harm avoidance scores have been associated with larger right anterior cingulate cortex surface area (Pujol et al. 2002), and with larger left amygdala volume in females, but not in males (Iidaka et al. 2006). Furthermore, higher harm avoidance scores have been negatively associated with right hippocampal volume, and in females, but not males with left anterior PFC volume. In addition, one DTI study reported that higher trait anxiety (as measured with the State Trait Anxiety Inventory) was associated with lower FA in the amygdalaventromedial PFC pathway (Kim and Whalen 2009). Finally, fMRI studies have also linked neuroticism to the limbic system. One study reported that higher neuroticism scores were associated with increased amygdala and subgenual anterior cortex activity in high vs. low emotional conflict trials (Haas et al. 2007). Interestingly, a recent fMRI study of negative emotional vs. neutral faces found that neuroticism was positively associated with right amygdala - dorsomedial PFC functional connectivity, but negatively correlated with left amygdala anterior cingulate connectivity (Cremers et al. 2010), implying that hemispheric asymmetry in the connectivity of these structures may be important for negative emotionality.

18

Chapter 3
This chapter presents brief descriptions of the motivations for conducting the studies included in this thesis along with the main hypotheses of each paper (for more detail see Papers 1-4). The major aim of the PhD project was to study relationships between brain microstructure and behavioural and neuroendocrinological phenotypes by employing diffusion-weighted imaging (DWI) in healthy human subjects.

Aims and Hypotheses

Hubu papers: Motivation and hypotheses

Cognitive control of thoughts, actions and emotions are important for normal behaviour and deficits in behavioural control are prominent in a variety of psychiatric disorders. Development of response inhibition and visuospatial motor function continues throughout childhood and adolescence. However, little is known about how response inhibition and visuospatial motor performance are correlated with microstructural brain development within the response inhibition and motor networks. In Papers 1 and 2, we aimed at finding such relationships by testing the below hypotheses in a relatively large cohort of healthy school-aged children.

Hypotheses
In Paper 1, we hypothesized that better response inhibition performance (lower SSRT) would be associated with higher FA in the white matter underlying the right IFG and the right preSMA, independently of age.

In Paper 2, we hypothesized that faster 5-choice RTs would be associated with lower MD in the corticospinal tracts and the neostriatum, and with higher FA in the corticospinal tracts, independently of age.

19

Aims and Hypotheses

Cimbi papers: Motivation and hypotheses

The limbic system plays an important role in regulating the HPA-axis as well as in aspects of emotion, and both neuroendocrine disturbance and increased negative emotionality have been associated with risk for developing affective disorders. To date, neuroimaging studies investigating the neurobiological correlates of HPA-axis activity and negative emotionality are sparse and have primarily used fMRI, and morphometric methods. No previous studies have investigated how the microstructural architecture of major limbic fibre tracts and the hippocampus relates to HPA-axis activity and neuroticism. Interestingly, studies in rodents indicate that the balance between left and right limbic structures is important for HPA-axis regulation and function. Surprisingly, such asymmetries have not been addressed in humans. Motivated by available evidence in humans and animals suggesting that asymmetry in limbic structures may be related with HPA-axis functioning and negative emotionality, we investigated the below hypotheses in a cohort of healthy adults in Papers 3 and 4.

Hypotheses
In Paper 3, we hypothesized that CAR and neuroticism would be associated with the extent of FA asymmetry in the cingulum and uncinate fasciculus. In Paper 4, we hypothesized that mean basal morning and afternoon/evening cortisol levels would be associated with the extent of hippocampal MD asymmetry.

20

Chapter 4
In this chapter, a brief description of the methods used in the thesis is provided. A more detailed description of most of the methods can be found in the papers. All studies were conducted as part of the Hubu and Cimbi projects.

Methods

Projects
Hubu
The Hubu project is an ongoing longitudinal project of children and adolescents that was initiated in 2007. With 6-months interval, subjects undergo structural MRI (including DWI) and behavioural assessments, including neuropsychological testing and questionnaires. In addition, genetic and biochemical determinants are assessed. In Autumn 2010, the seventh assessment round was completed. Data included in this thesis is from the baseline of this project.

Cimbi
Cimbi is founded on collaboration between several research institutions in the Copenhagen area, among which the DRCMR and the Neurobiology Research Unit at Rigshospitalet. One of the key areas of research in Cimbi is the investigation of neural correlates of personality dimensions and HPA-axis activity that predispose individuals for developing affective disorders. Healthy adult volunteers undergo an extensive assessment battery, including neuroimaging (structural MRI, DWI, fMRI and PET), behavioural assessments, such as personality and neuropsychological assessments, as well as assessments of genetic and biochemical (e.g. cortisol) determinants. This thesis includes the DWI, personality and cortisol data collected within the framework of Cimbi.

21

Methods

Subjects
Hubu
Ninety-two typically-developing children aged 7 to 13 from three schools (1st6th graders) in the Copenhagen suburban area (Kge kommune) were enrolled in the study. According to a parent report, no subjects had any known history of neurological or psychiatric disorders or significant brain injury. The age and gender distribution of the whole subject group is given in Table 1. Prior to participation, all children assented to the procedures and informed written consent was obtained from the parents/guardians after oral and written explanation of the study aims and study procedures. The study was approved by the local Danish Committee for Biomedical Research Ethics (protocol H-KF-01-131/03). Based on exclusion criteria (specified in Papers 1 and 2), 27 subjects were excluded for further studies in Paper 1, and 17 subjects in Paper 2. Thus, a total of 65 subjects were included in Paper 1 and 75 subjects in Paper 2. Table 1. Age and gender distribution of subjects participating in the Hubu study*. Age (mean SD) Gender (female/male) 1st/2nd graders 3rd/4th graders 5th/6th graders 8.2 0.5 18 / 14 10.1 0.4 19 / 15 12.2 0.4 16 /10 All subjects 10.0 1.6 53 / 39

*Children enrolled in the study were scanned either in the months just before (when in 1st, 3rd or 5th grade) or just after (when in 2nd, 4th or 6th grade) the summer holiday in July SD = standard deviation.

Cimbi
Sixty-nine healthy adults aged 19 to 86 years (mean standard deviation (SD) = 38.219.2) with no history of neurological or psychiatric disorders or brain injury and no MR contraindications as assessed with screening questionnaires were included. No subjects took psychoactive drugs, steroids, hormone replacements or drugs of abuse, including anabolic steroids. Six subjects received antihypertensive treatment with thiazide diuretics, ACE inhibitors, or angiotensin II receptor antagonists and all were normotensive. No participants received beta-blocking agents. The study was approved by the local Danish Committee for Biomedical Research Ethics (protocol number (KF) 01 2006-20) and performed in accordance with the Declaration of Helsinki. Informed written consent was obtained from all subjects prior to participation. Fifty-eight of the initial 69 subjects also agreed to collect salivary cortisol samples at home. Of these, 11 subjects were excluded (for details on exclusion criteria see Papers 3 and 4), and, thus, 48 subjects were included in analyses involving cortisol measures. Furthermore, 1 of the 48 subjects was excluded from analyses involving PM cortisol levels (see Paper 4).

22

Methods

Behavioural assessments
Stop-signal task
Response inhibition performance was assessed using the stop-signal task (SST), which provides a measure of a subjects ability to inhibit a prepotent manual response, i.e., the stop-signal reaction time (SSRT). The SST consists of frequent go and infrequent stop trials. In go trials, subjects have to respond as quickly and accurately to a go stimulus as possible. In stop trials, an acoustic stop signal is presented at some delay after the go stimulus, and subjects must attempt to withhold or cancel their motor responses. By adaptively varying the delay between the go and the stop signals during the task, a stop signal delay can be found for each individual that results in successful inhibition of the motor response on 50% of the stop trials. The SSRT is then estimated by subtracting this delay from the subjects median response latency on go trials. A detailed description of the task can be found in Paper 1.

Reaction time task

Choice reaction time (RT) performance was assessed using a basic computerized choice RT task requiring detection of a spatially-varying visual stimulus and selection, initiation, and execution of a manual response. Subjects sat in front of a touch screen, and were instructed to press a button on a pad in front of them using the index finger of the dominant hand. After a random delay, a yellow dot appeared in a circle displayed on the screen, and subjects had to let go of the button and touch the yellow dot as fast as possible. In an initial block of trials, the yellow dot appeared in a single circle located centrally on the screen. In the subsequent block of 5-choice trials, the yellow dot appeared (randomly) in one of five predefined circles. The behavioural measure used to test the hypotheses was the button release time (referred to as 5-choice RT) for the 5-choice test trials, which estimates the time taken to select and initiate the movement. A detailed description of the task can be found in Paper 2.

NEO-PI-R
Personality was assessed using the Danish version of the 240-item self-report questionnaire NEO-PI-R (NEO Personality Inventory Revised), which assesses the broad dimensions of the 5 fundamental personality traits: Neuroticism, Extraversion, Agreeableness, Conscientiousness and Openness. The Danish translation of NEO-PI-R has been psychometrically evaluated in a cohort of 600 subjects (Hansen et al. 2004). In Paper 3, we only focused on neuroticism, and, thus, the rest of this section will do so as well. Neuroticism scores are based on the added data of 48 items/statements (for examples see Table 2), which subjects evaluate on how well each of the statements fit their personality by using a 5-point Likert scale (Costa and McCrae 1992).

23

Methods

Table 2. Examples of NEO-PI-R neuroticism items I can handle myself pretty well in a crisis I often get angry at the way people treat me I often worry about things that might go wrong Too often, when things go wrong, I get discouraged and feel like giving up I rarely feel fearful or anxious Sometimes I do things on impulse that I later regret Neuroticism consists of 6 sub-scales (facets): anxiety, angry hostility, depression, self- conscientiousness, vulnerability, and impulsiveness. In the Danish version of NEO-PI-R and in line with the American version, the first 5 facets show high factor loadings on neuroticism (0.67- 0.84), while the loading for the latter (0.37) is lower (Hansen et al. 2004). Neuroticism reflects an individuals tendency to experience negative emotions, such as anger, anxiety, cheerlessness, guilt, sadness and worries, as well as susceptibility to stress.

Salivary cortisol assessment

Subjects collected 8 serial saliva samples at home at 0, 15, 30, 45 and 60 minutes after awakening and at 12, 6 and 11 PM using a cotton swab (for further details on the sampling protocol, see Papers 3 and 4). Salivary cortisol levels have been shown to be highly correlated with plasma cortisol levels (Lo et al. 1992), and, thus, salivary cortisol sampling provides an easy and non-invasive way of assessing cortisol variations in humans. Cortisol concentrations were determined at Copenhagen University Hospital, Rigshospitalet by a commercially available electrochemiluminescence immunoassay on Modular Analytics E170 equipment. The entire batch of samples was analyzed in one step, ensuring stability of laboratory dependent conditions. The lower limit of detection was 0.5 nmol/L, and samples below this level were scored as 0.25 nmol/L, when calculating the cortisol estimates. Values below 0.5 nmol/L were only found in the 6 PM (N=6) and 11 PM (N=15) samples, but not in any of the other sampling times. Averages of the morning and PM cortisol samples were used as estimates of HPA-axis tonus. Mean morning cortisol level was based on available samples collected within the first hour after awakening, and mean PM cortisol level was based on available samples collected in the afternoon/evening. The cortisol awakening response (CAR) was used as an estimate of HPA-axis reactivity. CAR was defined as the percentage difference (from mean cortisol) between maximum and mean morning cortisol level: CAR = ((MaximumMorning MeanMorning) / MeanMorning) * 100)

24

Methods Importantly, this estimate reflects the dynamics of the CAR, while it is not significantly correlated with the mean morning cortisol level (p=0.67), nor highly dependent on the first awakening sample (p=0.79).

MRI acquisition
All Hubu and Cimbi subjects were scanned on the same 3 Tesla Siemens Magnetom Trio MR scanner (Siemens, Erlangen, Germany) with an eight-channel head coil (Invivo, FL, USA) at the DRCMR. All subjects were scanned using the same MRI protocol consisting of (not in acquisition order): high-resolution T1-weighted and T2-weighted images of the whole head were acquired using, respectively, 3D magnetization prepared rapid gradient echo (MPRAGE) and 3D turbo spin echo sequences, whole brain DW images were acquired using a twice-refocused balanced spin echo sequence that minimised eddy current distortion (Reese et al. 2003), and a gradient echo based field map sequence (for sequence parameters, see papers). The total acquisition time was around 45 minutes.

Image evaluation
Before initiating the MR-image analyses, and blind to the behavioural and neuroendocrinological data, raw images from all subjects were visually inspected to ascertain the data quality. Further, an experienced neuroradiologist (Per keson, DRCMR) evaluated baseline images from all Hubu subjects, and based on this evaluation one subject was excluded due to incidental findings. Based on the quality inspection of the Hubu data, 17 subjects in Paper 1, and 12 subjects in Paper 2 were excluded from further analysis, because of reduced DW image quality due to movement or susceptibility artefacts. Fewer subjects were excluded in Paper 2 than in Paper 1 due to optimisation of the noise threshold used in RESTORE to fit our data (see RESTORE section below). No Cimbi subjects with NEO-PI-R or cortisol data were excluded due to poor image quality.

MRI data processing

Preprocessing
Images were preprocessed using pipelines implemented in Matlab using mainly SPM5 (SPM2 in Paper 1) routines. The data processing has been described in detail in the methods sections in Papers 1-4. A schematized overview of the image preprocessing pipeline used in the papers are given in Figure 7.

25

Methods

Figure 7. Schematic overview of the data preprocessing pipeline. MR-images are illustrated as black lined boxes. Methods applied to the images are depicted as blue boxes. The main preprocessing steps involved aligning the images to the same space, correcting the images for geometric distortions, movement correction of the DW images, estimation of the diffusion tensor, and skull stripping of the DTI images (see methods sections in Papers 1-4 for a written description of the data preprocessing). Abbreviations: Coreg = coregistered, Corr = corrected, DW = diffusion weighted, MNI = Montreal Neurological Institute (space), RESTORE = robust estimation of tensors by outlier detection, T1W = T1-weighted, T2W = T2-weighted.

RESTORE
In the data preprocessing pipeline, the diffusion tensor was estimated using the RESTORE (robust estimation of tensors by outlier detection) algorithm, which uses iteratively reweighted least-squares regression to identify potential outliers and subsequently exclude them (Chang et al. 2005). The current implementation of RESTORE requires setting a noise standard deviation (SD) threshold. In an attempt to optimize the outlier detection for our data, outlier maps were calculated based on varying noise SD thresholds on both good quality and corrupted (with slice or local dropouts) DWI data. A threshold, in which slice and local dropouts were detected as outliers without rejecting good quality data, was used in the preprocessing pipeline in Papers 2- 4). Because of this optimization, subjects who previously had been excluded in Paper 1 due to semi-poor data (e.g. a few slice dropouts in different slices) were included in Paper 2. It is recognized, however, that our current approach does not take into account that the noise in the data most certainly varies across the images.

26

Methods

Inter-subject spatial normalization of white matter

Spatial normalization and alignment of fibre tracts across subjects was achieved using the Tract- Based Spatial Statistics (TBSS) module (Smith et al. 2006) in FSL. A detailed description of TBSS can be found in Papers 1-3. In TBSS, a cross-subject mean FA image is created and thinned to create a mean FA skeleton, representing the centres of all tracts common to the group. To exclude areas with low FA and/or high inter-subject variability, the mean FA skeleton is then thresholded at a user determined FA threshold (Fig. 8). The threshold should be determined based on what regions one is interested in, and/or how liberal or conservative one wants to be with including different areas. In general, more conservative thresholds (Hubu (children): FA>0.25, Cimbi (adults): FA>0.3) were used in Papers 1-3.

Figure 8. TBSS skeleton at varying thresholds ranging from non to FA > 0.3 overlaid on the mean FA map. Skeleton voxels with high FA values are coloured yellow, while lower FA values are coloured orange - red. With increasing FA threshold more areas with low FA and/or high inter-subject variability gets excluded.

Next, each subjects aligned FA image is projected onto the mean skeleton by locating the highest local FA value in the direction perpendicular to the skeleton tracts (Fig. 9). ROIs were drawn onto the study-specific skeleton. Since the TBSS procedure projects individual FA values and other DTI derived parameters onto a common framework, i.e., the mean study-specific skeleton, one can extract each subjects individual measures directly by delineating a ROI once.

Figure 9. TBSS skeleton in red overlaid on an individual subjects FA map. For each subject, the aligned FA image is projected onto the study-specific skeleton by locating the voxels with the highest local FA value in the direction perpendicular to the skeleton tracts and assigning the value of these voxels to the skeleton at this standardized location. This results in a mapping of each voxel location in the skeleton to a specific voxel in the individual FA maps (Figure from FSL web page: www.fmrib.ox.ac.uk/fslcourse/lectures/fdt.pdf ).

27

Methods

Inter-subject spatial normalization and MD extraction from grey matter regions

To extract diffusivity measures and volumes from grey matter structures, different tools were used to move between image spaces (native T1, native DWI, and average T1 template). Spatial normalization of grey and white matter and generation of an average T1 template were achieved using DARTEL (Ashburner 2007). Alignment of native T1-weighted images and native DW images were achieved using Fnirt. The data processing has been described in detail in the methods sections in Papers 2 and 4. A schematized overview of the image processing used to extract diffusivity measures volumes from grey matter ROIs in the papers are given in Figure 10.

Figure 10. Schematic overview of the data processing used for extracting diffusivity measures and volumes from grey matter ROIs. MR-images are depicted as black lined boxes. Methods applied to the images are illustrated as blue boxes. Different tools were used to move between spaces. DARTEL was used to move between native T1 space and DARTEL space (Average T1 template), while Fnirt was used to move between native T1 and native DWI space. The main processing steps involved warping the native T1- weighted images into DARTEL space, using calculated deformation flow fields, creating an average study specific T1 template, drawing of ROIs in template space, projecting the ROIs back to native T1 space and then to native DWI space, thresholding the ROI images to exclude potential partial volume voxels, and to extract mean diffusion parameters from the ROIs for each individual (see methods sections in Papers 2 and 4 for a written description of the data processing). Abbreviations: Coreg = coregistered, Corr = corrected, DARTEL = Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra, DTI = diffusion tensor images, DWI = diffusion weighted image, Fnirt = FMRIBs nonlinear image registration tool, GM = grey matter, ROIs = regions-of-interest, T1W = T1-weighted, T2W = T2-weighted, WM = white matter.

28

Methods

Regions-of-interest
All studies included in this thesis were designed to test a priori hypotheses about the relation- ships between the microstructure of specific brain regions, and behavioural or neuroendocrine- logical phenotypes. Given the regional specific a priory hypotheses and to gain power, a region- of-interest (ROI) approach was used in all studies. Neither a whole brain, voxel-wise analysis of the effects (appropriately adjusted for test-multiplicity), nor a restricted voxel-wise analysis with small volume correction was deemed appropriate for testing our a priori hypotheses. To test our hypotheses, mean diffusion parameters (FA, MD, || and ) were extracted from specific ROIs, which are mentioned below for each study. Detailed descriptions of the anatomical location of the ROIs can be found in the methods sections in the papers. In Paper 1, mean FA, || and were extracted from ROIs in the white matter skeleton segments underlying the right IFG (pars opercularis) and right preSMA to test our a priory hypotheses. To assess the anatomical specificity of observed effects in the right ROIs, mean diffusion parameters were extracted from the whole skeleton (used as a measure of global white matter) as well as from ROIs in the white matter underlying the left IFG and left preSMA. In Paper 2, mean MD (and FA) was extracted from skeleton segments within the left and right CSTs and from left and right putamen and caudate nucleus. To assess the anatomical specificity of observed effects, mean MD was extracted from the whole skeleton, and from a combined ROI covering the left and right nucleus accumbens (nAcc) and amygdala. Moreover, left and right putamen and caudate volumes were extracted to determine the degree to which putamen and caudate volumes may have contributed to the observed effects. In Paper 3, mean FA, || and were extracted from skeleton segments within the left and right cingulum and uncinate fasciculus. The data from the left and right ROIs were then used to calculate the asymmetry of these structures: Asymmetry (%) = ((2 * (Left - Right)) / (Left + Right)) * 100 The cingulum and uncinate fasciculus asymmetries were used to test our main hypotheses. To assess the anatomical specificity of observed effect, mean FA was extracted from skeleton segments in the whole left or right hemisphere and used to calculate global hemispheric white matter asymmetry. In Paper 4, mean MD was extracted from left and right hippocampus, and hippocampal MD asymmetry was calculated to test our main hypothesis. To assess the anatomical specificity of observed effects, mean MD was extracted from the left and right amygdala to calculate amygdala MD asymmetry. Furthermore, left and right hippocampal volumes were extracted to determine the degree to which hippocampal volumes may have contributed to the observed effects.

29

Methods

Statistical analyses
Statistical analyses were performed with SPSS software. In all papers, specific hypotheses were tested using multiple linear regression models. The a priori hypotheses for each paper can be found in the Hypotheses section of this thesis, as well as in the papers. Main models were checked for the assumptions of normality (Shapiro-Wilk tests), linearity (scatter plots), no or little multicollinearity (tolerance, variance inflation factor), no outliers (standardised residuals within 3 SDs), homoscedasticity (scatter plots), and absence of autocorrelation (Durbin-Watson tests). All assumptions of multiple linear regression were fulfilled, expect for the assumption of normality in some cases. In these cases, appropriate data transformation yielded normality of the residuals (for details, see Papers 2-4). The statistical tests were performed hierarchically, meaning that we first tested our a priori hypotheses, and next, contingent on observing significant effects in our a priori hypotheses, conducted several planned follow-up analyses to explore the nature of the observed associations. The follow-up analyses were performed to: a) assess the anatomical specificity of observed effects, i.e. the extent to which the phenotypes exhibited a relatively specific relationship to diffusion parameters within the hypothesized ROIs relative to other regions. These analyses were performed by including diffusion parameters from a control region as an additional covariate (together with the hypothesized ROI) in the models predicting the phenotype.

b) assess the relative contribution of the left and right ROIs. This was particularly important in Papers 3 and 4, where the a priori hypotheses concerned asymmetry, since observed effects may have been driven entirely by either the left or the right ROI. These analyses were performed by including both the left and the right ROI simultaneously in the models as two separate predictors of the phenotype, as well as by modelling the left or right ROI individually.

c) assess the degree to which the volume of the grey matter structures may have contributed to the observed MD effects, since the MD effect may have been driven by variability in the volume of a given grey matter structure. These analyses were performed by including the volume of the hypothesized ROI as an additional covariate (together with the MD of this ROI) in the models predicting the phenotype. In addition, the volumes themselves were also assessed in separate models. These follow-up analyses were only performed in Papers 2 and 4.

In addition to these analyses, several other follow-up analyses were performed in each paper (for details see Papers 1-4).

30

Chapter 5
In this chapter, the main findings in the 4 papers will be summarized for each paper separately. For a more detailed description of the results please see Papers 1-4.

Main results

Paper 1: Response inhibition and WM microstructure in children

In this study, we examined the relationship between response inhibition performance and white matter microstructure underlying the right IFG and right preSMA. Results for the a priori hypotheses are summarized in Table 3, where each row represent a separate planned model predicting SSRT. As hypothesized, lower SSRT was significantly associated with higher FA in the right IFG (Fig. 11a) and right preSMA (Fig. 11b) ROIs, adjusted for age (Table 3(1,2)). Moreover, when modelling the right IFG and right preSMA as simultaneous predictors of SSRT, both remained significant predictors of SSRT (Table 3(3)), suggesting that the effects of these ROIs may be additive. Table 3. Linear regression models predicting SSRT from FA: a priori hypotheses* Model 1 2 3 R2 0.192 0.367 0.441 Right IFG FA -0.396 -0.330 p 0.00024 0.0022 Right preSMA FA -0.319 -0.221 p 0.0038 0.037 -0.420 -0.416 -0.396 Age p 0.00011 0.00023 0.00018

*Linear regression models are presented in rows. Predictors are presented in columns. Each model is referred to with a number in the leftmost column, representing the models: 1) Right IFG FA, adjusted for age; 2) Right preSMA FA, adjusted for age; 3) Right IFG and right preSMA, adjusted for age. Abbreviations: SSRT = stopsignal reaction time, IFG = inferior frontal gyrus, PreSMA= presupplementary motor area, FA = fractional anisotropy.

31

Main results

Figure 11. Partial regression plots of the stop-signal reaction time (SSRT) as a function of a) right inferior frontal gyrus (IFG) FA, adjusted for age and b) right presupplementary motor area (preSMA) FA, adjusted for age. Note that residualized measures are plotted. Dotted lines indicate 95% confidence intervals.

Given the significant associations with FA, follow-up analyses were done to assess the

anatomical specificity of the effects by adjusting for either whole skeleton (global white matter) FA, or for FA within the corresponding left ROI. The effects of both right ROIs remained significant in these extended models (Table 4), suggesting the relationships between SSRT and white matter underlying the right IFG and preSMA are not mediated by global increases in FA. The and || were investigated to further explore the nature of the observed FA effects. A

similar relationship with SSRT as found for FA was observed for , but not for || (see Paper 1 for more details). Thus, FA and of the white matter underlying the right IFG and right preSMA exhibit associations with SSRT that are not attributable to age or more global white matter measures of these parameters. Table 4. Linear regression models predicting SSRT from FA: follow-up analyses* Model 1 2

R2 0.383 0.372 R2 0.334 0.328

Right IFG p -0.339 0.0032 -0.372 0.0010 Right PreSMA -0.243 p 0.044

Age p 0.0019 0.0002 -0.361 -0.409

Whole Skeleton -0.153 Whole Skeleton p -0.183 p 0.162 p 0.214

Left IFG -0.076 p 0.486

Model 3 4

Age -0.349 -0.431 0.0037 0.0002

Left PreSMA 0.137 p 0.231

-0.367 0.0019

* Multiple linear regression models are presented in rows. Predictors are presented in columns. Each model is referred to with a number in the leftmost column, representing the models: [1,3] Right ROI, adjusted for age and whole skeleton FA; [2,4] Right ROI, adjusted for age and corresponding left ROI FA. Abbreviations: SSRT = stop signal reaction time, IFG = inferior frontal gyrus, PreSMA =presupplementary motor area, FA = fractional anisotropy.

32

Main results

Paper 2: Microstructure and visuospatial choice RT in children

In this study, we examined the relationship between a simple measure of visuospatial motor function and the microstructure of core motor brain structures. Results from the a priori hypotheses are summarized in Table 5. As hypothesized, faster choice RT was significantly associated with lower MD within the CSTs (Fig. 12a) and the neostriatum (Fig. 12b), adjusted for age, gender and handedness (Table 5(1,2)). The associations appeared stronger, when excluding age from the models (Table 5(1a,2a)). No significant relations were observed for CST FA. Given the significant associations with MD, follow-up analyses were done to assess the anatomical specificity of the effects by including either whole skeleton MD in the model together with the CST MD, or by including a combined nucleus accumbens (nAcc)/amygdala MD measure in the model together with neostriatal MD. Both CST MD and neostriatal MD remained significant predictors of 5-choice RT (Table 5(1b,2b)), suggesting that the effects are not mediated by general decreases in MD. Additional follow-up analyses were performed to assess the relative contributions of putamen and caudate nucleus MD, as well as the degree to which putamen and caudate volumes may have contributed to the observed effects. Faster 5-choice RT was significantly associated with lower MD in both the caudate (=0.271, p=0.017), and the putamen (=0.234, p=0.037). _
__

Table 5. Linear regression models predicting 5-choice reaction time (RT) with mean diffusivity (MD) of the corticospinal tracts (CST) and neostriatum* Model 1 1a 1b Model 2 2a 2b R2 0.400 0.238 0.402 R2 0.424 0.276 0.455 0.310 0.510 0.715 CST MD p 0.004 0.000004 0.002 -0.433 -0.457 Age -0.451 -0.450 p 0.0001 0.0001 Age p <0.0001 <0.0001 Whole skeleton -0.451 p 0.049

Striatal MD 0.282 0.485 0.311 p 0.012 0.00003 0.019

nAcc/amygdala MD -0.052 p 0.672

*Each row in the table represents a separate regression model. Models 1 and 2 were adjusted for age, gender and handedness. Models with a were adjusted for gender and handedness. Models with b were adjusted for age, gender, handedness, and whole skeleton MD or nucleus accumbens (nAcc)/amygdala MD. For each model, R2 is given in the leftmost column, and the regression coefficient () and the significance level (p) are given for each variable included in the model. Gender and handedness effects were non-significant (ps 0.27), and are not included in the table.

33

Main results

Figure 12. Partial regression plots of the 5-choice reaction time (RT) as a function of MD within the (a) corticospinal tracts or (b) neostriatum, adjusted for age, gender and handedness. The regression coefficient () and the significance level (p) for each MD variable are given in the lower right corner of each plot. Note that residualized measures are plotted. Dotted lines indicate 95% confidence intervals.

Moreover, both caudate MD (=0.261, p=0.018) and putamen MD (=0.231, p=0.046) remained significant predictors of 5-choice RT when including, respectively, caudate volume (p=0.51) and putamen volume (p=0.91) in the models, suggesting that the association between neostriatal MD and 5-choice RT was not mediated by differences in striatal volumes. By themselves, neither caudate volume, nor putamen volume was significantly associated with 5-choice RT (ps0.49, see Paper 2 for more details). Thus, MD in the CSTs, putamen, and caudate nuclei exhibit associations with 5-choice RT that are not attributable to age or more general maturational

decrease in MD across the brain.

Paper 3: CAR and neuroticism links to limbic fibre bundle asymmetry

In this study, we examined the relationship between CAR, neuroticism and major limbic fibre bundle asymmetry. Results from the a priori hypotheses regarding CA are presented in Table 6. As predicted, CAR was significantly associated with cingulum FA asymmetry and with uncinate fasciculus FA asymmetry (Table 6). Follow-up analyses assessed the relative contributions of left and right ROI FA. When including both the left and right ROIs in the same model, left cingulum FA was negatively (=-0.662, p=0.050) and right cingulum FA was positively (=0.687, p=0.039) associated with CAR. In contrast, left uncinate fasciculus ROI FA was positively (=0.546, p=0.035) and right ROI FA was negatively (=-0.441, p=0.108) associated with CAR, when modelled together. Notably, no significant correlations with CAR were found when modelling the left or right ROIs separately (ps0.159), indicating that it is the relationship between the left and right ROIs that exhibited associations with CAR, and not the left or right ROIs individually.

34

Main results Table 6. Results for the a priori hypotheses predicting cortisol awakening response or neuroticism with cingulum or uncinate fasciculus fractional anisotropy (FA) asymmetry. Cortisol Awakening Response (n=48)* Cingulum FA asymmetry Uncinate FA asymmetry Neuroticism (n=69)** Cingulum FA asymmetry Uncinate FA asymmetry -0.357 -0.063 0.002 0.591 -0.288 0.293 0.045 0.039 p

*Models predicting cortisol measures were adjusted for age, gender, time-interval between awakening and the first cortisol sample, and sample position of the cortisol morning peak (1st vs. 2nd-4th position). **Models predicting neuroticism were adjusted for age, and gender.

In an additional follow-up analysis, in which cingulum and uncinate fasciculus FA asymmetries were modelled simultaneously to test for possible additive effects, both ROI FA asymmetries exhibited relatively independent relationships with CAR that approached significance (R2=0.298, cingulum: =-0.256, p=0.067; uncinate fasciculus: =0.262, p=0.059). Follow-up analysis, testing the anatomical specificity of the observed FA asymmetry effects, showed that both cingulum (=-0.314, p=0.034) and uncinate fasciculus (=0.320, p=0.040) FA asymmetries remained significant predictors of CAR, when including global white matter hemispheric FA asymmetry (ps0.40) in the models, suggesting that the associations between the limbic FA asymmetries and CAR were not mediated by a general hemispheric asymmetry Results from the a priori hypothesis regarding neuroticism are presented in Table 6. As hypothesized, neuroticism was significantly associated with cingulum FA asymmetry, adjusted for age and gender (Table 6). Such a relationship, however, was not observed for uncinate fasciculus FA asymmetry (Table 6). Follow-up analyses showed that left cingulum FA was negatively (=-0.800, p=0.002) and right cingulum FA was positively (=0.722, p=0.003) associated with neuroticism, when modelled simultaneously. Models including only left or right cingulum FA yielded no significant effects on neuroticism (ps0.264), again suggesting that it is the relationship between left and right cingulum FA that mediated the association with neuroticism, and not FA in left or right cingulum individually. When assessing the anatomical specificity of the cingulum effect, cingulum FA asymmetry (=-0.363, p=0.002) continued to be significantly associated with neuroticism with global hemispheric FA asymmetry (p=0.83) included in the model. Finally, corroborating previous findings (Portella et al. 2005), higher neuroticism scores were found to be significantly associated with higher CAR (=0.309, p=0.029). An overview of the results is given in Figure 13.

35

Main results

Figure 13. Schematic overview of the direction of the significant associations between neuroticism, cortisol awakening response (CAR), and FA asymmetry in the cingulum (blue) and uncinate fasciculus (magenta). Positive associations are depicted with a + imbedded in green lines. Negative associations are depicted with a - imbedded in red lines. Neuroticism was negatively associated with cingulum FA asymmetry, with higher neuroticism scores correlating with higher right cingulum FA relative to left cingulum FA. Similar to neuroticism, CAR was negatively associated with cingulum FA asymmetry, i.e. higher CAR correlated with higher right cingulum FA relative to left cingulum FA. In contrast, CAR was associated positively with uncinate fasciculus FA asymmetry, with higher CAR correlating with lower right relative to left uncinate fasciculus FA. Finally, neuroticism and CAR exhibited a positive relationship.

Paper 4: HPA-axis tonus and hippocampal microstructural asymmetry

In this study, we examined the relationship between mean morning and afternoon/evening (PM) cortisol levels, and hippocampal microstructural asymmetry. Results from the a priori hypotheses are presented in Table 7. As hypothesized, both mean morning (Fig. 14a) and PM (Fig. 14b) cortisol levels were significantly associated with hippocampus MD asymmetry (Table 7). Follow-up analyses assessed the relative contributions of left and right hippocampus MD. When entered in the same model, left hippocampus MD was positively (=0.572, p=0.011) and right hippocampus MD was negatively (=-0.459, p=0.029) associated with mean morning cortisol level. Likewise, mean PM cortisol level was positively associated with left hippocampus MD (=0.656, p=0.005) and negatively associated with right hippocampus MD (=-0.445, p=0.039), when the left and right were modelled simultaneously. No significant correlations with mean morning or PM cortisol levels were found when left or right hippocampus MD were modelled separately (ps0.14), though a positive association between mean PM cortisol level and left hippocampus MD (=0.352, p=0.051) approached significance. These findings indicate that it is the relationship between left and right hippocampus MD that exhibited an association with mean morning and PM cortisol levels, and not left or right hippocampus MD individually.

36

Main results Table 7. Results from a priori hypotheses and follow-up analyses predicting cortisol measures. R2 0.286 0.295 0.217 0.234 Hippocampus MD asymmetry 0.373 0.366 0.418 0.409 P 0.008 0.010 0.005 0.006 Hippocampal volume asymmetry 0.106 0.139 P 0.455 0.348

Mean morning cortisol level (N=48)*

Mean PM cortisol level (N=47)**

*In models predicting mean morning cortisol level, effects of hippocampus MD asymmetry were tested after adjusting for age, age2, gender, time interval between awakening and taking the first cortisol sample (and hippocampal volume asymmetry). **In models predicting mean PM cortisol level, effects of hippocampus MD asymmetry were adjusted for age, age2, and gender (and hippocampal volume asymmetry).

Assessing the anatomical specificity of the hippocampus MD asymmetry effects, revealed that hippocampus MD asymmetry remained a significant predictor of mean morning (=0.426, p=0.003) and PM (=0.426, p=0.006) cortisol levels when including amygdala MD asymmetry (morning: =-0.265, p=0.057; PM: =-0.043, p=0.77) in the models, suggesting that the hippocampus MD asymmetry effects are not mediated by a general limbic MD asymmetry. Follow-up analyses also assessed the degree to which hippocampal volume may have contributed to the observed effects. The associations between hippocampus MD asymmetry and the mean morning and PM cortisol levels remained significant when including hippocampal volume asymmetry in the models (Table 7). No significant associations were observed between hippocampal volumes and mean cortisol levels (ps0.11, see Paper 4 for further details).

Figure 14. Partial regression plots of the main findings. Left) Mean morning cortisol level is plotted as a function of hippocampus mean diffusivity (MD) asymmetry, adjusted for age, age2, gender, time-interval between awakening and first cortisol sample. Right) The logarithm-transformed mean afternoon/evening (PM) cortisol level plotted as a function of hippocampus MD asymmetry, adjusted for age, age2, and gender. Note that residualized measures are plotted. Dotted lines indicate 95% confidence intervals.

37

38

Chapter 6
The main findings in the papers will be discussed in general terms in the following. A more detailed discussion of the findings can be found in the individual papers. The two Hubu papers (1 and 2) will be discussed together as will the two Cimbi papers (3 and 4), in that the discussions of the findings within the two projects are highly related.

General discussion

Hubu papers
In the two Hubu papers (Papers 1 and 2), behavioural performance on two different tasks was found to be associated with regional specific microstructure in children aged 7 to 13 years. Specifically, in Paper 1 we found that better response inhibition performance (lower SSRT) was significantly associated with higher FA in the white matter underlying the right IFG and right preSMA, after adjusting for age effects. In Paper 2, we found that lower CST and neostriatal MD, adjusted for age, gender, and handedness, were significantly associated with faster 5-choice RTs on a basic visuospatial motor task.

The importance of adjusting for age effects

Both microstructure (Eluvathingal et al. 2007; Lebel et al. 2008; Snook et al. 2005), response inhibition (Williams et al., 1999) and choice RTs (Kail 1991) continue to develop throughout childhood and adolescence. Therefore, a correlation between these measures could represent a nonspecific association attributable to unmeasured factors indexed by chronological age. For example, since height also increases in children over this age range, one might expect to find a simple correlation between height and response inhibition within a group of children with varying ages. However, as height has little direct relationship to cognitive functioning in children, one would expect the correlation between these variables to be substantially reduced after controlling for age statistically. Thus, all our hypotheses were based on the assumption that

39

General discussion

associations between behavioural performance and FA/MD within specific regions would remain significant after controlling for age; i.e., that even among children of similar age, those with higher FA or lower MD in the specific hypothesized regions would exhibit stronger response inhibition and 5-choice RT performance. These hypotheses were confirmed.

Anatomical specificity
In follow-up analyses, we addressed the question whether the associations between behavioural performance and regional microstructure (response inhibition: right IFG and preSMA, 5-choice RT: CSTs and neostriatum) were anatomically specific, i.e., the extent to which response inhibition and 5-choice RT performance exhibited a relatively specific relationship to FA/MD in these regions relative to others. This is an important question, since FA increases and MD decreases concurrently in many brain regions during childhood and adolescence (Colby et al. 2011; Eluvathingal et al. 2007; Lebel et al. 2008; Snook et al. 2005), and, thus the observed effects could be reflecting more global maturational differences in FA/MD. However, the function-structure relationships remained significant when including control regions (measures of global white matter or other grey matter regions) as additional covariates in the models. Therefore, the observed relationships between behavioural performance and regional FA/MD within the motor and response inhibition networks are not likely to be driven by global FA/MD differences, but may be more strongly related to the microstructure of the specific regions.

Neurobiological interpretation of the findings

Our findings suggest that even among children of similar age, higher FA in the white matter underlying the right IFG and preSMA is associated with better response inhibition performance, and lower MD within core motor structures is associated with faster 5-choice RTs. Questions remain regarding the neurobiological interpretation of such age-adjusted relationships between individual variability in regional microstructure and behavioural performance. The observed structure-function relationships might reflect individual differences in the maturational trajectories of the neural system subserving response inhibition and visuospatial motor function. In other words, there may be differences in the phase of fibre tract and neostriatal maturation among children of similar age (see Fig. 15). It is plausible that such a variability could mediate, at least to some extent, the associations found in the two Hubu studies, since both microstructure (Eluvathingal et al. 2007; Lebel et al. 2008; Snook et al. 2005) and RTs (Kail 1991) continue to develop across this age range. Alternatively, the variability in FA and MD may reflect individual differences in the underlying neural system (perhaps differences in the underlying connectivity) that emerge earlier during brain development and remain stable in spite of superimposed biological changes

40

General discussion associated with maturation (Fig. 15). This is plausible since individual differences in behavioural performance have been associated with FA and MD variability in adults (Forstmann et al. 2008; Gold et al. 2007; Konrad et al. 2009; Piras et al. 2010; Tuch et al. 2005) as well as children. It is also possible that dynamic processes, perhaps associated with activity levels in the neural circuits throughout childhood could influence the microstructure within the motor and response inhibition networks. A recent study of adults examined changes in DTI parameters in connection with 6 weeks of practice in a complex whole-body balancing task, observing negative correlations between improvements in motor performance and MD changes in bilateral anterior centrum semiovale, left brain stem, and right internal capsule (Taubert et al. 2010). Moreover, another DTI study found accelerated white matter development (higher FA) in the sagittal stratum in preterm relative to full-term infants at term equivalent age, possibly as a result of increased intensity of sensorimotor stimulation associated with extrauterine life (Gimenez et al. 2008). Thus, the individual differences observed in the present study could partly reflect variability in the experiences, learning and skill acquisition of the children.
Figure 15. Hypothetic examples of three different maturational trajectories of regional FA. Our findings that even among children of similar age, those with higher regional FA have better response inhibition performance which may reflect differences in the phase of maturation within these regions. This is illustrated with the black and blue lines that have the same starting and ending points but where the black line has a steeper maturational trajectory than the blue. Our findings may also reflect stable individual differences, as illustrated with the parallel black and red lines where the red line always has higher FA than the black.

Longitudinal studies are needed to help distinguish between these, and other explanations.

Cimbi papers
In the two Cimbi papers (Papers 3 and 4), we present novel findings relating cortisol measures and neuroticism with microstructural asymmetry in limbic structures. Specifically, higher CAR was negatively associated with cingulum FA asymmetry, i.e., increased right relative to left FA, and positively associated with uncinate fasciculus FA asymmetry, i.e., decreased right relative to left FA. Higher neuroticism scores were correlated with higher CAR, and with diminished cingulum FA asymmetry. Moreover, higher mean morning and PM cortisol levels were positively associated with hippocampus MD asymmetry, i.e., decreased right relative to left MD. Importantly, these associations were mediated by a discrepancy between diffusion parameters of the left and right limbic structures, and not driven by diffusion parameters in left or right

41

General discussion

limbic structures individually.

Neurobiological interpretation of the findings

At present, we can only speculate about the neurobiological significance of the observed relationships between the cortisol measures and the limbic microstructural asymmetries. The present asymmetry findings may reflect a) differences in the functional effect of left and right limbic structures on HPA-axis regulation, b) differences in how varying levels of secreted cortisol affect the left and right limbic microstructure, c) an indirect association that is mediated by other biological factors, d) some combination of these possibilities, or e) that both limbic asymmetries and HPA-axis activity are associated with unrelated, but correlated, factors. There is some evidence consistently with possibility a that left and right limbic structures may have differential regulatory effects on the HPA-axis. Prefrontal and limbic structures have high densities of corticoid receptors, and are thus, thought to be implicated in exerting negative feedback of the HPA-axis (Reul et al. 2000; Sanchez et al. 2000). Some studies support the interpretation of differential regulatory effects of left- and right-sided structures. Stroke patients with left-sided, but not right-sided, infarctions have been found to have increased morning cortisol levels compared to controls (Lueken et al. 2009). Another study found that healthy subjects with impaired cortisol suppression in response to dexamethasone had smaller left anterior cingulate gyrus volumes and substantially larger right than left anterior cingulate gyrus volumes as compared to suppressors (MacLullich et al. 2006). Furthermore, a lesion study in rodents found that right but not left medial prefrontal cortex lesions significantly suppressed stress-induced increases in plasma corticosterone levels in repeatedly restrained animals (Sullivan and Gratton 1999). Further, in mice the density of mineralocorticoid receptors (MRs) was observed to be higher in the right than in the left hippocampus (Neveu et al. 1998). Note that the latter is not only consistent with possibility a, but may also relate to possibility b. Conversely, there is also evidence consistently with the b possibility that limbic microstructural asymmetry could result from differential effects on left- and right-sided limbic structures of prolonged periods of elevated cortisol secretion. Some previous studies support this interpretation. Higher afternoon salivary cortisol levels have been associated with decreased volume and neuron number in the left, but not the right, dentate gyrus in chronic- restrained (stressed) pigs (van der Beek et al. 2004). Further, in a recent study (Zach et al. 2010) 3 weeks of corticosterone treatment in rats was associated with alterations of hippocampal volume asymmetry as well as measures of the number of neurons in hippocampal subfields. Moreover, a study in rats found that chronic treatment with high doses of dexamethasone significantly reduced the volume of the left anterior cingulate gyrus, and to a lesser extent the right, suggesting that left anterior cingulate may be more vulnerable to prolonged high cortisol

42

General discussion levels than the right (Cerqueira et al. 2005). It may well be that these effects also are reflected in the white matter.

Differences in brain microstructural correlates of tonic and phasic HPA-axis activity

The findings in Paper 3 suggest that cingulum and uncinate fasciculus microstructural asymmetries are mainly associated with the dynamics of CAR, but not with HPA-axis tonus, while the findings in Paper 4 suggest that hippocampal MD asymmetry may be specifically related to tonic and not phasic HPA-axis activity (for details see Papers 3 and 4). It is not clear what biological mechanisms might underlie these different observations. Cortisol primarily acts through MRs and glucocorticoid receptors (GRs). MRs have a 5-10 fold higher affinity for cortisol than the GRs, hence during basal cortisol levels mainly the MRs are occupied (De Kloet et al. 1998; Herman et al. 2005; Reul et al. 2000). Thus, tonic inhibitory control of the HPA-axis is thought to be mediated by MRs (De Kloet et al. 1998; Herman et al. 2005; Reul et al. 2000). High densities of MRs are only found in the hippocampus and dorsolateral septum (Reul et al. 2000; Sanchez et al. 2000; Seckl et al. 1991), suggesting that hippocampus is one of the key brain structures in regulating the basal cortisol levels. The GRs become increasingly involved as cortisol levels increase in relation with the circadian peak or in association with stress (De Kloet et al. 1998; Herman et al. 2005; Reul et al. 2000). GRs are expressed in a wide distribution throughout the brain, e.g., in hippocampus, pituitary, paraventricular nucleus, and neocortex (Sanchez et al. 2000), suggesting that several different brain structures may be involved in exerting negative feedback of the HPA-axis when cortisol levels are high. Our observation of a relationship between hippocampal MD asymmetries and tonic, but not phasic, HPA-axis activity, and between cingulum and uncinate fasciculus asymmetries and phasic, but not tonic, activity may relate to differences in the anatomical distribution of MRs and GRs; however further research is needed to investigate this possibility.

Link to affective disorders?

The two Cimbi papers showed that limbic microstructural asymmetries were associated with higher HPA-axis activity and neuroticism. There is some evidence that asymmetries in the limbic system are related to affective disorders (Drevets et al. 2008; Frodl et al. 2002; Kim et al. 2006; Versace et al. 2008). A recent study found that subjects with bipolar disorder had higher left and lower right uncinate fasciculus FA relative to control subjects (Versace et al. 2008). Another study found that subjects with PTSD had lower left cingulum FA compared to controls. Further, on average the PTSD group had higher right than left cingulum FA, while the control group had higher left relative to right cingulum FA (Kim et al. 2006). Moreover, a study found that patients with first episode major depression had a smaller left than right hippocampus, while no

43

General discussion

significant differences were observed in the healthy control subjects (Frodl et al. 2002). In addition, disruption of the diurnal profile of the HPA-axis has been found in psychiatric disorders, such as post-traumatic stress disorder (Chida and Steptoe 2009; Handwerger 2009), anxiety disorder (Vreeburg et al. 2010) and major depression (Aubry et al. 2010; Chida and Steptoe 2009; Handwerger 2009; Vreeburg et al. 2009). Further, higher CAR and basal cortisol levels as well as higher neuroticism scores have been linked to an increased risk of developing affective disorders (Adam et al. 2010; Kendler et al. 2006; Kendler et al. 2004). Since both increased HPA-axis activity and negative emotionality may increase the risk of developing affective disorders, one might speculate whether limbic microstructural asymmetries are brain structural markers for a predisposition or risk for developing affective disorders. However, further studies are needed to address this question.

44

Chapter 7
In this final chapter, it is discussed how data collected in the Hubu cohort can be used to address some of the questions that have arisen from the studies in the present thesis.

Future work

Future work
So far, Hubu has collected data from 7 time-points with 6-months interval in approximately 70 children and adolescents. This longitudinal dataset will help us to determine whether individual differences among children in the course of their cognitive development are mirrored by individual differences in the maturational trajectories of relevant neural systems, and how such associations in developmental trajectories may relate to and modify associations with other, more stable, neuroarchitectural attributes. Adolescence is associated with an increased incidence of neuropsychiatric disorders, such

as mood, anxiety and substance use disorders (for review see Paus et al. 2008). In Hubu, we have collected cortisol data from one time point and longitudinal data on neuroticism and basic emotional processing. This allows us to examine whether the relationships between limbic microstructural asymmetries, and HPA-axis activity and negative emotionality are present in children and adolescents, or whether these relationships emerge during this age period. In relation to this, longitudinal data on stressful life events have been collected, allowing us to examine how environmental stress may relate to and modify such associations. Finally, in Hubu we also collected data on genetics, and sex hormones. This allows us to

examine how genetic variability as well as variability in hormonal levels link to differences in individual maturational trajectories and developing cognitive and emotional functions.

45

46

References
Adam EK, Doane LD, Zinbarg RE, Mineka S, Craske MG, Griffith JW (2010) Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 35:921-31. Aron AR, Behrens TE, Smith S, Frank MJ, Poldrack RA (2007a) Triangulating a cognitive control network using diffusion-weighted magnetic resonance imaging (MRI) and functional MRI. J.Neurosci. 27:3743-52. Aron AR, Durston S, Eagle DM, Logan GD, Stinear CM, Stuphorn V (2007b) Converging evidence for a fronto-basal-ganglia network for inhibitory control of action and cognition. J.Neurosci. 27:11860-4. Aron AR, Fletcher PC, Bullmore ET, Sahakian BJ, Robbins TW (2003) Stop-signal inhibition disrupted by damage to right inferior frontal gyrus in humans. Nat.Neurosci. 6:115-6. Aron AR, Poldrack RA (2006) Cortical and subcortical contributions to Stop signal response inhibition: role of the subthalamic nucleus. J.Neurosci. 26:2424-33. Aron AR, Robbins TW, Poldrack RA (2004) Inhibition and the right inferior frontal cortex. Trends Cogn Sci. 8:170-7. Ashburner J (2007) A fast diffeomorphic image registration algorithm. Neuroimage 38:95-113. Aubry JM, Jermann F, Gex-Fabry M, Bockhorn L, Van der Linden M, Gervasoni N, Bertschy G, Rossier MF, Bondolfi G (2010) The cortisol awakening response in patients remitted from depression. J Psychiatr Res 44:1199-204. Barnea-Goraly N, Menon V, Eckert M, Tamm L, Bammer R, Karchemskiy A, Dant CC, Reiss AL (2005) White matter development during childhood and adolescence: a cross- sectional diffusion tensor imaging study. Cereb.Cortex 15:1848-54. Basser PJ, Mattiello J, LeBihan D (1994) MR diffusion tensor spectroscopy and imaging. Biophys.J. 66:259-67. Beaulieu C (2002) The basis of anisotropic water diffusion in the nervous system - a technical review. NMR Biomed. 15:435-55. Beaulieu C. 2009. The biological basis of diffusion anisotropy. In Diffusion MRI: from quantitative measurement to in-vivo neuroanatomy, ed. H Johansen-Berg, TEJ Behrens. London, UK: Elsevier Bienvenu OJ, Nestadt G, Samuels JF, Costa PT, Howard WT, Eaton WW (2001) Phobic, panic, and major depressive disorders and the five-factor model of personality. J Nerv Ment Dis 189:154-61. Boehler CN, Appelbaum LG, Krebs RM, Hopf JM, Woldorff MG (2010) Pinning down response inhibition in the brain--conjunction analyses of the Stop-signal task. Neuroimage 52:1621-32. Boucher L, Palmeri TJ, Logan GD, Schall JD (2007) Inhibitory control in mind and brain: an interactive race model of countermanding saccades. Psychol.Rev. 114:376-97. Buchanan TW, Kern S, Allen JS, Tranel D, Kirschbaum C (2004) Circadian regulation of cortisol after hippocampal damage in humans. Biol Psychiatry 56:651-6. Cerqueira JJ, Almeida OF, Sousa N (2008) The stressed prefrontal cortex. Left? Right! Brain Behav Immun 22:630-8. Cerqueira JJ, Catania C, Sotiropoulos I, Schubert M, Kalisch R, Almeida OF, Auer DP, Sousa N (2005) Corticosteroid status influences the volume of the rat cingulate cortex - a magnetic resonance imaging study. J Psychiatr Res 39:451-60.

47

References

Chambers CD, Bellgrove MA, Stokes MG, Henderson TR, Garavan H, Robertson IH, Morris AP, Mattingley JB (2006) Executive "brake failure" following deactivation of human frontal lobe. J.Cogn Neurosci. 18:444-55. Chambers CD, Garavan H, Bellgrove MA (2009) Insights into the neural basis of response inhibition from cognitive and clinical neuroscience. Neurosci.Biobehav.Rev. 33:631-46. Chang LC, Jones DK, Pierpaoli C (2005) RESTORE: robust estimation of tensors by outlier rejection. Magn Reson.Med. 53:1088-95. Chenevert TL, Sundgren PC, Ross BD (2006) Diffusion imaging: insight to cell status and cytoarchitecture. Neuroimaging Clin N Am 16:619-32, viii-ix. Chida Y, Steptoe A (2009) Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biol Psychol 80:265-78. Cohen JR, Asarnow RF, Sabb FW, Bilder RM, Bookheimer SY, Knowlton BJ, Poldrack RA (2010) Decoding developmental differences and individual variability in response inhibition through predictive analyses across individuals. Front Hum Neurosci 4:47. Colby JB, Van Horn JD, Sowell ER (2011) Quantitative in vivo evidence for broad regional gradients in the timing of white matter maturation during adolescence. Neuroimage 54:25-31. Costa PT, McCrae RR 1992 Professional Manual for Revised NEO Personality Inventory. Odessa, Florida: Psychological Assessment Resources Cremers HR, Demenescu LR, Aleman A, Renken R, van Tol MJ, van der Wee NJ, Veltman DJ, Roelofs K (2010) Neuroticism modulates amygdala-prefrontal connectivity in response to negative emotional facial expressions. Neuroimage 49:963-70. Czeh B, Muller-Keuker JI, Rygula R, Abumaria N, Hiemke C, Domenici E, Fuchs E (2007) Chronic social stress inhibits cell proliferation in the adult medial prefrontal cortex: hemispheric asymmetry and reversal by fluoxetine treatment. Neuropsychopharmacology 32:1490-503. De Kloet ER, Vreugdenhil E, Oitzl MS, Joels M (1998) Brain corticosteroid receptor balance in health and disease. Endocr Rev 19:269-301. Dedovic K, Duchesne A, Andrews J, Engert V, Pruessner JC (2009) The brain and the stress axis: the neural correlates of cortisol regulation in response to stress. Neuroimage 47:864-71. DeYoung CG, Hirsh JB, Shane MS, Papademetris X, Rajeevan N, Gray JR (2010) Testing predictions from personality neuroscience: brain structure and the big five. Psychol Sci 21:820-8. Drevets WC, Savitz J, Trimble M (2008) The subgenual anterior cingulate cortex in mood disorders. CNS Spectr 13:663-81. Dubois J, Dehaene-Lambertz G, Soares C, Cointepas Y, Le Bihan D, Hertz-Pannier L (2008) Microstructural correlates of infant functional development: example of the visual pathways. J.Neurosci. 28:1943-8. Durston S, Casey BJ (2006) What have we learned about cognitive development from neuroimaging? Neuropsychologia 44:2149-57. Eagle DM, Baunez C, Hutcheson DM, Lehmann O, Shah AP, Robbins TW (2008) Stop- signal reaction-time task performance: role of prefrontal cortex and subthalamic nucleus. Cereb.Cortex 18:178-88. Edgar JM, Griffiths IR. 2009. White matter structure: A microscopist's view. In Diffusion MRI: From quantitative measurements to in vivo neuroanatomy, ed. H Johansen- Berg, TEJ Behrens, pp. 75-104. London, UK: Elsevier

48

References

Eluvathingal TJ, Hasan KM, Kramer L, Fletcher JM, Ewing-Cobbs L (2007) Quantitative diffusion tensor tractography of association and projection fibers in normally developing children and adolescents. Cereb.Cortex 17:2760-8. Floden D, Stuss DT (2006) Inhibitory control is slowed in patients with right superior medial frontal damage. J.Cogn Neurosci. 18:1843-9. Forstmann BU, Jahfari S, Scholte HS, Wolfensteller U, van den Wildenberg WP, Ridderinkhof KR (2008) Function and structure of the right inferior frontal cortex predict individual differences in response inhibition: a model-based approach. J.Neurosci. 28:9790-6. Fries E, Dettenborn L, Kirschbaum C (2009) The cortisol awakening response (CAR): facts and future directions. Int J Psychophysiol 72:67-73. Frodl T, Meisenzahl EM, Zetzsche T, Born C, Groll C, Jager M, Leinsinger G, Bottlender R, Hahn K, Moller HJ (2002) Hippocampal changes in patients with a first episode of major depression. Am J Psychiatry 159:1112-8. Frokjaer VG, Mortensen EL, Nielsen FA, Haugbol S, Pinborg LH, Adams KH, Svarer C, Hasselbalch SG, Holm S, Paulson OB, Knudsen GM (2008) Frontolimbic serotonin 2A receptor binding in healthy subjects is associated with personality risk factors for affective disorder. Biol Psychiatry 63:569-76. Gibbs P, Liney GP, Pickles MD, Zelhof B, Rodrigues G, Turnbull LW (2009) Correlation of ADC and T2 measurements with cell density in prostate cancer at 3.0 Tesla. Invest Radiol 44:572-6. Giedd JN, Blumenthal J, Jeffries NO, Castellanos FX, Liu H, Zijdenbos A, Paus T, Evans AC, Rapoport JL (1999) Brain development during childhood and adolescence: a longitudinal MRI study. Nat.Neurosci. 2:861-3. Gimenez M, Miranda MJ, Born AP, Nagy Z, Rostrup E, Jernigan TL (2008) Accelerated cerebral white matter development in preterm infants: a voxel-based morphometry study with diffusion tensor MR imaging. Neuroimage. 41:728-34. Gogtay N, Giedd JN, Lusk L, Hayashi KM, Greenstein D, Vaituzis AC, Nugent TF, III, Herman DH, Clasen LS, Toga AW, Rapoport JL, Thompson PM (2004) Dynamic mapping of human cortical development during childhood through early adulthood. Proc.Natl.Acad.Sci.U.S.A 101:8174-9. Gold BT, Powell DK, Xuan L, Jiang Y, Hardy PA (2007) Speed of lexical decision correlates with diffusion anisotropy in left parietal and frontal white matter: evidence from diffusion tensor imaging. Neuropsychologia 45:2439-46. Haas BW, Omura K, Constable RT, Canli T (2007) Emotional conflict and neuroticism: personality-dependent activation in the amygdala and subgenual anterior cingulate. Behav Neurosci 121:249-56. Haber SN, Gdowski MJ. 2004. The basal ganglia. In The human nervous system, ed. G Paxinos, JK Mai. London: Elsevier Handwerger K (2009) Differential patterns of HPA activity and reactivity in adult posttraumatic stress disorder and major depressive disorder. Harv Rev Psychiatry 17:184-205. Hansen HS, Mortensen EL, Schioez HK 2004 Dokumentation for den danske udgace af NEO PI-R og NEO-FFI. Copenhagen, Denmark: Dansk psykologisk forlag Herman JP, Ostrander MM, Mueller NK, Figueiredo H (2005) Limbic system mechanisms of stress regulation: hypothalamo-pituitary-adrenocortical axis. Prog Neuropsychopharmacol Biol Psychiatry 29:1201-13.

49

References

Huang L, Smith A, Badaut J, Obenaus A (2010) Dynamic characteristics of 56Fe-particle radiation-induced alterations in the rat brain: magnetic resonance imaging and histological assessments. Radiat Res 173:729-37. Iidaka T, Matsumoto A, Ozaki N, Suzuki T, Iwata N, Yamamoto Y, Okada T, Sadato N (2006) Volume of left amygdala subregion predicted temperamental trait of harm avoidance in female young subjects. A voxel-based morphometry study. Brain Res 1125:85-93. Inase M, Tokuno H, Nambu A, Akazawa T, Takada M (1999) Corticostriatal and corticosubthalamic input zones from the presupplementary motor area in the macaque monkey: comparison with the input zones from the supplementary motor area. Brain Res. 833:191-201. Jang KL, Livesley WJ, Vernon PA (1996) Heritability of the big five personality dimensions and their facets: a twin study. J Pers 64:577-91. Jankord R, Herman JP (2008) Limbic regulation of hypothalamo-pituitary-adrenocortical function during acute and chronic stress. Ann N Y Acad Sci 1148:64-73. Jauregui-Huerta F, Ruvalcaba-Delgadillo Y, Gonzalez-Castaneda R, Garcia-Estrada J, Gonzalez-Perez O, Luquin S (2010) Responses of glial cells to stress and glucocorticoids. Curr Immunol Rev 6:195-204. Jernigan TL, Baare WFC, Stiles J, Madsen KS (2011) Postnatal brain development: structural imaging of dynamic neurodevelopmental processes. Prog Brain Res Accepted Jernigan TL, Trauner DA, Hesselink JR, Tallal PA (1991) Maturation of human cerebrum observed in vivo during adolescence. Brain 114 ( Pt 5):2037-49. Johansen-Berg H, Behrens TE, Robson MD, Drobnjak I, Rushworth MF, Brady JM, Smith SM, Higham DJ, Matthews PM (2004) Changes in connectivity profiles define functionally distinct regions in human medial frontal cortex. Proc.Natl.Acad.Sci.U.S.A 101:13335-40. Jones DK. 2009. Gaussian modeling of the diffusion signal. In Diffusion MRI: From quantitative measurement to in vivo neuroanatomy ed. H Johansen-Berg, TEJ Behrens, pp. 37-54. London, UK: Elsevier Kail R (1991) Developmental change in speed of processing during childhood and adolescence. Psychol Bull 109:490-501. Kalin NH, Larson C, Shelton SE, Davidson RJ (1998) Asymmetric frontal brain activity, cortisol, and behavior associated with fearful temperament in rhesus monkeys. Behav Neurosci 112:286-92. Kendler KS, Gatz M, Gardner CO, Pedersen NL (2006) Personality and major depression: a Swedish longitudinal, population-based twin study. Arch Gen Psychiatry 63:1113-20. Kendler KS, Kuhn J, Prescott CA (2004) The interrelationship of neuroticism, sex, and stressful life events in the prediction of episodes of major depression. Am J Psychiatry 161:631-6. Kim MJ, Whalen PJ (2009) The structural integrity of an amygdala-prefrontal pathway predicts trait anxiety. J Neurosci 29:11614-8. Kim SJ, Jeong DU, Sim ME, Bae SC, Chung A, Kim MJ, Chang KH, Ryu J, Renshaw PF, Lyoo IK (2006) Asymmetrically altered integrity of cingulum bundle in posttraumatic stress disorder. Neuropsychobiology 54:120-5. Knoops AJ, Gerritsen L, van der Graaf Y, Mali WP, Geerlings MI (2010) Basal hypothalamic pituitary adrenal axis activity and hippocampal volumes: the SMART-Medea study. Biol Psychiatry 67:1191-8.

50

References

Konrad A, Vucurevic G, Musso F, Stoeter P, Winterer G (2009) Correlation of brain white matter diffusion anisotropy and mean diffusivity with reaction time in an oddball task. Neuropsychobiology 60:55-66. Kremen WS et al (2010) Salivary cortisol and prefrontal cortical thickness in middle- aged men: A twin study. Neuroimage Lebel C, Walker L, Leemans A, Phillips L, Beaulieu C (2008) Microstructural maturation of the human brain from childhood to adulthood. Neuroimage. 40:1044-55. Lo MS, Ng ML, Azmy BS, Khalid BA (1992) Clinical applications of salivary cortisol measurements. Singapore Med J 33:170-3. Logan GD, Cowan WB (1984) On the Ability to Inhibit Thought and Action: A theory of an Act of Control Psychological Review 91:295-327. Lueken U, Leisse M, Mattes K, Naumann D, Wittling W, Schweiger E (2009) Altered tonic and phasic cortisol secretion following unilateral stroke. Psychoneuroendocrinology 34:402-12. Lupien SJ, de Leon M, de Santi S, Convit A, Tarshish C, Nair NP, Thakur M, McEwen BS, Hauger RL, Meaney MJ (1998) Cortisol levels during human aging predict hippocampal atrophy and memory deficits. Nat Neurosci 1:69-73. MacLullich AM, Ferguson KJ, Wardlaw JM, Starr JM, Deary IJ, Seckl JR (2006) Smaller left anterior cingulate cortex volumes are associated with impaired hypothalamic- pituitary-adrenal axis regulation in healthy elderly men. J Clin Endocrinol Metab 91:1591-4. Madden DJ, Whiting WL, Huettel SA, White LE, MacFall JR, Provenzale JM (2004) Diffusion tensor imaging of adult age differences in cerebral white matter: relation to response time. Neuroimage 21:1174-81. Madler B, Drabycz SA, Kolind SH, Whittall KP, MacKay AL (2008) Is diffusion anisotropy an accurate monitor of myelination? Correlation of multicomponent T2 relaxation and diffusion tensor anisotropy in human brain. Magn Reson.Imaging 26:874-88. Miller JO, Low K (2001) Motor processes in simple, go/no-go, and choice reaction time tasks: a psychophysiological analysis. J Exp Psychol Hum Percept Perform 27:266-89. Nachev P, Wydell H, O'Neill K, Husain M, Kennard C (2007) The role of the pre- supplementary motor area in the control of action. Neuroimage. 36 Suppl 2:T155-T63. Nambu A, Tokuno H, Takada M (2002) Functional significance of the cortico- subthalamo-pallidal 'hyperdirect' pathway. Neurosci.Res. 43:111-7. Neveu PJ, Liege S, Sarrieau A (1998) Asymmetrical distribution of hippocampal mineralocorticoid receptors depends on lateralization in mice. Neuroimmunomodulation 5:16-21. Omura K, Todd Constable R, Canli T (2005) Amygdala gray matter concentration is associated with extraversion and neuroticism. Neuroreport 16:1905-8. Paus T, Collins DL, Evans AC, Leonard G, Pike B, Zijdenbos A (2001) Maturation of white matter in the human brain: a review of magnetic resonance studies. Brain Res.Bull. 54:255-66. Paus T, Keshavan M, Giedd JN (2008) Why do many psychiatric disorders emerge during adolescence? Nat Rev Neurosci 9:947-57. Paus T, Zijdenbos A, Worsley K, Collins DL, Blumenthal J, Giedd JN, Rapoport JL, Evans AC (1999) Structural maturation of neural pathways in children and adolescents: in vivo study. Science 283:1908-11.

51

References

Piras F, Caltagirone C, Spalletta G (2010) Working memory performance and thalamus microstructure in healthy subjects. Neuroscience 171:496-505. Pliszka SR, Glahn DC, Semrud-Clikeman M, Franklin C, Perez R, III, Xiong J, Liotti M (2006) Neuroimaging of inhibitory control areas in children with attention deficit hyperactivity disorder who were treatment naive or in long-term treatment. Am.J.Psychiatry 163:1052-60. Portella MJ, Harmer CJ, Flint J, Cowen P, Goodwin GM (2005) Enhanced early morning salivary cortisol in neuroticism. Am.J.Psychiatry 162:807-9. Pruessner JC, Wolf OT, Hellhammer DH, Buske-Kirschbaum A, von Auer K, Jobst S, Kaspers F, Kirschbaum C (1997) Free cortisol levels after awakening: a reliable biological marker for the assessment of adrenocortical activity. Life Sci 61:2539- 49. Pruessner M, Pruessner JC, Hellhammer DH, Bruce Pike G, Lupien SJ (2007) The associations among hippocampal volume, cortisol reactivity, and memory performance in healthy young men. Psychiatry Res 155:1-10. Pujol J, Lopez A, Deus J, Cardoner N, Vallejo J, Capdevila A, Paus T (2002) Anatomical variability of the anterior cingulate gyrus and basic dimensions of human personality. Neuroimage 15:847-55. Rajkowska G, Miguel-Hidalgo JJ (2007) Gliogenesis and glial pathology in depression. CNS Neurol Disord Drug Targets 6:219-33. Reese TG, Heid O, Weisskoff RM, Wedeen VJ (2003) Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson.Med. 49:177-82. Reul JM, Gesing A, Droste S, Stec IS, Weber A, Bachmann C, Bilang-Bleuel A, Holsboer F, Linthorst AC (2000) The brain mineralocorticoid receptor: greedy for ligand, mysterious in function. Eur J Pharmacol 405:235-49. Rieger M, Gauggel S, Burmeister K (2003) Inhibition of ongoing responses following frontal, nonfrontal, and basal ganglia lesions. Neuropsychology. 17:272-82. Sanchez MM, Young LJ, Plotsky PM, Insel TR (2000) Distribution of corticosteroid receptors in the rhesus brain: relative absence of glucocorticoid receptors in the hippocampal formation. J Neurosci 20:4657-68. Sapolsky RM (2000) Glucocorticoids and hippocampal atrophy in neuropsychiatric disorders. Arch Gen Psychiatry 57:925-35. Schwartz ED, Cooper ET, Fan Y, Jawad AF, Chin CL, Nissanov J, Hackney DB (2005) MRI diffusion coefficients in spinal cord correlate with axon morphometry. Neuroreport 16:73-6. Seckl JR, Dickson KL, Yates C, Fink G (1991) Distribution of glucocorticoid and mineralocorticoid receptor messenger RNA expression in human postmortem hippocampus. Brain Res 561:332-7. Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE, Watkins KE, Ciccarelli O, Cader MZ, Matthews PM, Behrens TE (2006) Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage. 31:1487-505. Snook L, Paulson LA, Roy D, Phillips L, Beaulieu C (2005) Diffusion tensor imaging of neurodevelopment in children and young adults. Neuroimage. 26:1164-73. Song SK, Sun SW, Ju WK, Lin SJ, Cross AH, Neufeld AH (2003) Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage. 20:1714-22.

52

References

Song SK, Yoshino J, Le TQ, Lin SJ, Sun SW, Cross AH, Armstrong RC (2005) Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage. 26:132-40. Sowell ER, Thompson PM, Leonard CM, Welcome SE, Kan E, Toga AW (2004) Longitudinal mapping of cortical thickness and brain growth in normal children. J.Neurosci. 24:8223-31. Starkman MN, Gebarski SS, Berent S, Schteingart DE (1992) Hippocampal formation volume, memory dysfunction, and cortisol levels in patients with Cushing's syndrome. Biol Psychiatry 32:756-65. Starkman MN, Giordani B, Gebarski SS, Berent S, Schork MA, Schteingart DE (1999) Decrease in cortisol reverses human hippocampal atrophy following treatment of Cushing's disease. Biol Psychiatry 46:1595-602. Sullivan RM, Gratton A (1999) Lateralized effects of medial prefrontal cortex lesions on neuroendocrine and autonomic stress responses in rats. J Neurosci 19:2834-40. Taubert M, Draganski B, Anwander A, Muller K, Horstmann A, Villringer A, Ragert P (2010) Dynamic properties of human brain structure: learning-related changes in cortical areas and associated fiber connections. J Neurosci 30:11670-7. Tillman CM, Thorell LB, Brocki KC, Bohlin G (2008) Motor response inhibition and execution in the stop-signal task: development and relation to ADHD behaviors. Child Neuropsychol 14:42-59. Tuch DS, Salat DH, Wisco JJ, Zaleta AK, Hevelone ND, Rosas HD (2005) Choice reaction time performance correlates with diffusion anisotropy in white matter pathways supporting visuospatial attention. Proc Natl Acad Sci U S A 102:12212-7. van den Wildenberg WP, van Boxtel GJ, van der Molen MW, Bosch DA, Speelman JD, Brunia CH (2006) Stimulation of the subthalamic region facilitates the selection and inhibition of motor responses in Parkinson's disease. J.Cogn Neurosci. 18:626-36. van der Beek EM, Wiegant VM, Schouten WG, van Eerdenburg FJ, Loijens LW, van der Plas C, Benning MA, de Vries H, de Kloet ER, Lucassen PJ (2004) Neuronal number, volume, and apoptosis of the left dentate gyrus of chronically stressed pigs correlate negatively with basal saliva cortisol levels. Hippocampus 14:688- 700. Versace A, Almeida JR, Hassel S, Walsh ND, Novelli M, Klein CR, Kupfer DJ, Phillips ML (2008) Elevated left and reduced right orbitomedial prefrontal fractional anisotropy in adults with bipolar disorder revealed by tract-based spatial statistics. Arch Gen Psychiatry 65:1041-52. Vreeburg SA, Hoogendijk WJ, van Pelt J, Derijk RH, Verhagen JC, van Dyck R, Smit JH, Zitman FG, Penninx BW (2009) Major depressive disorder and hypothalamic- pituitary-adrenal axis activity: results from a large cohort study. Arch Gen Psychiatry 66:617-26. Vreeburg SA, Zitman FG, van Pelt J, Derijk RH, Verhagen JC, van Dyck R, Hoogendijk WJ, Smit JH, Penninx BW (2010) Salivary cortisol levels in persons with and without different anxiety disorders. Psychosom Med 72:340-7. Wilhelm I, Born J, Kudielka BM, Schlotz W, Wust S (2007) Is the cortisol awakening rise a response to awakening? Psychoneuroendocrinology 32:358-66. Williams BR, Ponesse JS, Schachar RJ, Logan GD, Tannock R (1999) Development of inhibitory control across the life span. Dev.Psychol. 35:205-13.

53

References

Wittling W, Pfluger M (1990) Neuroendocrine hemisphere asymmetries: salivary cortisol secretion during lateralized viewing of emotion-related and neutral films. Brain Cogn 14:243-65. Wolf OT, Convit A, de Leon MJ, Caraos C, Qadri SF (2002) Basal hypothalamo-pituitary- adrenal axis activity and corticotropin feedback in young and older men: relationships to magnetic resonance imaging-derived hippocampus and cingulate gyrus volumes. Neuroendocrinology 75:241-9. Wright CI, Williams D, Feczko E, Barrett LF, Dickerson BC, Schwartz CE, Wedig MM (2006) Neuroanatomical correlates of extraversion and neuroticism. Cereb Cortex 16:1809-19. Wust S, Federenko I, Hellhammer DH, Kirschbaum C (2000) Genetic factors, perceived chronic stress, and the free cortisol response to awakening. Psychoneuroendocrinology 25:707-20. Yakovlev PI, Lecours AR. 1967. The myelingenetic cycles of regional maturation of the brain. In Regional Development of the Brain Early in Life, ed. A Minkowski, pp. 3- 70. Boston: Blackwell Scientific Publications Inc. Zach P, Mrzilkova J, Rezacova L, Stuchlik A, Vales K (2010) Delayed effects of elevated corticosterone level on volume of hippocampal formation in laboratory rat. Physiol Res 59:985-96. Zandbelt BB, Vink M (2010) On the role of the striatum in response inhibition. PLoS One 5:e13848.

54

55

56

Papers

Paper 1
Madsen KS, Baar WFC, Vestergaard M, Skimminge A, Ejersboe LR, Ramsy TZ, Gerlach C, keson P, Paulson OB, and Jernigan TL (2009) Response inhibition is associated with white matter microstructure in children. Neuropsychologia 48(4): 854-862.

Paper 2
Madsen KS, Baar WFC, Skimminge A, Vestergaard M, Siebner HR, and Jernigan TL. Brain microstructural correlates of visuospatial choice reaction time in children. Submitted

Paper 3
Madsen KS, Jernigan TL, Iversen P, Frokjaer VG, Mortensen EL, Knudsen GM, and Baar WFC. Cortisol awakening response and negative emotionality linked to asymmetry in major limbic fibre bundle architecture. Submitted

Paper 4
Madsen KS, Jernigan TL, Iversen P, Frokjaer VG, Knudsen GM, Siebner HR and Baar WFC. Hypothalamic-pituitary-adrenal axis tonus is associated with hippocampal microstructural asymmetry. Submitted

57

Paper 1

Neuropsychologia 48 (2010) 854862

Contents lists available at ScienceDirect

Neuropsychologia
journal homepage: www.elsevier.com/locate/neuropsychologia

Response inhibition is associated with white matter microstructure in children

Kathrine Skak Madsen a,b,c, , William F.C. Baar a,b , Martin Vestergaard a , Arnold Skimminge a , Lisser Rye Ejersbo e , Thomas Z. Ramsy a , Christian Gerlach e , Per keson a , Olaf B. Paulson a,b,c , Terry L. Jernigan a,b,c,d
a

Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, Copenhagen, Denmark Faculty of Health Sciences, University of Copenhagen, Denmark d Center for Human Development, University of California, San Diego, CA, United States e Learning Lab Denmark, Danish School of Education, University of Aarhus, Copenhagen, Denmark
b c

a r t i c l e

i n f o

a b s t r a c t
Cognitive control of thoughts, actions and emotions is important for normal behaviour and the development of such control continues throughout childhood and adolescence. Several lines of evidence suggest that response inhibition is primarily mediated by a right-lateralized network involving inferior frontal gyrus (IFG), presupplementary motor cortex (preSMA), and subthalamic nucleus. Though the brains bre tracts are known to develop during childhood, little is known about how bre tract development within this network relates to developing behavioural control. Here we examined the relationship between response inhibition, as measured with the stop-signal task, and indices of regional white matter microstructure in typically-developing children. We hypothesized that better response inhibition performance would be associated with higher fractional anisotropy (FA) in bre tracts within right IFG and preSMA after controlling for age. Mean FA and diffusivity values were extracted from right and left IFG and preSMA. As hypothesized, faster response inhibition was signicantly associated with higher FA and lower perpendicular diffusivity in both the right IFG and the right preSMA, possibly reecting faster speed of neural conduction within more densely packed or better myelinated bre tracts. Moreover, both of these effects remained signicant after controlling for age and whole brain estimates of these DTI parameters. Interestingly, right IFG and preSMA FA contributed additively to the prediction of performance variability. Observed associations may be related to variation in phase of maturation, to activity-dependent alterations in the network subserving response inhibition, or to stable individual differences in underlying neural system connectivity. 2009 Elsevier Ltd. All rights reserved.

Article history: Received 24 March 2009 Received in revised form 27 August 2009 Accepted 4 November 2009 Available online 10 November 2009 Keywords: Brain maturation Cognitive development Diffusion tensor imaging Executive control Fractional anisotropy MRI

1. Introduction Development of cognitive control of behaviour continues throughout childhood and adolescence (Williams, Ponesse, Schachar, Logan, & Tannock, 1999). Cognitive control of thoughts, actions, and emotions is important for normal behaviour, and decits in behavioural control are prominent in a variety of psychiatric disorders, e.g., attention decit/hyperactive disorder (Pliszka et al., 2006) and obsessivecompulsive disorder (Enright & Beech, 1993). In recent years, investigators have developed experimental paradigms designed to measure motor control, or more specically the capacity to inhibit primed, or prepotent, motor

Corresponding author at: Danish Research Centre for Magnetic Resonance, MRDepartment, Section 340, Copenhagen University Hospital, Hvidovre, Kettegaard All 30, 2650 Hvidovre, Denmark. Tel.: +45 3632 3323; fax: +45 3647 0302. E-mail address: kathrine@drcmr.dk (K.S. Madsen). 0028-3932/$ see front matter 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuropsychologia.2009.11.001

responses (Chambers, Garavan, & Bellgrove, 2009). Among these is the Go/NoGo task. It is relatively easy to measure variability in the speed of a motor response using response latency; however, it is more difcult to assess the time a subject needs to inhibit a response, since no response occurs on successful NoGo trials. Thus, the primary measure of inhibitory function on Go/NoGo tasks is the number of errors of commission during inhibit conditions. The stop-signal task (SST) provides a more continuous measure of a subjects ability to inhibit a prepotent manual response (Logan & Cowan, 1984). The subjects task in the SST is to make a motor response to a visual target as rapidly as possible; however, on infrequent trials an acoustic stop signal occurs at some delay after the visual target. When the stop signal is detected, the subject must attempt to withhold, or cancel, the motor response. Obviously when the stop signal occurs very late in the trial (long stop-signal delay), the subject will not be able to inhibit the response, whereas when it occurs very soon after the visual target (short stop-signal delay), success in inhibiting the response is much more likely. The

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

855

stop-signal delay is varied dynamically during the task so that the subject succeeds in inhibiting the response on stop trials approximately 50% of the time. A stop-signal reaction time, or SSRT, is then computed for each subject by subtracting this average stopsignal delay from the median response latency for Go trials (on which no stop-signal occurs). In this way, a continuous measure is constructed of the estimated time needed to inhibit a response. Studies using the SST have implicated several brain structures in a neural network subserving response inhibition. Responding is thought to be mediated by a premotor-striatal-pallidal-motor cortical network, whereas a primarily right-lateralized network, involving prominently the inferior frontal gyrus (IFG), as well as the presupplementary motor area (preSMA), and the subthalamic nucleus (STN), has been implicated in response inhibition (Aron, Behrens, Smith, Frank, & Poldrack, 2007; Aron & Poldrack, 2006). Several functional magnetic resonance imaging (fMRI) studies have shown that inhibiting a prepotent response consistently activates prefrontal regions, particularly the right IFG, in adults (Aron et al., 2007; Aron & Poldrack, 2006; Chevrier, Noseworthy, & Schachar, 2007; Rubia, Smith, Brammer, & Taylor, 2003) as well as in children (Cohen et al., 2007). Moreover, human lesion studies suggest that the right IFG is critical for response inhibition (Rieger, Gauggel, & Burmeister, 2003), i.e., lesions in the IFG sub-region pars opercularis impair response inhibition (Aron, Fletcher, Bullmore, Sahakian, & Robbins, 2003; Aron, Robbins, & Poldrack, 2004). The latter has been conrmed in a study using transcranial magnetic stimulation, where temporary deactivation of the right pars opercularis selectively impaired the ability to stop an initiated response (Chambers et al., 2006). In addition, a recent study found that FA within the right IFG correlated with response inhibition measured with the Simon task in a small group of adults (Forstmann et al., 2008). Further, fMRI studies have found the right preSMA to be activated during stop trials (Aron et al., 2007; Aron & Poldrack, 2006). The involvement of the right preSMA in response inhibition has also been corroborated by human lesion studies (Floden & Stuss, 2006; Nachev, Wydell, Oneill, Husain, & Kennard, 2007). Moreover, weak microstimulation of neurons in the supplementary eye eld improves stop task performance by delaying saccadic initiation in monkeys (Stuphorn & Schall, 2006). Another monkey study, using a switching task designed to examine control over automatic responses, found that successful switching from an automatic to a volitionally controlled response selectively increased the activity of preSMA neurons. In addition, electrical stimulation in the preSMA increased the proportion of successful switch trials (Isoda & Hikosaka, 2007). The role of the STN region in response inhibition is evidenced by fMRI studies showing this region to be activated during stop trials (Aron et al., 2007; Aron & Poldrack, 2006), and by the nding that deep-brain stimulation of the STN improves response inhibition in patients with Parkinsons disease (van den Wildenberg et al., 2006). Consistently, STN lesions impair inhibition in rodents (Eagle et al., 2008). Finally and importantly, evidence from a tractography study (Aron et al., 2007) suggests that the three regions implicated in response inhibition are interconnected. A possible preSMA-IFG connection is further supported by another tractography study in humans (Johansen-Berg et al., 2004), whereas the existence of direct bre connections between the preSMA and STN is supported by tract tracing in monkeys (Inase, Tokuno, Nambu, Akazawa, & Takada, 1999). Brain maturation is a complex ongoing process during childhood and early adulthood. Conventional structural MRI studies have demonstrated morphological changes in grey and white matter structures during childhood and adolescence consistent with cellular maturational processes, i.e., synaptic pruning and myelination (Giedd et al., 1999; Gogtay et al., 2004; Jernigan, Trauner, Hesselink, & Tallal, 1991; Paus et al., 1999; Sowell et al., 2004); and different grey matter structures exhibit distinct maturational tra-

jectories (Gogtay et al., 2004; Jernigan et al., 1991; Sowell et al., 2004). Furthermore, gradual increases in global white matter volume as well as regional white matter density, have been reported in children and adolescents, possibly reecting age-related increase in axonal diameter and ongoing myelination across this age range (Giedd et al., 1999; Paus et al., 1999, 2001). However, conventional structural MRI provides limited information about the underlying white matter microstructural properties. Diffusion tensor imaging (DTI) measures the diffusion of water molecules in tissue. In white matter, diffusion perpendicular to highly organised bre bundles is hindered relative to diffusion parallel to the bres, causing diffusion anisotropy (Beaulieu, 2002). An estimate of the degree of diffusion directionality, fractional anisotropy (FA), as well as diffusivity parallel ( ) and perpendicular ( ) to the principal diffusion direction can be derived by tting the diffusion measurements of each voxel to the diffusion tensor model (Basser, Mattiello, & LeBihan, 1994; Beaulieu, 2002). In recent years, DTI has been applied in studies of typicallydeveloping children and adolescents (Eluvathingal, Hasan, Kramer, Fletcher, & Ewing-Cobbs, 2007; Lebel, Walker, Leemans, Phillips, & Beaulieu, 2008; Snook, Paulson, Roy, Phillips, & Beaulieu, 2005). Age-related increases in FA have been reported in multiple locations within white matter, reecting a disproportionate decrease in relative to , possibly due to ongoing myelination and/or an increase in bre density (Eluvathingal et al., 2007; Lebel et al., 2008; Snook et al., 2005). Although the physiological signicance of these changes in diffusion parameters during childhood are still not fully understood, a previous study of infants correlated increased FA (and decreased ) with apparently increased neural conduction speed, as reected in decreased latency of the rst positive wave of the visual evoked potential (Dubois et al., 2008). Different white matter tracts exhibit distinct maturational patterns, with diffusion parameters in some tracts approaching adult levels earlier than others (Lebel et al., 2008). The relationship between age-related changes in white matter microstructural properties and the concurrent development of cognitive control in children is poorly understood. White matter microstructural changes in the fronto-striatal network have been correlated with faster reaction times in the Go/NoGo task, particularly in conditions more demanding of inhibition, in a small group of children (Liston et al., 2006). However, processing factors other than inhibitory function may also inuence reaction times on Go trials of the Go/NoGo task. There are currently no reports of associations between white matter microstructure and the specic measure of response inhibition from the SST, the stop-signal reaction time, in children. Here we report associations between response inhibition performance and regional white matter microstructure in typically-developing children. The major hypothesis of the study was that better response inhibition, adjusted for age, would be associated with higher FA in ber tracts within the right IFG. Secondary hypotheses were that such relationships might also be observed for tracts in the right preSMA region. The parallel and perpendicular diffusivities were investigated to further explore the observed effects, since these diffusion parameters may provide additional information about the underlying white matter microstructure. Post-hoc analyses, including whole brain or left hemisphere region-of-interest estimates of FA and , were also performed to further explore the anatomical specicity of the effects.
2. Materials and methods 2.1. Subjects Ninety-two typically-developing children aged 713 from three schools (1st6th graders) in the Copenhagen area were inducted into the study. Prior to participation, all children assented to the procedures and informed written con-

856 Table 1 Demographic data for the included subjects.

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

1st/2nd gradersa Age (mean SD) Gender (female/male) Handedness (right/left) Parents average years of education (mean SD) 8.3 0.5 12/10 19/3 13.9 1.7

3rd/4th gradersa 10.1 0.3 14/10 22/2 13.8 2.0

5th/6th gradersa 12.2 0.4 10/9 17/2 13.7 2.1

All subjects 10.1 1.6 36/29 58/7 13.8 1.9

Abbreviation: SD = standard deviation. a Children enrolled in the study were scanned either in the months just before (when in 1st, 3rd or 5th grade) or just after (when in 2nd, 4th or 6th grade) the summer holiday.

sent was obtained from the parents/guardians after oral and written explanation of the study aims and study procedures. The study was approved by the local Danish Committee for Biomedical Research Ethics (H-KF-01-131/03), and conducted in accordance with the Declaration of Helsinki. Twenty-seven of the participating subjects (17 girls, 10 boys) were subsequently excluded for the following reasons: incidental ndings on MRI (1 subject), not completing the scanning session (3 subjects), not assessed on the SST due to failure of the response box (6 subjects), and reduced image quality as described below (17 subjects). Thus, 65 subjects were included in the present study (mean age std: 10.1 1.6, 36 girls, 29 boys). According to a parent report, no subjects had any known history of neurological or psychiatric disorders or signicant brain injury. There were no signicant differences between the included and the excluded subjects on age, gender, parental education, or handedness (as assessed with the Edinburgh handedness inventory). Demographic data from the included subjects are presented in Table 1. 2.2. Stop-signal task The stop-signal task (SST) was administered using the Cambridge Neuropsychological Test Automated Battery (Cambridge Cognition Ltd., Cambridge, UK). The SST consists of go and stop trials (Fig. 1). Subjects sit in front of a computer monitor, with each index nger resting on a response button. A circle is presented for 500 ms, followed by an arrow pointing either left or right. Subjects are instructed to press the left or right button (i.e., with the left or right index nger), depending on the direction of the arrow, without making any mistakes, and to press as fast as possible. The test consists of two parts. In the rst part, there are 16 Go training trials without an auditory stop-signal to introduce the subjects to the press pad. In the second and longer part, an auditory stop-signal occurs in 25% of the trials. When the tone occurs, subjects must try to withhold their responses. The time between the onset of the arrow and the auditory stop signal, i.e., the stop-signal delay (SSD), changes adaptively throughout the test, depending on the subjects past performance, so that responses are inhibited successfully approximately 50% of the time for each subject. The shorter the SSD, the more likely it is that the subject will be able to

inhibit his or her response. The SST is administered in 5 blocks of 64 trials. Each block is divided into four sub-blocks of 16 trials (12 go trials and 4 stop trials in random order). There is no gap between the sub-blocks and they are not evident to the subject. After each block, a feedback screen is displayed showing a histogram representation of the subjects reaction time on Go trials. The histogram shown after the rst block is identical for all subjects. The test administrator explains to the subject that if he/she can go faster, it will show in the next histogram, before encouraging him/her to go faster. The feedback after each of the last four blocks is the subjects go reaction time in relation to the rst block, and consists of a histogram containing the rst block and the relative performance of all previous blocks. The primary behavioural outcome measure is the stop-signal reaction time (SSRT), which measures how fast subjects can inhibit a prepotent response. The SSRT is estimated using the race model (Logan & Cowan, 1984) by subtracting the SSD50, where subjects are able to inhibit 50% of their responses, from the median go RT. The race model assumes that the go and stop processes are in a race with each other and are (mainly) independent (Boucher, Palmeri, Logan, & Schall, 2007; Logan & Cowan, 1984). To get a stable estimate of the SSRT, the trials in the rst half of the dataset were treated as training trials to familiarize subjects with the auditory stop-signal, and the SSRT used for statistical analysis was estimated from trials in the last half of the task. 2.3. Image acquisition All subjects were scanned using a 3T Siemens Magnetom Trio MR scanner (Siemens, Erlangen, Germany) with an eight-channel head coil (Invivo, FL, USA). Subjects were scanned the same day as the SST was administered. All acquired scans were aligned parallel to the anterior commisureposterior commisure (ACPC) line. Diffusion-weighted (DW) images of the whole brain were acquired using a twicerefocused balanced spin echo sequence that minimised eddy current distortion (Reese, Heid, Weisskoff, & Wedeen, 2003). Ten non-DW images (b = 0) and 61 DW images, encoded along independent collinear diffusion gradient orientations (Cook, Symms, Boulby, & Alexander, 2007; Jansons & Alexander, 2003), were acquired with a b value of 1200 s/mm2 (TR = 8200 ms; TE = 100 ms, FOV = 220 mm 220 mm, matrix = 96 96, GRAPPA: acceleration factor = 2; number of reference lines = 48, 61 transverse slices with no gap, 2.3 mm 2.3 mm 2.3 mm voxels, NEX = 1, acquisition time = 9.50 min). A gradient echo based eld map sequence (TR = 530 ms, TE[1] = 5.19 ms and TE[2] = 7.65 ms, FOV = 256 mm 256 mm; matrix = 128 128, 47 transverse slices with no gap, voxel size = 2 mm 2 mm 3 mm, NEX = 1, acquisition time = 2.18 min) was acquired to correct geometric distortions caused by B0 magnetic eld inhomogeneities. T2 -weighted images of the whole head were acquired using a 3D turbo spin echo sequence (TR = 3000 ms, TE = 354 ms, FOV = 282 mm 282 mm, matrix = 256 256, 192 sagittal slices with no gap, 1.1 mm 1.1 mm 1.1 mm voxels, NEX = 1, acquisition time = 8.29 min) for generating brain masks. T1 -weighted and proton density weighted images were also acquired in the imaging session, but these images were not used in the analysis reported here. 2.4. Image evaluation All subjects images were evaluated by an experienced neuroradiologist. Prior to image analysis, and blind to behavioural data, the raw images from all subjects were visually checked to ascertain the quality of the data. As described above, based on this inspection, 17 subjects had signicantly reduced image quality due to movement or susceptibility artefacts and were excluded from further analysis.

Fig. 1. The stop-signal task consists of go and stop-signal trials. A circle is presented for 500 ms, followed by a presentation of an arrow pointing either left or right. Subjects are instructed to respond as fast as possible by pressing a left or right button, depending on the direction of the arrow. In the stop trials, an auditory stopsignal occurs after the presentation of the arrow, and on these trials subjects must try to withhold their responses. The latency to the sound (the stop-signal delay [SSD]) varies dynamically throughout the study to produce the SSD50, where subjects are able to inhibit approximately 50% of their responses. The stop-signal reaction time (SSRT) is calculated as the median go RT minus the SSD50, according to the race model (Logan & Cowan, 1984).

2.5. Image analysis Images were preprocessed using pipelines implemented in Matlab, using mainly SPM2 routines. DW images were oriented to the MNI coordinate system and corrected for geometric distortions due to B0 inhomogeneities. The rst B0 image was coregistered to MNI space using a rigid transformation (6 parameter mutual information), after which all DW images were coregistered (no reslicing) to the rst B0 image. Next, all coregistered images were corrected for geometric distortions using the acquired B0 eld map (Andersson, Hutton, Ashburner, Turner, & Friston, 2001) and resliced into the MNI coordinate system using trilinear interpolation. Note that

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

857

Fig. 2. (a) Coronal and sagittal slices of the mean skeleton (yellow) overlaid on the target FA image. The blue segments are the IFG (inferior frontal gyrus, pars opercularis) ROIs and the red segments are the preSMA (presupplementary motor area) ROIs. (b) Midsagittal view showing the planes (in green) of the three coronal sections in the top row. The planes indicated in white and green represent the coronal views of the effect-size maps shown in Figs. 3 and 4.

morphology of the pars opercularis (Tomaiuolo et al., 1999). The boundaries of the IFG ROIs were found by using anatomical information visible in the target FA map, where the vertical ramus of the lateral ssure and the inferior part of the precentral sulcus, indicated by low FA values, were used as landmarks to dene the anterior and posterior boundaries of the pars opercularis. The skeleton dened the superior boundary, in that the lateral part of the pars opercularis segment was perpendicular to the segment going into the pars triangularis. The right IFG ROI contained 379 voxels and the left contained 322 voxels. The location of the preSMA ROIs was based on MNI coordinates derived from published functional and structural studies (Behrens, Jenkinson, Robson, Smith, & Johansen-Berg, 2006; Johansen-Berg et al., 2004), and dened between MNI x = 0 and x = 20. However, based on the anatomical information provided by the target FA map, the posterior and anterior boundaries were set to MNI x = 5 and x = 14 for the right and x = 6 and x = 15 for the left preSMA ROI, respectively. The right preSMA ROI included 295 voxels and the left ROI included and values from all four ROIs and the whole skeleton 311 voxels. Mean FA, were extracted for each subject for statistical analyses. We initially attempted to include a ROI in the internal capsule, given that the subthalamic nucleus (STN) has been implicated in response inhibition (Aron & Poldrack, 2006; van den Wildenberg et al., 2006) and bre tracts connecting the IFG and preSMA with the STN have been found (Aron et al., 2007). However, due to the lack of clear landmarks delimiting the relevant segments of the internal capsule, it was deemed not feasible to dene an appropriate ROI on the mean skeleton. 2.7. Statistical analysis The statistical analyses were performed with SPSS 15.0 using multiple linear regression models predicting SSRT. Multicollinearity between the predictors was assessed for all models. The statistical tests were performed hierarchically. The major hypotheses that right IFG FA and right preSMA, adjusted for age, would be signicantly and negatively associated with SSRT were tested with = 0.025. Planned follow-up analyses were contingent on observing signicant associations with FA in the primary analyses. Follow-up analyses were done to assess the anatomical specicity of the effects, either by adjusting for the whole skeleton FA or by adjusting for the corresponding left hemisphere ROI FA. The parallel and perpendicular diffusivities were investigated to further explore the nature of these observed effects by following the same statistical analysis protocol as for the FA data. An additional follow-up analysis was conducted to predict SSRT with the right IFG FA and the right preSMA FA measures simultaneously. Exploratory models including either gender or age by gender interaction effects revealed no signicant gender effects and, thus, models presented do not include gender. Two effect-size maps are presented to provide further anatomical information about the association between FA and SSRT (Jernigan, Gamst, Fennema-Notestine, & Ostergaard, 2003). The effect-size maps are t-maps of the correlation between FA and SSRT, however the sign of the correlations are reversed so that correlations between high FA and better response inhibition performance (lower reaction time) are shown in warm colours (red to yellow). The maps are presented with and without adjusting for age, respectively. The t-maps were generated using the Monte Carlo permutation test with 10,000 permutations implemented in the randomise program within FSL (Nichols & Holmes, 2002).

this procedure involves only one reslicing step. The diffusion tensor was tted using the RESTORE algorithm (Chang, Jones, & Pierpaoli, 2005) implemented in Camino ( 1 ) and (( 2 + 3 )/2) were calculated. A brain mask (Cook et al., 2006), and FA, based on the T2 -weighted image was automatically created using SPM2 segmentation routines and morphological operations and applied to the FA and diffusivity images. Because our hypotheses relate to the degree of anisotropy in specic bre tracts, we used a region-of-interest (ROI) approach to extract FA and diffusivity measures from specic ROIs for each subject. However, we rst used tract-based spatial statistics (TBSS) (Smith et al., 2006) to accomplish spatial normalisation and to align the bre tracts across children. Specically, FSL 4.0.1 (Smith et al., 2004) was used to align all subjects FA images into a common space using the nonlinear registration IRTK (Rueckert et al., 1999). A study-specic target, the groups most representative FA image, was identied by aligning each subjects FA image to every other subjects FA image. The target FA image was aligned to 1 mm3 MNI space using afne registration, and subsequently all subjects FA images were then aligned to this study-specic target FA image. A cross-subject mean FA image was created and thinned to create a mean FA skeleton, representing the centres of all tracts common to the group. The mean FA skeleton was thresholded at FA > 0.25, and contained 103,503 1 mm3 interpolated isotropic voxels, corresponding to approximately one quarter of the voxels with FA above 0.25. Each subjects aligned FA image was then projected onto the study-specic skeleton by locating the voxels with the highest local FA value in the direction perpendicular to the skeleton tracts and assigning the value of these voxels to the skeleton at this standardised location. Note that this results in a mapping of each voxel location in the skeleton to a specic voxel in the individual FA maps. In addition, the nonlinear warps and the skeleton projections and data. were applied to the 2.6. Regions-of-interest Regions-of-interest (ROIs) were drawn onto the study-specic skeleton overlaid on the target FA map. Since the TBSS procedure projects individual FA values and other DTI derived parameters onto a common framework, i.e., the mean studyspecic skeleton, one can extract each subjects individual measures directly by delineating a ROI once. The ROIs were placed in the right and left IFG (pars opercularis) and in the right and left preSMA (Fig. 2). The right and left pars opercularis were located using a brain atlas (Duvernoy, 1999) and published information about the

3. Results Behavioural measures for the SST and the FA measures for the four ROIs are presented in Table 2. The subjects inhibited approximately 50% of their responses in the stop trials, indicating that the adaptive tracking algorithm performed as expected. Separate statistics are provided for boys and girls, but no gender differences approached signicance (Ps > 0.37).

Table 2 Behavioural measures of the SST and fractional anisotropy measures of the region-of-interest.* . Behavioural measure Median correct Go RT (ms) Mean SSD 50% (ms) Percentage inhibition SSRT (ms) Region-of-interest Right IFG Right preSMA Left IFG Left preSMA All subjects 504.4 287.1 50.5 217.3 103.5 104.9 0.1 60.4 Girls 510.3 297.5 50.9 212.9 114.1 109.0 0.1 59.1 Boys 497.0 274.1 49.9 222.8 90.2 99.9 0.1 62.5

All subjects FA 0.405 0.435 0.411 0.444 0.025 0.034 0.031 0.024

Girls FA 0.404 0.435 0.414 0.442 0.026 0.034 0.030 0.027

Boys FA 0.407 0.436 0.407 0.446 0.024 0.034 0.331 0.021

Abbreviations: SST = stop-signal task, SD = standard deviation, RT = reaction time, SSRT = stop-signal reaction time, SSD = stop-signal delay. * All values are Mean SD.

858

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

Table 3 Linear regression models predicting SSRT from FA: a priori hypotheses.a . Right IFG Model [1] [2] [3] [4] [5] [6] R
2

Right preSMA P 0.00024 0.319 0.0038 0.002 0.023 0.037 0.00026 0.0029 0.0022 0.377 0.268 0.221 P

Age 0.420 0.416 P 0.00011 0.00023

0.396 0.438 0.356 0.330

0.192 0.367 0.142 0.312 0.257 0.441

0.396

0.00018

Each model is referred to with a number in the leftmost column, representing the models: [1] right IFG FA, adjusted for age; [2] right preSMA FA, adjusted for age; [3] right IFG FA; [4] right preSMA FA; [5] right IFG and right preSMA FA; [6] right IFG and right preSMA, adjusted for age. Models [1] and [2] test the a priori hypotheses (see text for details). Abbreviations: SSRT = stop-signal reaction time, IFG = inferior frontal gyrus, PreSMA = presupplementary motor area, and FA = fractional anisotropy. a Linear regression models are presented in rows. Predictors are presented in columns.

3.1. Associations between SSRT and DTI parameters within IFG and preSMA Results for the primary and secondary hypotheses are presented in Table 3, where each row represents a separate, planned model predicting SSRT. As hypothesized, right IFG FA, adjusted for age, was signicantly and negatively associated with SSRT (Fig. 3a and Table 3, model 1), indicating that better response inhibition is associated with higher mean FA values within the right IFG. A similar signicant relationship was observed for right preSMA FA (Fig. 3b and Table 3, model 2). When not controlling for age, which itself was signicantly and negatively correlated with SSRT (R2 = 0.212, = 0.460, P = 0.0001) the effects were slightly larger (Table 3, models 3 and 4). Interestingly, when entered as predictors simultaneously, either alone or adjusting for age, both right IFG and preSMA FA remained signicant predictors of SSRT (Table 3, models 5 and 6), suggesting that these effects are additive. Results from follow-up analyses of the FA data, testing for the anatomical specicity of the above-mentioned effects, are presented in Table 4. When adjusting for age and whole skeleton FA, right IFG FA as well as right preSMA FA remained signicant predictors of SSRT (Table 4, models 1 and 3), suggesting that the association between SSRT and right IFG and preSMA FA were not mediated by global increase in FA. In both models, the additional effect of whole skeleton FA did not reach signicance, though whole skeleton FA, adjusted for age, was itself signicantly and negatively associated with SSRT (R2 = 0.288, = 0.304, P = 0.013). Furthermore, the right hemisphere FA values remained signicant when controlling for age and the corresponding left hemisphere ROI FA values (Table 4, models 2 and 4). The additional contributions of the latter did not approach signicance in either case. Although left IFG FA by itself was correlated with SSRT (R2 = 0.075, P = 0.028), this association was not signicant after controlling for age (P = 0.09),

and left preSMA FA did not exhibit a signicant association with SSRT (P = 0.62). The parallel and perpendicular diffusivities were investigated to further explore the nature of these effects. Both right IFG and preSMA were signicantly and positively associated with SSRT alone (Table 5, models 1 and 5) or when controlling for age (Table 5, models 2 and 6) or for left hemisphere ROI (Table 5, models 4 and 8). When controlling for both age and whole skeleton , the effect of right IFG remained signicant, though the effect of right preSMA did not (Table 5, models 3 and 7, respectively). Whole skeleton , adjusted for age, was signicantly and positively related with SSRT (R2 = 0.307, = 0.333, P = 0.005). These effects contrasted with those for parallel diffusivity. Neither right IFG nor right preSMA was signicantly associated with SSRT, with or without adjusting for age (P > 0.5). However, whole skeleadjusted for age was signicantly and positively correlated ton with SSRT (R2 = 0.267, = 0.242, P = 0.035). Collinearity diagnostics were performed for all of the regression models and multicollinearity among the explanatory variables in the models was low (tolerance = 0.460.99, variance ination factor = 1.012.17). 3.2. Effect-size map The present study was designed to test specic anatomical hypotheses about the relationship between right IFG and preSMA bre tract microstructure and variability in SST performance in children. Therefore neither a whole brain, voxel-wise analysis of the effects (appropriately adjusted for test-multiplicity), nor a restricted voxel-wise analysis with small volume correction was deemed appropriate for testing these a priori hypotheses. However, since the analysis we used produces estimates of the effect size at each voxel, we have provided visualisation of the effect-size

Fig. 3. Partial regression plots of the stop-signal reaction time (SSRT) as a function of (a) right inferior frontal gyrus (IFG) FA, adjusted for age and (b) right presupplementary motor area (preSMA) FA, adjusted for age.

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862 Table 4 Linear regression models predicting SSRT from FA: follow-up analyses.a . Right IFG Model [1] [2] R
2

859

Age P 0.0032 0.00099 0.361 0.409 Age P 0.044 0.0019 0.349 0.431 P 0.0037 0.00015 P 0.0019 0.00020

Whole skeleton 0.153 P 0.214

Left IFG 0.076 P 0.486

0.339 0.372 Right PreSMA

0.383 0.372

Whole skeleton 0.183 P 0.162

Left PreSMA 0.137 P 0.231

Model [3] [4]

0.243 0.367

0.334 0.328

Each model is referred to with a number in the leftmost column, representing the models: [1,3] right ROI, adjusted for age and whole skeleton FA; [2,4] right ROI, adjusted for age and corresponding left ROI FA. Abbreviations: SSRT = stop-signal reaction time, IFG = inferior frontal gyrus, PreSMA = presupplementary motor area, and FA = fractional anisotropy. a Multiple linear regression models are presented in rows. Predictors are presented in columns.

maps. These maps show the distribution of t-values in skeleton voxels, and are presented to provide additional information about the anatomical distribution of associations between response inhibition performance and FA in white matter. Two maps are provided, the rst (Fig. 4) displaying the association between FA and SSRT controlling for age, and the second (Fig. 5) displaying the direct association of FA and SSRT. The contrast has been coded so that association of lower SSRT (better response inhibition) with higher FA yields positive t-values. Increasing positive t-values are shown in warm colours ranging from red to yellow, whereas decreasing negative t-values are shown in cool colours ranging from dark blue to light blue. The position of the depicted coronal slices is indicated in Fig. 2b. Slices 5, 10, and 13 are through the ROI locations and are the same as those shown in Fig. 2a, top row. Slice 18 shows the effects in the hand area in the motor cortex. In general, the associations between higher FA and better response inhibition appear to be modestly stronger when not correcting for age. In Fig. 6, sagittal views of the effect-size map show clusters of voxels with relatively high t-values within the internal capsule. Some clusters appear to be located in the posterior limb of the internal capsule. 4. Discussion The present study examined associations between response inhibition performance and white matter microstructure within IFG (pars opercularis) and preSMA in children aged 713 years. As hypothesized, higher FA in bre tracts within right IFG and preSMA was signicantly associated with better response inhibition. Since both FA (Lebel et al., 2008) and response inhibition (Williams et
Table 5 Linear regression models predicting SSRT from Right IFG Model [1] [2] [3] [4] R
2

al., 1999) increase throughout childhood and adolescence, a correlation between these measures could represent a nonspecic association attributable to unmeasured factors indexed by chronological age. For example, since height also increases in children over this age range, one might expect to nd a simple correlation between height and response inhibition within a group of children with varying ages. However, since height has little direct relationship to cognitive functioning in children, one would expect the correlation between these variables to be substantially reduced after controlling for age statistically. Thus, we hypothesized that the associations between SSRT and FA within the right IFG and preSMA would remain signicant after controlling for age; i.e., that even among children of similar age, those with higher FA in right IFG and right preSMA would exhibit stronger response inhibition performance. These hypotheses were conrmed. Several planned follow-up analyses were conducted to explore the nature of these associations. We were interested to assess the anatomical specicity of the associations between response inhibition performance and bre structure within the right IFG and preSMA, i.e., the extent to which response inhibition exhibited a relatively specic relationship to FA in this bre tract relative to others. This is an important question, since it is known that FA increases concurrently in many bre tracts during childhood (Lebel et al., 2008). Mean FA for the whole skeleton, adjusted for age, was modestly correlated with SSRT, suggesting that global individual differences in white matter microstructure between children of similar age are mediating this association. However, the effects on SSRT of FA in the right IFG and in the right preSMA remained signicant when this global measure of FA was included as a

follow-up analyses.a . Age P 0.000017 0.00014 0.0097 0.0013 0.360 0.346 0.355 Age P 0.0016 0.0059 0.145 0.011 0.403 0.342 0.409 P 0.00042 0.0037 0.00042 P 0.00090 0.0024 0.0014 Whole skeleton P Left IFG P

0.506 0.419 0.379 0.404 Right PreSMA

0.244 0.378 0.380 0.378

0.063

0.674 0.251 0.803

Whole skeleton P

Left PreSMA P

Model [5] [6] [7] [8]

0.383 0.308 0.191 0.337

0.147 0.303 0.331 0.305

0.218

0.118 0.055 0.672

Each model is referred to with a number in the leftmost column, representing the models: [1,5] right ROI ; [2,6] right ROI , adjusted for age; [3,7] right ROI , adjusted for age and whole skeleton ; [4,8] right ROI , adjusted for age and corresponding left ROI . Abbreviations: SSRT = stop-signal reaction time, IFG = inferior frontal gyrus, PreSMA = presupplementary motor area, and = perpendicular diffusivity. a Linear regression models are presented in rows. Predictors are presented in columns.

860

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

Fig. 6. Effect-size map of the negative association between SSRT and FA with (a) and without (b) adjusting for age. The effect-size map is overlaid on the target FA image. The colour bar shows the mapping of colours to the values of the t-statistics. All voxels on the skeleton are coloured. The sagittal slices were selected to show clusters of voxels with relatively high t-values within the internal capsule. The MNI coordinates for the images are given above the images.

Fig. 4. Effect-size map of the negative association between SSRT and FA adjusted for age. The effect-size map is overlaid on the target FA image. The colour bar shows the mapping of colours to the values of the t-statistics. All voxels on the skeleton are coloured. Images are shown with the subjects right side on the right side (neurological convention). The MNI coordinates for the coronal images are given under each image. The IFG and preSMA ROIs were located in slice 5, 10 and 13 (see Fig. 2a, top row).

Fig. 5. Effect-size map of the negative association between SSRT and FA. The effectsize map is overlaid on the target FA image. The colour bar shows the mapping of colours to the values of the t-statistics. All voxels on the skeleton are coloured. Images are shown with the subjects right side on the right side (neurological convention). The MNI coordinates for the coronal images are given under each image. The IFG and preSMA ROIs were located in slice 5, 10 and 13 (see Fig. 2a, top row).

covariate, and in these models the effect of the global measure no longer approached signicance. This suggests that the relationships between response inhibition and FA within the right IFG and preSMA are not likely to be mediated by global white matter FA differences and may be more strongly related to the structure in specic tracts. Moreover, results from similar analysis controlling for FA in the left hemisphere ROIs, suggested that the effects of structure in the right hemisphere tracts were more robust than effects of structure in the left hemisphere tracts. This pattern is consistent with a signicant degree of anatomical specicity of the effects relating FA and SSRT performance; and it is consistent with previous studies suggesting that a primarily right-lateralized network involving IFG and preSMA mediates response inhibition (Aron et al., 2007, 2003; Chambers et al., 2006; Floden & Stuss, 2006; Nachev et al., 2007). Interestingly, when predicting SSRT performance with right IFG and right preSMA FA simultaneously (with or without adjusting for age), both regions remained signicant predictors. This suggests that increased FA in bre tracts within the right IFG and right preSMA may contribute additively to better response inhibition performance. TBSS improves inter-subject registration of brain bre tracts relative to prior methods for performing spatial normalisation of FA images (Smith et al., 2006) and the method obviates the need for extensive spatial smoothing. Thus, the use of TBSS in the present study aimed to increase the sensitivity of our ROI approach relative to ROI methods used in previous DTI studies in children. We averaged FA over the segments of the tract skeleton that corresponded to the expected location of the targeted tracts after applying TBSS to align the tracts across subjects. This was considered to be a more objective way of estimating FA within a comparable portion of the tract within each individual than subjective (manual) delineation of regions within the white matter, and a more powerful test of our hypothesis than a voxel-wise approach. However, the method produces only an approximation of FA in the targeted tracts, and in future studies it may be possible to employ better validated tractography methods that dene the tracts based on connectivity, or new methods using probabilistic atlases (Hagler et al., 2009) to more denitively assess the structure in particular bre tracts. We performed additional analyses of parallel and perpendicular diffusivities to further investigate the effects found in the right IFG and preSMA, since these diffusion parameters may provide additional information about the underlying white matter microstructure. We found a similar relationship between SSRT and to that observed with FA in the bre tracts within the right appeared weaker and were IFG and preSMA. The effects of

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862

861

not signicant. This suggests that the increase in FA associated with better response inhibition performance is mainly driven by decreased . Although interpretation of changes in DTI parameters is not straightforward, previous studies suggest that may be more sensitive to changes in myelination (Song et al., 2003, 2005). Using a mouse model of retinal ischemia, a DTI and histology study revealed a gradual decrease in relative anisotropy, caused by, at rst, decrease in , corresponding with the timing of axonal degeneration, followed by increase in , associated with optic nerve demyelination (Song et al., 2003). Another rodent study, using a model of experimental de- and remyelination of the corpus callosum, found that continuous cuprizone (neurotoxin) treatment caused demyelination of the callosal bres, reected by increased . When the treatment was discontinued, the effects were reversed, and the progression of bre remyelination was consistent with decrease in (Song et al., 2005). Although does not measure myelin directly, these ndings suggest that the degree of myelination may be contributing to the observed effects in the present study. Age-related FA increases in bre tracts have been linked to decreases in in previous studies (Lebel et al., 2008). However, other tissue parameters, such as axonal diameter, packing density, spacing, or number; extracellular volume fraction; or tract geometry may also contribute to changes in FA and (Beaulieu, 2002; Madler, Drabycz, Kolind, Whittall, & MacKay, 2008; Schwartz et al., 2005). Previous studies have linked the STN to response inhibition (Aron & Poldrack, 2006; van den Wildenberg et al., 2006), and bre tracts connecting the IFG and preSMA with STN pass through the internal capsule (Aron et al., 2007). Due to the lack of clear landmarks delimiting these bres within the internal capsule, we were not able to dene an appropriate ROI in the internal capsule that distinguished these tracts. However, in the effect-size maps displaying the associations between higher FA and better response inhibition performance, clusters with relatively high t-values were observed in the internal capsule (Fig. 6). Though it is not possible to conclude whether or not these connections are contained in the observed clusters, the pattern observed in the effect-size maps suggests that the internal capsule may be an important part of the neural circuit mediating variability in response inhibition in this study. Future studies employing tractography may provide additional evidence. As mentioned above, these ndings suggest that even among children of similar age, higher FA and lower in right IFG and preSMA are associated with better response inhibition performance. Many questions remain about what such age-adjusted variability in FA and might represent. It may reect individual differences in trajectories of bre tract maturation of the neural system subserving response inhibition, i.e., differences in the phase of IFG and preSMA development among children of similar age. It is plausible that this variability could mediate, at least to some extent, the association between response inhibition performance and white matter microstructure found in the present study, since both white matter structure (Lebel et al., 2008) and inhibitory control (Williams et al., 1999) continue to develop across this age range. Alternatively, it could represent individual differences in the structure of the bre tracts (perhaps reecting differences in the underlying neural system connectivity) that emerge earlier during brain development and remain stable in spite of superimposed biological changes associated with development. This is plausible since individual differences in behavioural performance have been associated with FA variability in adults (Forstmann et al., 2008; Gold, Powell, Xuan, Jiang, & Hardy, 2007; Wolbers, Schoell, & Buchel, 2006) as well as children. It is also plausible that the difference in microstructure of the bre tracts are inuenced by dynamic processes, possibly associated with activity levels in the neural circuits. In a recent DTI study, accelerated white matter development (higher FA) in the sagittal

stratum was found in preterm relative to full-term infants at term equivalent age, possibly as a result of increased intensity of sensorimotor stimulation associated with extrauterine life (Gimenez et al., 2008). Thus, the differences observed in the present study could reect variability in the experiences and learning of the children. Finally, genetic variability may play a role in mediating these differences. 5. Conclusion FA and within the right IFG and preSMA exhibit associations with SSRT that are not attributable to age and the associations do not appear to reect behavioural effects of global white matter development. Further, the contributions of right IFG and preSMA FA to the prediction of response inhibition appear to be additive. Children may vary in the phase of maturation in the network subserving response inhibition, and this variability may mediate the associations. Alternatively, the associations could be mediated by individual differences among the children in underlying neural system connectivity, or to more transient differences associated with dynamic (perhaps activity-dependent) processes. Longitudinal observations are needed to help distinguish between these, and other, possibilities. Acknowledgements This work was supported by the Danish Medical Research Council, University of Copenhagens Research Priority Area Body and Mind, and the Lundbeck Foundation. References
Andersson, J. L., Hutton, C., Ashburner, J., Turner, R., & Friston, K. (2001). Modeling geometric deformations in EPI time series. Neuroimage, 13, 903919. Aron, A. R., Behrens, T. E., Smith, S., Frank, M. J., & Poldrack, R. A. (2007). Triangulating a cognitive control network using diffusion-weighted magnetic resonance imaging (MRI) and functional MRI. Journal of Neuroscience, 27, 37433752. Aron, A. R., Fletcher, P. C., Bullmore, E. T., Sahakian, B. J., & Robbins, T. W. (2003). Stopsignal inhibition disrupted by damage to right inferior frontal gyrus in humans. Nature Neuroscience, 6, 115116. Aron, A. R., & Poldrack, R. A. (2006). Cortical and subcortical contributions to Stop signal response inhibition: Role of the subthalamic nucleus. Journal of Neuroscience, 26, 24242433. Aron, A. R., Robbins, T. W., & Poldrack, R. A. (2004). Inhibition and the right inferior frontal cortex. Trends in Cognitive Sciences, 8, 170177. Basser, P. J., Mattiello, J., & LeBihan, D. (1994). MR diffusion tensor spectroscopy and imaging. Biophysical Journal, 66, 259267. Beaulieu, C. (2002). The basis of anisotropic water diffusion in the nervous systemA technical review. NMR in Biomedicine, 15, 435455. Behrens, T. E., Jenkinson, M., Robson, M. D., Smith, S. M., & Johansen-Berg, H. (2006). A consistent relationship between local white matter architecture and functional specialisation in medial frontal cortex. Neuroimage, 30, 220227. Boucher, L., Palmeri, T. J., Logan, G. D., & Schall, J. D. (2007). Inhibitory control in mind and brain: An interactive race model of countermanding saccades. Psychological Review, 114, 376397. Chambers, C. D., Bellgrove, M. A., Stokes, M. G., Henderson, T. R., Garavan, H., Robertson, I., et al. (2006). Executive brake failure following deactivation of human frontal lobe. Journal of Cognitive Neuroscience, 18, 444455. Chambers, C. D., Garavan, H., & Bellgrove, M. A. (2009). Insights into the neural basis of response inhibition from cognitive and clinical neuroscience. Neuroscience and Biobehavioral Review, 33, 631646. Chang, L. C., Jones, D. K., & Pierpaoli, C. (2005). RESTORE: Robust estimation of tensors by outlier rejection. Magnetic Resonance in Medicine, 53, 10881095. Chevrier, A. D., Noseworthy, M. D., & Schachar, R. (2007). Dissociation of response inhibition and performance monitoring in the stop signal task using eventrelated fMRI. Human Brain Mapping, 28, 13471358. Cohen, J. R., Sabb, F. W., Bilder, R. M., Bookheimer, S. Y., Knowlton, B. J., Asarnow, R. F., et al. (2007). Neural substrates of response inhibition in children examined with fMRI. 13th annual meeting of the organization of human brain mapping. Neuroimage, 36(Suppl. 1). Cook, P. A., Bai, Y., Nedjati-Gilani, S., Seunarine, K. K., Hall, M. G., Parker, G. J., et al. (2006). Camino: Open-source diffusion-MRI reconstruction and processing. In 14th scientic meeting of the international society for magnetic resonance in medicine Seattle, WA, USA, (p. 2759). Cook, P. A., Symms, M., Boulby, P. A., & Alexander, D. C. (2007). Optimal acquisition orders of diffusion-weighted MRI measurements. Journal of Magnetic Resonance Imaging, 25, 10511058.

862

K.S. Madsen et al. / Neuropsychologia 48 (2010) 854862 ticomponent T2 relaxation and diffusion tensor anisotropy in human brain. Magnetic Resonance Imaging, 26, 874888. Nachev, P., Wydell, H., Oneill, K., Husain, M., & Kennard, C. (2007). The role of the pre-supplementary motor area in the control of action. Neuroimage, 36(Suppl. 2), T155T163. Nichols, T. E., & Holmes, A. P. (2002). Nonparametric permutation tests for functional neuroimaging: A primer with examples. Human Brain Mapping, 15, 125. Paus, T., Collins, D. L., Evans, A. C., Leonard, G., Pike, B., & Zijdenbos, A. (2001). Maturation of white matter in the human brain: A review of magnetic resonance studies. Brain Research Bulletin, 54, 255266. Paus, T., Zijdenbos, A., Worsley, K., Collins, D. L., Blumenthal, J., Giedd, J. N., et al. (1999). Structural maturation of neural pathways in children and adolescents: In vivo study. Science, 283, 19081911. Pliszka, S. R., Glahn, D. C., Semrud-Clikeman, M., Franklin, C., Perez, R., III, Xiong, J., et al. (2006). Neuroimaging of inhibitory control areas in children with attention decit hyperactivity disorder who were treatment naive or in long-term treatment. American Journal of Psychiatry, 163, 10521060. Reese, T. G., Heid, O., Weisskoff, R. M., & Wedeen, V. J. (2003). Reduction of eddycurrent-induced distortion in diffusion MRI using a twice-refocused spin echo. Magnetic Resonance in Medicine, 49, 177182. Rieger, M., Gauggel, S., & Burmeister, K. (2003). Inhibition of ongoing responses following frontal, nonfrontal, and basal ganglia lesions. Neuropsychology, 17, 272282. Rubia, K., Smith, A. B., Brammer, M. J., & Taylor, E. (2003). Right inferior prefrontal cortex mediates response inhibition while mesial prefrontal cortex is responsible for error detection. Neuroimage, 20, 351358. Rueckert, D., Sonoda, L. I., Hayes, C., Hill, D. L., Leach, M. O., & Hawkes, D. J. (1999). Nonrigid registration using free-form deformations: Application to breast MR images. IEEE Transactions on Medical Imaging, 18, 712721. Schwartz, E. D., Cooper, E. T., Fan, Y., Jawad, A. F., Chin, C. L., Nissanov, J., et al. (2005). MRI diffusion coefcients in spinal cord correlate with axon morphometry. Neuroreport, 16, 7376. Smith, S. M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T. E., Mackay, C. E., et al. (2006). Tract-based spatial statistics: Voxelwise analysis of multi-subject diffusion data. Neuroimage, 31, 14871505. Smith, S. M., Jenkinson, M., Woolrich, M. W., Beckmann, C. F., Behrens, T. E., JohansenBerg, H., et al. (2004). Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage, 23(Suppl. 1), S208S219. Snook, L., Paulson, L. A., Roy, D., Phillips, L., & Beaulieu, C. (2005). Diffusion tensor imaging of neurodevelopment in children and young adults. Neuroimage, 26, 11641173. Song, S. K., Sun, S. W., Ju, W. K., Lin, S. J., Cross, A. H., & Neufeld, A. H. (2003). Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage, 20, 17141722. Song, S. K., Yoshino, J., Le, T. Q., Lin, S. J., Sun, S. W., Cross, A. H., et al. (2005). Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage, 26, 132140. Sowell, E. R., Thompson, P. M., Leonard, C. M., Welcome, S. E., Kan, E., & Toga, A. W. (2004). Longitudinal mapping of cortical thickness and brain growth in normal children. Journal of Neuroscience, 24, 82238231. Stuphorn, V., & Schall, J. D. (2006). Executive control of countermanding saccades by the supplementary eye eld. Nature Neuroscience, 9, 925931. Tomaiuolo, F., MacDonald, J. D., Caramanos, Z., Posner, G., Chiavaras, M., Evans, A. C., et al. (1999). Morphology, morphometry and probability mapping of the pars opercularis of the inferior frontal gyrus: An in vivo MRI analysis. European Journal of Neuroscience, 11, 30333046. van den Wildenberg, W. P., van Boxtel, G. J., van der Molen, M. W., Bosch, D. A., Speelman, J. D., & Brunia, C. H. (2006). Stimulation of the subthalamic region facilitates the selection and inhibition of motor responses in Parkinsons disease. Journal of Cognitive Neuroscience, 18, 626636. Williams, B. R., Ponesse, J. S., Schachar, R. J., Logan, G. D., & Tannock, R. (1999). Development of inhibitory control across the life span. Developmental Psychology, 35, 205213. Wolbers, T., Schoell, E. D., & Buchel, C. (2006). The predictive value of white matter organization in posterior parietal cortex for spatial visualization ability. Neuroimage, 32, 14501455.

Dubois, J., Dehaene-Lambertz, G., Soares, C., Cointepas, Y., Le Bihan, D., & HertzPannier, L. (2008). Microstructural correlates of infant functional development: Example of the visual pathways. Journal of Neuroscience, 28, 19431948. Duvernoy, H. M. (1999). The human brain; surface, blood supply and three-dimensional section anatomy (2nd ed.). Wien: Springer. Eagle, D. M., Baunez, C., Hutcheson, D. M., Lehmann, O., Shah, A. P., & Robbins, T. W. (2008). Stop-signal reaction-time task performance: Role of prefrontal cortex and subthalamic nucleus. Cerebral Cortex, 18, 178188. Eluvathingal, T. J., Hasan, K. M., Kramer, L., Fletcher, J. M., & Ewing-Cobbs, L. (2007). Quantitative diffusion tensor tractography of association and projection bers in normally developing children and adolescents. Cerebral Cortex, 17, 27602768. Enright, S. J., & Beech, A. R. (1993). Reduced cognitive inhibition in obsessivecompulsive disorder. British Journal of Clinical Psychology, 32(Pt 1), 6774. Floden, D., & Stuss, D. T. (2006). Inhibitory control is slowed in patients with right superior medial frontal damage. Journal of Cognitive Neuroscience, 18, 18431849. Forstmann, B. U., Jahfari, S., Scholte, H. S., Wolfensteller, U., van den Wildenberg, W. P., & Ridderinkhof, K. R. (2008). Function and structure of the right inferior frontal cortex predict individual differences in response inhibition: A modelbased approach. Journal of Neuroscience, 28, 97909796. Giedd, J. N., Blumenthal, J., Jeffries, N. O., Castellanos, F. X., Liu, H., Zijdenbos, A., et al. (1999). Brain development during childhood and adolescence: A longitudinal MRI study. Nature Neuroscience, 2, 861863. Gimenez, M., Miranda, M. J., Born, A. P., Nagy, Z., Rostrup, E., & Jernigan, T. L. (2008). Accelerated cerebral white matter development in preterm infants: A voxelbased morphometry study with diffusion tensor MR imaging. Neuroimage, 41, 728734. Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D., Vaituzis, A. C., et al. (2004). Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences United States of America, 101, 81748179. Gold, B. T., Powell, D. K., Xuan, L., Jiang, Y., & Hardy, P. A. (2007). Speed of lexical decision correlates with diffusion anisotropy in left parietal and frontal white matter: Evidence from diffusion tensor imaging. Neuropsychologia, 45, 24392446. Hagler, D. J., Jr., Ahmadi, M. E., Kuperman, J., Holland, D., McDonald, C. R., Halgren, E., et al. (2009). Automated white-matter tractography using a probabilistic diffusion tensor atlas: Application to temporal lobe epilepsy. Human Brain Mapping, 30(5), 15351547. Inase, M., Tokuno, H., Nambu, A., Akazawa, T., & Takada, M. (1999). Corticostriatal and corticosubthalamic input zones from the presupplementary motor area in the macaque monkey: Comparison with the input zones from the supplementary motor area. Brain Research, 833, 191201. Isoda, M., & Hikosaka, O. (2007). Switching from automatic to controlled action by monkey medial frontal cortex. Nature Neuroscience, 10, 240248. Jansons, K. M., & Alexander, D. C. (2003). Persistent angular structure: New insights from diffusion MRI data. Information Processing in Medical Imaging, 18, 672683. Dummy version. Jernigan, T. L., Gamst, A. C., Fennema-Notestine, C., & Ostergaard, A. L. (2003). More mapping in brain mapping: Statistical comparison of effects. Human Brain Mapping, 19, 9095. Jernigan, T. L., Trauner, D. A., Hesselink, J. R., & Tallal, P. A. (1991). Maturation of human cerebrum observed in vivo during adolescence. Brain, 114(Pt 5), 20372049. Johansen-Berg, H., Behrens, T. E., Robson, M. D., Drobnjak, I., Rushworth, M. F., Brady, J. M., et al. (2004). Changes in connectivity proles dene functionally distinct regions in human medial frontal cortex. Proceedings of the National Academy of Sciences United States of America, 101, 1333513340. Lebel, C., Walker, L., Leemans, A., Phillips, L., & Beaulieu, C. (2008). Microstructural maturation of the human brain from childhood to adulthood. Neuroimage, 40, 10441055. Liston, C., Watts, R., Tottenham, N., Davidson, M. C., Niogi, S., Ulug, A. M., et al. (2006). Frontostriatal microstructure modulates efcient recruitment of cognitive control. Cerebral Cortex, 16, 553560. Logan, G. D., & Cowan, W. B. (1984). On the ability to inhibit thought and action: A theory of an act of control. Psychological Review, 91, 295327. Madler, B., Drabycz, S. A., Kolind, S. H., Whittall, K. P., & MacKay, A. L. (2008). Is diffusion anisotropy an accurate monitor of myelination? Correlation of mul-

Paper 2

Brain microstructural correlates of visuospatial choice reaction time in children

Kathrine Skak Madsen1,2,3, William F. C. Baar1,2, Arnold Skimminge1, Martin Vestergaard1, Hartwig R. Siebner1,2,3 and Terry L. Jernigan1,2,3,4
1 2 3 4

Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, Copenhagen, Denmark Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark Center for Human Development, University of California, San Diego, La Jolla, CA, US

Abbreviated title: Microstructure linked to reaction time

Corresponding Author Kathrine Skak Madsen Danish Research Centre for Magnetic Resonance Copenhagen University Hospital, Hvidovre Kettegaard All 30 DK-2650 Hvidovre, Denmark Email: Kathrine@drcmr.dk, Phone: +45 38623323, Fax: +45 36470302

The authors have no conflict of interests

Acknowledgements This work was supported by the Danish Medical Research Council and the Lundbeck Foundation.

Abstract
The corticospinal tracts and the basal ganglia continue to develop during childhood and adolescence, and indices of their maturation can be obtained using diffusion-weighted imaging. Here we show that a simple measure of visuomotor function is correlated with diffusion parameters in the corticospinal tracts and neostriatum. In a cohort of 75 typically-developing children aged 7 to 13 years, mean 5choice reaction times (RTs) were assessed. We hypothesized that children with faster choice RTs would show lower mean diffusivity (MD) in the corticospinal tracts and neostriatum and higher fractional anisotropy (FA) in the corticospinal tracts, after controlling for age, gender, and handedness. Mean MD and/or FA were extracted from the right and left corticospinal tracts, putamen, and caudate nuclei. As predicted, faster 5-choice RTs were associated with lower MD in the corticospinal tracts, putamen, and caudate. MD effects on RT were bilateral in the corticospinal tracts and putamen, while right caudate MD was more strongly related to performance than was left caudate MD. Our results suggest a link between motor performance variability in children and diffusivity in the motor system, which may be related to: individual differences in the phase of fibre tract and neostriatal maturation in children of similar age, individual differences in motor experience during childhood (i.e., usedependent plasticity), and/or more stable individual differences in the architecture of the motor system.

Introduction
Perceptual-motor speed continues to develop throughout childhood and adolescence (Kail 1991), yet little is known about how increasing perceptual-motor expertise is related to structural brain development. Increasing speed of processing can involve a wide range of sensory, cognitive and motor functions, and has been studied at the behavioural level with many different choice reaction time (RT) tasks (Konrad et al. 2009; Madden et al. 2004; Miller and Low 2001). In this study, we wished to clarify whether choice RT on a simple behavioural task was correlated with microstructure within the motor network. Our focus was on core motor brain structures, namely the major motor output pathway, the corticospinal tract (CST), and the neostriatum (putamen and caudate nuclei), which has been implicated in movement selection, initiation and execution (Bohr et al. 2007; Gerardin et al. 2004; Haber and Gdowski 2004). The CSTs and neostriatum continue to develop during childhood and adolescence (Jernigan et al. 1991; Lebel et al. 2008; Sowell et al. 1999), and indices of their maturation can be obtained using diffusion-weighted imaging (DWI). DWI is sensitive to the diffusion of water molecules. In white matter, water diffusion is less hindered parallel than perpendicular to the fibre bundles, ostensibly due to occluding cellular structures, i.e., axonal membranes and surrounding myelin sheaths, causing diffusion anisotropy. In grey matter, diffusion is also hindered by cellular structures, but it is relatively isotropic (at clinical voxel sizes) (Beaulieu 2009). By employing the diffusion tensor model, useful measures can be derived, including fractional anisotropy (FA) estimating the degree of diffusion directionality, mean diffusivity (MD), and diffusivity parallel (||) and perpendicular () to the principal diffusion direction (Basser et al. 1994; Beaulieu 2009). Studies applying DWI in children and adolescents have reported age-related increases in FA and/or decreases in MD within both white and grey matter structures (Eluvathingal et al. 2007; Lebel et al. 2008; Snook et al. 2005). Lebel et al. (2008) observed that diffusion parameters within the motor network, including the CST, putamen, and caudate nuclei, exhibit protracted age-related changes continuing into early adulthood. Focusing on the subjects studied by Lebel et al. (2008) within the current samples age range (7 to 13 years), strong age-related reductions in MD are apparent in both CSTs and neostriatal structures across this range. Age-related increases in FA were also prominent in CSTs, but neostriatal FA changes were more variable, and less strongly related to age, in this age group. A growing number of DWI studies in children have linked regional white matter microstructure to performance levels on, e.g., reading (Niogi and McCandliss 2006), response inhibition (Madsen et al. 2010), and working memory (Vestergaard et al. 2010). In healthy young right-handed adults, faster right- relative to left-handed responses have been correlated with higher FA in the left corticospinal tract in an auditory simple reaction time task (Bohr et al. 2007). However, no studies have examined the relationship between the microstructure of motor pathways and perceptual-motor speed as

reflected by choice RT in developing children. Here we tested the hypotheses that in typicallydeveloping children faster 5-choice RT would be associated with lower MD in the CSTs and neostriatum, and higher FA in the CSTs, independently of age.

Materials and Methods

Subjects
Ninety-two typically-developing children and adolescents (53 girls, 39 boys) aged 7 to 13 years from three schools (1st-6th graders) in the Copenhagen suburban area participated in the study. Prior to participation and after receiving oral and written explanation of the study aims and study procedures, all children assented to the procedures and informed written consent was obtained from the parents/guardians of all subjects. According to parent reports, no subjects had any known history of neurological or psychiatric disorders or significant brain injury. The study was approved by the local Danish Committee for Biomedical Research Ethics (H-KF-01-131/03) and conducted in accordance with the Declaration of Helsinki. Seventeen (10 girls, 7 boys) of the 92 recruited children were subsequently excluded from further analyses because of technical problems with the behavioural task (1 subject), incidental findings on MRI (1 subject), reduced DWI image quality as described below (12 subjects) or because they did not complete the scanning session (3 subjects). There were no significant differences between the 75 included and the 17 excluded subjects with respect to age, gender, parental education, or handedness (as assessed with the Edinburgh handedness inventory). Demographic data from the included subjects are presented in Table 1. Table 1. Demographic data for the included subjects. 1st/2nd graders Age (mean std) Gender (female/male) Handedness (right/left) Parents average years of education (mean std) 8.26 0.46 13 / 10 20 / 3 14.0 1.7 3rd/4th graders 10.11 0.37 16 / 13 25 / 4 13.4 2.0 5th/6th graders 12.1 0.42 14 / 9 21 / 2 13.7 2.0 All 10.17 1.59 43 /32 66 / 9 13.7 1.9

* Children enrolled in the study were scanned either in the months just before (when in 1st, 3rd or 5th grade) or just after (when in 2nd, 4th or 6th grade) the summer holiday.

Reaction time task

The experimental task consisted of a computerized visuospatial 5-choice RT task implemented in the Cambridge Neuropsychological Test Automated Battery (Cambridge Cognition Ltd., Cambridge, UK), and was the first test out of a series of tests within a 1.5 hour test battery. The task required detection of a spatially-varying visual target stimulus and selection, initiation, and execution of a

manual response spatially matched to that target. The children were sitting in front of a touch screen and were instructed to press a button on a response pad with the index finger of the dominant hand. After a random delay, between 750 and 2250 ms, a yellow dot was presented in one of the circles displayed on the screen, and subjects were required to release the button and touch the yellow dot on the screen as fast as possible. In an initial block of trials, the yellow dot appeared in a single circle located centrally on the screen. In the subsequent block of 5-choice trials, the yellow dot appeared (randomly) in one of five circles visible on the screen (Figure 1). Both blocks were divided into a practice and a test phase. The training phases each consisted of 5 trials. Subjects who failed to perform 4 of the 5 practice trials correctly received a second practice phase before the test phase. Each test phase consisted of 15 trials. For each block, button release time, movement time, and accuracy scores were recorded. The behavioural measure used to test the hypotheses was button release time for the 5-choice test trials (referred to as 5-choice RT), which estimates the time taken to select and initiate the movement. Error rates were very low (47 of 75 subjects produced no errors). However, data from error trials were excluded from the calculations. Z-scores for trial RTs were computed within each subject (by block) and trials with a Z-score above 3 were excluded (as outliers) from the calculation of the mean RTs. Movement times were not used for the analyses, because this measure clearly reflected not only individual differences in movement speed, but also individual differences in response style. The subjects were required to physically touch the display screen. Some subjects consistently decelerated their movements as they approached the screen in order to touch it softly, while others elected to stab the screen with more force.

Figure 1. Set-up of the 5-choice reaction time (RT) task. Subjects sit in front of a touch screen and are instructed to press a button on a pad in front of them. After a random delay, a yellow dot appears in one of five circles. Subjects must release the button as quickly as possible and touch the yellow dot. The 5-choice RT was defined as the average time it took to release the button on accurate choice RT trials.

Image acquisition
All subjects were scanned using a 3T Siemens Magnetom Trio MR scanner (Siemens, Erlangen, Germany) with an eight-channel head coil (Invivo, FL, USA). All acquired scans were aligned parallel

to the anterior commissureposterior commissure line. T1-weighted images of the whole head were acquired using a 3D MPRAGE sequence (TR = 1550 ms, TE = 3.04 ms, matrix 256 x 256, 192 sagittal slices, 1 x 1 x 1 mm3 voxels, acquisition time = 6:38). T2-weighted images of the whole head were acquired using a 3D turbo spin echo sequence (TR = 3000 ms, TE = 354 ms, FOV = 282 x 216, matrix = 256 x 196, 192 sagittal slices, 1.1 x 1.1 x 1.1 mm3 voxels, acquisition time = 8:29). Whole brain diffusion-weighted (DW) images were acquired using a twice-refocused balanced spin echo sequence that minimized eddy current distortion (Reese et al. 2003). Ten non-DW images (b = 0) and 61 DW images (b = 1200 s/mm2), encoded along independent collinear diffusion gradient orientations, were acquired (TR = 8200 ms, TE = 100 ms, FOV = 220 x 220, matrix = 96 x 96, GRAPPA: factor = 2, 48 lines, 61 transverse slices with no gap, 2.3 x 2.3 x 2.3 mm3 voxels, acquisition time = 9:50). A gradient echo field map was acquired to correct B0 field distortions (TR = 530 ms, TE[1] = 5.19 ms and TE[2] = 7.65 ms, FOV = 256 x 256; matrix = 128 x 128, 47 transverse slices with no gap, voxel size = 2 x 2 x 3 mm3, acquisition time = 2:18).

Image evaluation
All subjects images were evaluated by an experienced neuroradiologist. Prior to image analysis, and blind to behavioural data, the raw images from all subjects were visually inspected to ascertain the quality of the data. Based on this inspection, 12 subjects were deemed to have significantly reduced DW image quality due to movement or susceptibility artefacts and were excluded from further analysis. An additional 6 subjects (3 boys and 3 girls, 3 right- and 3 left-handers) had reduced T1weighted image quality and were not included in analyses involving grey matter regions-of-interest (ROIs), but were included in all other analyses.

Image preprocessing
Images were preprocessed using pipelines implemented in Matlab, using mainly SPM5 (Wellcome Department of Cognitive Neurology, University College London, UK) routines. T1-weighted and T2weighted images were corrected for spatial distortions due to non-linearity in the gradient system of the scanner (Jovicich et al. 2006). The T2-image was coregistered (no reslicing), using a 6-parameter mutual information rigid transformation to the T1-image. In the DWI analysis, each subjects mean b0 image was coregistered (no reslicing), to the T2-image, after which all DW images were coregistered (no reslicing) to the mean b0 image. Next, all coregistered images were corrected for geometric distortions using a voxel displacement map based on both the acquired B0 field map (Andersson et al. 2001) and the scanner specific gradient non-linearities (Jovicich et al. 2006). Finally, all images were resliced using trilinear interpolation. Note that this procedure involves only one reslicing step. The diffusion gradient orientations were adjusted to account for any rotation applied during registration. The diffusion tensor was fitted using the RESTORE algorithm with a noise standard deviation of 30 (Chang et al. 2005) implemented in Camino, and fractional anisotropy (FA), mean diffusivity (MD =

(1 + 2 + 3) / 3) as well as diffusivity parallel (|| = 1) and perpendicular ( = (2 + 3) / 2) to the principal diffusion direction were calculated. A brain mask based on the T2-weighted image was applied to the FA and diffusivity images.

Inter-subject spatial normalization of fibre tracts

In the present study, we extracted FA and diffusivity measures from regions-of-interest (ROIs) to test specific hypotheses and to determine the anatomical specificity of observed associations (see below). Spatial normalization and alignment of fibre tracts across subjects was achieved using the Tract-Based Spatial Statistics (TBSS) module (Smith et al. 2006), part of FSL 4.1.4 (Smith et al. 2004). At first, all subjects FA images were aligned into a common space using the non-linear registration tool FNIRT (Andersson et al., 2007). A study-specific target, the groups most representative FA image, was then identified after non-linearly registering each subjects FA image to every other subjects FA image. Next, the target FA image was aligned to MNI space using affine registration and subsequently the entire aligned dataset was transformed into 1 mm3 MNI space. A cross-subject mean FA image was created and thinned to create a mean FA skeleton, representing the centres of all tracts common to the group. The mean FA skeleton was thresholded at FA > 0.25, and contained 103,588 1 mm3 interpolated isotropic voxels, corresponding to approximately one quarter of the voxels with FA above 0.25. Each subjects aligned FA image was then projected onto the mean skeleton by locating the highest local FA value in the direction perpendicular to the skeleton tracts and assigning this value to the skeleton. Finally, a symmetric mean FA map based on the average of the mean FA map and its left-right flipped counter part was generated, and subsequently thinned and masked with a (one voxel) dilated version of the original skeleton in order to generate a symmetric mean skeleton. The symmetric mean skeleton contained 102,154 voxels. In addition, the nonlinear warps, skeleton projections and symmetric skeleton generation were applied to the MD, || and data.

Inter-subject spatial normalization of grey matter

Both T1 and T2 images were processed using the VBM5 toolbox in SPM5 (http://dbm.neuro.unijena.de/vbm/vbm5-for-spm5/), which includes a unified segmentation algorithm (Ashburner and Friston 2005) and a Hidden Markov Random Field (HMRF) method (Cuadra et al. 2005). T2-weighted images were used to create brain masks in native space. Resulting brain masked grey and white matter tissue maps in native space, the affine part of the spatial transformation from native to MNI space (obtained from the VBM5 analysis), and their left-right flipped counter parts were subsequently used in DARTEL (Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra) using default settings (Ashburner 2007), for high-dimensional inter-subject registration. Including the leftright flipped tissue maps ensured that resulting flow fields that parameterize the deformations were symmetric. Using the final symmetric ow elds, the native T1-images were warped into average symmetric image space (DARTEL space).

Regions-of interest (ROIs)

White matter ROIs ROIs were drawn onto the mean symmetric skeleton overlaid on the mean symmetric FA image using FSLview. Corticospinal tract (CST) ROIs (Figure 2a,c) were delineated using information from the probabilistic fibre atlas (Hua et al. 2008) implemented in FSLview. The ROIs each contained 2,434 voxels. Mean FA, MD, || and values from the left and right CST, and from the whole skeleton were extracted for each subject for statistical analyses. Note that for the left-handed subjects, the left and right CST ROI values were flipped after extraction, so that data corresponded with the contra- and ipsilateral CST to the dominant hand, which was used in the RT task, i.e. left CST for the right-handed subjects and right CST for the left-handed subjects corresponds to the contralateral CST.

Figure 2. (a) The corticospinal tract ROIs in red and the symmetric TBSS skeleton in yellow overlaid on the group mean symmetric FA map. (b) In the top images, DARTEL space ROIs in the putamen (magenta), caudate nucleus (blue), and the combined nucleus accumbens and amygdala (cyan) overlaid on the symmetric average of the DARTEL-warped T1-images of all subjects. The four images in the bottom right illustrate the thresholded ROIs in native DWI space of two individuals. (c) Smoothed 3D versions of the ROIs depicted together with a representation of the right hemisphere of the symmetric average DARTEL-warped grey and white matter segmentations.

Grey matter ROIs ROIs of the caudate nuclei and putamen were constructed to extract the MD and volume of these structures. In order to examine the anatomical specificity of the neostriatal effects, we constructed an ROI of nearby, but functionally distinct, subcortical grey matter, namely the combined nucleus

accumbens (nAcc) and amygdala (nAcc/amygdala). This ROI thus serves as a control region for comparison to the neostriatal ROIs. ROIs were drawn onto the symmetric average of the DARTELwarped T1-images of all subjects (Fig. 2b,c). To ensure that the striatal ROIs for which MD was extracted included only fully-volumed grey matter, an approximately one-voxel thick layer of grey matter at the outer border was not included. Resulting binary ROIs were resliced into native T1-space using inverse deformation fields and trilinear interpolation, and then binarized using a threshold of 0.5. To transform ROIs from native T1-space to native DTI space all subjects FA images were warped to the subjects own native T1-image using FNIRT. Using the inverse warp field, ROIs were resliced into native DTI space using trilinear interpolation and binarized. The binary ROIs were multiplied with the subjects own FA image thresholded at FA < 0.25 to exclude potential white matter (partially-volumed) voxels. The mean ( standard deviation (std)) volume of the ROIs, from which MD were extracted, were: left putamen = 4,311 (432), right putamen = 4,011 (444), left caudate nucleus = 2,750 (348), right caudate nucleus = 2,918 (372), left nAcc/amygdala = 1,093 (120), and right nAcc/amygdala = 1,093 (120). Mean MD values were extracted from all ROIs for each subject and used in statistical analyses. Four ROIs including the whole left and right putamen and caudate nuclei were delineated to estimate the volumes of these structures. The mean ( std) number of 1 mm3 voxels in the ROIs were: left putamen = 4,544 (460), right putamen = 4,562 (455), left caudate nucleus = 3,954 (424), right caudate nucleus = 3,954 (443). Further, total brain volume was extracted from all subjects (mean std = 1,300,289 118,353). For use in statistical analyses, relative volumes of the putamen and caudate were calculated: (ROI / brain volume). Note that for the left-handed subjects, the values of the left and right ROIs were flipped after extraction, to yield ROIs that were contra- and ipsilateral to the dominant hand.

Statistical analysis
Statistical analyses were performed in SPSS software (SPSS 19, SPSS Inc., Chicago, USA). Analyses involving white matter ROIs were based on 75 individuals, while analyses involving grey matter ROIs included data from 69 individuals. Two-tailed t-tests were used to compare boys and girls on the behavioural and ROI measures. Age effects on the behavioural and ROI measures, adjusted for gender and handedness, were examined using linear regression models. Specific hypotheses were tested using multiple linear regression models. To fulfil assumptions of normality of the residuals, mean RTs were transformed using the equation: 1 (1 / mean RT). After transforming the behavioural data, all data were approximately normally distributed according to Shapiro-Wilk tests. All other assumptions for linear regression were fulfilled. The statistical tests were performed hierarchically. Our main hypotheses were that faster 5-choice RT would be associated with lower MD in the CSTs and neostriatum as well as with higher FA in the CSTs, after adjusting for age, gender and

handedness. These hypotheses were tested at a statistical threshold of p 0.017 (Bonferroni corrected for 3 tests), while follow-up analyses were considered significant when p 0.05. Planned follow-up analyses were contingent on observing significant associations with MD or FA in the primary analyses. Follow-up analyses were performed 1) to assess the anatomical specificity of the effects, either by adjusting for whole skeleton MD in models with CST MD, or by adjusting for nAcc/amygdala MD in models with neostriatal MD; 2) to assess the contribution of the ipsi- and contralateral ROIs to the effects; and 3) to determine whether a measure of simple RT mediated the effects on choice RT. To further explore the nature of the observed CST MD effects, || and were examined, since these diffusion parameters sometimes provide additional information about the underlying white matter microstructure. Additional planned follow-up analyses were performed to assess the relative contributions of putamen and caudate nucleus MD, as well as the degree to which putamen and caudate volumes may have contributed to the observed effects. Finally, based on the results, post hoc analyses were performed on caudate MD to assess whether the effect was lateralized, by modelling ipsi- and contralateral caudate MD simultaneously, at first for the whole subject group, and then for the right- and left-handers separately. Two effect-size maps are presented to provide further anatomical information about the association between MD and 5-choice RT across the white matter skeleton and the neostriatum (Jernigan et al. 2003). Both maps were only based on the right-handed subjects, and included data from, respectively, 66 and 63 subjects for the skeleton map and the neostriatum map. The effect-size maps are t-maps of the correlation between MD and 5-choice RT, adjusted for age and gender. Positive correlations, i.e., lower MD with faster 5-choice RTs, are shown in warm colours (red to yellow). The effect-size map of the neostriatum was based on the warped MD image smoothed with a 4 mm FWHM Gaussian kernel.

Results
Simple and 5-choice RTs as well as diffusion parameter measures of the ROIs are summarized in Table 2. Separate statistics are provided for boys and girls, but no gender differences approached significance (ps > 0.17), except for putamen MD (p = 0.043) and nAcc/amygdala MD (p = 0.055). Scatter plots of the relationship between age and the main variables are shown in Figure 3. Significant age effects, after adjusting for gender and handedness, were observed on the behavioural measures (5choice RT: p = 10-7; simple RT: p = 10-6) and on the ROI MD (CST: p = 0.00004; neostriatal: p = 0.00008, nAcc/amygdala: p = 0.05) and FA (CST: p = 0.0004) measures, but not on the neostriatal volumes (ps 0.15). All models below were adjusted for age, gender and handedness, unless otherwise noted.

10

Table 2. Mean standard deviation of the behavioural measures and DTI measures within the regionsof-interest. All subjects Behavioural measure Mean 5-choice RT (ms) a Mean simple RT (ms) Regions-of-interest CST FA a CST MD (10 mm /s)
-3 -3 -3 2 a 2 b a

Girls 361.2 65.8 330.0 59.3 0.568 0.021 0.738 0.020 0.810 0.017 0.796 0.018 0.830 0.020 0.852 0.019

Boys 352.0 51.6 317.6 46.4 0.572 0.022 0.734 0.022 0.807 0.019 0.786 0.020 0.838 0.024 0.862 0.024

357.3 59.9 324.7 54.2 0.570 0.022 0.736 0.021 0.809 0.018 0.791 0.019 0.833 0.022
2 b b b

Neostriatal MD (10 mm /s) Putamen MD (10 mm /s) Caudate MD (10 mm /s)

-3 a
b

-3

nAcc/amygdala MD (10 mm /s)

0.858 0.022

Data from 75 subjects; Data from 69 subjects. Abbreviations: RT = reaction time, CST = corticospinal tract, FA = fractional anisotropy, MD = mean diffusivity, nAcc = nucleus accumbens.

Figure 3. Scatter plots of a) 5-choice reaction time (RT transformed values), b) corticospinal tract (CST) FA, c) neostriatal MD and d) CST MD by age (in years). To visualize the relationships across age, a model-free smoothing spline ( =70) was applied to the data.

Associations between 5-choice RT and diffusion parameters of the corticospinal tract

Results obtained from the regression models that tested our main hypotheses regarding diffusion parameters of the CST are listed in Table 3, where each row represents a separate planned model predicting 5-choice RT. As hypothesized, faster 5-choice RT was significantly associated with lower CST MD (Table 3(2), Figure 4a). The association became stronger when excluding age from the model (Table 3(2a)). No significant association was found between 5-choice RT and CST FA (Table 3(1)).

11

Table 3. Multiple linear regression models predicting 5-choice reaction time with diffusion parameters within the corticospinal tracts Corticospinal tracts Model 1 2 2a 2b R2 0.374 0.424 0.276 0.455 -0.176 0.310 0.510 0.715 p 0.095 0.004 0.000004 0.002 -0.505 -0.433 . -0.457 Age p <10-5 <0.0001 . <0.0001 Whole skeleton . . . -0.451 p . . . 0.049

Fractional anisotropy Mean diffusivity

Each row in the table represents a separate regression model. Model 1 and 2 were adjusted for age, gender and handedness. Model 2a were adjusted for gender and handedness. Model 2b was adjusted for age, gender, handedness and whole skeleton MD. For each model, R2 is given in the leftmost column, and the regression coefficient () and the significance level (p) are given for each variable included in the model. Gender and handedness effects were non-significant (ps 0.27), and are not included in the table.

Follow-up analysis, testing the anatomical specificity of the observed MD effect, showed that CST MD remained significant when including mean MD of the whole skeleton in the model (Table 3(2b)), suggesting that the association between 5-choice RT and MD in the CSTs is not mediated by a global decrease in MD. Assessing the relative contribution of the ipsi- and contralateral CST to the observed effect revealed that MD in both the ipsilateral ( = 0.326, p = 0.005) and contralateral CST ( = 0.339, p = 0.004) were significantly and positively associated with 5-choice RT, suggesting a bilateral CST effect. Moreover, the CST MD remained a significant predictor of 5-choice RT ( = 0.204, p = 0.003) when including simple RT ( = 0.701, p = 10-14) as a covariate in the model. By itself, simple RT did not exhibit a significant relationship with CST MD ( = 0.15, p = 0.18). To further explore the nature of the observed CST MD effect, the || and were examined, revealing that faster 5-choice RT was significantly associated with both lower CST || ( = 0.222, p = 0.023) and lower CST ( = 0.244, p = 0.023).

Figure 4. Partial regression plots of the 5-choice reaction time (RT) as a function of MD within the (a) corticospinal tracts or (b) neostriatum, adjusted for age, gender and handedness. The regression coefficient () and the significance level (p) for each MD variable are given in the lower right corner of each plot. Note that residuals are plotted.

12

Associations between 5-choice RT and mean diffusivity of the neostriatum

Testing our main hypothesis regarding neostriatal MD revealed that faster 5-choice RT was significantly associated with lower neostriatal MD (Table 4(4), Figure 4b). Further, this association became stronger when excluding age from the model (Table 4(4a)). Follow-up analysis showed that the association between 5-choice RT and neostriatal MD remained significant (Table 4(4b)) when including nAcc/amygdala MD in the model as a control region, suggesting that the effects are relatively specific to the neostriatum. Further, MD in both the ipsilateral ( = 0.319, p = 0.005) and the contralateral neostriatum ( = 0.379, p = 0.044) were significantly and positively associated with 5-choice RT, when modelled separately. Moreover, neostriatal MD ( = 0.157, p = 0.031) remained a significant predictor of 5-choice RT when including simple RT ( = 0.720, p = 10-14) as an additional covariate in the model. By itself, simple RT was not significantly associated with neostriatal MD ( = 0.183, p = 0.122).

Table 4. Multiple linear regression models predicting 5-choice RT with mean diffusivity (MD) of striatal regions-of-interest (ROIs)

Model 3 3a 3b 4 4a 4b 5 5a 5b R2 0.400 0.238 0.402 0.395 0.227 0.397 0.381 0.178 0.384

MD in ROI* 0.282 0.485 0.311 0.271 0.478 0.260 0.234 0.429 0.284 p 0.012 0.00003 0.019 0.017 0.00005 0.027 0.037 0.0005 0.057 *Putamen *Caudate nuclei

Age p 0.0001 . 0.0001 <0.0001 . 0.0001 <0.0001 . <0.0001

nAcc/amygdala MD . . -0.052 . . -0.044 . . -0.073 p . . 0.672 . . 0.682 . . 0.600

*Neostriatum -0.451 . -0.450 -0.458 . -0.452 -0.489 . -0.487

Each row in the table represents a separate regression model. Models 3, 4 and 5 were adjusted for age, gender and handedness. Models with a were adjusted for gender and handedness. Models with b were adjusted for age, gender, handedness, and nucleus accumbens (nAcc)/amygdala MD. For each model, R2 is given in the leftmost column, and the regression coefficient () and the significance level (p) are given for each variable included in the model. Gender and handedness effects were non-significant (ps 0.29), and are not included in the table.

Relative contributions of caudate nucleus and putamen in predicting 5-choice RT

13

Results from the follow-up analyses of regional MD in the putamen and caudate nuclei are given in Table 4. Faster 5-choice RT was significantly associated with lower MD in both the caudate (Table 4(4)) and the putamen (Table 4(5)), and the effects became stronger when excluding age as a covariate in the models (Table 4(4a,5a)). Further, effects of caudate and putamen MD remained significant when including nAcc/amygdalar MD as an additional covariate in the models (Table 4(4b,5b)). Follow-up analyses were performed to assess the potential influence of individual variation in striatal volumes on the observed effects by including the relative volume of the putamen or caudate nucleus as an additional covariate in the models. Both caudate MD ( = 0.261, p = 0.018) and putamen MD ( = 0.231, p = 0.046) remained significant predictors of 5-choice RT when including, respectively, caudate volume ( = -0.066, p = 0.51) and putamen volume ( = 0.012, p = 0.91) in the models, suggesting that the association between neostriatal MD and 5-choice RT is not mediated by differences in striatal volumes. By themselves, neither caudate volume nor putamen volume was significantly associated with 5-choice RT (ps 0.49). Moreover, neither caudate MD, nor putamen MD was significantly correlated with, respectively, caudate volume ( = -0.030, p = 0.83) or putamen volume ( = 0.288, p = 0.095), indicating that volume and MD may reflect distinct biological properties of the neostriatum. Follow-up analyses assessed the relative contributions of MD in the ipsi- and contralateral putamen and caudate on 5-choice RT. When modelled individually, the effects of ipsilateral ( = 0.221, p = 0.050) and contralateral ( = 0.215, p = 0.051) putamen MD on 5-choice RT were of similar size, suggesting that the effect of putamen MD was bilateral. However, for caudate MD, the association between 5-choice RT and MD in the ipsilateral caudate ( = 0.310, p = 0.008) appeared to be stronger than that for the contralateral caudate ( = 0.180, p = 0.097), which only approached significance. To further investigate the laterality of the caudate MD effect, post hoc analyses were conducted, in which ipsi- and contralateral caudate MD were included in the same model. When modelled simultaneously, ipsilateral ( = 0.312, p = 0.038), but not contralateral caudate MD ( = 0.003, p = 0.98) exhibited a significant relationship with 5-choice RT. Analyzing the right-handers (n=63) and left-handers (n=6) separately revealed that for the right-handers, MD in the right caudate ( = 0.309, p = 0.050), but not the left caudate ( = -0.013, p = 0.93) exhibited a relationship with 5choice RT. For the left-handed subjects, 5-choice RT appeared to exhibit a stronger, though not significant, relationship with MD in the right caudate ( = 1.244, p = 0.23) relative to that of the left caudate ( = -0.048, p = 0.92). Thus, the association between 5-choice RT and caudate MD may be, to some degree, lateralized.

Effect-size maps
The primary goal of this study was to test specific anatomical hypotheses about the relationship between microstructure of the CST and neostriatum and variability in 5-choice RT performance in children. Therefore neither a whole brain, voxel-wise analysis of the effects (appropriately adjusted for

14

test-multiplicity), nor a restricted voxel-wise analysis with small volume correction was deemed appropriate for testing these a priori hypotheses. However, since the analyses we used produce estimates of the effect size at each voxel, we have provided visualization of the effect-size maps. The maps show the distribution of t-values in skeleton (Figure 5) or neostriatal (Figure 6) voxels and thus, provide complementary information about the regional distribution of associations between 5-choice RT performance and MD in the brain, adjusted for age and gender.

Figure 5. Effect-size map based on data from the right-handed subjects (N=66) of the association between 5choice reaction time and mean diffusivity, adjusted for age and gender. The effect-size map is overlaid on the mean symmetric FA images. The association of faster 5-choice RT with lower MD values yields positive tvalues shown in warm colours, while negative t-values are shown in cool colours. The colour bar shows the colour mapping of all skeleton voxels to the values of the uncorrected t-statistics (t = 0 = p = 1; t = 1.294 = p = 0.2; t = 1.995 = p = 0.05; t = 2.648 = p = 0.01). The MNI coordinates for the coronal images are given under each image. In the lower right corner, the coronal slices shown are depicted on the midsagittal image with the coronal slice 35 displayed to the far left and coronal image -78 to the far right. Images are shown with the subjects right side on the right (neurological convention). The CST ROIs were located between slice -10 and 35.

15

For the right-handed children only, the map of the TBSS skeleton revealed some areas with apparent relationships between MD and 5-choice RT performance comparable with those in the CSTs (Figure 4), although it should be emphasized that these maps show uncorrected t-values. The symmetric skeleton as a whole contained 102,154 voxels, of which 10,308 voxels, or 10.1 %, displayed t-values greater than 1.995 (a value associated with an uncorrected p < 0.05). In comparison, 38.0 % of the voxels within the left CST ROI and 34.0 % of the voxels within the right CST ROI had t-values exceeding this value. Additionally, to further compare the results from the ROI analyses with the number of uncorrected significant voxels within the CSTs, a t-map based on all 75 subjects (lefthanders left-right flipped) included in the CST ROI analyses was generated (not shown). In this map, 9.6 % of the skeleton voxels had associated t-values greater than 1.995, while 43.3 % of the voxels within the ipsilateral CST ROI and 62.0 % of the voxels within the contralateral CST ROI had t-values exceeding this value.

Figure 6. Effect-size map based on the right-handed subjects (N=63) of the association between 5-choice reaction time and mean diffusivity (MD), adjusted for age and gender. The effect-size map is overlaid on the symmetric average of the DARTEL-warped T1-images of all subjects. The association of faster 5-choice RT with lower MD values yields positive t-values shown in warm colours, while negative t-values are shown in cool colours. The colour bar shows the colour mapping of voxels within caudate nucleus, putamen and nucleus accumbens to the values of the uncorrected t-statistics (t = 0 = p = 1; t = 1.294 = p = 0.2; t = 1.995 = p = 0.05; t = 2.648 = p = 0.01). Two sagittal slices through the left and right putamen are shown in the top row, and 8 axial slices are shown in the lower 2 rows. Coordinates in DARTEL space are given under each image. Top right, the axial slices shown are depicted on the midsagittal image with the axial slice 20 representing the most superior slice shown and axial slice -8 the most inferior. Axial images are shown with the subjects right side on the right (neurological convention).

16

From the map of the skeleton voxels (Figure 5), several clusters exceeding t-values greater than 1.995 were apparent. Among these were clusters located in the right inferior fronto-occipital fasciculus/uncinate fasciculus (cluster not shown), bilateral cingulum (slices -15 to -25), left superior longitudinal fasciculus (slice -25), and in the white matter underlying the precuneus (slice -65) and cuneus (slice -78) cortices bilaterally, in addition to a number of smaller clusters visible in the effectsize map. Interestingly, the effect-size map of the neostriatum revealed that some regions of the neostriatum appear to exhibit a closer relationship with 5-choice RT than others (Figure 6). An apparent ventrorostral to dorsocaudal gradient of low to high t-values was observed in the putamen, most notable in the two sagittal images.

Discussion
The present study examined associations between 5-choice RT performance and microstructure within the motor network in children aged 7 to 13 years. As hypothesized, faster choice RT was significantly associated with lower CST and neostriatal MD, adjusted for age, gender and handedness. Effects of CST and putamen MD were bilateral, while MD in right caudate nucleus exhibited a stronger relationship with 5-choice RT than did MD in the left caudate nucleus. Because both RTs (Kail 1991) and MD (Lebel et al. 2008) decrease throughout childhood and adolescence, a simple correlation between these measures could represent a non-specific association attributable to unmeasured factors that are indexed by chronological age (for further discussion see Madsen et al. 2010). Thus, we hypothesized that the association between 5-choice RT and motor system MD would remain significant after controlling for age, i.e., that even among children of similar age, those with lower CST and neostriatal MD would exhibit faster choice RT. These hypotheses were confirmed. Our follow-up analyses provided further clues about the nature of these associations. First, we addressed the question whether the associations between 5-choice RT and microstructure within the CSTs and neostriatum were anatomically specific, i.e., the extent to which 5-choice RT exhibited a relatively specific relationship to MD in these regions relative to others. This is an important question, since MD decreases concurrently in many brain regions during childhood and adolescence (Eluvathingal et al. 2007; Lebel et al. 2008). The relationship between 5-choice RT of CST and neostriatal MD remained significant when including, respectively, whole skeleton MD or nAcc/amygdalar MD as covariates in the models. Therefore, we infer that the relationship between 5choice RT and MD within these regions is not likely to be driven by global MD differences, but may be more strongly related to the microstructure of specific regions. Analyses of || and showed that faster 5-choice RT was associated with both lower || and lower in the CSTs. These findings may explain why we only observed effects of CST MD, and not of CST FA. Although interpretation of differences in DTI parameters is not straightforward, some studies suggest that may be more sensitive to changes in myelination (Schwartz et al. 2005; Song et

17

al. 2003; Song et al. 2005), while || may be more sensitive to differences in axonal diameter (Schwartz et al. 2005). Yet it should be noted that other tissue parameters, such as axon density, spacing, and number, and crossing fibres and tract geometry may also contribute to differences in and || (Beaulieu 2009; Schwartz et al. 2005). Whatever the neuroantomical correlate of or || might be, our observation of similar effects of and || provides no further evidence regarding the cellular basis of these effects. We also sought evidence that variability in the volumes of putamen and caudate might be relevant to our effects, and we found none. The relationship between choice RT and neostriatal MD remained significant when including the relative volumes as additional covariates in the models, suggesting that the effects are more closely linked to the microstructure than to the volumes of these structures. Little is known about the biological bases of variability in grey matter MD, particularly in healthy individuals. Lower MD has been related to higher cell density in tumours (Chenevert et al. 2006; Gibbs et al. 2009), suggesting that differences in cell (neuron and glia) density may be contributing to the observed neostriatal effects in the present study. Other tissue parameters, such as cell size, myelination or diameter of axons originating or terminating in the neostriatum, or membrane permeability, may also contribute to differences in MD (Chenevert et al. 2006). We also examined the mediating role of simple RT on the observed effects. The MD effects on 5-choice RT remained significant when including simple RT in the models, suggesting that the increased task demand for response selection may be particularly relevant to the association between choice RT and motor system MD in normal children. Alternatively, this pattern may be related to more general differences in task difficulty. Previous studies of young adults examining relationships between choice RT and white matter microstructure generally employed more complex choice RT tasks that also placed high demands on visual discrimination, response inhibition, and/or working memory (Konrad et al. 2009; Madden et al. 2004; Tuch et al. 2005). These studies mainly focused on regions outside the motor system and found effects in pathways supporting visuospatial attention. The choice RT task used in the present study mainly requires response selection and initiation, but little/no visual discrimination, response inhibition, or working memory. Thus, we focused on relationships with regions within the motor system known to be involved in movement selection and execution (Gerardin et al. 2004). However, in the TBSS effect-size map (Figure 5), some clusters exhibiting an apparent relationship with choice RT are present in regions similar to those found in adults (Konrad et al. 2009; Tuch et al. 2005), suggesting that regions supporting visuospatial attention may also be related to childrens performance on the present task. Future studies are necessary to address this in more detail. In the effect-size map of the neostriatum (Figure 5), an apparent ventrorostral to dorsocaudal gradient of low to high t-values may be observed in the putamen. Previous studies have shown a rostrocaudal functional gradient of inputs in the neostriatum, where motor and premotor cortices mainly project to dorsal and caudal areas of the putamen (Draganski et al. 2008; Haber 2003; Haber

18

and Gdowski 2004; Lehericy et al. 2004), suggesting that the putamen voxels with high t-values observed in the present study may be within the regions involved in motor control. However, further studies are needed to elucidate the relationship between choice RT performance and MD within specific neostriatal regions. Our findings suggest that even among children of similar age, lower MD within core motor structures is associated with faster 5-choice RTs. Questions remain regarding the neurobiological interpretation of such age-adjusted relationships between individual variability in MD and choice RT. This structure-function relationship may reflect individual differences in the maturational trajectories of the motor system, i.e., differences in the phase of CST and neostriatal maturation among children of similar age. It is plausible that this variability could mediate, at least to some extent, the associations found in the present study, since both microstructure (Lebel et al. 2008) and RTs (Kail 1991) continue to develop across this age range. Alternatively, the variability in motor system MD may reflect individual differences in the architecture of the motor system (perhaps differences in the underlying connectivity) that emerge earlier during brain development and remain stable in spite of superimposed biological changes associated with maturation. This is plausible since individual differences in behavioural performance have been associated with MD variability in adults (Konrad et al. 2009; Piras et al. 2010) as well as children. It is also possible that dynamic processes, perhaps associated with activity levels in the neural circuits throughout childhood could influence the microstructure within the motor network. A recent study of adults examined changes in DTI parameters in connection with 6 weeks of practice in a complex whole-body balancing task, observing negative correlations between improvements in motor performance and MD changes in bilateral anterior centrum semiovale, left brain stem, and right internal capsule (Taubert et al. 2010). Thus, the individual differences observed in the present study could reflect variability in sensorimotor experiences and skill acquisition of the children. In conclusion, individual differences in 5-choice RT performance were associated with individual differences in MD within the CSTs and neostriatum. These relationships were not attributable to age and did not appear to reflect behavioural effects of a more general maturational decrease in MD across the brain. Children of similar age may vary in the phase of maturation in the motor network, and this variability may mediate the observed associations. The associations could also be mediated by more stable individual differences in the architecture of the motor system, or to more transient differences associated with dynamic (perhaps activity-dependent) processes. Longitudinal observations are needed to further clarify these important questions.

References
Andersson JL, Hutton C, Ashburner J, Turner R, Friston K (2001) Modeling geometric deformations in EPI time series. Neuroimage. 13:903-19.

19

Ashburner J (2007) A fast diffeomorphic image registration algorithm. Neuroimage 38:95-113. Ashburner J, Friston KJ (2005) Unified segmentation. Neuroimage 26:839-51. Basser PJ, Mattiello J, LeBihan D (1994) MR diffusion tensor spectroscopy and imaging. Biophys.J. 66:259-67. Beaulieu C. 2009. Gaussian modeling of the diffusion signal. In Diffusion MRI: from quantitative measurement to in-vivo neuroanatomy, ed. H Johansen-Berg, TEJ Behrens, pp. 105-26. London, UK: Elsevier Bohr S, Gullmar D, Knab R, Reichenbach JR, Witte OW, Haueisen J (2007) Fractional anisotropy correlates with auditory simple reaction time performance. Brain Res 1186:194-202. Chang LC, Jones DK, Pierpaoli C (2005) RESTORE: robust estimation of tensors by outlier rejection. Magn Reson.Med. 53:1088-95. Chenevert TL, Sundgren PC, Ross BD (2006) Diffusion imaging: insight to cell status and cytoarchitecture. Neuroimaging Clin N Am 16:619-32, viii-ix. Cuadra MB, Cammoun L, Butz T, Cuisenaire O, Thiran JP (2005) Comparison and validation of tissue modelization and statistical classification methods in T1-weighted MR brain images. IEEE Trans Med Imaging 24:1548-65. Deary I, Der G, Ford G (2001) Reaction times and intelligence differences. A population-based cohort study. Intelligence 29:389-99. Draganski B, Kherif F, Kloppel S, Cook PA, Alexander DC, Parker GJ, Deichmann R, Ashburner J, Frackowiak RS (2008) Evidence for segregated and integrative connectivity patterns in the human Basal Ganglia. J Neurosci 28:7143-52. Eluvathingal TJ, Hasan KM, Kramer L, Fletcher JM, Ewing-Cobbs L (2007) Quantitative diffusion tensor tractography of association and projection fibers in normally developing children and adolescents. Cereb.Cortex 17:2760-8. Gerardin E, Pochon JB, Poline JB, Tremblay L, Van de Moortele PF, Levy R, Dubois B, Le Bihan D, Lehericy S (2004) Distinct striatal regions support movement selection, preparation and execution. Neuroreport 15:2327-31. Gibbs P, Liney GP, Pickles MD, Zelhof B, Rodrigues G, Turnbull LW (2009) Correlation of ADC and T2 measurements with cell density in prostate cancer at 3.0 Tesla. Invest Radiol 44:572-6. Haber SN (2003) The primate basal ganglia: parallel and integrative networks. J Chem Neuroanat 26:317-30. Haber SN, Gdowski MJ. 2004. The basal ganglia. In The human nervous system, ed. G Paxinos, JK Mai. London: Elsevier Hua K, Zhang J, Wakana S, Jiang H, Li X, Reich DS, Calabresi PA, Pekar JJ, van Zijl PC, Mori S (2008) Tract probability maps in stereotaxic spaces: analyses of white matter anatomy and tract-specific quantification. Neuroimage 39:336-47.

20

Jernigan TL, Gamst AC, Fennema-Notestine C, Ostergaard AL (2003) More "mapping" in brain mapping: statistical comparison of effects. Hum.Brain Mapp. 19:90-5. Jernigan TL, Trauner DA, Hesselink JR, Tallal PA (1991) Maturation of human cerebrum observed in vivo during adolescence. Brain 114 ( Pt 5):2037-49. Jovicich J, Czanner S, Greve D, Haley E, van der Kouwe A, Gollub R, Kennedy D, Schmitt F, Brown G, Macfall J, Fischl B, Dale A (2006) Reliability in multi-site structural MRI studies: effects of gradient non-linearity correction on phantom and human data. Neuroimage 30:436-43. Kail R (1991) Developmental change in speed of processing during childhood and adolescence. Psychol Bull 109:490-501. Konrad A, Vucurevic G, Musso F, Stoeter P, Winterer G (2009) Correlation of brain white matter diffusion anisotropy and mean diffusivity with reaction time in an oddball task. Neuropsychobiology 60:55-66. Lebel C, Walker L, Leemans A, Phillips L, Beaulieu C (2008) Microstructural maturation of the human brain from childhood to adulthood. Neuroimage. 40:1044-55. Lehericy S, Ducros M, Krainik A, Francois C, Van de Moortele PF, Ugurbil K, Kim DS (2004) 3-D diffusion tensor axonal tracking shows distinct SMA and pre-SMA projections to the human striatum. Cereb Cortex 14:1302-9. Madden DJ, Whiting WL, Huettel SA, White LE, MacFall JR, Provenzale JM (2004) Diffusion tensor imaging of adult age differences in cerebral white matter: relation to response time. Neuroimage 21:1174-81. Madsen KS, Baare WF, Vestergaard M, Skimminge A, Ejersbo LR, Ramsoy TZ, Gerlach C, Akeson P, Paulson OB, Jernigan TL (2010) Response inhibition is associated with white matter microstructure in children. Neuropsychologia 48:854-62. Miller JO, Low K (2001) Motor processes in simple, go/no-go, and choice reaction time tasks: a psychophysiological analysis. J Exp Psychol Hum Percept Perform 27:266-89. Niogi SN, McCandliss BD (2006) Left lateralized white matter microstructure accounts for individual differences in reading ability and disability. Neuropsychologia 44:2178-88. Piras F, Caltagirone C, Spalletta G (2010) Working memory performance and thalamus microstructure in healthy subjects. Neuroscience 171:496-505. Reese TG, Heid O, Weisskoff RM, Wedeen VJ (2003) Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson.Med. 49:177-82. Schwartz ED, Cooper ET, Fan Y, Jawad AF, Chin CL, Nissanov J, Hackney DB (2005) MRI diffusion coefficients in spinal cord correlate with axon morphometry. Neuroreport 16:73-6. Smith SM, Jenkinson M, Johansen-Berg H, Rueckert D, Nichols TE, Mackay CE, Watkins KE, Ciccarelli O, Cader MZ, Matthews PM, Behrens TE (2006) Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage. 31:1487-505.

21

Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE, Johansen-Berg H, Bannister PR, De Luca M, Drobnjak I, Flitney DE, Niazy RK, Saunders J, Vickers J, Zhang Y, De Stefano N, Brady JM, Matthews PM (2004) Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage. 23 Suppl 1:S208-S19. Snook L, Paulson LA, Roy D, Phillips L, Beaulieu C (2005) Diffusion tensor imaging of neurodevelopment in children and young adults. Neuroimage. 26:1164-73. Song SK, Sun SW, Ju WK, Lin SJ, Cross AH, Neufeld AH (2003) Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage. 20:1714-22. Song SK, Yoshino J, Le TQ, Lin SJ, Sun SW, Cross AH, Armstrong RC (2005) Demyelination increases radial diffusivity in corpus callosum of mouse brain. Neuroimage. 26:132-40. Sowell ER, Thompson PM, Holmes CJ, Jernigan TL, Toga AW (1999) In vivo evidence for postadolescent brain maturation in frontal and striatal regions. Nat Neurosci 2:859-61. Taubert M, Draganski B, Anwander A, Muller K, Horstmann A, Villringer A, Ragert P (2010) Dynamic properties of human brain structure: learning-related changes in cortical areas and associated fiber connections. J Neurosci 30:11670-7. Tuch DS, Salat DH, Wisco JJ, Zaleta AK, Hevelone ND, Rosas HD (2005) Choice reaction time performance correlates with diffusion anisotropy in white matter pathways supporting visuospatial attention. Proc Natl Acad Sci U S A 102:12212-7. Vestergaard M, Madsen KS, Baare WF, Skimminge A, Ejersbo LR, Ramsoy TZ, Gerlach C, Akeson P, Paulson OB, Jernigan TL (2010) White Matter Microstructure in Superior Longitudinal Fasciculus Associated with Spatial Working Memory Performance in Children. J Cogn Neurosci

22

Paper 3

Cortisol awakening response and negative emotionality linked to asymmetry in major limbic fibre bundle architecture
Kathrine Skak Madsena,b,c*; Terry L. Jernigana,b,c,d; Pernille Iversena,b; Vibe G. Frokjaerb,e; Erik Lykke Mortensenb,f; Gitte M. Knudsenb,c,e; William F.C. Baara,b

a b c d e f

Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, Copenhagen, Denmark Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark Veterans Affairs Center of Excellence for Stress and Mental Health, La Jolla, CA, United States Neurobiology Research Unit, Rigshospitalet, Copenhagen, Denmark Institute of Public Health and Center for Healthy Aging, University of Copenhagen, Copenhagen,

Denmark

Corresponding author
Kathrine Skak Madsen Danish Research Centre for Magnetic Resonance Copenhagen University Hospital, Hvidovre Kettegaard All 30 2650 Hvidovre, Denmark Phone: +45 3632 3323, Fax: +45 3647 0302, E-mail: kathrine@drcmr.dk

Abstract
The limbic system plays an important role in regulating the hypothalamic-pituitary-adrenal (HPA) axis as well as aspects of emotion, and both neuroendocrine disturbance and increased negative emotionality are associated with risk for developing affective disorders. However, the extent to which the architecture of connections between limbic structures may be linked to individual differences in basal HPA-axis reactivity and negative emotionality is unknown. Here we tested the hypotheses that microstructural asymmetry of the major limbic fibre bundles would be associated with cortisol awakening response (CAR) and neuroticism, a personality trait associated with the tendency to experience negative emotions. Sixty-nine healthy adults were studied with diffusion-weighted imaging, and fractional anisotropy (FA) was extracted from the cingulum and uncinate fasciculus. Higher neuroticism scores, which were associated with higher CAR, were also correlated with higher right relative to left cingulum FA. Elevated CAR was associated with the degree of FA asymmetry within both the cingulum and the uncinate fasciculus, but in opposing directions. These results suggest that the balance between left- and right-sided limbic circuits may bear an important relationship to hypothalamic-pituitary-adrenal axis reactivity, and to the tendency to experience negative emotions, and they raise important questions about the significance of limbic system architecture.

Key words: diffusion tensor imaging, HPA-axis, laterality, MRI, personality

1. Introduction
The limbic system plays a central role in emotionality and stress regulation. Limbic structures, such as the medial prefrontal cortex (PFC), amygdala, and hippocampus, have been associated with negative affect (Drevets et al. 2008, Phillips et al. 2008, Cremers et al. 2010) and have been implicated in modulating the response to stress by regulating the hypothalamic-pituitary-adrenal (HPA)-axis (Sullivan and Gratton 1999, Jankord and Herman 2008, Dedovic et al. 2009). Basal HPA-axis activity follows a distinct diurnal profile, with increasing plasma cortisol levels during the second half of the night, peaking approximately 30 minutes after morning awakening, i.e. the cortisol awakening response (CAR), followed by a gradual decrease thereafter (Pruessner et al. 1997, Wilhelm et al. 2007, Fries et al. 2009). Disruption of the diurnal profile of the HPA-axis has been found in psychiatric disorders, such as post-traumatic stress disorder (PTSD) (Chida and Steptoe 2009), anxiety disorder (Vreeburg et al. 2010) and major depression (Chida and Steptoe 2009, Vreeburg et al. 2009). Moreover, high CAR and mean morning cortisol secretion have been associated with high scores on measures of neuroticism (Portella et al. 2005), a personality trait reflecting a tendency towards experiencing negative emotions, e.g., anxiousness, sadness, and difficulties coping with stress. Like CAR, high neuroticism scores have been linked to mood and anxiety disorders, such as phobias, panic disorder, and major depression (Bienvenu et al. 2001). Moreover, both variables show substantial heritability (Jang et al. 1996, Wust et al. 2000, Kendler et al. 2006). Prospective studies indicate that both high CAR and high neuroticism scores increase the risk of developing major depression (Kendler et al. 2004, Kendler et al. 2006, Adam et al. 2010). The relationship between diurnal cortisol profiles and human limbic brain structures is unclear and available data are sparse. Higher mean cortisol levels have been associated with regionally reduced frontal cortical thickness (Kremen et al. 2010), and inconsistently with hippocampal volume (Pruessner et al. 2007, Knoops et al. 2010, Kremen et al. 2010). Interestingly, a growing body of data suggests that HPAaxis functioning may be related to hemispheric lateralization. Indeed, stroke patients with left-sided, but not right-sided, infarctions showed increased morning cortisol levels compared to controls (Lueken et al. 2009). Moreover, healthy subjects with impaired cortisol suppression in response to dexamethasone had smaller left and substantially larger right than left anterior cingulate gyrus volumes as compared to suppressors (MacLullich et al. 2006). Furthermore, presenting an emotionally adverse film to the right hemisphere in healthy controls resulted in a consistently higher increase in phasic cortisol secretion than when presented to the left hemisphere (Wittling and Pfluger 1990). Additionally, findings from animal studies suggest that the left and right hemisphere may be differentially related to HPA-axis functioning (for review, see Cerqueira et al. 2008). In rodents right but not left medial PFC lesions significantly suppressed stress-induced increases in corticosterone levels in repeatedly restrained animals, suggesting that the right mPFC may be activating physiological stress responses (Sullivan and Gratton 1999). It has

been suggested that the left medial PFC may be modulating the neuroendocrine stress responses through interhemispheric inhibition (Sullivan 2004). Finally, an EEG study in monkeys found that animals with greater right relative to left frontal activity had higher plasma cortisol levels (Kalin et al. 1998). Overall, the balance between left and right prefrontal structures appears to be important for HPA-axis regulation and function. Studies investigating the relationship between limbic structures and neuroticism reported negative associations between neuroticism and right amygdala grey matter concentration (Omura et al. 2005), and right dorsomedial PFC and left medial temporal lobe volume (DeYoung et al. 2010). Moreover, neuroticism has been positively correlated with frontolimbic serotonin 2A receptor binding (Frokjaer et al. 2008). Interestingly, a recent functional magnetic resonance imaging (fMRI) study of negative emotional vs. neutral faces found that neuroticism was positively associated with right amygdala - dorsomedial PFC functional connectivity, but negatively correlated with left amygdala anterior cingulate connectivity (Cremers et al. 2010), implying that hemispheric asymmetry in the connectivity of these structures may be important for negative emotionality. Until now, MRI studies investigating how structural connections between limbic structures may underlie HPA-axis function and neuroticism are lacking. The cingulum bundle and the uncinate fasciculus are two major limbic fibre tracts that connect areas including amygdala and parahippocampal region with anterior cingulate and medial PFC, and orbitofrontal cortex, respectively (Schmahmann and Pandya 2006). Diffusion-weighted imaging (DWI), which is sensitive to the diffusion of water molecules, allows investigation of the microstructure of the brain in vivo. In white matter, cellular structures, like the axonal membrane and surrounding myelin sheath, hinder diffusion of water perpendicular relative to parallel to a fibre bundle, thereby causing diffusion anisotropy (Beaulieu 2009). The diffusion tensor model is used to estimate parameters such as fractional anisotropy (FA), and diffusivities parallel (||) and perpendicular () to the principal diffusion direction. FA reflects the degree of diffusion directionality, which in white matter can be influenced by microstructural properties, such as axonal density, organization, and myelination (Beaulieu 2009). Diffusion tensor imaging (DTI) studies show that cingulum and uncinate fasciculus reach maturity relatively late with FA values peaking around the mid-thirties (Lebel et al. 2008, Westlye et al. 2009) followed by an at first slow, and subsequently more rapid, decrease in FA around the 6th decade (Westlye et al. 2009). Studies in healthy subjects consistently report higher FA in the left relative to the right cingulum (Park et al. 2004, Gong et al. 2005), whereas uncinate fasciculus asymmetry findings have been inconsistent (Park et al. 2004, Hasan et al. 2009). Nevertheless, in a post mortem study, the right uncinate fasciculus was relatively larger and contained more axons than the left (Highley et al. 2002). The functional importance of these asymmetries is illustrated by findings that subjects with bipolar disorder

had higher FA in the left and lower FA in the right uncinate fasciculus than controls (Versace et al. 2008). Moreover, subjects with PTSD had lower left cingulum FA compared to controls (Kim et al. 2006). Overall, available evidence suggests that the balance between left and right-sided limbic structures may be relevant for HPA-axis reactivity, and negative emotionality. To our knowledge, there are currently no DTI studies linking CAR and neuroticism with characteristics of fibre tracts in the brain. Here we examined the relationship between neuroticism, cortisol measures, and major limbic fibre bundle asymmetry. Specifically, we tested the hypotheses that CAR and neuroticism would be associated with FA asymmetry in the cingulum and uncinate fasciculus. Follow-up analyses were conducted to test the specificity of observed relationships.

2. Methods
2.1. Subjects
Sixty-nine healthy adults aged 19-86 years with no history of neurological or psychiatric disorders or brain injury and no MR contraindications were included. No subjects took psychoactive drugs, steroids, hormone replacements or drugs of abuse, including anabolic steroids. Subjects had a body-mass-index (BMI) below 35. Six subjects received antihypertensive treatment with thiazide diuretics, ACE inhibitors, or angiotensin II receptor antagonists and all were normotensive. No participants received beta-blocking agents. Fifty-eight of the 69 subjects agreed to collect salivary cortisol samples. Of these, ten subjects were excluded because of less than four samples within the first hour after awakening (N=7), not following the saliva-sampling schedule (N=2), and reporting intense exam related stress (N=1). Thus, 48 subjects were included in analyses involving cortisol measures. Demographic data for the whole and the cortisolcollecting group are presented in Table 1.

Table 1. Demographic data of the whole group and the subgroup in whom awakening cortisol was measured * Whole group (N=69) Age (mean std) Gender (male / female) Education (mean std) Handedness (right / left) Antihypertensive treatment 38.2 19.2 45 / 24 3.7 1.5 64 / 5 6 Cortisol group (N=48) 37.4 19.3 34 / 14 3.9 1.3 45 / 3 4

* Subgroups with or without cortisol data did not differ significantly on any of the demographic variables (Ps0.16)

The study was approved by the local Danish Committee for Biomedical Research Ethics (protocol (KF) 01-2006-20) and performed in accordance with the Declaration of Helsinki. Informed written consent was obtained from all subjects prior to participation.

2.2. Personality assessment

All subjects completed the Danish version of the 240-item self-report NEO-PI-R personality questionnaire, which assesses the personality traits Neuroticism, Extraversion, Agreeableness, Conscientiousness and Openness (Hansen et al. 2004). The number of days between the NEO-PI-R assessment and the MR-scan ranged from 1-966 days (median=19 days), with 23 subjects having more than 1-month interval between the assessments. Here we focused only on neuroticism.

2.3. Salivary cortisol assessment

Subjects collected 5 serial morning saliva samples using Salivette sampling devices (Sarstedt, Neubringen, Germany) at 0, 15, 30, 45 and 60 minutes after awakening. All subjects were trained in the sampling technique before. Subjects were instructed to rinse their mouths before sampling, refrain from eating, smoking, drinking caffeinated beverages and brushing teeth until after collecting the last morning sample, and to register the exact times of falling asleep the night before and awakening on the sampling day, and the actual sampling times. The time-interval between reported awakening and the first sample ranged from 0-10 minutes [mean(std) = 2.8(2.9)]. The number of days between cortisol sampling and MRI ranged from 1-279 days (median=15 days) with 10 subjects having more than 1-month interval between assessments. The interval between cortisol and NEO-PI-R assessments ranged from 1-403 days (median=6 days), with 8 subjects having more than 1-month between assessments. At home, samples were stored 5OC for up to 3 days. Samples were sent by 24-hour mail to the Cimbi laboratory, and stored at -80OC. Cortisol concentrations were determined using an electrochemiluminescence immunoassay (Roche, Mannheim, Germany) on Modular Analytics E170 equipment. The entire batch of samples was analyzed in one step. Intra-assay coefficients of variance have routinely been found to be below 6.1%. Mean morning cortisol level was based on available samples collected within the first hour after awakening. CAR was defined as the percentage difference between maximum and mean morning cortisol level [CAR = ((MaximumMorning MeanMorning) / MeanMorning) * 100)]. This estimate was not significantly correlated with mean morning cortisol level (P=0.67), nor highly dependent on the awakening sample (P=0.79).

2.4. Image acquisition

All subjects were scanned using a 3T Trio scanner (Siemens, Erlangen, Germany) with an eight-channel head coil (Invivo, FL, USA). Scans were acquired parallel to the anterior commisureposterior commisure line. Whole brain diffusion-weighted (DW) images were acquired using a twice-refocused balanced spin echo sequence that minimized eddy current distortion (Reese et al. 2003). Ten non-DW images (b=0) and 61 DW images (b=1200), encoded along independent collinear diffusion gradient orientations, were acquired (TR=8200 ms; TE=100 ms; FOV=220x220, matrix=96x96; GRAPPA: factor 2, 48 lines; 61 transverse slices; no gap; 2.3x2.3x2.3 mm3 voxels). A gradient echo field map was acquired to correct B0 field distortions (Andersson et al. 2001) (TR=530 ms; TE[1]=5.19 ms and TE[2]=7.65 ms; FOV=256x256; matrix=128x128; 47 transverse slices; no gap, 2x2x3 mm3 voxels). T2-weighted images of the whole head were acquired using a 3D turbo spin echo sequence (TR=3000 ms; TE=354 ms; FOV=282x216; matrix=256x196; 192 sagittal slices; 1.1x1.1x1.1 mm3 voxels).

2.5. Image preprocessing

Raw images were visually inspected to ascertain data quality. Images were preprocessed using pipelines implemented in Matlab using mainly SPM5 routines. T2-weighted images were corrected for gradient nonlinearity distortions (Jovicich et al. 2006). Each subjects mean b0 image was coregistered to the T2-image using a mutual information rigid transformation, after which all DW images were coregistered to the mean b0 image (no reslicing). Resulting images were corrected for geometric distortions due to B0 field (Andersson et al. 2001) and gradient non-linearities and resliced using a single trilinear interpolation step. Diffusion gradient orientations were adjusted to account for any rotation applied during registration. The diffusion tensor was fitted using the RESTORE algorithm (Chang et al. 2005), and FA, || (1) and [(2 + 3) / 2] were calculated. A brain mask automatically generated on the T2-weighted image was applied to the FA and diffusivity images.

2.6. Inter-subject spatial normalization of fibre tracts

Tract-Based Spatial Statistics was used for spatial normalization and alignment of white matter fibre tracts across subjects (Smith et al. 2006) (FSL 4.1.2). All subjects FA images were warped to the FMRIB58_FA_1mm template in MNI space using nonlinear registration (Andersson et al. 2007). A crosssubject mean FA image was generated and thinned to create a mean FA skeleton, representing the centres of all tracts common to the group. To suppress areas with low mean FA and/or with high inter-subject variability, the mean FA skeleton was thresholded at FA>0.3, and contained 92,757 1 mm3 interpolated isotropic voxels. Each subjects aligned FA image was projected onto the study-specific skeleton by locating the voxels with the highest local FA value in the direction perpendicular to the skeleton tracts,

and assigning the values of these voxels to the skeleton at this standardized location. Note that this results in a mapping of each voxel location in the skeleton to a specific voxel in the individual FA maps. Finally, the nonlinear warps and skeleton projections were applied to the || and data.

2.7. Regions-of-interest (ROIs)

To test our hypotheses, mean FA, || and values were extracted from right and left cingulum and uncinate fasciculus ROIs (Figure 1). ROIs were manually drawn onto the mean skeleton overlaid on the mean FA image using FSLview. Cingulum ROIs included all skeleton segments within the cingulum, and excluded segments bordering the cingulum. Note that the skeleton did not include segments underlying the subgenual cingulate cortex (Brodmann area 25). Right and left cingulum ROIs contained 1,024 and 963 voxels, respectively. Uncinate fasciculus ROIs were delineated using the JHU White-Matter Tractography Atlas (Hua et al. 2008) implemented in FSLview for guidance. Central uncinate fasciculus segments, and segments extending towards the ventromedial prefrontal cortex were included, while segments extending towards the temporal pole and the inferior frontal gyrus were excluded. Right and left uncinate fasciculus ROIs included 810 and 785 voxels, respectively. Smoothed 3D-projections of the ROIs are depicted in Figure 3. To assess the anatomical specificity of observed associations, right (46,410 voxels) and left (46,054 voxels) total hemispheric skeleton ROIs were delineated using the mid-sagittal plane (not included).

Figure 1. Regions-of-interest in the uncinate fasciculus (blue) and cingulum (red), and TBSS skeleton (yellow) overlaid on the group mean fractional anisotropy (FA) map. Axial (z), coronal (y) and sagittal (x) views of the ROIs with corresponding MNI coordinates for each slice.

2.8. Asymmetry measures

ROI diffusion parameter asymmetries were calculated as the difference between left and right ROI values expressed as a percentage of the bilateral mean: [(2 * (Left Right)) / (Left + Right) * 100].

2.9. Statistical analysis

Statistical analyses were performed in SPSS18. Analyses involving cortisol measures were based on 48 individuals. All other analyses included data from 69 individuals, unless otherwise noted. Two-tailed ttests were used to compare left and right ROI values (paired-tests) and the cortisol-collecting subgroup vs. the non-collecting subgroup. Age and gender effects were examined using linear regression models. Multiple linear regression was used to test our hypotheses. Shapiro-Wilk tests showed that all data were normally distributed except for CAR. Logarithm-transformation of CAR normalized the residuals. All other assumptions for linear regression were fulfilled. A p-value below 0.05 was considered significant. We hypothesized that CAR would be associated with cingulum FA asymmetry and uncinate fasciculus FA asymmetry after adjusting for age, gender, time-interval between awakening and first cortisol sample, and sample position of the cortisol peak (i.e., first vs. second-fourth sample). Neuroticism was hypothesized to be associated with cingulum FA asymmetry and uncinate fasciculus FA asymmetry, adjusted for age and gender. In separate exploratory post-hoc analyses, we examined whether cingulum or uncinate FA asymmetries or neuroticism were associated with HPA-axis tonus, i.e., mean morning cortisol level, after adjusting for age, gender, time-interval between awakening and first cortisol sample, and cortisol peak position. Planned follow-up analyses were contingent on observing significant associations in our primary analyses and 1) assessed left and right ROI contributions to observed effects, 2) assessed the anatomical specificity of the effects by including global white matter hemispheric FA asymmetry in the models, and 3) ascertained that observed effects were not mediated by other factors that may affect the cortisol, neuroticism or diffusion measures, e.g., awakening time, sleep duration, interval between sunrise and awakening, hours of daylight, weekday vs. weekend cortisol sampling, BMI, handedness, educational level, smoking, antihypertensive treatment, interval between assessments (MRI, personality assessment and salivary cortisol sampling) and age2 (Clow et al. 2004, Hansen et al. 2004, Westlye et al. 2009). To corroborate previous findings (Portella et al. 2005), we examined whether higher neuroticism scores were associated with higher CAR, adjusted for age, gender, time-interval between awakening and first cortisol sample, and cortisol peak position. Based on our results, CAR was modelled with cingulum and uncinate fasciculus FA asymmetry measures simultaneously to assess if their effects were additive. Moreover, neuroticism was modelled with

CAR and cingulum FA asymmetry simultaneously to assess whether their contributions to neuroticism were shared or independent. The || and were investigated to further explore the nature of observed FA findings, since higher FA could be due to increased || and/or decreased .

3. Results
Neuroticism, cortisol and ROI FA measures are summarized in Table 2. Left cingulum FA was significantly larger than right cingulum FA (P<10-23). The opposite was true for uncinate fasciculus FA (P=0.05). Further, women had significantly higher neuroticism (P=0.002) and lower CAR (P=0.037) levels than men, after adjusting for age. Table 2. Mean standard deviation values for the neuroticism, region-of-interest (ROI) and cortisol measures Whole group (N=69) Personality Neuroticism score Left cingulum Right cingulum Left uncinate fasciculus Right uncinate fasciculus ROI FA asymmetry measures Cingulum asymmetry Uncinate asymmetry Cortisol measures Cortisol Awakening Response (%) MeanMorning cortisol level (nmol/L) N/A N/A 49.1 27.0 8.64 4.20 53.1 27.2 8.17 4.20 39.4 24.8 9.79 4.13 5.8 2.8 -1.1 4.5 6.2 2.9 -1.0 4.3 6.2 2.6 -0.4 4.5 5.1 3.2 -2.3 4.5 73.4 17.6 0.559 0.034 0.526 0.029 0.461 0.032 0.466 0.032 70.5 17.9 0.562 0.037 0.528 0.031 0.462 0.034 0.467 0.033 68.7 17.3 0.566 0.034 0.531 0.029 0.467 0.030 0.469 0.031 82.1 15.1 0.547 0.031 0.520 0.029 0.451 0.033 0.462 0.033 ROI fractional anisotropy (FA) measures Cortisol group (N=48) Men (N=45/34)* Women (N=24/14)*

* Note that for the personality and ROI measures there were 45 males and 24 females, whereas for the cortisol measures there were 34 males and 14 females. Abbreviations: FA = Fractional Anisotropy.

3.1. Associations between CAR and limbic fibre bundle asymmetries

As hypothesized, CAR was significantly associated with cingulum FA asymmetry and with uncinate fasciculus FA asymmetry (Table 3, Figure 2a,b). Relative contributions of left and right ROI FA are

10

presented in Table 4. When entering left and right cingulum FA simultaneously as predictors of CAR, left was negatively and right was positively associated with CAR (Table 4(a). In contrast, left uncinate fasciculus FA was positively and right was negatively associated with CAR, when modelled simultaneously (Table 4(b)). Notably, no significant correlations with CAR were found when modelling the left or right ROIs separately (Table 4(b)), indicating that it is the relationship between the left and right ROIs that exhibited associations with CAR, and not the left or right ROIs individually.

Table 3. Results for the a priori hypotheses predicting cortisol awakening response or neuroticism with cingulum or uncinate fasciculus fractional anisotropy (FA) asymmetry. * Cortisol Awakening Response (n=48)** Cingulum FA asymmetry Uncinate FA asymmetry Neuroticism (n=69)*** Cingulum FA asymmetry Uncinate FA asymmetry -0.357 -0.063 0.002 0.591 -0.288 0.293 0.045 0.039 P

* The standardized regression coefficient for the variables of interest controlling for model covariates. ** Models predicting cortisol measures were adjusted for age, gender, time-interval between awakening and the first cortisol sample, and sample position of the cortisol morning peak (first vs. second-fourth position). *** Models predicting neuroticism were adjusted for age, and gender.

Follow-up analysis, testing the anatomical specificity of the observed FA asymmetry effects, showed that both cingulum (=-0.314, P=0.034) and uncinate fasciculus (=0.320, P=0.040) FA asymmetries remained significant predictors of CAR, when including global white matter hemispheric FA asymmetry (Ps0.40) in the models, suggesting that the associations between the limbic FA asymmetries and CAR are not mediated by a general hemispheric asymmetry. Moreover, observed associations did not significantly change when expanding the models to include one of the additional covariates mentioned in point 3 in the statistical section. To further investigate that the number of days between the cortisol sampling and MR-scan was not a cofounding factor, additional analyses were made in a subgroup with less that 1-month interval between these assessments (N=38). In this subgroup, both cingulum (=-0.375, P=0.021) and uncinate (=0.320, P=0.042) FA asymmetries remained significant predictors of CAR, demonstrating that the effects are not mediated by varying intervals between the assessments. When entering cingulum and uncinate fasciculus FA asymmetries simultaneously as predictors, both asymmetries exhibited relatively independent relationships with CAR (cingulum: =-0.256, P=0.067; uncinate fasciculus: =0.262, P=0.059).

11

Figure 2. Partial regression plots of the main findings. In the top row, the logarithm-transformed cortisol awakening response (CAR) is plotted as a function of a) cingulum and b) uncinate fasciculus fractional anisotropy (FA) asymmetry, adjusted for age, gender, time-interval between awakening and first cortisol sample, and cortisol peak sample position (first vs. second through fourth samples). Both fibre tract FA asymmetries exhibited significant but opposite relationships with CAR. In plot c), neuroticism was significantly and negatively associated with cingulum FA asymmetry, adjusted for age and gender. In plot d), neuroticism exhibited a significant positive relationship with CAR, adjusted for age, gender, time-interval between awakening and first cortisol sample, and cortisol peak sample position (first vs. other morning samples). Note that residualized measures are plotted. Dotted lines indicate 95% confidence intervals.

Further exploration of the nature of the observed FA asymmetry effects revealed that cingulum || asymmetry (=-0.335, P=0.039), but not cingulum asymmetry (=0.165, P=0.266), was significantly associated with CAR. In contrast, uncinate fasciculus asymmetry (=-0.323, P=0.024), but not || asymmetry (=0.141, P=0.379) was significantly associated with CAR. In 10 of 48 subjects, the cortisol peak occurred in the awakening sample, making it uncertain whether the actual peak was captured or occurred before the awakening sample. Post hoc comparison of peakafter-awakening (N=38) and peak-at-awakening subgroups (N=10) revealed that in the peak-afterawakening subgroup, CAR exhibited significant relationships with cingulum (=-0.348, P=0.025) and uncinate fasciculus FA asymmetry (=0.369, P=0.016). No significant associations were observed in the peak-at-awakening subgroup (Ps0.46). Exploratory analyses of mean morning cortisol levels showed no significant association with cingulum FA asymmetry (P=0.97), though a positive association with uncinate fasciculus FA asymmetry approached significance (=0.259; P=0.062).

12

Table 4. Follow-up analyses assessing the relative contribution of fractional anisotropy (FA) in left and right regions-of-interest (ROIs) on cortisol awakening response or neuroticism Left ROI FA * Cortisol awakening response (n=48)** Cingulum a Cingulum Cingulum
b b

Right ROI FA P * P

-0.662 -0.059

0.050 0.734

0.687 0.127

0.039 0.458

Uncinate fasciculus a Uncinate fasciculus b Uncinate fasciculus

b

0.546 0.275

0.035 0.159

-0.441 -0.054

0.108 0.800

Neuroticism (n=69)*** Cingulum a Cingulum b Cingulum b -0.800 -0.144 0.002 0.264 0.053 0.669 0.722 0.003

* The standardized regression coefficient for the variables of interests controlling for model covariates. ** Models predicting cortisol awakening response either by modelling left and right ROI FA simultaneously (a) or individually (b) were adjusted for age, gender, time-interval between awakening and the first cortisol sample, and sample position of the cortisol morning peak (first vs. second-fourth position). *** Models predicting neuroticism response either by modelling left and right ROI FA simultaneously (a) or individually (b) were adjusted for age, and gender.

3.2. Associations between neuroticism and limbic fibre bundle asymmetries

As hypothesized, neuroticism was significantly associated with cingulum FA asymmetry, adjusted for age and gender (Table 3, Figure 2c). A similar effect size (=-0.377, P=0.003) was also observed in the smaller cohort with cortisol measures (N=48). Such a relationship, however, was neither observed for uncinate fasciculus FA asymmetry in the large group of 69 subjects (Table 3), nor in the smaller group of 48 subjects (P=0.34). Follow-up analyses showed that left cingulum FA was negatively and right cingulum FA was positively associated with neuroticism, when they were modelled simultaneously (Table 4(a)). Models including only left or right cingulum FA yielded no significant effects on neuroticism (Table 4(b)), suggesting that the association with neuroticism is mediated by the relationship between left and right cingulum FA.

13

When assessing the anatomical specificity of the cingulum effect, cingulum FA asymmetry (=0.363, P=0.002) continued to be significantly associated with neuroticism with global hemispheric FA asymmetry (P=0.83) included in the model. Furthermore, the observed association between neuroticism and cingulum FA asymmetry remained unchanged when including potentially mediating factors (see point 3 in statistical section) as additional covariates. Moreover, when analysing the subgroup with less than 1month interval between the NEO-PI-R assessment and MR-scan (N=46), cingulum FA asymmetry (=0.460, P=0.001) remained significantly associated with neuroticism. Exploring the diffusivities showed that cingulum asymmetry (=0.438, P=0.0008), but not cingulum || asymmetry (=-0.206, P=0.183) was significantly associated with neuroticism.

3.3. Associations between neuroticism, CAR and cingulum FA asymmetry

Corroborating previous findings (Portella et al. 2005), higher neuroticism scores were significantly associated with higher CAR (=0.309, P=0.029, Figure 2d). However, exploratory analyses of mean morning cortisol levels showed no significant association with neuroticism (P=0.19). In a final follow-up analysis, cingulum FA asymmetry continued to exhibit a significant relationship with neuroticism (=-0.329, P=0.014), while the CAR association was reduced (=0.204, P=0.139), when both predictors were modelled simultaneously.

4. Discussion
In the current study, we present novel findings relating CAR and neuroticism with microstructural asymmetry in major limbic fibre tracts. An overview of the main findings is given in Figure 3. Specifically, higher CAR was negatively associated with cingulum FA asymmetry, i.e., increased right relative to left FA, and positively associated with uncinate fasciculus FA asymmetry, i.e., decreased right relative to left FA. Higher neuroticism scores were correlated with diminished cingulum FA asymmetry. Importantly, these associations were mediated by the relationship between left and right fibre tract FA and not driven by FA in either the left or right fibre tract individually. Moreover, including a global white matter hemispheric FA asymmetry measure in the models did not change the asymmetry effects, indicating that the effects are likely to be anatomically specific and not associated with general hemispheric asymmetry. Further, effects remained when including factors that potentially could influence CAR, neuroticism and/or the diffusion parameters (Clow et al. 2004, Hansen et al. 2004, Westlye et al. 2009).

14

Figure 3. Schematic overview of the direction of the significant associations between neuroticism, cortisol awakening response (CAR), and fractional anisotropy (FA) asymmetry in the two major limbic fibre tracts: cingulum (blue) and uncinate fasciculus (magenta). Positive associations are depicted with a + imbedded in green lines. Negative associations are depicted with a - imbedded in red lines. Neuroticism was negatively associated with cingulum FA asymmetry, with higher neuroticism scores correlating with higher right cingulum FA relative to left cingulum FA. Similar to neuroticism, CAR was negatively associated with cingulum FA asymmetry, i.e., higher CAR correlated with higher right cingulum FA relative to left cingulum FA. In contrast, CAR was associated positively with uncinate fasciculus FA asymmetry, with higher CAR correlating with lower right relative to left uncinate fasciculus FA. Finally, neuroticism and CAR exhibited a positive relationship. Note that the depicted fibre tracts are smoothed versions of the ROIs used in the present study.

Interestingly, the associations between cingulum and uncinate fasciculus FA asymmetries and CAR were statistically in opposite directions. Moreover, the effects of the two fibre tract asymmetries on CAR appeared to be relatively independent of each other. While increased right relative to left cingulum FA was expected to be associated with higher CAR, we did not anticipate an opposite relationship for uncinate fasciculus FA asymmetry (i.e., decreased right relative to left FA with higher CAR). The direction of the cingulum FA asymmetry association corresponds with findings showing that nonsuppressors to dexamethasone had substantially larger right than left anterior cingulate gyrus volume, while suppressors exhibited symmetry in this region (MacLullich et al. 2006). It is also congruent with animal studies suggesting that functions of the right medial PFC may enhance neuroendocrine responses (Sullivan and Gratton 1999, Sullivan 2004, Cerqueira et al. 2008), and that greater right relative to left frontal activity may increase cortisol levels (Kalin et al. 1998). Previous data relevant to the observed opposite effects for the cingulum and the uncinate fasciculus are very sparse. A recent study found that subjects responding to a stressful task with increased cortisol levels showed decreased activation in left anterior cingulate (innervated by left cingulum) and right orbitofrontal cortex (receiving connections from the right uncinate) relative to non-responders (Pruessner et al. 2008). Moreover, findings from animal

15

studies suggest that prelimbic and infralimbic cortices may have differential regulatory influences on HPA-axis activity, in that these regions have different efferent projection patterns to regions implicated in, respectively, stress inhibition and excitation (for review, see Herman et al. 2005). However, this awaits additional support. Further studies need to elucidate the opposite effects of cingulum and uncinate fasciculus FA asymmetries may have on CAR. At present, we can only speculate about the neurobiological significance of the observed relationship between CAR and limbic fibre tract asymmetries. Prefrontal and limbic structures have high densities of corticoid receptors, and are thus, thought to be implicated in exerting negative feedback of the HPA-axis (Reul et al. 2000, Sanchez et al. 2000). Although the medial- and orbitofrontal cortices (potentially through connections with limbic structures, e.g. hippocampus and amygdala, via the cingulum and the uncinate fasciculus) may be involved in regulation of the HPA-axis, these structures may also be affected by cortisol. A study in rats found that chronic treatment with high doses of dexamethasone significantly reduced the volume of the left anterior cingulate gyrus, and to a lesser extent the right, suggesting that left anterior cingulate may be more vulnerable to prolonged high cortisol levels than the right (Cerqueira et al. 2005). It may well be that these effects also are reflected in the white matter. Overall, the present findings could reflect a) the effects of limbic connectivity on functional regulation of the HPA-axis, b) effects on the neural architecture of varying levels of secreted hormones, c) some combination of these factors, or d) a third factor influencing both CAR and structural properties of the fibre tracts. Our observation that higher neuroticism scores were associated with increased right relative to left cingulum FA is consistent with findings from a recent fMRI study, in which higher neuroticism scores were associated with increased right amygdala - dorsomedial PFC, and decreased left amygdala anterior cingulate cortex functional connectivity (Cremers et al. 2010). The present findings suggest that asymmetry in the structural connectivity may underlie the above observed asymmetry in functional connectivity. We found that asymmetry in uncinate fasciculus , but not ||, was significantly associated with CAR, suggesting that this effect may be related to asymmetry in axonal density, organization and/or myelination of the uncinate fasciculus (Schwartz et al. 2005, Beaulieu 2009). The cingulum findings were more complex, in that || asymmetry was associated with CAR, while asymmetry was associated with neuroticism. || may reflect increased axonal diameter, though many other factors, like axonal number and organization, may also contribute to differences in both and || (Schwartz et al. 2005, Beaulieu 2009). Cingulum diffusion parameters may also be affected by bundle size, e.g., increased partial volume effects may reduce FA in smaller fibre bundles (Vos et al. 2011). The present cingulum findings could therefore be affected by the structural properties intrinsic to the tracts or to the size of the cingulum, which could affect the underlying diffusivities in the cingulum differently.

16

Microstructural asymmetries of major limbic fibre tracts have been related to affective disorders (Kim et al. 2006, Drevets et al. 2008, Versace et al. 2008). A recent study found that subjects with bipolar disorder had higher left and lower right uncinate fasciculus FA relative to control subjects (Versace et al. 2008). Another study found that subjects with PTSD had lower left cingulum FA compared to controls. Moreover, on average the PTSD group had higher right than left cingulum FA, while the control group had higher left relative to right cingulum FA (Kim et al. 2006). Since both higher CAR and neuroticism scores have been linked to an increased risk of developing affective disorders (Kendler et al. 2004, Kendler et al. 2006, Adam et al. 2010), we speculate whether uncinate fasciculus and cingulum microstructural asymmetries are brain structural markers for a predisposition or risk for developing affective disorders. However, further studies are needed to address this question. In the present study, we estimated the dynamics of CAR as the difference between the maximum and mean morning cortisol level relative to the mean morning cortisol level. Frequently used other modelling approaches are estimates of the area under the curve increase (AUCi) relative to the awakening sample (Chida and Steptoe 2009, Vreeburg et al. 2010), the absolute difference score (Adif) between the peak sample and awakening sample (Chida and Steptoe 2009, Knoops et al. 2010), and the mean increase (awakening sample subtracted from the mean of the following morning samples) (Wust et al. 2000, Chida and Steptoe 2009). The present estimate of CAR was significantly correlated with Adif (P=0.02), but not with AUCi or mean increase (Ps0.29). Unlike the present estimate, AUCi and Adif show substantial correlations with the mean morning cortisol level. Moreover, AUCi, Adif and mean increase estimates of CAR highly rely on the awakening sample, which may be influenced by discrepancies in the time-interval between reported awakening and the first sample (Clow et al. 2004). To further minimize the influence of these discrepancies, we included the time-interval between reported awakening and the first sample as a covariate in all models containing cortisol measures. In around 20% of the subjects the maximum cortisol level was observed in the awakening sample. For these subjects, it is not possible to infer whether the awakening samples include the peak, or whether the cortisol levels peaked before the awakening samples were taken, for instance due to discrepancies in actual awakening and reported awakening, or because they may have a different response to awakening. We attempted to control for this uncertainty by including an additional covariate in all models describing whether the peak was in the awakening sample or in the subsequent samples. Further, when splitting subjects into peak-at-awakening and peak-after-awakening subgroups, the observed associations between CAR and ROI asymmetries became stronger in the peakafter-awakening subgroup. In the peak-at-awakening subgroup no associations were observed, potentially due to uncertainty of the peak or lack of statistical power. Some limitations to the present study should be noted. First, cortisol samples were collected over a single day, which might give a less precise estimate of CAR than if samples were collected over two or

17

more days. Second, the number of days between the different assessments (MRI, NEO-PI-R, and cortisol sampling) differed between subjects. Notably, including the number of days between assessments as a covariate, or excluding subjects with more than 1 month between assessments did not change the results. Moreover, neuroticism generally remains stable in adulthood (Caspi et al. 2005), and CAR shows relatively high intra-subject stability (Pruessner et al. 1997), though factors such as job stress and general life stress may influence CAR (Chida and Steptoe 2009). Although, these limitations may have added noise to the data, it is unlikely that they mediated the observed results, and effects may have been stronger without these limitations. In conclusion, our findings provide a new perspective on the biology of basal HPA-axis reactivity and negative emotionality in humans. Our study significantly extends previous animal work by showing that in humans the balance between left- and right-sided limbic circuits is significantly associated with individual variations in HPA-axis reactivity, and with the tendency to experience negative emotions. Critically, our findings raise important questions how limbic circuits are related to negative emotionality and neuroendocrine functions.

Acknowledgements
The study was performed within the framework of the Center for Integrated Molecular Brain Imaging. This work was supported by the Lundbeck Foundation and the Dagmar Marshalls Foundation. TLJ receives support for this work from the Veterans Affairs Center of Excellence for Stress and Mental Health, La Jolla, CA, United States.

18

References
Adam, E.K., Doane L.D., Zinbarg R.E., Mineka S., Craske M.G., Griffith J.W., 2010. Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 35, 921-931. Andersson, J.L., Hutton C., Ashburner J., Turner R., Friston K., 2001. Modeling geometric deformations in EPI time series. Neuroimage 13, 903-919. Andersson, J.L.R., Jenkinon M., Smith S., 2007. Non-linear optimisation and Non-linear registration, aka Spatial Normalisation. FMRIB technical report TR07JA1 and TR07JA2, Beaulieu, C. 2009. The biological basis of diffusion anisotropy. In: Johansen-Berg, H., Behrens, T.E.J., (Eds.), Diffusion MRI: from quantitative measurement to in-vivo neuroanatomy Elsevier, London, pp. 105-126. Bienvenu, O.J., Nestadt G., Samuels J.F., Costa P.T., Howard W.T., Eaton W.W., 2001. Phobic, panic, and major depressive disorders and the five-factor model of personality. Journal of Nervous and Mental Disease 189, 154-161. Caspi, A., Roberts B.W., Shiner R.L., 2005. Personality development: stability and change. Annual Review of Psychology 56, 453-484. Cerqueira, J.J., Almeida O.F., Sousa N., 2008. The stressed prefrontal cortex. Left? Right! Brain, Behavior, and Immunity 22, 630-638. Cerqueira, J.J., Catania C., Sotiropoulos I., Schubert M., Kalisch R., Almeida O.F., Auer D.P., Sousa N., 2005. Corticosteroid status influences the volume of the rat cingulate cortex - a magnetic resonance imaging study. Journal of Psychiatric Research 39, 451-460. Chang, L.C., Jones D.K., Pierpaoli C., 2005. RESTORE: robust estimation of tensors by outlier rejection. Magnetic Resonance in Medicine 53, 1088-1095. Chida, Y., Steptoe A., 2009. Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biological Psychology 80, 265-278. Clow, A., Thorn L., Evans P., Hucklebridge F., 2004. The awakening cortisol response: methodological issues and significance. Stress 7, 29-37. Cremers, H.R., Demenescu L.R., Aleman A., Renken R., van Tol M.J., van der Wee N.J., Veltman D.J., Roelofs K., 2010. Neuroticism modulates amygdala-prefrontal connectivity in response to negative emotional facial expressions. Neuroimage 49, 963-970. Dedovic, K., Duchesne A., Andrews J., Engert V., Pruessner J.C., 2009. The brain and the stress axis: the neural correlates of cortisol regulation in response to stress. Neuroimage 47, 864-871.

19

DeYoung, C.G., Hirsh J.B., Shane M.S., Papademetris X., Rajeevan N., Gray J.R., 2010. Testing predictions from personality neuroscience: brain structure and the big five. Psychological Science 21, 820-828. Drevets, W.C., Savitz J., Trimble M., 2008. The subgenual anterior cingulate cortex in mood disorders. CNS Spectrums 13, 663-681. Fries, E., Dettenborn L., Kirschbaum C., 2009. The cortisol awakening response (CAR): facts and future directions. International Journal of Psychophysiology 72, 67-73. Frokjaer, V.G., Mortensen E.L., Nielsen F.A., Haugbol S., Pinborg L.H., Adams K.H., Svarer C., Hasselbalch S.G., Holm S., Paulson O.B., Knudsen G.M., 2008. Frontolimbic serotonin 2A receptor binding in healthy subjects is associated with personality risk factors for affective disorder. Biological Psychiatry 63, 569-576. Gong, G., Jiang T., Zhu C., Zang Y., He Y., Xie S., Xiao J., 2005. Side and handedness effects on the cingulum from diffusion tensor imaging. Neuroreport 16, 1701-1705. Hansen, H.S., Mortensen E.L., Schitz H.K. 2004. NEO-PI-R, manual - klinisk. editors. Dokumentation for den danske udgave af NEO PI-R og NEO PI-R Kort Version. Dansk Psykologisk Forlag, Copenhagen. Hasan, K.M., Iftikhar A., Kamali A., Kramer L.A., Ashtari M., Cirino P.T., Papanicolaou A.C., Fletcher J.M., Ewing-Cobbs L., 2009. Development and aging of the healthy human brain uncinate fasciculus across the lifespan using diffusion tensor tractography. Brain Research 1276, 67-76. Herman, J.P., Ostrander M.M., Mueller N.K., Figueiredo H., 2005. Limbic system mechanisms of stress regulation: hypothalamo-pituitary-adrenocortical axis. Progress in Neuropsychopharmacology and Biological Psychiatry 29, 1201-1213. Highley, J.R., Walker M.A., Esiri M.M., Crow T.J., Harrison P.J., 2002. Asymmetry of the uncinate fasciculus: a post-mortem study of normal subjects and patients with schizophrenia. Cerebral Cortex 12, 1218-1224. Hua, K., Zhang J., Wakana S., Jiang H., Li X., Reich D.S., Calabresi P.A., Pekar J.J., van Zijl P.C., Mori S., 2008. Tract probability maps in stereotaxic spaces: analyses of white matter anatomy and tractspecific quantification. Neuroimage 39, 336-347. Jang, K.L., Livesley W.J., Vernon P.A., 1996. Heritability of the big five personality dimensions and their facets: a twin study. Journal of Personality 64, 577-591. Jankord, R., Herman J.P., 2008. Limbic regulation of hypothalamo-pituitary-adrenocortical function during acute and chronic stress. Annals of the New York Academy of Science 1148, 64-73.

20

Jovicich, J., Czanner S., Greve D., Haley E., van der Kouwe A., Gollub R., Kennedy D., Schmitt F., Brown G., Macfall J., Fischl B., Dale A., 2006. Reliability in multi-site structural MRI studies: effects of gradient non-linearity correction on phantom and human data. Neuroimage 30, 436-443. Kalin, N.H., Larson C., Shelton S.E., Davidson R.J., 1998. Asymmetric frontal brain activity, cortisol, and behavior associated with fearful temperament in rhesus monkeys. Behavioral Neuroscience 112, 286292. Kendler, K.S., Gatz M., Gardner C.O., Pedersen N.L., 2006. Personality and major depression: a Swedish longitudinal, population-based twin study. Archives of General Psychiatry 63, 1113-1120. Kendler, K.S., Kuhn J., Prescott C.A., 2004. The interrelationship of neuroticism, sex, and stressful life events in the prediction of episodes of major depression. American Journal of Psychiatry 161, 631636. Kim, S.J., Jeong D.U., Sim M.E., Bae S.C., Chung A., Kim M.J., Chang K.H., Ryu J., Renshaw P.F., Lyoo I.K., 2006. Asymmetrically altered integrity of cingulum bundle in posttraumatic stress disorder. Neuropsychobiology 54, 120-125. Knoops, A.J., Gerritsen L., van der Graaf Y., Mali W.P., Geerlings M.I., 2010. Basal hypothalamic pituitary adrenal axis activity and hippocampal volumes: the SMART-Medea study. Biological Psychiatry 67, 1191-1198. Kremen, W.S., O'Brien R.C., Panizzon M.S., Prom-Wormley E., Eaves L.J., Eisen S.A., Eyler L.T., Hauger R.L., Fennema-Notestine C., Fischl B., Grant M.D., Hellhammer D.H., Jak A.J., Jacobson K.C., Jernigan T.L., Lupien S.J., Lyons M.J., Mendoza S.P., Neale M.C., Seidman L.J., Thermenos H.W., Tsuang M.T., Dale A.M., Franz C.E., 2010. Salivary cortisol and prefrontal cortical thickness in middle-aged men: A twin study. Neuroimage 53, 1093-1102. Lebel, C., Walker L., Leemans A., Phillips L., Beaulieu C., 2008. Microstructural maturation of the human brain from childhood to adulthood. Neuroimage 40, 1044-1055. Lueken, U., Leisse M., Mattes K., Naumann D., Wittling W., Schweiger E., 2009. Altered tonic and phasic cortisol secretion following unilateral stroke. Psychoneuroendocrinology 34, 402-412. MacLullich, A.M., Ferguson K.J., Wardlaw J.M., Starr J.M., Deary I.J., Seckl J.R., 2006. Smaller left anterior cingulate cortex volumes are associated with impaired hypothalamic-pituitary-adrenal axis regulation in healthy elderly men. Journal of Clinical Endocrinology & Metabolism 91, 1591-1594. Omura, K., Todd Constable R., Canli T., 2005. Amygdala gray matter concentration is associated with extraversion and neuroticism. Neuroreport 16, 1905-1908. Park, H.J., Westin C.F., Kubicki M., Maier S.E., Niznikiewicz M., Baer A., Frumin M., Kikinis R., Jolesz F.A., McCarley R.W., Shenton M.E., 2004. White matter hemisphere asymmetries in healthy subjects and in schizophrenia: a diffusion tensor MRI study. Neuroimage 23, 213-223.

21

Phillips, M.L., Ladouceur C.D., Drevets W.C., 2008. A neural model of voluntary and automatic emotion regulation: implications for understanding the pathophysiology and neurodevelopment of bipolar disorder. Molecular Psychiatry 13, 829, 33-57. Portella, M.J., Harmer C.J., Flint J., Cowen P., Goodwin G.M., 2005. Enhanced early morning salivary cortisol in neuroticism. American Journal of Psychiatry 162, 807-809. Pruessner, J.C., Dedovic K., Khalili-Mahani N., Engert V., Pruessner M., Buss C., Renwick R., Dagher A., Meaney M.J., Lupien S., 2008. Deactivation of the limbic system during acute psychosocial stress: evidence from positron emission tomography and functional magnetic resonance imaging studies. Biological Psychiatry 63, 234-240. Pruessner, J.C., Wolf O.T., Hellhammer D.H., Buske-Kirschbaum A., von Auer K., Jobst S., Kaspers F., Kirschbaum C., 1997. Free cortisol levels after awakening: a reliable biological marker for the assessment of adrenocortical activity. Life Sciences 61, 2539-2549. Pruessner, M., Pruessner J.C., Hellhammer D.H., Bruce Pike G., Lupien S.J., 2007. The associations among hippocampal volume, cortisol reactivity, and memory performance in healthy young men. Psychiatry Research 155, 1-10. Reese, T.G., Heid O., Weisskoff R.M., Wedeen V.J., 2003. Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo. Magnetic Resonance in Medicine 49, 177-182. Reul, J.M., Gesing A., Droste S., Stec I.S., Weber A., Bachmann C., Bilang-Bleuel A., Holsboer F., Linthorst A.C., 2000. The brain mineralocorticoid receptor: greedy for ligand, mysterious in function. European Journal of Pharmacology 405, 235-249. Sanchez, M.M., Young L.J., Plotsky P.M., Insel T.R., 2000. Distribution of corticosteroid receptors in the rhesus brain: relative absence of glucocorticoid receptors in the hippocampal formation. Journal of Neuroscience 20, 4657-4668. Schmahmann, J.D., Pandya D.N. 2006. Fiber Pathways of the Brain. Oxford University Press, Oxford. Schwartz, E.D., Cooper E.T., Fan Y., Jawad A.F., Chin C.L., Nissanov J., Hackney D.B., 2005. MRI diffusion coefficients in spinal cord correlate with axon morphometry. Neuroreport 16, 73-76. Smith, S.M., Jenkinson M., Johansen-Berg H., Rueckert D., Nichols T.E., Mackay C.E., Watkins K.E., Ciccarelli O., Cader M.Z., Matthews P.M., Behrens T.E., 2006. Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data. Neuroimage 31, 1487-1505. Sullivan, R.M., 2004. Hemispheric asymmetry in stress processing in rat prefrontal cortex and the role of mesocortical dopamine. Stress 7, 131-143. Sullivan, R.M., Gratton A., 1999. Lateralized effects of medial prefrontal cortex lesions on neuroendocrine and autonomic stress responses in rats. Journal of Neuroscience 19, 2834-2840.

22

Versace, A., Almeida J.R., Hassel S., Walsh N.D., Novelli M., Klein C.R., Kupfer D.J., Phillips M.L., 2008. Elevated left and reduced right orbitomedial prefrontal fractional anisotropy in adults with bipolar disorder revealed by tract-based spatial statistics. Archives of General Psychiatry 65, 10411052. Vos, S.B., Jones D.K., Viergever M.A., Leemans A., 2011. Partial volume effect as a hidden covariate in DTI analyses. Neuroimage, In Press. Vreeburg, S.A., Hoogendijk W.J., van Pelt J., Derijk R.H., Verhagen J.C., van Dyck R., Smit J.H., Zitman F.G., Penninx B.W., 2009. Major depressive disorder and hypothalamic-pituitary-adrenal axis activity: results from a large cohort study. Archives of General Psychiatry 66, 617-626. Vreeburg, S.A., Zitman F.G., van Pelt J., Derijk R.H., Verhagen J.C., van Dyck R., Hoogendijk W.J., Smit J.H., Penninx B.W., 2010. Salivary cortisol levels in persons with and without different anxiety disorders. Psychosomatic Medicine 72, 340-347. Westlye, L.T., Walhovd K.B., Dale A.M., Bjornerud A., Due-Tonnessen P., Engvig A., Grydeland H., Tamnes C.K., Ostby Y., Fjell A.M., 2009. Life-Span Changes of the Human Brain White Matter: Diffusion Tensor Imaging (DTI) and Volumetry. Cerebral Cortex 20, 2055-2068. Wilhelm, I., Born J., Kudielka B.M., Schlotz W., Wust S., 2007. Is the cortisol awakening rise a response to awakening? Psychoneuroendocrinology 32, 358-366. Wittling, W., Pfluger M., 1990. Neuroendocrine hemisphere asymmetries: salivary cortisol secretion during lateralized viewing of emotion-related and neutral films. Brain and Cognition 14, 243-265. Wust, S., Federenko I., Hellhammer D.H., Kirschbaum C., 2000. Genetic factors, perceived chronic stress, and the free cortisol response to awakening. Psychoneuroendocrinology 25, 707-720.

23

Paper 4

Hypothalamic-pituitary-adrenal axis tonus is associated with hippocampal microstructural asymmetry

Kathrine Skak Madsen1,2,3, Terry L. Jernigan1,2,3,4, Pernille Iversen1,2, Vibe G. Frokjaer2,5, Gitte M. Knudsen2,3,5, Hartwig R. Siebner1,2,3, and William F.C. Baar1,2

Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital, Hvidovre, Denmark Center for Integrated Molecular Brain Imaging, Copenhagen, Denmark Faculty of Health Sciences, University of Copenhagen, Denmark Veterans Affairs Center of Excellence for Stress and Mental Health, San Diego, La Jolla, CA, United

States
5

Neurobiology Research Unit, Copenhagen University Hospital, Rigshospitalet, Copenhagen, Denmark

Short title: HPA-axis tonus and hippocampal asymmetry

Keywords: Diffusion tensor imaging, cortisol, hippocampus, MRI, laterality

Corresponding author Kathrine Skak Madsen Danish Research Centre for Magnetic Resonance MR-department, section 340 Copenhagen University Hospital, Hvidovre Kettegaard All 30 2650 Hvidovre Denmark Phone: +45 3662 3323, Fax: +45 3647 0302, E-mail: kathrine@drcmr.dk

Abstract
The hippocampus plays a central role in hypothalamic-pituitary-adrenal (HPA) axis regulation. Yet, morphometric studies linking hippocampal volumes to basal HPA-axis activity have produced inconsistent results. In the light of growing evidence for hemispheric lateralization in brain systems regulating arousal and emotion, we used diffusion-weighted imaging in healthy adults to test the hypotheses that individual variations in mean morning and afternoon/evening cortisol levels would be associated with the degree of hemispheric asymmetry in hippocampal mean diffusivity (MD). As predicted, higher left relative to right hippocampus MD was associated with higher basal cortisol levels. Associations were not attributable to hippocampal volume asymmetry, and no correlation between volume and MD was observed, suggesting that MD and volume index distinct biological properties of the hippocampus. The associations between hippocampal MD asymmetry and basal cortisol levels raises a number of possibilities, among them a potential asymmetric role of the hippocampus on HPA-axis regulation, or conversely, that individual variations in secreted cortisol, perhaps associated with stress, may have lateralized effects on hippocampal microstructure. The relationship between hippocampal MD asymmetry and HPA-axis tonus underscores that hemispheric asymmetries should be taken into account when trying to unravel links between the architecture of the limbic system and neuroendocrine functions.

Introduction
The hippocampus plays an important role in stress regulation by modulating the hypothalamic-pituitaryadrenal (HPA)-axis (Jankord and Herman 2008, Dedovic et al. 2009). The HPA-axis regulates the secretion of cortisol in response to stress and is critically involved in coping with bodily and environmental challenges (Fries et al. 2009). Under normal conditions, basal HPA-axis activity follows a distinct diurnal profile, with plasma cortisol levels peaking approximately 30 minutes after morning awakening, commonly referred to as cortisol awakening response (CAR), followed by a gradual decrease in cortisol levels during the rest of the day (Pruessner et al. 1997, Clow et al. 2009, Fries et al. 2009). Disruption of the normal diurnal profile of the HPA-axis has been found in several psychiatric disorders. Although not consistently, lower basal cortisol levels and lower CAR have been associated with fatigue, burnout, and post-traumatic stress disorder (Chida and Steptoe 2009, Handwerger 2009), while higher morning cortisol levels have been associated with anxiety disorder (Vreeburg et al. 2010) and major depression (Chida and Steptoe 2009, Handwerger 2009, Vreeburg et al. 2009). Moreover, individuals with high CAR may have an increased risk of developing major depression (Adam et al. 2010). Finally, prolonged increases in cortisol level, for instance induced by chronic stress, have adverse effects on hippocampal neurons and glial cells (Starkman et al. 1992, Sapolsky 2000, Rajkowska and MiguelHidalgo 2007, Jauregui-Huerta et al. 2010). Previous research on the relationship between human brain structure and cortisol measures as an index of HPA-axis function has typically focused on the hippocampus in healthy subjects and pathological conditions (Wolf et al. 2002, Buchanan et al. 2004, Pruessner et al. 2007, Knoops et al. 2010). Bilateral and unilateral hippocampal damage has been reported to abolish the morning response, while preserving overall cortisol levels throughout the day (Buchanan et al. 2004). In patients with elevated cortisol levels due to Cushings syndrome, hippocampal volume was negatively correlated with mean plasma cortisol levels (Starkman et al. 1992), and following treatment to reduce glucocorticoid levels, patients had a significant increase in hippocampal volume, which correlated with a decrease in 24-hour urinary cortisol levels (Starkman et al. 1999). In healthy individuals, studies relating cortisol measures to hippocampal volume have yielded inconsistent results. In a small cohort of 13 young subjects, higher morning cortisol levels were associated with larger hippocampal volumes (Pruessner et al. 2007). By contrast, two studies in large, middle-aged cohorts (N>380) did not find such a relationship (Knoops et al. 2010, Kremen et al. 2010) but one of these reported that smaller hippocampal volumes were correlated with higher evening cortisol levels, as well as with higher morning cortisol levels after dexamethasone administration (Knoops

et al. 2010). Another study found that smaller hippocampi were associated with higher morning adrenocorticotropin levels and higher 24-hour urine cortisol levels (Wolf et al. 2002). Further, prolonged high basal cortisol levels have also been associated with age-related hippocampal atrophy in elderly subjects (Lupien et al. 1998). Magnetic resonance imaging (MRI) studies of HPA-axis function in humans have mainly focused on brain activity, or on gray matter volumes or cortical thickness. In recent years, diffusion-weighted imaging (DWI), which is sensitive to the diffusion of water molecules, has been increasingly used to study microstructural properties of the human grey and white matter in vivo. In white matter, water diffuses more readily parallel than perpendicular to fibre bundles, ostensibly because of the hindering effects of cellular structures such as the axonal membranes and surrounding myelin sheaths, causing diffusion anisotropy. In grey matter, diffusion is also hindered by cellular structures, but the diffusion directionality is generally isotropic because of the relatively large imaging voxels that are employed (Beaulieu 2009). By fitting the DWI measurements of each voxel to the diffusion tensor model, several useful measures can be derived, such as fractional anisotropy (FA), estimating the degree of diffusion directionality, as well as mean diffusivity (MD), estimating the overall magnitude of water diffusion in a given voxel (Basser et al. 1994, Beaulieu 2009). Some DWI studies have studied hippocampal MD in healthy individuals. Hippocampal MD remains relatively stable in adulthood until the fifties and then progressively increases with age (Carlesimo et al. 2010). Two recent studies in healthy subjects found lower MD in the hippocampus in healthy adults who had a higher level of education (Piras et al. 2010) and showed better memory performance (Carlesimo et al. 2010), suggesting a link between hippocampal MD and function. Moreover, increased hippocampal MD has been observed in patients with unipolar major depression (Abe et al. 2010) and schizophrenia (Spoletini et al. 2011) relative to healthy control subjects. A single DWI study reported on hemispheric asymmetry of hippocampal microstructure, where healthy subjects and subjects with mild cognitive impairment showed higher MD in the right relative to left hippocampus (Muller et al. 2007). This observation ties in with a growing body of data suggesting that hemispheric lateralization is linked to the HPA-axis. Stroke patients with left-sided, but not right-sided, infarctions showed increased morning cortisol levels compared to controls (Lueken et al. 2009). In chronically-restrained (stressed) pigs, higher basal afternoon cortisol levels were associated with decreased volume and neuron number in the left, but not the right dentate gyrus (van der Beek et al. 2004). Using DWI, we recently found that the degree of microstructural asymmetry in the two major limbic fibre bundles, the cingulum and uncinate fasciculus, were associated with individual differences in the CAR (Madsen et al., Submitted). This raises the possibility that the individual expression of structural asymmetry between left and right limbic structures bear a relationship to HPA-axis functioning.

Following-up on our previous work, we used DWI to test the hypothesis that MD asymmetry in the human hippocampus predicts individual variations in tonic HPA-axis output as reflected by mean morning and mean afternoon/evening (PM) cortisol levels.

Materials and methods

Participants
Fifty-eight healthy adults aged 19-86 years were included. None of them had any self-reported history of neurological or psychiatric disorders or brain injury or contraindications for MR. None of them took psychoactive drugs, steroids, hormone replacements, or drugs of abuse, including anabolic steroids. Subjects had a body-mass-index (BMI) below 35 kg/m2 (i.e., were not severely obese). Four subjects received antihypertensive treatment with thiazide diuretics, ACE inhibitors, or angiotensin II receptor antagonists and all were normotensive. No participants received beta-blocking agents. Handedness was assessed with the Edinburgh Handedness Inventory. Educational level after basic school was scored on a five-point scale (1 - No education; 2 - Special worker; 3 - Skilled worker; 4 - Longer theoretical education, 5 Academic education). Ten of the 58 subjects were excluded because less than four saliva samples were available within the first hour after awakening (N=7), not following the saliva sampling schedule (N=2), or because the participant reported intense exam related stress (N=1). In total, 48 subjects aged 19-86 years were included in the statistical analyses. For one of those subjects, PM cortisol levels were not available. Demographic data of the 48 subjects are presented in Table 1. The study was approved by the local (Copenhagen and Frederiksberg) Committee for Biomedical Research Ethics (protocol (KF) 01-2006-20, and performed in accordance with the Declaration of Helsinki. Informed written consent was obtained from all subjects prior to participation. Table 1. Demographic data of the participants (N=48). Age (years)* Gender (male / female) BMI (kg/m )* Education (group score)* Handedness (right / left)
*Values are given as mean standard deviation
2

37.4 19.3 34 / 14 25.1 2.6 3.9 1.3 45 / 3

Salivary cortisol assessment

Participants collected 8 serial saliva samples at home using Salivette sampling devices (Sarstedt, Neubringen, Germany) 0, 15, 30, 45 and 60 minutes after morning awakening, and at noon, 6 PM and 11 PM, after having been trained in the sampling technique in the lab. Participants were instructed to rinse their mouths before sampling, refrain from eating, smoking, drinking caffeinated beverages and brushing teeth until after collecting the last morning sample, and to register the exact times of falling asleep the night before and awakening on the sampling day, as well as the actual time for each sampling. The timeinterval between awakening and taking the first sample ranged from 0-10 minutes (mean standard deviation (SD) = 2.8 2.9). The number of days between cortisol sampling and MRI ranged from 1-279 days (median = 15 days) with 10 participants having more than 1 month delay between the assessments. At home, samples were stored 5OC for up to 3 days. Samples were sent by 24-hour mail to the Cimbi laboratory, and stored there at -80OC. Cortisol concentrations were determined by a commercially available electrochemiluminescence immunoassay (ECLIA) (Roche, Mannheim, Germany) on Modular Analytics E170 equipment. The entire batch of samples was analyzed in one step, ensuring stability of laboratory dependent conditions. Intra-assay coefficients of variance have routinely been found to be below 6.1%. The lower limit of detection was 0.5 nmol/L, and values below this were scored as 0.25 nmol/L. Values below 0.5 nmol/L were only found in PM samples (6 PM (N=6) and 11 PM (N=15)). Averages of the morning and PM cortisol samples were used as estimates of HPA-axis tonus. The mean morning cortisol level was based on available samples collected within the first hour after awakening, and mean PM cortisol level was based on available samples collected in the afternoon/evening. The cortisol awakening response (CAR) was used as an estimate of HPA-axis reactivity. CAR was defined as the percentage difference between maximum and mean morning cortisol level [CAR = ((MaximumMorning MeanMorning) / MeanMorning) * 100)].

Image acquisition
All subjects were scanned using a 3T Siemens scanner (Siemens, Erlangen, Germany) with an eightchannel head coil (Invivo, FL, USA). All acquired scans were aligned parallel to the anterior commisure posterior commisure line. T1-weighted images of the whole head were acquired using a 3D MPRAGE sequence (TR=1550 ms, TE=3.04 ms, matrix=256x256, 192 sagittal slices with no gap, 1 mm3 isotropic voxels). T2-weighted images of the whole head were acquired using a 3D turbo spin echo sequence (TR=3000 ms, TE=354 ms, matrix=256x196, 192 sagittal slices, 1.1x1.1x1.1 mm3 isotropic voxels). Whole brain diffusion weighted (DW) images were acquired using a twice-refocused balanced spin echo sequence that minimised eddy current distortion (Reese et al. 2003). Ten non-DW images (b=0) and 61 DW images (b=1200), encoded along independent collinear diffusion gradient orientations, were acquired

(TR=8200 ms; TE=100 ms; FOV=220x220, matrix=96x96; GRAPPA: factor 2, 48 lines; 61 transverse slices; no gap; 2.3x2.3x2.3 mm3 voxels). A gradient echo field map was acquired to correct B0 field distortions (Andersson et al. 2001) (TR=530 ms; TE[1]=5.19 ms and TE[2]=7.65 ms; FOV=256x256; matrix=128x128; 47 transverse slices; no gap, 2x2x3 mm3 voxels).

Image preprocessing
Raw images were visually inspected to ascertain data quality. Images were preprocessed using pipelines implemented in Matlab, using mainly SPM5 routines. T1-weighted and T2-weighted images were corrected for spatial distortions due to non-linearity in the gradient system of the scanner (Jovicich et al. 2006). The T2-image was coregistered (no reslicing), using a 6 parameter mutual information rigid transformation to the T1-image. Both T1 and T2 images were processed using the VBM5 toolbox in SPM5 (http://dbm.neuro.uni-jena.de/vbm/vbm5-for spm5/), which includes a unified segmentation algorithm (Ashburner and Friston 2005) and a Hidden Markov Random Field (HMRF) method (Cuadra et al. 2005). T2-weighted images were used to automatically create brain masks in native space. Resulting brain masked grey and white matter tissue maps in native space, the affine part of the spatial transformation from native to MNI space (obtained from the VBM5 analysis), and their left-right flipped counterparts were subsequently used in DARTEL (Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra), using default settings (Ashburner 2007), for high-dimensional inter-subject registration. Including the left-right flipped tissue maps ensured that resulting flow fields that parameterize the deformations were symmetric. Using the final symmetric ow elds, the native T1-images were warped into average symmetric image space (DARTEL space). In the DWI analysis, each subjects mean b0 image was coregistered to the T2-image (no reslicing), after which all DW images were coregistered to the mean b0 image (no reslicing). Next, all coregistered images were corrected for geometric distortions using a voxel displacement map based on both the acquired B0 field map (Andersson et al. 2001) and the scanner specific gradient non-linearities (Jovicich et al. 2006). Finally, all images were resliced using trilinear interpolation. Note that this procedure involves only one reslicing step. The diffusion gradient orientations were adjusted to account for any rotation applied during registration. The diffusion tensor was fitted using the RESTORE algorithm (Chang et al. 2005) with a noise standard deviation of 30, and FA, mean diffusivity (MD = (1 + 2 + 3) / 3) as well as perpendicular diffusivity ( = (2 + 3) / 2)) were calculated. A brain mask automatically generated on the T2-weighted image was applied to the MD images.

Grey matter regions-of-interest

Mean diffusivity measures were obtained for left and right hippocampus and amygdala for each subject for statistical analyses. Regions-of-interest (ROIs) were drawn onto the symmetric average of the DARTEL-warped T1-images of all subjects using FSLview. As a first step, to ensure that the ROIs included only grey matter, an approximately one-voxel thick layer of grey matter at the white matter/CSF grey matter borders of the hippocampus was excluded (see Fig. 1). Resulting binary ROIs were resliced into native T1-space using inverse deformation fields using a trilinear interpolation. Next, ROIs were binarized using a threshold of 0.5, and overlaid on native T1-images to visually inspect the anatomical fit of the ROIs for each individual subject. To transform ROIs from native T1-space to native DTI space, all subjects FA images were warped to the subjects own native T1-image using FNIRT. Using the inverse warp field, ROIs were resliced into native DTI space using trilinear interpolation and binarized using a threshold of 0.5. Again, the anatomical fit of the ROIs was visually inspected. Finally, binary ROIs were multiplied with subjects own FA images thresholded at FA<0.25 to exclude potential white matter (partial-volume) voxels, and subsequently multiplied with subjects own image, thresholded at < 0.0012, to exclude potential CSF (partial volume) voxels. The mean (SD) number of 12 mm3 voxels in the ROIs were: left hippocampus = 143.4 (24.9), right hippocampus = 137.6 (22.7) voxels, left amygdala = 79.6 (15.4), and right amygdala = 80.2 (14.8).

Figure 1. Hippocampal regions-of-interest (ROIs) in DARTEL space. The hippocampal ROIs used to extract mean diffusivity (MD - red) and volume (blue) are overlaid on the symmetric average of the DARTEL-warped T1-images depicted in (a) sagittal, (b) coronal, and (c) axial view. In (d), smoothed 3-D versions of the MD hippocampal ROIs depicted together with a representation of the symmetric average DARTEL-warped grey and white matter segmentations.

The left and right hippocampal volumes were extracted for post hoc analyses. The hippocampus ROIs were drawn onto the symmetric average DARTEL template (see Fig. 1) using established criteria (Maller et al. 2006), and subsequently hand edited to fit each subjects DARTEL-warped image. For all subjects, the number of voxels and mean intensity values within the left and right hippocampus ROIs were extracted and multiplied to get the hippocampal volumes: (mean (SD): left hippocampus = 3013.6 (383.7), right hippocampus = 3045.5 (376.8) 1 mm3 voxels). Further, total brain volume (including both cerebrum and cerebellum) was extracted from all subjects by calculating and adding GM and WM volumes from respective tissue maps (mean SD = 1,247,547 152,552). Relative volumes of the left and right hippocampus (ROI / brain volume) were used in statistical analyses.

Asymmetry measures
Hippocampus and amygdala MD asymmetries were calculated as the percentage difference between left and right ROI values expressed as a percentage of the bilateral mean: ((2 * (Left Right)) / (Left + Right)) * 100.

Statistical analysis
Statistical analyses were performed in SPSS19 (SPSS Inc., Chicago, USA). Analyses using mean morning cortisol levels or CAR as dependent variable were based on 48 individuals, while analyses involving mean PM cortisol levels included data from 47 individuals. Two-tailed paired t-tests were used to compare the left and right ROI values. Age and gender effects were examined using linear regression models. Multiple linear regression was used to test our hypotheses. Shapiro-Wilk tests showed that all data were normally distributed except for mean PM cortisol level and CAR. Logarithm-transformation of the mean PM cortisol levels and CAR data normalized the residuals. All other assumptions for linear regression were fulfilled. The statistical tests were performed hierarchically. Major hypotheses were tested with 0.025 (Bonferroni corrected for 2 tests), while follow-up analyses were considered significant when 0.05. The major hypotheses were that mean morning cortisol level would be associated with hippocampus MD asymmetry, adjusted for age, age2, gender and time-interval between awakening and taking the first cortisol sample and that mean PM cortisol level was hypothesized to be associated with hippocampus MD asymmetry, after adjusting for age, age2, and gender effects. Planned follow-up analyses were contingent on observing significant associations in our primary analyses. Follow-up analyses assessed the relative contribution of left and right hippocampus MD to the observed effects, and ascertained that observed effects were not mediated by other known factors that may affect the cortisol or MD measures, e.g., sleep duration, awakening time, interval between sunrise and

awakening, hours of daylight, weekday vs. weekend cortisol sampling, BMI, handedness, educational level, smoking, antihypertensive treatment, or interval between MR-scan and cortisol sampling. Furthermore, we assessed the anatomical specificity of hippocampus MD asymmetry effects on mean cortisol levels by including amygdala MD asymmetry as an additional covariate in the models. Additional follow-up analyses were conducted to examine the influence of hippocampal volume on observed associations between hippocampal MD asymmetry and mean morning and PM cortisol levels by including hippocampal volume asymmetry as an additional covariate in the models. Finally, exploratory analyses were conducted to examine whether the hippocampus and amygdala MD asymmetries were associated with CAR, a measure of HPA-axis reactivity.

Results
Cortisol and ROI MD measures are summarized in Table 2. MD in the right hippocampus was significantly higher than MD in left hippocampus (p<10-8), while no right-left difference in MD was found in amygdala (p=0.72). Women had significant lower CAR (p=0.037) than men, after correcting for age effects. There was a trend towards a positive age effect on the mean morning cortisol level (p=0.055), adjusted for gender. Table 2. Mean standard deviation values for regions-of-interest (ROIs) and cortisol measures Whole group (N=48) ROI mean diffusivity (MD - 10-3 mm2/s) Right hippocampus MD Left hippocampus MD Right amygdala MD Left amygdala MD ROI MD asymmetry (%)* Hippocampus Amygdala Cortisol measures** Mean morning level (nmol/L) Mean PM level (nmol/L)*** Awakening Response 8.64 4.20 2.87 2.53 49.1 27.0 8.17 4.20 2.77 2.72 53.1 27.2 9.79 4.13 3.12 2.07 39.4 24.8 -1.77 1.94 -0.20 2.97 -1.95 2.08 -0.30 3.04 -1.33 1.56 0.05 2.89 0.881 0.021 0.866 0.026 0.825 0.024 0.824 0.033 0.877 0.020 0.860 0.025 0.823 0.025 0.821 0.034 0.891 0.021 0.879 0.023 0.830 0.019 0.830 0.032 Men (N=34) Women (N=14)

* Percentage MD asymmetry in ROIs. Negative values indicate that MD in the left is lower than the right ROI. ** Mean PM cortisol level and cortisol awakening response values are raw values and not logarithm-transformed values that were used in the statistical analyses. *** Mean PM cortisol level was based on 47 subjects (33 Males, 14 Females).

10

Table 3. Main results from the a priori hypotheses and follow-up analyses predicting cortisol measures Hippocampus MD asymmetry R2 0.286 0.295 0.217 0.234 * 0.373 0.366 0.418 0.409 P 0.008 0.010 0.005 0.006 0.139 0.348 0.106 0.455 Mean morning cortisol level** Hippocampal volume asymmetry * P

Mean PM cortisol level***

* The standardized regression coefficient for the variables of interest controlling for model covariates. ** In models predicting mean morning cortisol level, effects of hippocampus MD asymmetry were tested after adjusting for age, age2, gender, time interval between awakening and the first cortisol sample (and hippocampal volume asymmetry). *** In models predicting mean PM cortisol level, effects of hippocampus MD asymmetry were adjusted for age, age2, and gender (and hippocampal volume asymmetry). Abbreviations: MD = Mean diffusivity

Associations between mean cortisol levels and hippocampal MD asymmetry

As hypothesized, hippocampus MD asymmetry was significantly associated with both mean morning cortisol level and mean PM cortisol level (see Table 3 and Fig. 2). The higher the right relative to left hippocampal MD, the lower were the morning cortisol levels. Follow-up analyses revealed that it was the ratio between regional MD in left and right hippocampus that exhibited an association with mean morning and PM cortisol levels, and not the left or right hippocampus MD individually. Using left and right hippocampus MD as simultaneous predictors of either mean morning or mean PM cortisol levels revealed that left hippocampus MD exhibited a significant positive and right hippocampus MD a significant negative relationship with both cortisol measures (Table 4). Notably, no significant correlations with mean morning or PM cortisol levels were found for left or right hippocampus MD when modelled separately (Table 4). The observed associations between mean cortisol levels and hippocampus MD asymmetry remained significant when expanding the models to include one of the additional covariates: sleep duration, awakening time, interval between sunrise and awakening, hours of daylight, weekday vs. weekend cortisol sampling, BMI, handedness, educational level, smoking, antihypertensive treatment, or interval between MR-scan and cortisol sampling. To further evaluate if the number of days between the cortisol sampling and MR-scan could be a cofounding factor, additional analyses were made in a subgroup with less that 1month interval between these assessments. In this subgroup (N=38), hippocampus MD asymmetry

11

remained a significant predictor of mean morning (=0.399, p=0.011) and PM (=0.524, p=0.002) cortisol levels, suggesting that the effects are not mediated by varying intervals between the assessments.

Figure 2. Partial regression plots of the main findings. Left) Mean morning cortisol level is plotted as a function of hippocampus mean diffusivity (MD) asymmetry, adjusted for age, age2, gender, time-interval between awakening and first cortisol sample. Right) The logarithm-transformed mean afternoon/evening (PM) cortisol level plotted as a function of hippocampus MD asymmetry, adjusted for age, age2, and gender. Note that residualized measures are plotted. Dotted lines indicate 95% confidence intervals.

The anatomical specificity of the associations between hippocampus MD asymmetry and mean cortisol levels was assessed by including amygdala MD asymmetry as an additional covariate in the models. The association between hippocampus MD asymmetry and mean morning and PM cortisol levels remained significant (morning: =0.426, p=0.003; PM: =0.426, p=0.006) when including amygdala MD asymmetry (morning: =-0.265, p=0.057; PM: =-0.043, p=0.77), suggesting that the hippocampus MD asymmetry effects are not mediated by a general limbic MD asymmetry. When considering the MD in the amygdala alone, neither mean morning cortisol level (p=0.23) nor mean PM cortisol level (p=0.79) were significantly associated with amygdala MD asymmetry.

No contribution of hippocampal volume to the observed associations

The follow-up analyses, which considered hippocampal volume, revealed that regional volume and MD may reflect distinct biological properties of the hippocampus: MD asymmetry of the hippocampus was not correlated with hippocampal volume asymmetry (p=0.42) adjusted for age and gender. Likewise, left or right hippocampus MD showed no linear relationship with respectively the left or right relative hippocampal volume (p0.44). Most importantly, the associations between hippocampus MD asymmetry and, respectively, mean morning cortisol level and mean PM cortisol level remained significant when including hippocampal volume asymmetry as an additional covariate in the statistical models (see Table

12

3). Neither hippocampal volume asymmetry (p0.30), nor left or right relative hippocampal volume alone (p0.11) were significantly associated with mean morning or mean PM cortisol levels.

Table 4. Follow-up analyses assessing the relative contribution of mean diffusivity (MD) in left and right hippocampus on mean morning or afternoon/evening (PM) cortisol levels. Left hippocampus MD R2 0.287 0.198 0.163 Mean PM cortisol level (n=47)*** 0.223 0.137 0.055 0.656 0.352 0.005 0.051 -0.043 0.803 -0.445 0.039 * 0.572 0.253 p 0.011 0.143 -0.102 0.536 Mean morning cortisol level (n=48)** -0.459 0.029 Right hippocampus MD * p

* The standardized regression coefficient for the variables of interests controlling for model covariates. ** Models predicting mean morning cortisol level either by modelling left and right hippocampus MD simultaneously (top) or individually (bottom two) were adjusted for age, age2, gender and time-interval between awakening and taking the first cortisol sample. *** Models predicting mean PM cortisol level either by modelling left and right hippocampus MD simultaneously (top) or individually (bottom two) were adjusted for age, age2, and gender.

Associations between CAR and ROI MD asymmetries

Exploratory analyses showed no significant associations between CAR and hippocampus MD asymmetry (p=0.89), indicating that MD asymmetry in the hippocampus is not related to HPA-axis reactivity to awakening. Further, no significant associations were observed between CAR and amygdala MD asymmetry (p=0.73).

Discussion
The main novel finding was a relationship between individual variations in the basal cortisol levels and the expression of microstructural asymmetry in the hippocampus. Specifically, higher mean morning and PM cortisol levels were associated with lower MD in the right relative to left hippocampus. Importantly, it was the ratio between left and right hippocampus MD and not the absolute MD values in left or right

13

hippocampus individually that was predicted basal cortisol levels. The function-structure relationship between morning and PM cortisol levels and hippocampus MD asymmetry remained significant when including factors that could potentially influence the cortisol measures or the diffusion parameters. There are some indications that left-hemispheric limbic structures may be more influenced by high cortisol levels than those on the right (van der Beek et al. 2004, Cerqueira et al. 2005). However, it is unlikely that the observed association between cortisol levels and hippocampal MD asymmetry reflects a general hemispheric asymmetry of the limbic system. Follow-up analyses revealed that the relationship between hippocampus MD asymmetry and mean morning and PM cortisol levels remained significant when including hemispheric MD asymmetry of the amygdala in the statistical model. Therefore, we argue that the relationship between the HPA-axis tonus and MD asymmetry in the hippocampus was anatomically specific to the hippocampus. Further studies are needed to more precisely map the extent of these effects. Pending questions are whether the association between the HPA-axis tonus and MD asymmetry is expressed to a similar degree in all regions of the hippocampus. We also assessed the potential effect of hippocampal volume, because hippocampal volume has previously, though somewhat inconsistently, been related to basal HPA-axis activity (Lupien et al. 1998, Wolf et al. 2002, Pruessner et al. 2007, Knoops et al. 2010, Kremen et al. 2010). The MD effects remained significant when relative volumes were included in the models. Moreover, no significant associations were observed between hippocampal volumes and mean cortisol levels. These findings suggest that measurements of regional MD may provide a more sensitive means to detect links between neuroendocrinological measures and regional brain structure, because microstructural changes in a brain region may well influence regional MD without necessarily causing volume changes at the macrostructural level. The neurobiological factors that account for individual variations of MD in grey matter structures are poorly understood. Lower MD has been related to higher cell density in tumours (Chenevert et al. 2006, Gibbs et al. 2009). Moreover, a recent study of rats found that over an 18-month period after radiation temporal changes in apparent diffusion coefficient were negatively correlated with changes in glial cell staining in the hippocampus (Huang et al. 2010). Although MD does not measure cell density directly, these findings suggest that differences in cell (neuron and glia) density may contribute to the observed hippocampus effects in the present study. However, many other tissue parameters may also influence the regional magnitude of water diffusion, including membrane permeability, cell number and size (Chenevert et al. 2006), myelination and axon diameter of fibres originating or terminating in the hippocampus, as well as number, size and tortuosity of axons, dendrites and astrocytic processes. Our findings suggest that hippocampal MD asymmetry is specifically related to tonic, but not phasic circadian HPA-axis activity, as we did not observe a significant association between hippocampal MD

14

asymmetry and the phasic estimate of CAR. In the same sample, we previously found a significant association between individual variability in CAR and the extent of microstructural asymmetry in the cingulum and uncinate fasciculus, but not with mean morning cortisol level (Madsen et al., Submitted). It is not clear what biological mechanisms might underlie these two observations. Cortisol primarily acts through mineralocorticoid (MR) and glucocorticoid (GR) receptors. Cortisol has a 5-10 fold higher affinity for MRs than GRs (De Kloet et al. 1998, Reul et al. 2000, Herman et al. 2005). Tonic inhibitory control of the HPA-axis is thought to be mediated primarily by MRs (De Kloet et al. 1998, Reul et al. 2000, Herman et al. 2005), which are densely expressed in only the hippocampus and dorsolateral septum (Seckl et al. 1991, Reul et al. 2000, Sanchez et al. 2000). The GRs become increasingly involved at the circadian cortisol peak or in association with stress (De Kloet et al. 1998, Reul et al. 2000, Herman et al. 2005). GRs are widely expressed throughout the brain, e.g., in hippocampus, pituitary, paraventricular nucleus, and neocortex (Sanchez et al. 2000). Our observation of a relationship between hippocampal MD asymmetries and tonic, but not phasic, HPA-axis activity, and between cingulum and uncinate fasciculus asymmetries and phasic, but not tonic, activity (Madsen et al., Submitted), may therefore relate to differences in the anatomical distribution of MRs and GRs. Further research is needed to investigate this possibility. At present, we can only speculate about the neurobiological significance of the observed relationship between mean cortisol levels and the asymmetry of mean MD in the hippocampus. Left and right hippocampi may have differential regulatory effects on the HPA-axis. If so, the asymmetry may reflect primary hemispheric differences in regulatory systems resulting in individual differences in basal cortisol levels. In accordance with this hypothesis, the density of MRs in mice was reported to be higher in the right than in the left hippocampus (Neveu et al. 1998). An alternative account is that the differences in cortisol levels themselves may have driven the MD asymmetry of the hippocampus. There is some evidence supporting the possibility that prolonged periods of elevated cortisol secretion exert differential effects on the microstructure of left and right hippocampus. Higher basal afternoon cortisol levels have been associated with decreased volume and neuron number in the left, but not the right, dentate gyrus in chronic-restrained (stressed) pigs (van der Beek et al. 2004). Further, in a recent study (Zach et al. 2010) 3 weeks of corticosterone treatment in rats was associated with alterations of hippocampal volume asymmetry as well as measures of the number of neurons in hippocampal subfields. Finally, the observed association may be mediated by other biological factors, or it may result from the association of both limbic MD asymmetries and basal cortisol levels with unrelated, but correlated, factors. Some limitations to the present study should be noted. First, cortisol samples were collected over a single day, which might give a less precise estimate of basal cortisol level than if samples were collected over two or more days. Second, the number of days between the MR-scan, and cortisol sampling differed

15

between subjects. Notably, the results remained significant when including the number of days between the MR-scan and cortisol sampling as a covariate, or when excluding the subjects with more than 1 month between assessments. Moreover, mean cortisol levels generally show relatively high intra-subject stability (r=0.39-0.67) across days or a week (Pruessner et al. 1997), though factors such as general life stress may influence mean cortisol levels (Chida and Steptoe 2009). Although, these limitations may have added noise to the data, it is unlikely that they mediated the observed results, and effects may have been stronger without these limitations. In conclusion, we found that the balance between left and right hippocampal diffusivity is significantly associated with individual variation in basal cortisol levels, providing a new perspective on the biology of tonic HPA-axis activity in humans. Possible explanations might be that the left and right hippocampi have differential regulatory effects on the HPA-axis, or conversely that hippocampal microstructural asymmetry results from differential effects on the left and right hippocampus of high levels of secreted cortisol. These findings raise important questions about how the architecture of the limbic system is related to neuroendocrine functions.

Acknowledgements
The study was performed within the framework of the Center for Integrated Molecular Brain Imaging. This work was supported by the Lundbeck Foundation and the Dagmar Marshalls Foundation. TLJ receives support for this work from the Veterans Affairs Center of Excellence for Stress and Mental Health, La Jolla, CA, United States.

Conflicts of interest
None to declare

References
Abe O, Yamasue H, Kasai K, Yamada H, Aoki S, Inoue H et al. 2010. Voxel-based analyses of gray/white matter volume and diffusion tensor data in major depression. Psychiatry Res 181(1):64-70. Adam EK, Doane LD, Zinbarg RE, Mineka S, Craske MG, Griffith JW. 2010. Prospective prediction of major depressive disorder from cortisol awakening responses in adolescence. Psychoneuroendocrinology 35(6):921-31. Andersson JL, Hutton C, Ashburner J, Turner R, Friston K. 2001. Modeling geometric deformations in EPI time series. Neuroimage. 13(5):903-19.

16

Ashburner J. 2007. A fast diffeomorphic image registration algorithm. Neuroimage 38(1):95-113. Ashburner J, Friston KJ. 2005. Unified segmentation. Neuroimage 26(3):839-51. Basser PJ, Mattiello J, LeBihan D. 1994. MR diffusion tensor spectroscopy and imaging. Biophys.J. 66(1):259-67. Beaulieu C. 2009. Gaussian modeling of the diffusion signal. In: H Johansen-Berg, TEJ Behrens, editors. Diffusion MRI: from quantitative measurement to in-vivo neuroanatomy. London, UK: Elsevier pp 105-26. Buchanan TW, Kern S, Allen JS, Tranel D, Kirschbaum C. 2004. Circadian regulation of cortisol after hippocampal damage in humans. Biol Psychiatry 56(9):651-6. Carlesimo GA, Cherubini A, Caltagirone C, Spalletta G. 2010. Hippocampal mean diffusivity and memory in healthy elderly individuals: a cross-sectional study. Neurology 74(3):194-200. Cerqueira JJ, Catania C, Sotiropoulos I, Schubert M, Kalisch R, Almeida OF et al. 2005. Corticosteroid status influences the volume of the rat cingulate cortex - a magnetic resonance imaging study. J Psychiatr Res 39(5):451-60. Chang LC, Jones DK, Pierpaoli C. 2005. RESTORE: robust estimation of tensors by outlier rejection. Magn Reson.Med. 53(5):1088-95. Chenevert TL, Sundgren PC, Ross BD. 2006. Diffusion imaging: insight to cell status and cytoarchitecture. Neuroimaging Clin N Am 16(4):619-32, viii-ix. Chida Y, Steptoe A. 2009. Cortisol awakening response and psychosocial factors: a systematic review and meta-analysis. Biol Psychol 80(3):265-78. Clow A, Hucklebridge F, Stalder T, Evans P, Thorn L. 2009. The cortisol awakening response: More than a measure of HPA axis function. Neurosci Biobehav Rev Cuadra MB, Cammoun L, Butz T, Cuisenaire O, Thiran JP. 2005. Comparison and validation of tissue modelization and statistical classification methods in T1-weighted MR brain images. IEEE Trans Med Imaging 24(12):1548-65. De Kloet ER, Vreugdenhil E, Oitzl MS, Joels M. 1998. Brain corticosteroid receptor balance in health and disease. Endocr Rev 19(3):269-301. Dedovic K, Duchesne A, Andrews J, Engert V, Pruessner JC. 2009. The brain and the stress axis: the neural correlates of cortisol regulation in response to stress. Neuroimage 47(3):864-71. Fries E, Dettenborn L, Kirschbaum C. 2009. The cortisol awakening response (CAR): facts and future directions. Int J Psychophysiol 72(1):67-73. Gibbs P, Liney GP, Pickles MD, Zelhof B, Rodrigues G, Turnbull LW. 2009. Correlation of ADC and T2 measurements with cell density in prostate cancer at 3.0 Tesla. Invest Radiol 44(9):572-6.

17

Handwerger K. 2009. Differential patterns of HPA activity and reactivity in adult posttraumatic stress disorder and major depressive disorder. Harv Rev Psychiatry 17(3):184-205. Herman JP, Ostrander MM, Mueller NK, Figueiredo H. 2005. Limbic system mechanisms of stress regulation: hypothalamo-pituitary-adrenocortical axis. Prog Neuropsychopharmacol Biol Psychiatry 29(8):1201-13. Huang L, Smith A, Badaut J, Obenaus A. 2010. Dynamic characteristics of 56Fe-particle radiationinduced alterations in the rat brain: magnetic resonance imaging and histological assessments. Radiat Res 173(6):729-37. Jankord R, Herman JP. 2008. Limbic regulation of hypothalamo-pituitary-adrenocortical function during acute and chronic stress. Ann N Y Acad Sci 114864-73. Jauregui-Huerta F, Ruvalcaba-Delgadillo Y, Gonzalez-Castaneda R, Garcia-Estrada J, Gonzalez-Perez O, Luquin S. 2010. Responses of glial cells to stress and glucocorticoids. Curr Immunol Rev 6(3):195-204. Jovicich J, Czanner S, Greve D, Haley E, van der Kouwe A, Gollub R et al. 2006. Reliability in multi-site structural MRI studies: effects of gradient non-linearity correction on phantom and human data. Neuroimage 30(2):436-43. Knoops AJ, Gerritsen L, van der Graaf Y, Mali WP, Geerlings MI. 2010. Basal hypothalamic pituitary adrenal axis activity and hippocampal volumes: the SMART-Medea study. Biol Psychiatry 67(12):1191-8. Kremen WS, O'Brien RC, Panizzon MS, Prom-Wormley E, Eaves LJ, Eisen SA et al. 2010. Salivary cortisol and prefrontal cortical thickness in middle-aged men: A twin study. Neuroimage Lueken U, Leisse M, Mattes K, Naumann D, Wittling W, Schweiger E. 2009. Altered tonic and phasic cortisol secretion following unilateral stroke. Psychoneuroendocrinology 34(3):402-12. Lupien SJ, de Leon M, de Santi S, Convit A, Tarshish C, Nair NP et al. 1998. Cortisol levels during human aging predict hippocampal atrophy and memory deficits. Nat Neurosci 1(1):69-73. Maller JJ, Reglade-Meslin C, Anstey KJ, Sachdev P. 2006. Sex and symmetry differences in hippocampal volumetrics: before and beyond the opening of the crus of the fornix. Hippocampus 16(1):80-90. Muller MJ, Greverus D, Weibrich C, Dellani PR, Scheurich A, Stoeter P et al. 2007. Diagnostic utility of hippocampal size and mean diffusivity in amnestic MCI. Neurobiol Aging 28(3):398-403. Neveu PJ, Liege S, Sarrieau A. 1998. Asymmetrical distribution of hippocampal mineralocorticoid receptors depends on lateralization in mice. Neuroimmunomodulation 5(1-2):16-21. Piras F, Cherubini A, Caltagirone C, Spalletta G. 2010. Education mediates microstructural changes in bilateral hippocampus. Hum Brain Mapp

18

Pruessner JC, Wolf OT, Hellhammer DH, Buske-Kirschbaum A, von Auer K, Jobst S et al. 1997. Free cortisol levels after awakening: a reliable biological marker for the assessment of adrenocortical activity. Life Sci 61(26):2539-49. Pruessner M, Pruessner JC, Hellhammer DH, Bruce Pike G, Lupien SJ. 2007. The associations among hippocampal volume, cortisol reactivity, and memory performance in healthy young men. Psychiatry Res 155(1):1-10. Rajkowska G, Miguel-Hidalgo JJ. 2007. Gliogenesis and glial pathology in depression. CNS Neurol Disord Drug Targets 6(3):219-33. Reese TG, Heid O, Weisskoff RM, Wedeen VJ. 2003. Reduction of eddy-current-induced distortion in diffusion MRI using a twice-refocused spin echo. Magn Reson.Med. 49(1):177-82. Reul JM, Gesing A, Droste S, Stec IS, Weber A, Bachmann C et al. 2000. The brain mineralocorticoid receptor: greedy for ligand, mysterious in function. Eur J Pharmacol 405(1-3):235-49. Sanchez MM, Young LJ, Plotsky PM, Insel TR. 2000. Distribution of corticosteroid receptors in the rhesus brain: relative absence of glucocorticoid receptors in the hippocampal formation. J Neurosci 20(12):4657-68. Sapolsky RM. 2000. Glucocorticoids and hippocampal atrophy in neuropsychiatric disorders. Arch Gen Psychiatry 57(10):925-35. Seckl JR, Dickson KL, Yates C, Fink G. 1991. Distribution of glucocorticoid and mineralocorticoid receptor messenger RNA expression in human postmortem hippocampus. Brain Res 561(2):3327. Spoletini I, Cherubini A, Banfi G, Rubino IA, Peran P, Caltagirone C et al. 2011. Hippocampi, thalami, and accumbens microstructural damage in schizophrenia: a volumetry, diffusivity, and neuropsychological study. Schizophr Bull 37(1):118-30. Starkman MN, Gebarski SS, Berent S, Schteingart DE. 1992. Hippocampal formation volume, memory dysfunction, and cortisol levels in patients with Cushing's syndrome. Biol Psychiatry 32(9):75665. Starkman MN, Giordani B, Gebarski SS, Berent S, Schork MA, Schteingart DE. 1999. Decrease in cortisol reverses human hippocampal atrophy following treatment of Cushing's disease. Biol Psychiatry 46(12):1595-602. van der Beek EM, Wiegant VM, Schouten WG, van Eerdenburg FJ, Loijens LW, van der Plas C et al. 2004. Neuronal number, volume, and apoptosis of the left dentate gyrus of chronically stressed pigs correlate negatively with basal saliva cortisol levels. Hippocampus 14(6):688-700.

19

Vreeburg SA, Hoogendijk WJ, van Pelt J, Derijk RH, Verhagen JC, van Dyck R et al. 2009. Major depressive disorder and hypothalamic-pituitary-adrenal axis activity: results from a large cohort study. Arch Gen Psychiatry 66(6):617-26. Vreeburg SA, Zitman FG, van Pelt J, Derijk RH, Verhagen JC, van Dyck R et al. 2010. Salivary cortisol levels in persons with and without different anxiety disorders. Psychosom Med 72(4):340-7. Wolf OT, Convit A, de Leon MJ, Caraos C, Qadri SF. 2002. Basal hypothalamo-pituitary-adrenal axis activity and corticotropin feedback in young and older men: relationships to magnetic resonance imaging-derived hippocampus and cingulate gyrus volumes. Neuroendocrinology 75(4):241-9. Zach P, Mrzilkova J, Rezacova L, Stuchlik A, Vales K. 2010. Delayed effects of elevated corticosterone level on volume of hippocampal formation in laboratory rat. Physiol Res 59(6):985-96.

20