Blockade of TNF- rapidly inhibits pain responses in the central nervous system

Andreas Hessa,1, Roland Axmannb,1, Juergen Rechb,1, Stefanie Finzelb, Cornelia Heindla, Silke Kreitza, Marina Sergeevaa, Marc Saakec, Meritxell Garciac, George Kolliasd, Rainer H. Straube, Olaf Spornsf, Arnd Doererc, Kay Brunea, and Georg Schettb,2
a Institute of Experimental and Clinical Pharmacology and Toxicology, bDepartment of Internal Medicine 3, and cDivision of Neuroradiology, University of Erlangen-Nuremberg, 91054 Erlangen, Germany; dInstitute of Immunology, Alexander Fleming Biomedical Sciences Research Center, 16672 Vari, Greece; e Department of Internal Medicine I, University of Regensburg, 93053 Regensburg, Germany; and fDepartment of Psychological and Brain Sciences, Programs in Neuroscience and Cognitive Science, Indiana University, Bloomington, IN 47405

Edited* by Charles A. Dinarello, University of Colorado Denver, Aurora, CO, and approved December 29, 2010 (received for review August 19, 2010)

There has been a consistent gap in understanding how TNF- neutralization affects the disease state of arthritis patients so rapidly, considering that joint inammation in rheumatoid arthritis is a chronic condition with structural changes. We thus hypothesized that neutralization of TNF- acts through the CNS before directly affecting joint inammation. Through use of functional MRI (fMRI), we demonstrate that within 24 h after neutralization of TNF-, nociceptive CNS activity in the thalamus and somatosensoric cortex, but also the activation of the limbic system, is blocked. Brain areas showing blood-oxygen level-dependent signals, a validated method to assess neuronal activity elicited by pain, were signicantly reduced as early as 24 h after an infusion of a monoclonal antibody to TNF-. In contrast, clinical and laboratory markers of inammation, such as joint swelling and acute phase reactants, were not affected by anti-TNF- at these early time points. Moreover, arthritic mice overexpressing human TNF- showed an altered pain behavior and a more intensive, widespread, and prolonged brain activity upon nociceptive stimuli compared with wild-type mice. Similar to humans, these changes, as well as the rewiring of CNS activity resulting in tight clustering in the thalamus, were rapidly reversed after neutralization of TNF-. These results suggest that neutralization of TNF- affects nociceptive brain activity in the context of arthritis, long before it achieves antiinammatory effects in the joints.
cytokines

| antiinammatory therapy

rthritis is one of the most disabling chronic human diseases. Rheumatoid arthritis (RA) affects up to 1% of the population and is characterized by pain, swelling, and stiffness of joints, leading to a serious decay of life quality. Pain is the initial and prevailing symptom of the disease, leading to immobility, which in turn causes complications such as osteoporosis and cardiovascular disease. During the last 10 y the pharmacologic treatment of diseases such as RA has substantially improved because of the development of cytokine blocking agents (1). Although inamed joints express a multitude of mediators, including cytokines, chemokines, and growth factors, which contribute to the pathogenesis of arthritis, inhibition of TNF- has emerged as a particularly successful therapeutic strategy (2, 3). The therapeutic success of TNF- blockade in RA is unique and has been largely considered to result from rapid and efcient neutralization of joint inammation based on breakdown of the inammatory cytokine network in the affected joint, which results in an improvement of the signs and symptoms of the disease (4). It has, however, always been stunning, how fast the blockade of TNF- improves the patient condition, in particular because diseases like RA are highly chronic, building up a vast amount of inammatory tissue, and leading to irreversible damage of the cartilage and the bone. Thus, rapid resolution of this highly organized inammatory tissue or tissue damage is very unlikely to explain the fast effect of TNF- blockade. Even more importantly, neutralization of other inammatory mediators, in particular IL-1,
www.pnas.org/cgi/doi/10.1073/pnas.1011774108

which is a central inducer of inammation and structural damage in arthritis (5), appears to have less-pronounced and less-rapid effects on the symptoms of RA, although its role in protecting structural changes in the joints is substantial (6, 7). These observations, and the fact that traditional antirheumatic drugs such as methotrexate achieve far slower clinical responses than TNF- blockers, have led us to suspect that blockade of TNF- could has additional effects that go beyond the sole inhibition of joint inammation. We hypothesized that TNF- blockade could inuence processes in the CNS, which control the patients perception of the disease state. Indeed, earlier observations in mice have suggested that TNF- can induce a depressive-like behavior in mice (8). Pain processing and sensation are key mechanisms in the CNS elicited by arthritis, and pain is the dominant symptom of disease, which by far has the strongest impact on the disease burden of arthritis. Importantly, effects on pain also drive the standard read-out parameters for measuring therapeutic response in arthritis. Thus, if TNF- blockade has unique pain-reducing properties, its efcacy in reducing clinical disease activity of RA will be high, as such disease-activity measures are strongly inuenced by pain. In consequence, other treatment strategies, lacking such CNS effects, will less strongly affect the patients disease state, even if showing similar anti-inammatory or structure-sparing effects. Interestingly, inammation is considered to lower the threshold for pain perception in the CNS, leading to hyperalgesia (9). In contrast to the well-documented role of TNF- as a proinammatory cytokine, its role as a mediator of pain is incompletely characterized. It is known that both receptors for TNF- are expressed in dorsal root ganglion neurons and that intra-articular injection of TNF- blocking agents reduces nociception elicited by joint inammation (10). TNF- is also involved in development of mechanical hyperalgesia following injury (11, 12). Hence, we hypothesized that TNF- leads to a major change in pain perception in the CNS during arthritis. To search for the rapid effects of TNF- blockade, we visualized cerebral nociception elicited by arthritis and its reversibility upon TNF- blockade using blood-oxygen level-dependent (BOLD) functional MRI (fMRI) as early as 24 h after initiation of the treatment. The BOLD signal reects changes in hemodynamics linked to increased or decreased neuronal metabolic ac-

Author contributions: K.B. and G.S. designed research; A.H., R.A., J.R., S.F., C.H., S.K., M. Sergeeva, M.G., and G.S. performed research; A.H., G.K., and G.S. contributed new reagents/analytic tools; A.H., R.A., J.R., M. Saake, R.H.S., O.S., A.D., K.B., and G.S. analyzed data; and A.H. and G.S. wrote the paper. The authors declare no conict of interest. *This Direct Submission article had a prearranged editor. See Commentary on page 3461.
1 2

A.H., R.A., and J.R. contributed equally to this work. To whom correspondence should be addressed. E-mail: georg.schett@uk-erlangen.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1011774108/-/DCSupplemental.

PNAS | March 1, 2011 | vol. 108 | no. 9 | 37313736

MEDICAL SCIENCES

SEE COMMENTARY

tivity in response to outer sensory stimulation, thus allowing localization of brain areas activated by stimuli (13, 14). As a proofof-concept, we rst evaluated whether TNF- blockade rapidly reverses the hypernociception in patients with RA. We then performed an in-depth analysis of this rapid effect of TNF- blockade in a mouse model of arthritis based on the overexpression of human TNF-. Results whether TNF- blockade indeed affects the CNS, we rst performed a proof-of-concept study in humans. We measured CNS activity in patients with RA by assessing the BOLD signal in the fMRI before and after exposure to an infusion of 3 mg/kg of the TNF- blocker iniximab (IFX) (15, 16). The BOLD signal was induced by compression of the metacarpophalangeal joints of the dominantly affected hand, which is a standard assessment procedure for RA. Interestingly, the size of the brain area with enhanced BOLD activity was signicantly reduced as early as 24 h after administration of the TNF- blocker IFX. Moreover, it remained low up to 6 wk after the rst infusion (Fig. 1A). In contrast, BOLD signals elicited by a standardized control procedure, such as nger tapping, were not affected by TNF- blockade (Fig. 1B). Importantly, clinical measures of disease activity, such as joint swelling and joint tenderness, composite disease activity scores (DAS), such as DAS28, and acute phase reactants, such as blood sedimentation rate, serum C-reactive protein levels, and serum IL-6 levels, did not change measurably within 24 h after TNF- blockade but improved later during the treatment (Fig. 1 CE and Table S1). In contrast, subjective rating of pain intensity using the visual analog scale (VAS) was decreased within 24 h after TNF- blockade (Fig. 1F). Thus, these data indeed indicate a fast effect of TNF- blockade on the pain responses in the CNS even before a measurable anti-inammatory effect is achieved.
Reduced Activity in CNS Regions Involved in Pain Perception and in the Limbic System After TNF- Blockade. A more detailed analysis Reduction of Nociceptive Responses in the CNS of Arthritis Patients by TNF- Blockade Preceding Its Anti-Inammatory Effects. To test

activated brain area size after TNF- blockade could be attributed to the CNS regions contralateral to the affected joint (Fig. 2 AC). No signicant side difference was found for nger tapping, which was used as a control procedure (Fig. 2 AC). TNF- blockade did particularly change the activity of CNS structures, which are involved in pain perception, such as the thalamus and the secondary somatosensoric cortex (Fig. 2 D and E). Interestingly, centers of the limbic system were also affected, such as the cingulate and insular cortex. The cingulate cortex is particularly involved in forming and processing of emotions and memory, whereas the insular region is important for the subjective sense of the inner body, including the experience for pain and emotions (Fig. 2D). These ndings suggest that not only pain responses are affected by TNF- blockade but also CNS structures relevant for creating pain perception, emotions, and body experience.
Nociceptive Responses in Arthritic Mice Overexpressing Human TNF- and Their Rapid Reversal by TNF- Blockade. To study the effect of

of the BOLD signal in arthritis patients showed that reduction of

TNF- blockade on CNS pain responses elicited by arthritis in more detail, we used an animal model, which is characterized by an overexpression of human TNF- in the peripheral joints because of knocking the human TNF- gene into the mouse genome (17, 18). These TNF- transgenic (TNFtg) mice spontaneously develop inammatory arthritis, which leads to progressive functional impairment. The disease itself manifests as progressive swelling of the peripheral joints associated with decreased grip strength, mirroring progressively impaired joint function (Fig. S1). We rst assessed the effect of TNF- overexpression on behavioral pain responses and their reversibility upon TNF- blockade by IFX. Spinal reex arches, as measured by the tailick test, were not affected in early arthritis of TNFtg mice but were increased in latency at later disease stages (Fig. S2A). When testing mechanical hyperalgesia using von Frey laments, the force (threshold) needed to elicit responses was signicantly lower in early (6 wk) and later stages of the disease (10 wk) (Fig. 2B). The Hargraves test, assessing thermal hyperalgesia, showed a similar pattern, as observed with mechanical hyperalgesia (i.e., increased sensitivity to nociception in early and late stages of the

Fig. 1. Rapid reversal of pain induced BOLD signals after TNF- blockade in patients with RA. (A and B) Area of BOLD signal based on fMRI scans elicited by joint compression (A) and nger tapping (B) in patients (n = 5) with RA before, 1, 14, and 42 d after administration of TNF- blocking agent IFX. (C ) VAS (reaching from 0 to 10) for arthritis-related pain before, 1, 14, and 42 d after administration of IFX; (D) DAS28 before, as well as 1, 14, and 42 d after administration of IFX; (E ) swollen and (F ) tender joint count based on the assessment of 28 joints. (G) Maps of BOLD activity in a single patient before (Top) as well as 1 (Middle) and 42 d (Bottom) after administration of IFX. Asterisks indicate signicant difference to baseline (P < 0.05).

3732 | www.pnas.org/cgi/doi/10.1073/pnas.1011774108

Hess et al.

Fig. 2. Mapping of BOLD changes after TNF- blocking therapy. (AC ) Changes in the area size of the BOLD signal elicited by joint compression (blue bars) or nger tapping (red bars) comparing day 0 with day 1 (W2/W1) and day 0 with day 42 (W3/W1) after IFX treatment: overall (A) and specic for the contralateral (B) and ipsilateral hemisphere (C ). (D and E ) Changes in the area size of the BOLD signal elicited by joint compression (D) or nger tapping (E ) comparing day 0 with day 1 (W2/W1) at the following brain regions: thalamus (Th), secondary (S2) and primary (S1), somatosensoric cortex, parietal cortex (Par), posterior cingulate cortex (PCC), anterior cingulate cortex (ACC), lateral (LPFC) and medial (MPFC) prefrontal cortex, anterior (AIns) and posterior (PIns) insular cortex, cerebellum (Cb), motor cortex (M1), and periaqueductal gray (PAG).

disease of TNFtg mice) (Fig. S2C). In contrast, visceral nociception, measured by intraperitoneal injection of MgSO4, was signicantly reduced rather than enhanced in TNFtg mice, both in early and late disease stages, suggesting a desensitization to visceral nociception in TNFtg mice (Fig. S2D). The Rotarod test measuring motor activity was not impaired at an early disease stage but signicantly decreased at later stages (Fig. S3A). Similar data were obtained when the suspended-tail test was applied, which showed a signicant decrease in overall activity in TNFtg mice but only in later stages of disease (Fig. S3B). To test whether this nociceptive sensitization is reversible, we treated TNFtg mice with the TNF- blocker IFX. Whereas the latency of spinal pain reex arches, as measured by the tail-ick test, were normalized 72 h after TNF- blockade (Fig. S2E, increased sensitivity to mechanical nociceptive stimuli in TNFtg mice was completely reversed as early as 24 h after injection (Fig. S2F). This profound and rapid antinociceptive effect of TNF- blockade was further substantiated by the Hargreaves test, which normalized within 24 h after TNF- blockade and remained at the level of WT mice over 72 h (Fig. S2G). Within this short time period after TNF- blockade, no apparent change of clinic-analog parameters (such as paw swelling or grip strength) or histopathologic signs of arthritis (such as synovitis) could be observed. Moderate effects on visceral nociception were also observed, as the reduced sensitivity of TNFtg mice to visceral nociception was reversed to WT levels within 72 h after TNF- blockade (Fig. S2H). TNF- blockade also completely restored motor activity within the rst 24 h (Rotarod test) (Fig. S3C). Similarly, but slightly less pronounced, decreased overall mobility and activity of TNFtg mice as measured by the suspended-tail test also improved upon TNF- blockade(Fig. S3D).
Mapping of Nociceptive Brain Responses Elicited by Arthritis and Their Modulation by TNF- Blockade. Comparative fMRI analysis of

TNFtg mice to the level of WT mice within 24 h. When quantifying the BOLD activity by measuring the area size and peak amplitude of the BOLD signal, we observed a consistent and signicant increase in the somatosensoric cortex of TNFtg mice, which was completely abolished after TNF- blockade (Fig. 3C). Smaller, but still signicant, increases (Fig. 3C) were also observed in the thalamus, as well as in the association cortex in TNFtg compared with WT mice, which were again completely reversed upon blockade of TNF- (Fig. 3D). On the basis of these observations, we aimed at precisely mapping the interaction of the temporal structure of the BOLD signals in arthritis (Fig. 4A). Again, amplitudes of BOLD signals were generally higher in TNFtg mice than in WT mice. Moreover, even low-impact subthreshold stimulation led to increased brain activity in TNFtg mice, which was not found in WT mice (red color in Fig. 4A at S1/S2 and peaks in Fig. 4B). This pattern was completely reversed upon TNF- blockade as early as 24 h after injection. Aside from higher amplitudes, brain activation was more widespread in TNFtg mice, affecting many more brain regions than in WT mice. In the latter, the nociceptive response was localized to several well-dened brain areas (vertical pattern in Fig. 4A, white circles). This spreading of brain activation in TNFtg mice was partially reversed upon blockade of TNF-. In addition, the increased BOLD activity was maintained even during the intervals between stimulations and did not completely regress to baseline levels (horizontal red lines in Fig. 4A; red line in Fig. 4B is permanently high above baseline). TNF- blockade rapidly reversed this prolonged activation, and the BOLD signal in the time periods between the stimulations dropped dramatically, at times below the baseline level. This effect was particularly pronounced in somatosensoric cortical and limbic regions of the brain (asterisks at dark blue intervals in Fig. 4A and green line Fig. 4B).
Rewiring of the Pain Matrix in the CNS by Chronic Arthritis and Its Reversibility by TNF- Blockade. Higher-order changes in brain

the brain-activity patterns elicited by heat stimulation of the hind paw revealed a signicant increase of the BOLD activity in TNFtg mice in the so-called pain matrix (19, 20) of the brain compared with the WT mice (Fig. 3 AC). TNF- blockaded to a fast and complete reversal of the increased brain activity in
Hess et al.

function may not only be reected in changes of response properties of single brain structures. Instead, dynamic changes of the complex network connectivity (i.e., within the pain matrix)

PNAS | March 1, 2011 | vol. 108 | no. 9 | 3733

MEDICAL SCIENCES

SEE COMMENTARY

Fig. 3. Reversible enhancement of central pain responses by TNF-mediated arthritis. Functional MRI of the brain of 10-wk-old WT and human TNFtg mice without and with human antiTNF- antibody IFX (each n = 10). (A and B) 2D axial scans with heatmap (A) or superimposed on the corresponding anatomical image (B) show enhanced neuronal activity in the primary (S1) and secondary (S2) somatosensoric cortex, as well as in association cortex (RS) and the thalamus (Th) of TNFtg mice, which is reversible 24 h after IFX administration. (C) 3D reconstruction of functional brain activity. (D ) Bar graphs showing area size (Left) and peak height (Right) of the BOLD signal in WT mice (blue) and TNFtg mice treated with either vehicle (red) or the human antiTNF- antibody IFX (green). Four key groups of brain regions (brainstem, thalamus, sensory cortex, and association cortex) involved in central nociception are shown. Asterisks indicate signicant differences of WT to TNFtg plus those of TNFtg/IFX to TNFtg, with * and + indicating a P value of < 0.05 and ** and ++ a P value of < 0.025) (n = 10).

might occur. Graph theory has proven to have a high impact on the quantitative analysis of complex networks (i.e., brain network organization) (19, 20). We therefore wondered if graph theoretical analysis explaining the functional connectivity of CNS regions provides insight into the dynamic of rewiring of the pain matrix in arthritis and after neutralization of TNF-. After removing the global (mainly task-driven) signals, cross-correlation of the fMRI time proles of all structures, which were signicantly activated, revealed distinct patterns of interactions within the pain matrix (Fig. S4A, lower triangular half). There was an increased degree of connectivity, clustering, and modularity in arthritic TNFtg mice compared with WT mice (Fig. S4A and Table S2). These changes largely resulted from the formation of a tight cluster of the thalamus, the periaqueductal gray, and the amygdala (Kamada-Kawai plots, Fig. S4B). TNF- blockade led to a rapid partial dissolution of this cluster, resulting in a more diffuse network, indexed by an overall decreased connectivity, clustering, and modularity similar to that observed in WT mice. Discussion In this study we show that neutralization of the proinammatory cytokine TNF- rapidly affects CNS pain responses elicited by arthritis. Administration of TNF- blockers rapidly abrogates CNS activity linked to nociceptive stimuli in patients with arthritis, as well as in mice with TNF-driven arthritis. This effect occurs immediately after antibody administration and well before anti-inammatory effects of TNF- blockade are observed. Our data suggest that TNF- appears to trigger hypernociception because of changed pain processing in the CNS and a strong linkage to activation of limbic areas involved in pain experience, emotions, and body sensation. Cytokine blockade has dramatically changed the treatment of inammatory diseases, particularly of inammatory arthritides, such as RA. Despite RA being the embodiment of a chronic disease, patients with RA experience a very rapid improvement of disease-related symptoms upon blockade of TNF-. Although, it takes several weeks to observe a consistent and objective reduction of joint inammation (16), many patients report on a
3734 | www.pnas.org/cgi/doi/10.1073/pnas.1011774108

rapid effect of TNF- blockade on their subjective disease state. However, the basis of this rapid effect has not been addressed convincingly so far, although CNS effects of TNF- blockade have always been suspected to play a role in this process. We were stunned by the fact that TNF- blockade triggered a profound and signicant change of CNS activity in RA patients as early as 24 h after its administration, compared with baseline CNS activity. In contrast, joint swelling, as well as composite DAS (DAS28) did not change signicantly within the rst 24 h, although they improved later. These observations suggest that functional CNS changes clearly precede the improvement of inammatory signs of arthritis in human patients. Nociceptive signals elicited by joint compression did induce the activity in CNS structures, which are typically involved in pain perception, such as the thalamus and the somatosensoric cortex, but also in parts of the limbic system, such as the cingulum and the insular region, which are relevant for the inner body sensation as well as negative emotional experience, including pain experience. Suppression of neuronal activity in these centers may likely explain the rapid improvement of the symptom state of patients with RA after TNF- blockade, because the perception of nociceptive stimuli in the CNS is impaired and activation of CNS centers involved in shaping the emotional state of the individualas well as body perceptionis blocked. In murine arthritis triggered by TNF- overexpression, we observed that behavioral measurements of nociceptive responses showed a sensitization to somatic nociception (hyperalgesia) during arthritis. Moreover, investigation of arthritic mice by fMRI unequivocally demonstrated an altered pattern of brain activation, as measured by an enhanced BOLD activity. Even low-impact subthreshold stimulation led to increased BOLD signals, which could not be depicted in nonarthritic controls. In support of this notion, TNF- blockade caused a signicant and rapid reduction of these nociceptive BOLD signals in the somatosensoric and the association cortex, as well as the thalamus, within 24 h. Upon neutralization of TNF-, both thermal and mechanical hyperalgesia induced by arthritis were completely reversed within 24 h, an observation that supports the
Hess et al.

Fig. 4. Temporal proles and reversibility of TNFinduced changes of central pain response. (A) Temporal proles showing the intensity of BOLD signals obtained by fMRI of 10-wk-old WT mice and human TNFtg mice treated either with vehicle or with human antiTNF- antibody IFX. Colors indicate peak height of the BOLD signal from very low (dark blue) to very high (red). The x axis indicates time with at total of three sequences, each of which comprises four pain stimuli (S1 to S4) with increasing intensity (40, 45, 50, 55 C). The y axis reects 32 different brain regions as follows: motor cortex (M1), cerebellum (Cb), ventral pallidum (VP), globus pallidus (GP), nucleus accumbens (Acb), striatum (CPu), periaqueductal gray (PAG), zona incerta (ZI), hypothalamus (HT), bed nucleus of stria terminalis (BST), amygdala (Amd), hippocampus (Hip), septal area (Sep), piriform cortex (Pir), perirhinal/ectorhinal cortex (Prh/Ect), entorhinal cortex (Ent), insular cortex (Ins), frontal association cortex (FrA), cingulate cortex (Cg), retrosplenial cortex (RS), secondary (S2) and primary somatosensory cortex (S1), ventral posterolateral/posteromedial thalamic nucleus (VPL/VPM), medial thalamus (MT), lateral posterior thalamic nucleus (LP), lateral (LG) and medial geniculate nucleus (MG), pretectal area (PTA), superior (SC) and inferior colliculus (IC), substantia nigra (SN), ventral tegmental area (VTA). (B) Curves showing average time-dependent changes of the amplitudes of BOLD signals in WT (blue), hTNFtg mice (red), and the latter treated with the human antiTNF- antibody IFX (green).

results obtained in RA patients. Within this short period no apparent effect on clinical analog parameters (such as paw swelling) or histopathologic signs of arthritis (such as synovitis) could be observed. This result indicates that the CNS effect of TNF- blockade precedes its anti-inammatory effect. Interestingly, we even observed a decrease in the BOLD signal below the normal levels observed in nonarthritic WT mice. This effect was again particularly pronounced in somatosensoric cortical and limbic regions of the brain. One explanation for this decrease of the BOLD signal below the baseline level could be an increased activity of the cerebral inhibitory network developed during this chronic pain state (21). In this context, it is worth mentioning that brain network analysis detected a tight cluster consisting of thalamic structures, the periaquaeductal gray, and limbic areas, such as the amygdala in TNFtg mice. This nding supports the concept of an increased efcacy of the cerebral inhibitory network in arthritis. This rewiring may have developed as an adaptation to chronic pain preventing the abnormally increased ow of nociceptive signals to higher CNS structures, especially to the cortex. Importantly, TNF- blockade rapidly induced partial dissolution of this cluster; this happened within 24 h after TNF- blockade, in parallel with the reduction of the nociceptive BOLD activity and the restoration of the normal nociceptive behavior. In summary, these data show that TNF- leads to more intensive, widespread, and prolonged brain activity upon nociceptive stimulation, which is reversible upon neutralization of TNF-. This reversal of pain sensitization occurs rapidly and is not primarily linked to the anti-inammatory effects of TNF- blockade. In patients with RA, a very similar rapid and profound downregulation of nociceptive brain activity can be observed, which did precede the alleviation of joint inammation. Although subclinical anti-inammatory effects of TNF- blockade within the rst 24 h after exposure to TNF- blockade cannot be ruled out completely, the lack of a drop in acute phase reactant and cytokine levels within this short time interval does at least not support such a concept. Our ndings emphasize that changes of nociceptive responses in the brain precede the anti-inammatory effect of TNF- blockers. Importantly, these ndings may be a good
Hess et al.

explanation for the rapid improvement of the symptom state induced by TNF- blocking therapy. These data also suggest that anticytokine therapy targeting TNF- may achieve its (fast or rapid) clinical benet remote from the inamed joint through inuencing CNS processes. Our ndings also extend previous observations on the effects of TNF- in the CNS (i.e., the regulation of expression of clock genes), which may explain the high prevalence of fatigue in inammatory diseases (22). Finally, our ndings also raise the possibility that the assessment of responsiveness of BOLD activity in the CNS of RA patients and patients with other forms of inammatory arthritis may be used as a surrogate marker for predicting clinical responses to therapeutic intervention in a much faster way than it is currently realized. Materials and Methods
Mice. The human TNF- and TNFtg mice (strain Tg197) were previously described (17). Treatment with IFX (Centocor), a neutralizing chimeric humanmurine monoclonal antibody directed against human TNF-, was done by a single intravenous injection of a dose of 10 mg/kg antibody 24 h before testing (15). Clinical evaluation was performed weekly, starting at 4 wk after birth. Arthritis was evaluated in a blinded manner, as described previously (18). All animal experiments were approved by the local ethics committee of the University of Erlangen-Nuremberg. Tests for Nociceptive Behavior. The tail-ick test was conducted using a tailick analgesia meter (Columbus Instruments), Hargreaves test by IITC Plantar Analgesia Meter, and measurement of paw-withdrawal latency upon heat stimulation. The mechanical nociception test was performed by applying an ascending series of von Frey hairs (23). In visceral nociception tests, mice were intraperitoneally injected with MgSO4 (120 mg/kg) (23). Depression was tested by the tail-suspension test and locomotor activitiy was tested by Ugo Basile 7750 accelerating Rotarod (Ugo Basile) (24). Functional MRI in Mice. WT and TNFtg mice (n = 10 per group, 10 wk) were anesthetized with isourane and placed on a cradle inside the magnetic resonance tomograph (Bruker BioSpec 47/40, quadrature head coil) (25). The contact heat stimuli sequences (40, 45, 50, and 55 C, plateau for 5 s, ramp 15 s) were presented at the right hind paw with 3-min and 25-s intervals, three times, using a custom-made computer-controlled Peltier heating device. A series of 750 sets of functional images (matrix 64 64, eld of view 15 15 mm, slice thickness 0.5 mm, axial, 22 slices) were sampled using gradient echo-based Echo Planar Imaging Technique (single shot: TR = 4,000 ms,

PNAS | March 1, 2011 | vol. 108 | no. 9 | 3735

MEDICAL SCIENCES

SEE COMMENTARY

TEef = 24.38 ms, NEX = 2) within 50 min. Finally, 22 corresponding anatomical T2 reference images (RARE, slice thickness 0.5 mm, eld of view 15 15 mm, matrix 256 128, TR = 2,000 ms, TEef =56 ms) were taken as previously described in detail (26). Patients. Five female patients with RA, failing on standard treatment with disease-modifying antirheumatic drugs and having active disease with joint tenderness and swelling, received the TNF- blocker IFX at a dose of 3 mg/kg as an intravenous infusion. Mean ( SD) age was 56.3 8.2 y and mean ( SD) disease duration was 8.5 3.3 y. The number of tender and swollen joints, joint pain (VAS ranging from 0 to 10), overall disease activity calculated by the DAS based on 28 joints (DAS28) (27), C-reactive protein, and IL-6 levels were assessed at baseline and 1, 14, and 42 d after the infusion. Functional MRI in Humans. All anatomical and fMRI data were acquired on a 3T scanner (Magnetom Trio; Siemens) using a standard eight-channel phasedarray head coil. The ethics committee of the University Clinic of Erlangen approved all procedures and written informed consent was obtained from all patients. For anatomic datasets, we used a T1-weighted MPRAGE-sequence (eld-of-view = 256 mm, matrix size = 256 256, voxel size = 1.0 1.0 1.0 mm3, slices = 176, slice thickness = 1 mm, TR = 1,900 ms, TE = 1.13 ms). For each subject, two experiments with different stimulation conditions (rst, nger-tapping and second, compression of the metacarpophalangeal joints) were performed. In each of them, 93 whole-brain images were obtained with a gradient-echo, echo-planar scanning sequence (TR = 3,000 ms, TE = 30 ms, ip angle 90; eld-of-view, 220 mm2, acquisition matrix 64 64, 36 axial slices, slice thickness 3 mm, gap 0.75 mm). Functional MRI Analysis. Functional analysis was performed for mice and humans using Brain Voyager QX (Version 10.3) and our own software MagnAn (25), as previously described (28). In summary, after preprocessing [motioncorrected using sinc interpolation Gaussian spatial (human: FWHM = 4 mm, mouse: 0.469 mm) and temporal (FWHM = 3 volumes) smoothing], general linear modeling analysis with separate predictors for each stimulus was performed. The statistical parametric mapping obtained were corrected for multiple comparisons, false-discovery rate threshold at z-score level of 3.3, and different groups of activated voxels were labeled as belonging to certain brain structures based on (i ) the mouse atlas from Paxinos (29) or (ii ) the Mai atlas of

the human brain (30). The voxels, which were signicantly activated by the above criterion, were counted as the activated volume per given brain region. The mean corresponding peak activity was determined for each stimulation temperature for mice and for tapping and compression in humans, averaged over all activated brain structures and nally over all subjects of one experimental group, respectively, to provide a global group comparison. Graph Theoretical Analysis. Functional connectivity patterns were computed as cross-correlations of the residuals after the global signal mean was removed by linear regression and represented as correlation matrices. Similarity across these correlation matrices was calculated as Spearman rank correlation. Modularity and clustering coefcients, as well as several centrality measures, were computed from the weighted pattern of positive correlations present between brain regions (31). The network modularity index was obtained by optimally subdividing the network into modules, such that most connections are made within modules and only few connections exist between modules (32, 33). Clustering was computed from the weighted functional-connectivity matrix (34) and scaled relative to a population of 100 random networks with identical degree distribution. Node-strength centrality was measured as the sum of each nodes positive cross-correlations. The networks are visualized using a force-based algorithm after KamadaKawai for achieving that all edges are of more or less equal length, and there exist as few crossing edges as possible (35). Statistical Analysis. Data are presented as mean SEM. Group mean values were compared by the Students t tests in a task-specic single-test framework to assess signicant differences between the different mouse strains. ACKNOWLEDGMENTS. This study was supported by the Deutsche Forschungsgemeinschaft FG 661/TP4 and SPP1468-Immunobone (to A.H. and G.S.); the Bundesministerium fr Bildung und Forschung projects 01EM0514, 01GQ0731, 0314102, Ankyloss and Immunopain (to G.S. and A.H.); the Masterswitch, Kinacept, and Adipoa projects of the European Union (to G.S.); the Interdisciplinary Centre for Clinical Research and the Erlanger Leistungsbezogene Anschubnanzierung und Nachwuchsfrderung (ELAN) fund of the University of Erlangen-Nuremberg (to G.S. and A.H.); K.B. is Doerenkamp Professor for Innovations in Animal and Consumer Protection.

1. Smolen JS, Steiner G (2003) Therapeutic strategies for rheumatoid arthritis. Nat Rev Drug Discov 2:473488. 2. Firestein GS (2003) Evolving concepts of rheumatoid arthritis. Nature 423:356361. 3. McInnes IB, Schett G (2007) Cytokines in the pathogenesis of rheumatoid arthritis. Nat Rev Immunol 7:429442. 4. Brennan FM, Jackson A, Chantry D, Maini R, Feldmann M (1989) Inhibitory effect of TNFa antibodies on synovial cell interleukin-1 production in rheumatoid arthritis. Lancet 334:244247. 5. Dinarello CA (2010) IL-1: Discoveries, controversies and future directions. Eur J Immunol 40:599606. 6. Singh JA, et al. (2009) Biologics for rheumatoid arthritis: An overview of Cochrane reviews. Cochrane Database Syst Rev 4:CD007848. 7. van den Berg WB (2001) Uncoupling of inammatory and destructive mechanisms in arthritis. Semin Arthritis Rheum 30(5, Suppl 2):716. 8. OConnor JC, et al. (2009) Interferon-gamma and tumor necrosis factor-alpha mediate the upregulation of indoleamine 2,3-dioxygenase and the induction of depressive-like behavior in mice in response to bacillus Calmette-Guerin. J Neurosci 29:42004209. 9. Reinold H, et al. (2005) Spinal inammatory hyperalgesia is mediated by prostaglandin E receptors of the EP2 subtype. J Clin Invest 115:673679. 10. Boettger MK, et al. (2008) Antinociceptive effects of tumor necrosis factor alpha neutralization in a rat model of antigen-induced arthritis: evidence of a neuronal target. Arthritis Rheum 58:23682378. 11. Marchand F, et al. (2009) Effects of Etanercept and Minocycline in a rat model of spinal cord injury. Eur J Pain 13:673681. 12. Schfers M, et al. (2008) Selective stimulation of either tumor necrosis factor receptor differentially induces pain behavior in vivo and ectopic activity in sensory neurons in vitro. Neuroscience 157:414423. 13. Ogawa S, Lee TM, Kay AR, Tank DW (1990) Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad USA 87:98689872. 14. Thulborn KR, Waterton JC, Matthews PM, Radda GK (1982) Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high eld. Biochim Biophys Acta 714:265270. 15. Knight DM, et al. (1993) Construction and initial characterization of a mouse-human chimeric anti-TNF antibody. Mol Immunol 30:14431453. 16. Maini R, et al.; ATTRACT Study Group (1999) Iniximab (chimeric anti-tumour necrosis factor alpha monoclonal antibody) versus placebo in rheumatoid arthritis patients receiving concomitant methotrexate: A randomised phase III trial. Lancet 354:19321939. 17. Keffer J, et al. (1991) Transgenic mice expressing human tumour necrosis factor: A predictive genetic model of arthritis. EMBO J 10:40254031.

18. Diarra D, et al. (2007) Dickkopf-1 is a master regulator of joint remodeling. Nat Med 13:156163. 19. Apkarian AV, Bushnell MC, Treede RD, Zubieta JK (2005) Human brain mechanisms of pain perception and regulation in health and disease. Eur J Pain 9:463484. 20. Borsook D, Becerra L (2007) Phenotyping central nervous system circuitry in chronic pain using functional MRI: considerations and potential implications in the clinic. Curr Pain Headache Rep 11:201207. 21. Shmuel A, Augath M, Oeltermann A, Logothetis NK (2006) Negative functional MRI response correlates with decreases in neuronal activity in monkey visual area V1. Nat Neurosci 9:569577. 22. Cavadini G, et al. (2007) TNF-alpha suppresses the expression of clock genes by interfering with E-box-mediated transcription. Proc Natl Acad USA 104:1284312848. 23. Whishaw IQ (1999) The Behavior of the Laboratory Rat: A Handbook with Tests, eds Whishaw IQ, Kolb B (Oxford University Press, Oxford), pp 478486. 24. Jones BJ, Roberts DJ (1968) The quantiative measurement of motor inco-ordination in naive mice using an acelerating rotarod. J Pharm Pharmacol 20:302304. 25. Hess A, Sergejeva M, Budinsky L, Zeilhofer HU, Brune K (2007) Imaging of hyperalgesia in rats by functional MRI. Eur J Pain 11:109119. 26. Hennig J, Nauerth A, Friedburg H (1986) RARE imaging: A fast imaging method for clinical MR. Magn Reson Med 3:823833. 27. Prevoo ML, et al. (1995) Modied disease activity scores that include twenty-eightjoint counts. Development and validation in a prospective longitudinal study of patients with rheumatoid arthritis. Arthritis Rheum 38:4448. 28. Knabl J, et al. (2008) Reversal of pathological pain through specic spinal GABAA receptor subtypes. Nature 451:330334. 29. Paxinos G, Franklin KBJ (2001) The Mouse Brain in Stereotactic Coordinates (Elsevier Academic Press, Burlington, MA). 30. Mai JK, Paxinos G, Voss T (2008) Atlas of the Human Brain (Elsevier Academic Press, Burlington, MA). 31. Rubinov M, Sporns O (2010) Complex network measures of brain connectivity: Uses and interpretations. Neuroimage 52:10591069. 32. Girvan M, Newman ME (2002) Community structure in social and biological networks. Proc Natl Acad Sci USA 99:78217826. 33. Newman ME (2004) Analysis of weighted networks. Phys Rev E Stat Nonlin Soft Matter Phys 70:056131. 34. Onnela JP, Saramki J, Kertsz J, Kaski K (2005) Intensity and coherence of motifs in weighted complex networks. Phys Rev E Stat Nonlin Soft Matter Phys 71:065103. 35. Kamada T, Kawai S (1989) An algorithm for drawing general undirected graphs. Inf Process Lett 31:715.

3736 | www.pnas.org/cgi/doi/10.1073/pnas.1011774108

Hess et al.

COMMENTARY

Mapping the immunological homunculus

Betty Diamond and Kevin J. Tracey1
The Feinstein Institute for Medical Research, Manhasset, NY 11030

uring the rst century, the Roman physician Cornelius Celsus dened four cardinal signs of inammation: redness, swelling, heat, and pain. These signs and symptoms occur during infection by invasive pathogens or as a consequence of trauma. Today, we understand the molecular basis of these physiological responses as mediated by cytokines and other factors produced by cells of the innate immune system. Cytokines are both necessary and sufcient to cause pathophysiological alterations manifested as the four cardinal signs. Importantly, this knowledge has enabled the development of highly selective therapeutical agents that target individual cytokines to prevent or reverse inammation. For example, selective inhibitors of TNF, a major inammatory cytokine, have revolutionized the therapy of rheumatoid arthritis, inammatory bowel disease, and other autoimmune and autoinammatory diseases affecting millions worldwide. Now, in PNAS, Hess et al. (1) use functional MRI to monitor brain activity and report that patients with rheumatoid arthritis who receive anti-TNF develop signicant changes in brain activity before resolution of inammation in the affected joints. To accomplish this, the authors measured blood oxygen level-dependent (BOLD) signals in the brain after compressing the metacarpal phalangeal joints of the arthritic hand. They observe enhanced activity in the brain regions associated with pain perception, including the thalamus, somatosensory cortex, and limbic system, regions known to process body sensations and emotions associated with the pain experience (1). Brain activity was signicantly reduced within 24 h after treatment with TNF inhibitors, a time frame that preceded any observable evidence of reduced signs of inammation in affected joints. Clinical composite scores, comprising measurements of C-reactive protein, a circulating marker of inammation severity, were not improved until after 24 h. This suggests that selective inhibition of TNF has a primary early effect on the nervous system pain centers. The authors further explore this result in genetically engineered transgenic mice that overexpress TNF and develop spontaneous arthritis. Administration of anti-TNF signicantly improved the development of mechanical hyperalgesia in the standardized von Frey lament test system and also improved the response to thermal hyperalgesia in the Hargreaves test. This clinical improvement occurred within a short pewww.pnas.org/cgi/doi/10.1073/pnas.1100329108

riod after administration of anti-TNF, before any signicant improvement in joint swelling or grip strength, and before histopathological improvement of synovitis. As in the human subjects, there was signicant abolishment of BOLD signals in the somatosensory cortex. Brain activation was more extensive in the transgenic mice and affected more brain regions compared with WT mice, and this spreading of brain activation was down-modulated by administration of the TNF inhibitors. Administration of anti-TNF reversed the prolonged activation of BOLD activity that was

The nervous system is hardwired to monitor the presence of cytokines and molecular products of invaders.
observed during the intervals between stimulations, and was particularly pronounced in the somatosensory cortex and limbic system. These intriguing ndings provide a glimpse into the future, when it is likely that sensory and motor brain regions associated with immune responses will be mapped as an immunological homunculus (2). Of particular signicance here is that this approach bridges two major elds of medicine, neurology and immunology. Reexes Regulate Immune Responses Recent advances at the intersection of these elds have provided direct evidence that neural circuits reexively control the onset and resolution of inammation, and previously undescribed molecular mechanisms have been dened (3). The fundamental principle of neurological reexes, as originally established by Sir Charles Scott Sherrington in the early 20th century, is that neural circuits cooperate to maintain the organisms homeostasis. He elucidated the role of the nervous system in integrating reexes to provide stable physiological systems. He famously remarked that Existence of an excited state is not a prerequisite for the production of inhibition; inhibition can exist apart from excitation no less than, when called forth against an excitation already in progress, it can suppress or moderate it. These principles were rst dened in relatively accessible organ sys-

tems, including cardiovascular, gastrointestinal, and respiratory, which enabled the mapping of reex units. For example, by recording signals to the heart, it was possible to dene a reex that controls heart rate: increased heart rate activates afferent arcs that in turn elicit ring of efferent arcs in the vagus nerve, which slows heart rate. More recently, these principles have been applied to mapping the somewhat more elusive immune system. The combined use of neurophysiological stimulating techniques and current molecular strategies to measure immune response established the inammatory reex (3). Cytokines activate afferent signals in the vagus nerve, which culminate on an efferent arc that inhibits further cytokine production. The molecular basis of this cytokineinhibiting action is attributable to acetylcholine, the principal neurotransmitter of the vagus nerve, interacting with 7 nicotinic acetylcholine receptor subunits expressed by cytokine-producing cells. Signal transduction through this receptorligand interaction down-regulates the activity of NF-B, a primary pathway for regulating cytokine transcription and synthesis. Accordingly, 7 KO mice rendered decient in this pathway are exquisitely sensitive to inammation caused by pathogen-associated molecules and experimental arthritis. The efferent arm of the inammatory reex, termed the cholinergic antiinammatory pathway, has enabled the development of experimental nerve stimulators and highly selective pharmacological agents that suppress cytokine-mediated disease in standardized preclinical models. Sensing Inammation and Infection Less is known about the mapping of the afferent or sensory arc of the inammatory reex, and there is signicant interest in the question of how the nervous system monitors the state of inammation. The seminal work of Watkins et al. (4) demonstrated that the fever-causing activities of the pyrogenic cytokine IL-1 required an intact vagus nerve, because administration of IL-1 into the abdomen failed to produce fever if the vagus nerve had been cut. These researchers proposed a critical role of the glomus cell, specialized to detect
Author contributions: B.D. and K.J.T. wrote the paper. The authors declare no conict of interest. See companion article on page 3731.
1

To whom correspondence should be addressed. E-mail: kjtracey@nshs.edu.

PNAS | March 1, 2011 | vol. 108 | no. 9 | 34613462

pH and other chemical changes in the internal milieu, in sensing IL-1 (5). Glomus cells express IL-1 receptors, and it is a plausible hypothesis that IL-1 binding to glomus cells stimulates the release of dopamine, which, in turn, triggers afferent signaling in the vagus nerve. Ascending action potentials culminate on the nucleus tractus solitarius and are then relayed to higher brain regions. Sensory neural circuits can also be activated by LPS, and Toll-like receptor 4 (TLR4), the principal endotoxin receptor, is expressed by neurons in the nodose ganglia (6). These neurons have been implicated in transmitting afferent signals in the vagus nerve. Another mechanism by which the nervous system can sense the presence of injury is dependent on the expression of TLR4 in microglial cells, which is required for the development of neuropathic pain in a murine model of nerve transection (7). This TLR4 pathway, which may be activated even in the absence of infection by HMGB1 released from injured cells (8), stimulates TNF release in the development of chronic pain states. Together, these results indicate that the nervous system is hardwired to monitor the presence of cytokines and molecular products of invaders. The presence of TNF can also be sensed by sensory neurons, which express type I and type II TNF receptors (9). Local or regional administration of TNF elicits pain, partially by inciting tissue damage and activating release of other molecules capable of activating sensory neurons. TNF receptorligand interaction on sensory neurons may activate pain bers directly, however, serving as a pathway that may activate the brains pain centers as described here. Others have shown that administration of anti-TNF antibodies into the cerebrospinal uid of rats subjected to experimental arthritis signicantly attenuated pain-related behavior and the development of inammation in the joints
1. Hess A, et al. (2011) Blockade of TNF- rapidly inhibits pain responses in the central nervous system. Proc Natl Acad Sci USA 108:37313736. 2. Tracey KJ (2007) Physiology and immunology of the cholinergic antiinammatory pathway. J Clin Invest 117:289296. 3. Tracey KJ (2009) Reex control of immunity. Nat Rev Immunol 9:418428. 4. Watkins LR, et al. (1995) Blockade of interleukin-1 induced hyperthermia by subdiaphragmatic vagotomy: Evidence for vagal mediation of immune-brain communication. Neurosci Lett 183:2731. 5. Goehler LE, et al. (1997) Vagal paraganglia bind biotinylated interleukin-1 receptor antagonist: A possible mechanism for immune-to-brain communication. Brain Res Bull 43:357364.

(10). Moreover, intrathecal administration of antibody was signicantly more effective compared with systemic anti-TNF, suggesting that the principal effects of anti-TNF are mediated through direct interaction with neurons. Considered in light of the current paper, it is plausible that local accumulation of TNF in the region of sensory pain circuit neurons drives activation of the brains pain centers. Thus, it is possible that the rapid early improvement from anti-TNF results from acutely neutralizing TNF and blocking its ability to stimulate pain bers. Predictably, this circuit would not require any signicant improvement in the severity of inammation in the arthritic joints in order for benet to be perceived. The current study also raises another possibility, that anti-TNF antibodies traversed the bloodbrain barrier and directly altered the activity of brain neurons in the pain centers. Until quite recently, it has been dogmatic that antibodies do not enter the brain, but this has been overturned by studies in a murine model of the autoimmune disease lupus (11). This work identied antibodies that bind to dsDNA (a common antibody phenotype in this syndrome) and to the NMDA receptor expressed on brain neurons. These antibodies readily access the brain, because the bloodbrain barrier can be opened during relatively mild stress caused by administration of endotoxin, cytokines, or even epinephrine. On entering the brain, the interaction of these antibodies with neurons in the hippocampus and other regions mediates cell damage and cognitive or behavioral dysfunction (12). In light of the present study, it is interesting to consider whether inammatory mediators produced during arthritis might open the bloodbrain barrier to enable anti-TNF direct access to brain neurons. TNF expressed by glial cells in brain regulates synaptic scaling, a mechanism that adjusts the strength of neuronal synapses
6. Hosoi T, Okuma Y, Matsuda T, Nomura Y (2005) Novel pathway for LPS-induced afferent vagus nerve activation: Possible role of nodose ganglion. Auton Neurosci 120:104107. 7. Tanga FY, Nutile-McMenemy N, DeLeo JA (2005) The CNS role of Toll-like receptor 4 in innate neuroimmunity and painful neuropathy. Proc Natl Acad Sci USA 102:58565861. 8. Yang H, et al. (2010) A critical cysteine is required for HMGB1 binding to Toll-like receptor 4 and activation of macrophage cytokine release. Proc Natl Acad Sci USA 107:1194211947. 9. Boettger MK, et al. (2008) Antinociceptive effects of tumor necrosis factor alpha neutralization in a rat model of antigen-induced arthritis: Evidence of a neuronal target. Arthritis Rheum 58:23682378.

under conditions of prolonged alterations of electrical activity (13). A testable hypothesis emerges from this reasoning: The systemic inammation associated with rheumatoid arthritis opens the barrier to anti-TNF, which inhibits glial cell-derived TNF and alters the strength of synapses in the brains pain regions. This might explain the observed reversal of prolonged activation of BOLD activity during intervals between stimulations in the arthritic mice. Immunological Homunculus Advances in brain imaging modalities and molecular biology are mapping brain networks and circuits with high resolution and precision. These maps reveal sensory and motor circuits that are somatotopically and functionally organized, and form the basis for understanding the neural circuits that coordinate physiological responses to the external and internal environment. Early advances were achieved by mapping the relatively accessible domains of body sensations and skeletal muscle motor function. Now, knowledge of the inammatory reex as a prototypical circuit that regulates inammation, combined with the ability to visualize brain activity in patients and mice with inammation, makes it possible to make additional advances. It is interesting to consider that in addition to defending the host from infection, the immune system functions as a sensory organ that transmits information in real time to the central nervous system about the tissue response to injury and infection. This presents important possibilities for understanding fundamental mechanisms that maintain physiological homeostasis. Progress in this eld is all but certain to reveal the identity of other circuits that reexively regulate the immune system and to produce maps that will guide understanding of the neurological basis of immunity and physiology.
10. Boettger MK, et al. (2010) Spinal tumor necrosis factor alpha neutralization reduces peripheral inammation and hyperalgesia and suppresses autonomic responses in experimental arthritis: A role for spinal tumor necrosis factor alpha during induction and maintenance of peripheral inammation. Arthritis Rheum 62:1308 1318. 11. Kowal C, et al. (2006) Human lupus autoantibodies against NMDA receptors mediate cognitive impairment. Proc Natl Acad Sci USA 103:1985419859. 12. Faust TW, et al. (2010) Neurotoxic lupus autoantibodies alter brain function through two distinct mechanisms. Proc Natl Acad Sci USA 107:18569 18574. 13. Stellwagen D, Malenka RC (2006) Synaptic scaling mediated by glial TNF-alpha. Nature 440:10541059.

3462 | www.pnas.org/cgi/doi/10.1073/pnas.1100329108

Diamond and Tracey