Trophic Structure and Mercury Biomagnification in Tropical Fish Assemblages, Itnez Ri er, Boli ia

Abstract
We examined mercury concentrations in three fish assemblages to estimate biomagnification rates in the Itnez main river, affected by anthropogenic activities, and two unperturbed rivers from the Itnez basin, Bolivian Amazon. ivers presented low to moderate water mercury concentrations !from ".#$ ng %&" to #.'( ng %&") and natural differences in terms of sediment load. *ercury biomagnification rates were confronted to trophic structure depicted by carbon and nitrogen stable isotopes composition !+"$,- +"./) of primary trophic sources, invertebrates and fishes. esults showed a slight fish contamination in the Itnez iver compared to the unperturbed rivers, with higher mercury concentrations in piscivore species !0."$ 1g g&" vs. 0."" 1g g&" in the unperturbed rivers) and a higher biomagnification rate. 2rophic structure analysis showed that the higher biomagnification rate in the Itnez iver could not be attributed to a longer food chain. ,evertheless, it revealed for the Itnez iver a higher contribution of periphyton to the diet of the primary consumers fish species- and more negative +"./ values for primary trophic sources, invertebrates and fishes that could indicate a higher contribution of methanotrophic bacteria. 2hese two factors may enhance methylation and methyl mercury transfer in the food web and thus, alternatively or complementarily to the impact of the anthropogenic activities, may explain mercury differences observed in fishes from the Itnez iver in comparison to the two other rivers. Citation: 3ouilly *, e4as 5, 3rez 2, 5uprey 67%, *olina /I, et al. !#0".) 2rophic 8tructure and *ercury Biomagnification in 2ropical 9ish Assemblages, Itnez iver, Bolivia. 3%o8 :,; <!$)= e($0$>. doi="0.".?"@4ournal.pone.00($0$> Editor: Aincent %audet, ;cole ,ormale 8uprieure de %yon, 9rance Received: 6anuary "?, #0".- Accepted: April ##, #0".- Published: *ay .", #0". Copyright: B #0". 3ouilly et al. 2his is an open7access article distributed under the terms of the /reative /ommons Attribution %icense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: 2his worC was part of the I 5 !http=@@www.ird.fr) funded pro4ect D2rophic ;cology of Amazonian AEuatic 8ystemsF !6;AI7;*AA). Its also received financial support from a WW9 Bolivia !http=@@bolivia.panda.org) funded pro4ect !G,"0 WW97I 5 agreement). 2he funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: 2he authors have declared that no competing interests exist.

Introduction

*ercury, and its organic form methyl mercury, that is easily assimilated and accumulated in aEuatic food chains, constitute a ma4or environmental and public health issue in the Amazonian context. *ercury inputs may originate from exogenous sources related to gold mining or industrial uses, but also come from natural sources of mercury accumulated and trapped in the soils along the geological history of the basin H"I. 2his endogen mercury is liberated by natural or anthropogenic erosions and transported by lixiviation towards the aEuatic systems. /ontamination is thereafter controlled by a set of biotic and abiotic conditions among which methylation rates H#IJH>I and amplification processes along the food chain H$I, H(I are Cey factors. 9ood uptaCe represents more than <$K of the methylmercury total uptaCe, well above passive uptaCe from water H?I, and amplification processes along the food chain may increase the mercury concentration several thousand fold from water to fish top predators. 2wo ma4or amplification processes, bioaccumulation and biomagnification, are liCely to control mercury concentrations in organisms H$I. Bioaccumulation refers to the increase of mercury concentrations along the lifetime of an individual while biomagnification is defined as the increment of mercury concentration between the successive consumer levels of the food chain. Biomagnification is assumed to be positively linCed to food chain length, that may be derived from ,itrogen stable isotope analysis H<I, H'I. 9ood source origin and pathway could also be determinant= sediment biofilm, phytoplanCton and periphyton are potential food sources and also support mercury methylation H#I, H.I, H"0I in relation to the activity of sulfate7reducing H""Iand methanogen H"#I bacteria. A previous study concluded that mercury concentration in fishes from the Itnez could not be completely explained by bioaccumulation processes H".I. In this study, we examined mercury concentrations in a fish assemblage to compare biomagnification rates in three rivers from the Itnez basin with low to moderate water mercury concentrations !from ".#$ ng % &" to #.'( ng %&"). 2hey also differ in their natural sediment load !clear vs. white waters) and anthropogenic activities !deforestation and gold mining activity). We hypothesized that these differences are liCely to affect biological production, food web structure and conseEuently mercury biomagnification rates. Accordingly, stable carbon and nitrogen isotopic composition !+"./- +"$,) were measured in trophic sources, invertebrates and fish in order to evaluate the relationship between biomagnification rates, food web sources and trophic chain length.

Methods ;thic 8tatement

L% A@L*88 laboratory is an Authorized 8cientific Institution !I/A) accredited by the Bolivian 5irecciMn Neneral de la Biodiversidad y Oreas 3rotegidas !5NBA3, Aiceministerio de *edio Ambiente, *inisterio de *edio Ambiente y Agua) to conduct biological scientific research within the Bolivian territory, including protected areas ! esoluciMn administrativa B*AB// 0#(@0'). I 5 is linCed to L% A@L*88 through cooperation agreements. 2his particular pro4ect has been approved and permissions for biological collects have been issued by 5NBA3, departmental 3refecture of Beni, Itnez departmental parC !357A*,I Itnez) and local authorities ! emanso, *ategua, Aersalles and Bella Aista villages). %ocal fishermen captured and manipulated fish according to procedures permitted by the Aiceministerio de *edio Ambiente. apidly after the capture, living fishes were manually sacrificed !by percussive stunning) or left in high doses of anaesthetic !phenoxy7ethanol) to minimize suffering. %ocal fish assemblage did not involve endangered or protected species.

8tudy Area

2he study was carried out in three rivers of the Itnez basin= 8an *artPn iver, Blanco iver and the main Itnez iver !9igure ", see H".I for further details on the basin, rivers and studied sites). 2hey present differences in river water chemistry mainly related to their sediment load and mercury concentration in water. Itnez and 8an *artPn rivers present clear, yellow to green waters characteristic of low sediment load !mean suspended particulate matter concentration H83*I of ?.. and "".> mg %&", respectively H".I). :n the contrary, Blanco iver drains white waters with higher sediment load !H83*I of #(." mg %&"). Itnez iver is affected by deforestation in the Brazilian territory and by a gold mine !8erranPa 8an 8imMn, Bolivia). Blanco and 8an *artPn rivers belong to the same catchment, mainly covered by tropical forest. 2hey present low human population densities and globally low anthropogenic impact. 9looding area and duration are liCely to be higher in the Itnez main river. 8atellite mapping of flood and vegetation !based on 8A and 67; 8 images, see method in H">I) indicated flooding areas of "$J#0K for the 8an *artin basin, #0J#$K for Blanco basin and .0J.$K for the main Itnez river basin in its central part H"$I. *ercury shows high affinity with sediment particles and, for that reason, its total concentration in water !HQgI) increases with sediment load H"(I, H"?I. As a conseEuence, waters from the Blanco iver and its floodplain laCes naturally present higher HQgI !mean HQgI R #.'( ng %&" in river- ".$# ng %&" in laCes) than those from the Itnez !".$> ng %&"- ".#( ng %&") and 8an *artPn rivers !".#$ ng %&"- 0.(> ng %&"), !H".I, 9igure "). All these mercury concentrations are low compared with the regional Amazonian context !HQgI total from " to .$ ng %&" H"<I). Qowever, higher values observed in the Itnez iver, compared to the other clear water 8an *artPn iver, suggest that this system is slightly perturbed H".I.

Figure 1. ydrological map o! the "olivian part o! the It#ne$ basin. 8ampling locations !blacC point) and water mercury concentrations !star) are indicated. 2he principal sampled locations were= "7 8an *artPn iver- #7 Blanco iver- .7 Itnez iver. ! edrawn from H".I). doi="0.".?"@4ournal.pone.00($0$>.g00" 5ifferent floodplain laCes of each river were visited but most of the samples were collected in two laCes of each river !9igure ")= /uricha !"#S.(T><U 8J(.S#(T""U W) and ,egra !"#S.?T><U 8J(.S#>T>0U W) in the Itnez iver- /ambarazal !".S"?T$<U 8J(.S.(T.?U W) and edonda !".S"<T"(U8J(.S..T">U W) in the 8an *artPn iver- Vacarias !".S"$T>?U 8J(.S>#T.'U W) and %a Nran4a !".S"(T0>U 8J (.S>#T$(U W) in the Blanco iver. :ther samples !W"$K) were collected in secondary localities to

complete the data set !see positions on 9igure "). 8ome fishes !W"0K) were collected during two previous field trips !,ovember #00> and August #00$) but most of the fish and all the trophic source samples were collected during the dry season of #00? in three dates in the Itnez iver !6une, 8eptember and ,ovember), and six dates in the Blanco and 8an *artPn iver !monthly from 6une to ,ovember). All the studied laCes were located near the river mainstream !between $0 and #00 m) so that they received water from the river during the high water season but remained isolated during the dry season. 2his ensured that fish sampled during the dry season had been living in the fishing site at least during the precedent two or three months.

*aterial
*aterial from potential fish food sources were collected in floodplain laCes in order to evaluate their isotopic signatures= terrestrial plants !tree leaves from the laCe banC), /. ! Eichhornia crassipes, Pistia stratioides, Polygonum sp. and Cyperus sp.) and /> !Paspalum repens) aEuatic macrophytes, periphyton !epiphytic biofilm), particulate organic matter !3:*, obtained by successive water filtration onto a #071m mesh and a pre7combusted glass fibre N9@9 filter), leaf litter !mainly decaying leaves of terrestrial plants collected from the bottom of the laCes) and common groups of aEuatic macroinvertebrates !:donata, 5ecapoda, ;phemeroptera, /oleoptera and Nasteropoda). 8amples were rinsed with ultra7pure !milli7X) water, stored in individual tubes or bags, and stored frozen until their analysis. 9ishes were captured with gill nets !#.$ m height Y #$ m long, mesh sizes of #0, #$, .0, .$, >0 and $0 mm). We collected specifically fishes of eight species and four trophic levels to represent the fish assemblage= 5etritivore@algivore= Curimatella cf. alburna and Psectrogastersp.- Qerbivore= Schizodon fasciatus- *icrocarnivore !insectivore)= Triportheus angulatus- Neneralist piscivore= Pseudoplatystoma sp. and Pygocentrus nattereri- 8pecialized piscivore=Acestrorhynchus sp. and Hoplias malabaricus. 9ishes were identified and measured !8tandard %ength, 8% in cm) and >J$ g of dorsal muscle tissue were extracted using an ultra clean sampling procedure H"'I and taCing care to exclude blood, sCin or bones. All the fish muscle samples were frozen in individual tubes. 8ize ranges of studied individuals were set to include only adults, less sub4ect to dietary shifts, and to obtain comparable size ranges between the three populations studied for each species. In the laboratory, samples were lyophilized to obtain a completely dry extract, and grounded to a fine powder to perform mercury and isotopic analysis.

*ercury Analyses
2he %aboratorio de /alidad Ambiental !%/A) from Instituto de ;cologPa of %a 3az Lniversity !Bolivia) carried out mercury analyses on fish muscle samples. *ercury was extracted by acid digestion and Euantified by cold vapour atomic fluorescence spectroscopy !/AA98, BrooCs and *odel III see H".I for further details on the protocol). esults were expressed as total mercury concentration in wet weight muscle !HQgIww in 1g g&"). A previous worC showed that some populations present a significant influence of fish size on mercury concentration H".I. 8o fish size was selected to be similar between populations and limited to adult range and then HQgI values were not corrected by fish size.

Isotopic Analysis

,itrogen !+"$,) and carbon !+ "./) stable isotope ratios of food sources, invertebrates and fishes were measured to describe food web structure in the three locations studied. + "$, was used to estimate consumer trophic position as consumers are constantly + "$, enriched in comparison to their preferred food source- on the contrary, the +"./ is relatively stable among trophic levels but varies in relation with the sources that support the food chain and rather indicates energy pathway H#0I. elative individual trophic position !23) was calculated by the formula= 23 R Z[!+ "$,fish&+"$,base)@\ !where Z is the trophic position of the organism used to estimate + "$,base and \ is the , isotopic fractionation in ] that occurs between each trophic level). 2he isotopic fractionation value \ was set to #.<] H#"I. +"$,base was estimated using mean +"$, of the detritivore species C. alburna and then Z was set to #. L/ 5avis 8table Isotope 9acility laboratory !Lniversity of /alifornia, 5avis, L8A, http=@@stableisotopefacility.ucdavis.edu@) performed the isotopic analyses.

8tatistical Analysis
In order to evaluate differences in isotopic signatures ") between source categories, #) between species, .) between the three rivers for each species and source categories and to test differences in mercury concentration between species, we employed GrusCal7Wallis !GJW, non parametric Anova) and 3ermanova tests !permutational multivariate Anova that may consider simultaneously the + "./ and +"$, values- available on the Aegan pacCage of the statistical computing freeware program http=@@www.r7pro4ect.org@, H##I). Qomogeneity of multivariate dispersion was tested with a permutation test prior to 3ermanova. elative contributions of primary food sources to isotopic signature of primary consumer fish species !detritivore and herbivore) were estimated applying a Bayesian mixing model !8IA 7pacCage H#.I) in order to depict differences in river food web source that may explain differences in biomagnification. 2his model allows to estimate probability distributions of multiple source contributions to an isotopic signature while accounting for the observed variability in source, mixture isotopic signatures and isotopic fractionation H#.I. ,evertheless the selection of a small set of sources is reEuired to provide a better resolution of the resultsH#>I. 8tomach contents information !based on Eualitative field trip observation and H#$I) was used to depict large diet categories of fish species and to select the sources. A biomagnification factor was calculated as the ratio between the maximum and minimum species HQgI mean values. 2his factor was completed by the evaluation of the slope of the 23 vs. HQgI relation !%og transformed). 9inally, a relative food chain length was evaluated for each river by mean trophic level of the four piscivore species. 5ifferences of food chain length values between rivers were tested by GrusCal7Wallis. 9or all tests, type I error was set to p R 0.0$.

Results 2rophic 8tructure

Isotopic signatures of primary food sources were significantly different !3ermanova, p R 0.00") between the six categories !terrestrial plants, /. and /> macrophytes, periphyton, leaf litter and 3:*)- but differences became non significant !3ermanova, p R 0.0?$) when excluding the /> macrophytes that presented the highest +"./ values !varying between &"..#] and &"#..]) in comparison to the other food source categories that oscillated between &.$] and &#$] !2able "). 2hese five categories were not significantly different among them for + "./ values !GrusCal7Wallis, GJ W, p R 0.0(>) nor for +"$, values !GJW, p R 0.0$(). 3eriphyton !3ermanova, p R 0.00#) and 3:* !3ermanova, p R 0.0"#) isotopic signatures presented significant variation between localities, being

more "./ depleted and "$, enriched in the Itnez iver in comparison to Blanco and 8an *artPn rivers !2able ", 9igure #a,b,c). 2he remaining sources presented no significant differences !3ermanova, p^0.0$) in +"./ and +"$, values.

Figure %. Isotopic signature o! sources& invertebrates and !ish in three rivers o! the It#ne$ basin 'Ama$on& "olivia(. Biplots display mean values !_ sd in error bars) of + "./ and +"$, values for sources !A&"&C) 3:* R 3articulate organic matter, %itt R %eaf litter, 3eri R 3eriphyton- /. R /. aEuatic macrophytes and 2Aeg R 2errestrial Aegetation- and invertebrates groups !)&E&F)= ;phe R ;phemeroptera, :do R :donata, 5ec R 5ecapoda, /ol R /oleoptera and Nas R Nasteropoda. ! *& &I) represented mean values !_ sd in error bars) of +"./ and 2rophic 3osition !derived from +"$,) of fish species= detritivore ! " Curimatella cf. alburna, # " Psectrogaster sp.), herbivore $ " Schizodon fasciatus), insectivore !> R 2riportheus angulatus) and piscivore !$ R Pseudoplatystoma sp., ( R Pygocentrusnattereri, ? R Hoplias malabaricus and % " Acestrorhynchus sp.). In biplots !A&"&C), /> aEuatic marcophytes !+"./ = &"#.. to &"..#]- +"$, = # to #.#]) were not plotted and +"./ and +"$, values of detritivore !", #) and herbivore !.) fish species were reported. doi="0.".?"@4ournal.pone.00($0$>.g00#

+able 1. Isotopic composition ',1-.& ,1/C( o! !ood sources and invertebrates in three rivers o! the It#ne$ basin. doi="0.".?"@4ournal.pone.00($0$>.t00" Isotopic signatures of the five invertebrate groups !:donata, 5ecapoda, ;phemeroptera, /oleoptera and Nasteropoda) were significantly different among them !3ermanova, p R 0.00"-2able "). 5ifferences between groups for the +"./ values !GJW, p R 0.0"$<) concerned principally the ;phemeroptera that were "./ depleted !+ "./ from &>$] to &..]) compared to the other groups !+ "./ oscillating between &.(] and &#?]). ;phemeroptera and Nasteropoda showed the lowest +"$, values !population means between ..0<] and ..??]), /oleoptera and :donata were intermediate !..$] J $.$>]) and 5ecapoda showed the highest values !(.><]J?."]). Isotopic compositions between the three rivers were significantly different for the 5ecapoda, ;phemeroptera and :donata !3ermanova, p R 0.00", 0.00. and 0.00#, respectively) but not significantly different for /oleoptera and Nasteropoda !3ermanova, p R 0.0$. and p R 0.0'#, respectively). 9or the first three groups +"./ values were significantly lower in the Itnez iver in relation to the other rivers !GJW, p R 0.000", 0.0"$ and 0.00", respectively, 9igure #d,e,f), although +"$, values were not significantly different between rivers !GJW, p^0.$). /arbon isotope ratios of /oleoptera and Nasteropoda tended to be "./ depleted in the Itnez iver as well !2able ", 9igure #d,e,f). 9or the Itnez iver, all the invertebrate groups presented more negative + "./ values !from &>..(' to &.".(?]) than primary food sources !&.".0> to &#'.$"], 2able ", 9igure #d,e,f). All the three rivers merged, significant differences in the isotopic signature between fish species were found !3ermanova, p R 0.00") and species were gradually positioned on the trophic position axis in accordance to their coarse diet regime !9igure #g,h,i). 2he eight fish species also showed significant

differences among rivers !3ermanova, p R 0.00", 2able #). 2rophic position !23) of piscivore species varied significantly between rivers !GJW, pW0.00$) and was higher in the 8an *artPn iver !between #.? and .) than in the two other sites !between #.. and #.?). :n the contrary, non7piscivore species did not present significant differences !GJW, p^0.#), except for Psectrogaster sp. !GJW, p R 0.0") that also showed a higher trophic level in the 8an *artPn iver.

+able %. 0tandard 1ength& mercury concentration and isotope signature ',1-. and relative +rophic Position 2 +P& ,1/C( o! eight !ish species populations sampled in three rivers o! the It#ne$ basin 'Ama$on& "olivia(. doi="0.".?"@4ournal.pone.00($0$>.t00# As for periphyton, 3:* and invertebrates, fish species globally tended to be more "./ depleted in the Itnez iver !9igure #). 9ish assemblage values ranged between &."] and &#(] in 8an *artPn iver, &..] and &#?] in Blanco iver and &.?] and &."] in Itnez iver. 2hese differences persisted and were significant for all species !GJW, pW0.000$). In the Itnez iver, as for the invertebrates, the two detritivore species !Psectrogaster sp. and C. alburna) also presented more negative +"./ than all the considered food sources !9igure #a,b,c). elative contribution of primary food sources to detritivore fish species may be biased because these species presented more negative +"./ than all the considered food sources- this was not the case for the herbivore S. fasciatus !2able .). Qowever, the three species showed a similar pattern, with a high contribution of periphyton in the Itnez iver !(<J<0K) and low contribution !#J"(K) in the two other rivers. 2he contribution of terrestrial vegetation followed a reverse pattern, being lower in Itnez iver !#J.K) than in the two remaining rivers !"<J?'K). 2he contribution of terrestrial vegetation was the highest for S. fasciatus and /. alburna in 8an *artPn iver !(? and ?'K, respectively). ,o dominant primary food source category appeared in the diet of the three species in the Blanco iver.

+able /. 0ource relative contributions 'mean 3 4 sd& estimated by 0IAR mi5ing model( to detritivore 'Psectrogaster sp. and Curimatella cf. alburna( and herbivore 'Schizodon fasciatus( !ish diet in three rivers o! the It#ne$ basin. doi="0.".?"@4ournal.pone.00($0$>.t00. elative food chain length, evaluated by mean trophic level of the four piscivore species, presented significant differences !GJW, pW0.000") with higher values in 8an *artPn iver !#.<() in comparison to Blanco iver !#.$$) and in the Itnez iver !#.$#).

9ish *ercury /oncentration and Biomagnification

9ish species presented significant differences in mercury concentrations !GJW, pW0.000") that could be related to their coarse diet regime in agreement with biomagnification processes !9igure .). At the assemblage level we found a significant global correlation !8pearman & R 0.$?', pW0.000") between individual mercury concentrations and trophic position that was still valid individually for each river !8an *artPn= & R 0.(?<- Blanco= & R 0.(.. and Itnez= & R 0.($>, all pW0.000").

Figure /. Mercury biomagni!ication in !ish assemblage o! three rivers !rom the It#ne$ basin 'Ama$on& "olivia(. Biplots display mean values !_ sd in error bars) of 2rophic 3osition !derived from + "$,) and HQgIww of detritivore ! " Curimatella cf. alburna, # " Psectrogaster sp.), herbivore $ " Schizodon fasciatus), insectivore !> R 2riportheus angulatus) and piscivore !$ R Pseudoplatystoma sp., ( R Pygocentrus nattereri, ? R Hoplias malabaricus and % " Acestrorhynchus sp.) fish species. 8lope of the relation correspond to biomagnification along the food chain. doi="0.".?"@4ournal.pone.00($0$>.g00. 3iscivore species showed significantly higher mercury concentrations in the Itnez iver !0."$" 1g g&"_0.0<, n R #><) than in 8an *artin and Blanco rivers !0."0( 1g g &"_0.0<, n R '0 and 0."0$ 1g g&"_0.0?, n R (> respectively) !GJW, pW0.000"). A similar difference !GJW, p R 0.00$) also occurred for detritivore and herbivore species with higher values in the Itnez iver !0.0$# 1g g &"_0.0., n R "#() than in the two others !0.0>( 1g g&"_0.0., n R .$ in Blanco iver and 0.0.' 1g g &"_0.0#, n R >? in 8an *artin iver). At the species level, four species showed significant differences in mercury concentrations between rivers !GJW= Acestrorhynchus sp., p R 0.00(- H. malabaricus, p R 0.000"- P. nattereri, pW0.000" and S. fasciatus, p R 0.0#>), all of them presented higher values in the Itnez iver !2able #). Biomagnification factor, calculated as the ratio between HQgI of P. nattereri !species with the highest mean HQgI R 0."( 1g g&") and C. alburna !lowest mean HQgI R 0.0$ 1g g&"), was #.$ in the 8an *artPn, # in the Blanco and ..< in the Itnez iver. 8imilarly, the slope of the relationship between + "$, and HQgI !%og transformed) was higher in Itnez iver !slope R 0.>., # R 0.<#, pW0.00") than in the Blanco iver !slope R 0..>, # R 0.?0, p R 0.0#) and in the 8an *artPn iver where the relation was not significant !slope R 0.##, # R 0.>$, p R 0.0?) !9igure .).

)iscussion
2he three studied rivers presented a similar general pattern of food source contribution that is in agreement with Cnowledge from previous studies in the Amazon H#(IJH.0I. In particular, the isotopic signature of />7macrophytes is clearly "./ enriched compared to the other primary sources and consumers, thus they are not a significant food source for consumers and can not sustain the food chains in the study sites. :n the other hand, the other food sources may all contribute to the food web, but remained widely overlapped. Qowever, although the three rivers are submitted to the same climatic conditions and belong to the same hydrographical basin, ma4or differences in carbon isotopic signatures and food chain length could be detected=

1.

Itnez iver differed from the two others mainly because primary sources, primary consumers and secondary consumers were all more "./ depleted than in 8an *artPn and Blanco rivers !9igure #)-

2.

Itnez iver also presented a higher contribution of periphyton to the diet of the detritivore and herbivore fishes !2able .)3. 8an *artPn iver showed a longer food chain than the two other rivers because of the higher trophic position of all piscivore species !9igure #g,h,i), while the three rivers presented similar +"$, values for the five primary source categories considered !2able "). We hypothesized that natural variations of water Euality !clear water with low sediment load vs. white water with high sediment load) would have an effect on trophic structure, as shown for instance in Aenezuelan rivers H#"I. In such a case, the two clear water rivers !Itnez and 8an *artin) would have shown a similar trophic structure and origin, and different from the one of the white water Blanco iver. 2he results did not follow this pattern= Itnez iver presented different carbon isotopic signature and periphyton contribution than the two other rivers- whereas, 8an *artin iver showed a longer food chain in comparison to Itnez and Blanco rivers. It thus appears that sediment load was not a dominant factor controlling trophic structure in the laCes studied. 2he more negative +"./ values for primary producers, invertebrates and fish from the Itnez iver compared to those from the two other rivers indicate differences in carbon sources between rivers. *oreover, fish +"./ values, especially those of the detritivore species C. alburna !&.(."] in Itnez iver) and Psectrogaster sp. !&.?]), as well as ;phemeroptera mayfly !&>..?]), were more "./ depleted than the sampled primary producers !&#'.$] to &."]). 2he low positive isotopic fractionation of carbon !_"]) that occurs between each trophic level H#0I could not explain this discrepancy, that then implies the contribution of an additional !not sampled) "./7depleted carbon source. 5etritivore fish species and ;phemeroptera are liCely to feed predominantly on the bottom near the sediment !see H."I for a discussion on ;phemeroptera feeding). *ethane production from anoxic sediments could provide such "./7depleted carbon source H.#I, H..I. 8everal studies demonstrated that methane7oxidizing bacteria !*:B) activity allows the transfer of this "./7depleted carbon to zooplanCton H.>I and fish H.$I. 2hus, more "./7depleted carbon could be an indicator of a contribution of methane carbon to benthic as well as pelagic laCe food webs in temperate H.(Iand tropical H.$I systems. Amazonian laCes and reservoirs can support a high methane production H.?I and several studies observed low +"./ values in fish from 8outh American tropical systems H#<I, H.$I, H.<I, H.'I. In the Ichilo iver !Bolivian Amazonian lowlands) e4asH#<I observed low +"./ values for algivore !Anodus elongatus, &.'] _0..) and detritivore fishes !Potamorhina altamazonica, &.(.>] _".#- Psectrogaster rutiloides &.$..] _".#) and even lower values for benthic invertebrates !/hironomidae, ;phemeroptera, &.'.?] _".#) than for the most "./ depleted primary food source !3:*, &.?] _0.(). Wantzen et al. H.<I suggested that seasonal variations in methane production, induced by water level in the Brazilian 3antanal, might explain lower + "./ values during the wet season for the detritivore fishPsectrogaster cur'i'entris( and 8anseverino et al. H.$I demonstrated that the "./ signature of fishes is related to *:B activity. %ower + "./ values for invertebrates and fish in Itnez iver than in the other rivers could then be tentatively interpreted as an effect of higher carbon production by metanotrophic bacteria. Qowever, *olina et al. H."I did not report such low values in the Beni iver !Bolivian Amazonian lowlands) where Campsurus mayfly Ephemeroptera)presented similar +"./ values !&.$.? to &.>.?]) to seston !&.$." to &...<]), revealing that this process is not a generality. 2he three studied rivers presented relatively low water mercury concentration, similar to mercury levels found in natural systems of the region H".I. 5ue to their lower sediment load, clear water rivers, liCe Itnez and 8an *artin, should have demonstrated a naturally lower mercury concentration in comparison to Blanco iver. 3revious results H".I and this study showed a slightly perturbed situation in the Itnez iver, with higher mercury concentrations in piscivore and herbivore species, compared to fish from non7perturbed rivers !Blanco and 8an *artPn). Based on a partially similar data set and sampling locations, 3ouilly et al. H".I concluded that bioaccumulation, defined as the increment of mercury concentrations during an organism`s lifetime, is

not the principal factor explaining increased mercury concentrations in fish from Itnez iver. /onversely, Itnez iver showed higher biomagnification factor !..<) than the two other rivers !Blanco R #, 8an *artPn R #.$), indicating that this process may partially explain higher mercury concentrations in fish from the Itnez iver. We hypothesized that the trophic structure and in particular food chain length could control the biomagnification rate, because freshwater systems generally demonstrate a positive relationship between mercury biomagnification rates and food chain length H'I, H.0I, H>0I. Qowever, the two clear water rivers studied showed an opposite relationship !9igure .), with higher mercury biomagnification factor !..<) and shorter food chain !#.$#) in Itnez iver, and lower biomagnification factor !#.$) longer food chain in !#.<() in the 8an *artPn iver. 2his discrepancy between the general pattern and the situation in the two studied clear water rivers could originate from a higher mercury bioavailability and@or a better efficiency in the transfer along the food web in the Itnez iver. It has been suggested that periphyton and macrophytes constitute the main pathway of mercury between primary producers and macro7invertebrates in /anadian temperate laCes H>"I. A strong linC between methanogenic bacteria and mercury methylation in the periphyton has been demonstrated H"#I and 5ominiEue et al. H.'I related the high methyl mercury concentrations found in detritivore fishes downstreams of a dam in 9rench Nuyana, to the export of methyl mercury from the reservoir and to the Euality of the biofilm which is characterized by low +"./ values, indicating *:B activity. In the Amazonian systems, periphyton associated to macrophyte roots is a ma4or mercury methylation site H.I, H>#I and higher biomagnification rates for invertebrates feeding on periphyton has been demonstrated H.0I. In our study, estimation of food source contribution by the mixing model showed that the contribution of periphyton to the diet of the detritivore and herbivore fishes was high in the Itnez iver and low in 8an *artPn and Blanco rivers, and that a higher contribution of terrestrial vegetation, in particular for S. fasciatus and C. alburna in the 8an *artPn iver. A scheme of higher methylation rates due to methanogenic bacteria activity within biofilms !as indicated by the more negative +"./ values observed) and higher contribution of periphyton in the food web may explain the higher biomagnification rates observed in fish from the Itnez iver in comparison to the two other rivers. Balance of internal !periphyton, phytoplanCton) vs. external !terrestrial vegetation) primary production as well as *:B activity may thus be critical factors in food web mercury contamination. 2he three rivers differ in their flooding regime, the main Itnez iver showing larger flooding area and longer flooding duration, therefore more laCe connectivity, than Blanco and 8an *artPn rivers H"$I. Apart from this difference, it remains unclear which other factors could generate a higher *:B activity and periphyton contribution in the Itnez iver. 2he observations reported correspond mainly to the #00? dry season and a generalisation based on several years of studied would be necessary. 9or this date we can conclude that, in the Itnez basin with low to moderate mercury concentrations in water, fish mercury contamination appeared mainly controlled by biomagnification enhanced by periphyton contribution to food web and probably environmental conditions, such as flooding, favourable to methylation and methanogenesis. 8urprisingly in these systems biomagnification rates were not related to food chain length, but rather to a methanogenic pathway. :ur results also suggest that biomagnification, favoured by trophic structure and biotic processes, may lead to critical contamination of fishes even at low rates of mercury input.

Ac6no7ledgments
We thanC the 2rinidad WW9 team, Itnez departmental parC !357A*,I Itnez) and Beni 3refecture !Bolivia) for the logistical support and all the people who helped in the field worC, especially the fishermen and boat pilots= Q. ibero odriguez, 2. 8uarez and 6. AasEuez. We greatly appreciated the warm welcome in the Itnez communities of Bella Aista, emanso, *ategua and Aersalles. We

are grateful to N. Abril !Bordeaux Lniversity, ;3:/ laboratory) to helpful comments and manuscript revision.

Author Contributions
/onceived and designed the experiments= *3 5 23 /I* 6 5N. 3erformed the experiments= *3 5 23 /I*. Analyzed the data= *3 5 23 /Q 6 5N. /ontributed reagents@materials@analysis tools= 6%5 /Q. Wrote the paper= *3 5 /Q 6 5N.

Re!erences
". ". oulet *, %ucotte *, 9arella ,, 8eriEue N, /oelho Q, et al. !"''') ;ffects of recent human colonization on the presence of mercury in Amazonian ecosystems. Water, Air, a 8oil 3ollution ""#= #'?J.".. doi= "0."0#.@a="00$0?.>.#0"$. /ross ef 3ub*ed@,/BI Noogle 8cholar #.Nilmour /, iedel N, ;derington *, Bell 6, Benoit 6, et al. !"''<) *ethylmercury concentrations and production rates across a trophic gradient in the northern ;verglades. Biogeochemistry >0= .#?J.>$. doi= "0."0#.@a="00$'?#?0<("(. /ross ef 3ub*ed@,/BI Noogle 8cholar ..Nuimaraes 6, *eili *, Qylander %, /astro da 8ilva ; de, oulet *, et al. !#000) *ercury net methylation in five tropical flood plain regions of Brazil= high in the root zone of floating macrophyte mats but low in surface sediments and flooded soils. 8cience of the 2otal ;nvironment #("= ''J"0?. doi= http=@@dx.doi.org@"0."0"(@s00><7'('?!00)00(#<7<. /ross ef 3ub*ed@,/BI Noogle 8cholar >./oelho78ouza 8, Nuimarbes 6 5, *iranda * , 3oirier Q, *auro 6B,, et al. !#0"") *ercury and flooding cycles in the 2apa4Ms iver basin, Brazilian Amazon= the role of periphyton of a floating macrophyte !3aspalum repens). 2he 8cience of the 2otal ;nvironment >0'= #?>(J#?$.. doi= http=@@dx.doi.org@"0."0"(@4.scitotenv.#0"".0..0#<. /ross ef 3ub*ed@,/BI Noogle 8cholar $./abana N, 2remblay A, Galff 6, asmussen 6B !"''>) 3elagic 9ood /hain 8tructure in :ntario %aCes= A 5eterminant of *ercury %evels in %aCe 2rout !8alvelinus namaycush ). /anadian 6ournal of 9isheries and AEuatic 8ciences $"= .<"J.<'. doi=http=@@dx.doi.org@"0."".'@f'>70.'. /ross ef 3ub*ed@,/BI Noogle 8cholar (.8ilva 58, %ucotte *, oulet *, 3oirier Q, *ergler 5, et al. !#00$) 2rophic 8tructure and Bioaccumulation of *ercury in 9ish of 2hree ,atural %aCes of the Brazilian Amazon. Water, Air, and 8oil 3ollution "($= ??J'>. doi= http=@@dx.doi.org@"0."00?@s""#?0700$7><""7<. /ross ef 3ub*ed@,/BI Noogle 8cholar ?.Qall B, Bodaly , 9udge , udd 6, osenberg 5 !"''?) 9ood as the dominant pathway of methylmercury uptaCe by fish. Water, Air, a 8oil 3ollution "00= ".J#>.

2.

3.

4.

5.

6.

7.

/ross ef 3ub*ed@,/BI Noogle 8cholar 8. <.Bowles G, Apte 8, *aher W, Gawei *, 8mith !#00") Bioaccumulation and biomagnification of mercury in %aCe *urray, 3apua ,ew Nuinea. /anadian 6ournal of 9isheries and AEuatic 8ciences $<= <<<J<'?. doi= http=@@dx.doi.org@"0."".'@c4fas7$<7$7<<<. /ross ef 3ub*ed@,/BI Noogle 8cholar 9. './abana N, asmussen 6 !"''>) *odelling food chain structure and contaminant bioaccumulation using stable nitrogen isotopes. ,ature .?#= #$$J#$?. doi=http=@@dx.doi.org@"0."0.<@.?##$$a0. /ross ef 3ub*ed@,/BI Noogle 8cholar 10. "0.Quguet %, /astelle 8, 8chcfer 6, Blanc N, *aury7Brachet , et al. !#0"0) *ercury methylation rates of biofilm and planCton microorganisms from a hydroelectric reservoir in 9rench Nuiana. 2he 8cience of the 2otal ;nvironment >0<= "..<J".><. doi=http=@@dx.doi.org@"0."0"(@4.scitotenv.#00'."0.0$<. /ross ef 3ub*ed@,/BI Noogle 8cholar 11. "".Achd 5, Qintelmann Q, eee 6 !#0"") Importance of sulfate reducing bacteria in mercury methylation and demethylation in periphyton from Bolivian Amazon region. /hemosphere <#= '""J '"(. doi= http=@@dx.doi.org@"0."0"(@4.chemosphere.#0"0."0.0$0. /ross ef 3ub*ed@,/BI Noogle 8cholar 12. "#.Qamelin 8, Amyot *, BarCay 2, Wang e, 3lanas 5 !#0"") *ethanogens= principal methylators of mercury in %aCe 3eriphyton. ;nvironmental 8cience a 2echnology >$= ?('.J??00. doi= http=@@dx.doi.org@"0."0#"@es#0"00?#. /ross ef 3ub*ed@,/BI Noogle 8cholar 13. "..3ouilly *, 3rez 2, e4as 5, Nuzman 9, /respo N, et al. !#0"#) *ercury bioaccumulation patterns in fish from the Itnez river basin, Bolivian Amazon. ;cotoxicology and environmental safety <.= <J"$. doi=http=@@dx.doi.org@"0."0"(@4.ecoenv.#0"#.0$.0"<. /ross ef 3ub*ed@,/BI Noogle 8cholar 14. ">.*artinez 6*, %e 2oan 2 !#00?) *apping of flood dynamics and spatial distribution of vegetation in the Amazon floodplain using multitemporal 8A data. emote 8ensing of ;nvironment "0<= #0'J##.. doi= http=@@dx.doi.org@"0."0"(@4.rse.#00(."".0"#. /ross ef 3ub*ed@,/BI Noogle 8cholar 15. "$.:vando A% !#00<) 2he pathways of mercury in the Itnez7Nuapor basin, mapping the potential deposition and transformation zones. *aster 2hesis, ;nschede, the ,etherlands= International Institute for geoinformation science and earth observation.

16.

"(.Brigham *;, Wentz 5A, AiCen N , Grabbenhoft 53 !#00') *ercury /ycling in 8tream ;cosystems. ". Water /olumn /hemistry and 2ransport. ;nvironmental 8cience a 2echnology >.= #?#0J#?#$. doi= http=@@dx.doi.org@"0."0#"@es<0#('>n. /ross ef 3ub*ed@,/BI Noogle 8cholar 17. "?.5avid ,, *cGee %6, BlacC 96, 9legal A , /onaway /Q, et al. !#00') *ercury concentrations and loads in a large river system tributary to 8an 9rancisco Bay, /alifornia, L8A. ;nvironmental 2oxicology and /hemistry #<= #0'"J#"00. doi=http=@@dx.doi.org@"0."<'?@0<7><#.". /ross ef 3ub*ed@,/BI Noogle 8cholar 18. "<.*aurice7Bourgoin % !#00") %e mercure dans les eaux de surface du bassin amazonien= transfert du mercure des sols aux milieux aEuatiEues, spciation, transport et sdimentation dans les rivifres et plaines d`inondation. In= /armouze 6, %ucotte *, Boudou A, editors. %e mercure en Amazonie= rgle de l`homme et de l`environnement, risEues sanitaires. 3aris= I 5, ;xpertise /ollgiale. "(?J#0". 19. "'.,ational esearch /ouncil !#000) 2oxicological ;ffects of *ethylmerc. Washington, 5./.= /ommittee on the 2oxicological ;ffects of *ethylmercury, Board on ;nvironmental 8tudies and 2oxicology. ,ational Academy 3ress. 20. #0.Aander Vanden *6, /abana N, asmussen 6B !"''?) /omparing trophic position of freshwater fish calculated using stable nitrogen isotope ratios !d"$,) and literature dietary data. /anadian 6ournal of 9isheries and AEuatic 8ciences $>= "">#J""$<. doi=http=@@dx.doi.org@"0."".'@c4fas7$>7$7"">#. /ross ef 3ub*ed@,/BI Noogle 8cholar 21. #".6epsen 5B, Winemiller G: !#00#) 8tructure of tropical river food webs revealed by stable isotope ratios. :iCos '(= >(J$$. doi= http=@@dx.doi.org@"0."0.>@4."(0070?0(.#00#.'(0"0$.x. /ross ef 3ub*ed@,/BI Noogle 8cholar 22. ##.Anderson *6 !#00") A new method for non7parametric multivariate analysis of variance. Austral ;cology #(= .#J>(. doi= http=@@dx.doi.org@"0.""""@4.">>#7'''..#00".0"0?0.pp.x. /ross ef 3ub*ed@,/BI Noogle 8cholar 23. #..3arnell A/, Inger , Bearhop 8, 6acCson A%/7A !#0"0) 8ource 3artitioning Lsing 8table Isotopes= /oping with 2oo *uch Aariation. 3los :ne $= e'(?#. doi=http=@@dx.doi.org@"0.".?"@4ournal.pone.000'(?#. /ross ef 3ub*ed@,/BI Noogle 8cholar 24. #>.3olito *6, 2rivelpiece WV, GarnovsCy ,6, ,g ;, 3atterson W3, et al. !#0"") Integrating stomach content and stable isotope analyses to Euantify the diets of pygoscelid penguins. 3lo8 one (= e#((>#. doi= http=@@dx.doi.org@"0.".?"@4ournal.pone.00#((>#. /ross ef 3ub*ed@,/BI Noogle 8cholar

25.

#$.3ouilly *, eunoCi 2, osales /, 2rophic 2% !#00>) 2rophic structure of fish assemblages from *amor iver floodplain laCes !Bolivia ). ;cology of 9reshwater 9ish ".= #>$J#$?. doi= http=@@dx.doi.org@"0.""""@4."(0070(...#00>.000$$.x. /ross ef 3ub*ed@,/BI Noogle 8cholar 26. #(.Arau4o7%ima /, 9orsberg B, Aictoria , *artinelli % !"'<() ;nergy sources for detritivorous fishes in the Amazon. 8cience #.>= "#$(J"#$<. doi=http=@@dx.doi.org@"0.""#(@science.#.>.>?<"."#$(. /ross ef 3ub*ed@,/BI Noogle 8cholar 27. #?.Qamilton 8, %ewis W, 8ippel 8 !"''#) ;nergy sources for aEuatic animals in the :rinoco iver floodplain= evidence from stable isotopes. :ecologia <'= .#>J..0. /ross ef 3ub*ed@,/BI Noogle 8cholar 28. #<. e4as 5 !#00>) 2rophic elations and ,utrient ecycling in a 2ropical 9loodplain %aCe. 3h5 2hesis, GatholieCe Lniversiteit %euven, Belgium. Available= http=@@lab7ulra.com@pdf@ e4asK#05. K#0K#03h.5K#02hesis.pdf. 29. #'.6epsen 5B, Winemiller G: !#00?) Basin geochemistry and isotopic ratios of fishes and basal production sources in four neotropical rivers. ;cology of 9reshwater 9ish "(= #(?J#<". doi= http=@@dx.doi.org@"0.""""@4."(0070(...#00(.00#"<.x. /ross ef 3ub*ed@,/BI Noogle 8cholar 30. .0.*olina /, Nibon 9, 5uprey 6, 5ominguez ;, Nuimarbes 6, et al. !#0"0) 2ransfer of mercury and methylmercury along macroinvertebrate food chains in a floodplain laCe of the Beni iver, Bolivian Amazonia. 8cience of the 2otal ;nvironment >0<= ..<#J..'". doi= http=@@dx.doi.org@"0."0"(@4.scitotenv.#0"0.0>.0"'. /ross ef 3ub*ed@,/BI Noogle 8cholar 31. .".*olina /, Nibon 9, :berdorff 2 !#0"") *acroinvertebrate food web structure in a floodplain laCe of the Bolivian Amazon. Qydrobiologia ((.= ".$J"$.. doi=http=@@dx.doi.org@"0."00?@s"0?$070"070$($7>. /ross ef 3ub*ed@,/BI Noogle 8cholar 32. .#.Bunn 8, Boon 3 !"''.) What sources of organic carbon drive food webs in billabongsh A study based on stable isotope analysis. :ecologia '(= <$J'>. doi=http=@@dx.doi.org@"0."00?@bf00."<0.>. /ross ef 3ub*ed@,/BI Noogle 8cholar 33. ...6ones , Nrey 6 !#00>) 8table isotope analysis of chironomid larvae from some 9innish forest laCes indicates dietary contribution from biogenic methane. Boreal ;nvironment esearch '= "?J#.. /ross ef 3ub*ed@,/BI Noogle 8cholar

34.

.>.GanCaala 3, 2aipale 8, Nrey 6 !#00() ;xperimental i"./ evidence for a contribution of methane to pelagic food webs in laCes. %imnology and :ceanography $"= #<#"J#<#?. doi= http=@@dx.doi.org@"0.>."'@lo.#00(.$".(.#<#". /ross ef 3ub*ed@,/BI Noogle 8cholar 35. .$.8anseverino A, BastviCen 5, 8undh I !#0"#) *ethane /arbon 8upports AEuatic 9ood Webs to the 9ish %evel. 3lo8 one ?= e>#?#.. doi=http=@@dx.doi.org@"0.".?"@4ournal.pone.00>#?#.. /ross ef 3ub*ed@,/BI Noogle 8cholar 36. .(.BastviCen 5, ;4lertsson 6, 8undh I, 2ranviC % !#00.) *ethane as a source of carbon and energy for laCe pelagic food webs. ;cology <>= '('J'<". doi=http=@@dx.doi.org@"0."<'0@00"#7 '($<!#00.)0<>H0'('=maasocI#.0.co-#. /ross ef 3ub*ed@,/BI Noogle 8cholar 37. .?.Bartlett G, /rill 3, 8ebacher 5, Qarriss , Wilson 6, et al. !"'<<) *ethane flux from the central Amazon floodplain. 6ournal of Neophysical research '.= "$?"J"$<#. doi=http=@@dx.doi.org@"0."0#'@4d0'.id0#p0"$?". /ross ef 3ub*ed@,/BI Noogle 8cholar 38. .<.Wantzen G, *achado 9, Aoss *, Boriss Q, 6unC W !#00#) 8easonal isotopic shifts in fish of the 3antanal wetland, Brazil. AEuatic 8ciences (>= #.'J#$". doi=http=@@dx.doi.org@"0."00?@pl000"."'(. /ross ef 3ub*ed@,/BI Noogle 8cholar 39. .'.5ominiEue e, *aury7Brachet , *uresan B, Aigouroux , ichard 8, et al. !#00?) Biofilm and mercury availability as Cey factors for mercury accumulation in fish !/urimata cyprinoides) from a disturbed Amazonian freshwater system. ;nvironmental toxicology and chemistry@8;2A/ #(= >$J$#. doi= http=@@dx.doi.org@"0."<'?@0$7(>'r.". /ross ef 3ub*ed@,/BI Noogle 8cholar 40. >0.8abo 6, 9inlay 6, 3ost 5 !#00') 9ood chains in freshwaters. Annal of the ,ew eorC Academy of 8ciences ""(#= "<?J##0. doi= http=@@dx.doi.org@"0.""""@4."?>'7((.#.#00'.0>>>$.x. /ross ef 3ub*ed@,/BI Noogle 8cholar 41. >"./remona 9, Qamelin 8, 3lanas 5, %ucotte * !#00') 8ources of organic matter and methylmercury in littoral macroinvertebrates= a stable isotope approach. Biogeochemistry '>= <"J'>. doi= http=@@dx.doi.org@"0."00?@s"0$..700'7'.0'7'. /ross ef 3ub*ed@,/BI Noogle 8cholar 42. >#./oelho78ouza 8, Nuimarbes 6 5, *auro 6B,, *iranda * , Azevedo 8*9: !#00() *ercury methylation and bacterial activity associated to tropical phytoplanCton. 2he 8cience of the 2otal ;nvironment .(>= "<<J"''. doi=http=@@dx.doi.org@"0."0"(@4.scitotenv.#00$.0?.0"0. /ross ef

3ub*ed@,/BI Noogle 8cholar