In situ feeding assay with Gammarus fossarum (Crustacea): Modelling the influence of confounding factors to improve water quality biomonitoring.

Abstract

In situ feeding assays implemented with transplanted crustacean gammarids have been claimed as promising tools for the diagnostic assessment of water quality. Nevertheless the implementation of such methodologies in biomonitoring programs is still limited. This is explained by the necessity to improve the reliability of these bioassays. The present study illustrates how modelling the influence of confounding factors could allow to improve the interpretation of in situ feeding assay with Gammarus fossarum. We proceeded in four steps: (i) we quantified the influence of body size, temperature and conductivity on feeding rate in laboratory conditions; (ii) based on these laboratory findings, we computed a feeding inhibition index, which proved to be robust to environmental conditions and allowed us to define a reference statistical distribution of feeding activity values through the data compilation of 24 in situ assays among diverse reference stations at different seasons; (iii) we tested the sensitivity of the feeding assay using this statistical framework by performing 41 in situ deployments in contaminated stations presenting a large range of contaminant profiles; and (iv) we illustrated in two site-specific studies how the proposed methodology improved the diagnosis of water quality by preventing false-positive and false-negative cases mainly induced by temperature confounding influence. Interestingly, the implementation of the developed protocol could permit to assess water quality without following an upstream/downstream procedure and to compare assays performed at different seasons as part of large-scale biomonitoring programs.

Full text

Page 1

In situ feeding assay with Gammarus fossarum (Crustacea):
Modelling the influence of confounding factors to improve
water quality biomonitoring
Romain Coulauda,b, Olivier Geffarda,*, Benoı ˆt Xuereba,1, Emilie Lacazea, Herve ´ Que ´aua,
Jeanne Garrica, Sandrine Charlesb, Arnaud Chaumota,*
aCemagref, UR MALY, 3 bis quai Chauveau-CP 220, F-69336 Lyon, France
bUniversite ´ de Lyon, F-69000, Lyon, Universite ´ Lyon 1, CNRS, UMR5558, Laboratoire de Biome ´trie et Biologie Evolutive,
F-69622 Villeurbanne, France
a r t i c l e i n f o
Article history:
Received 11 March 2011
Received in revised form
31 August 2011
Accepted 15 September 2011
Available online 22 September 2011
Keywords:
In situ assay
Feeding rate
Gammarus
Temperature
Biomonitoring
Modelling
a b s t r a c t
In situ feeding assays implemented with transplanted crustacean gammarids have been
claimed as promising tools for the diagnostic assessment of water quality. Nevertheless
the implementation of such methodologies in biomonitoring programs is still limited. This
is explained by the necessity to improve the reliability of these bioassays. The present
study illustrates how modelling the influence of confounding factors could allow to
improve the interpretation of in situ feeding assay with Gammarus fossarum. We proceeded
in four steps: (i) we quantified the influence of body size, temperature and conductivity on
feeding rate in laboratory conditions; (ii) based on these laboratory findings, we computed
a feeding inhibition index, which proved to be robust to environmental conditions and
allowed us to define a reference statistical distribution of feeding activity values through
the data compilation of 24 in situ assays among diverse reference stations at different
seasons; (iii) we tested the sensitivity of the feeding assay using this statistical framework
by performing 41 in situ deployments in contaminated stations presenting a large range of
contaminant profiles; and (iv) we illustrated in two site-specific studies how the proposed
methodology improved the diagnosis of water quality by preventing false-positive and
false-negative cases mainly induced by temperature confounding influence. Interestingly,
the implementation of the developed protocol could permit to assess water quality without
following an upstream/downstream procedure and to compare assays performed at
different seasons as part of large-scale biomonitoring programs.
ª 2011 Elsevier Ltd. All rights reserved.
1.Introduction
In aquatic ecosystems, organisms are constantly exposed to
different levels of physical and chemical stressors. To
estimate and predict their biological effects, the need for
relevant tools has increased considerably in the last decades,
which is of broad importance in the regulatory framework for
the diagnosis of ecological impacts of chemicals (e.g. EU Water
Abbreviations: FR, feeding rate; FI, feeding inhibition index.
* Corresponding authors. Tel.: þ33 4 72208788; fax: þ33 4 78477875.
E-mail addresses: olivier.geffard@cemagref.fr (O. Geffard), arnaud.chaumot@cemagref.fr (A. Chaumot).
1Present address: Laboratoire d’Ecotoxicologie - Milieux Aquatiques (LEMA: EA 3222), Universite ´ du Havre, 76058 Le Havre Cedex,
France.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.09.035
journal homepage: www.elsevier.com/locate/watres
water research 45 (2011) 6417e6429

Page 2

Framework Directive, 2000/60/EC). Up to now, water quality
has been monitored using both chemical and biological
measures. Concerning biological measures, several biotic
indices have been developed. Because these methods referred
to changes in community structure, the established diagnosis
of ecosystem quality reflects integrative effects from diverse
sources of degradation. That is why the identification of
pressure/impacts relationships, is often difficult. To disen-
tangle the role of chemical contaminations in the degradation
ofenvironmentalquality,
consists of methods based on lower levels of biological orga-
nization for assessing biological impacts (Chapman, 2007;
Dagnino et al., 2008; Dama ´sio et al., 2008), e.g. measuring
sublethal responses of single species (Maltby et al., 2002).
These methods are expected to be more specific and sensitive
to the toxic effects of contaminants, and thus to supply early
warning indicators of pollution impacts. Nevertheless, the use
of individual responses still remains limited because their
interpretation under non-controlled environmental condi-
tions often lacks the definition of relevant reference values
(Maltby et al., 2002; Hagger et al., 2008).
Individual responses can supply ecologically relevant
endpoints because some of them constitute or can at least be
related to fitness traits (survival, reproduction, growth). In the
diagnosticcontext,they are
measurement of such physiological or demographic rates
necessitates the adaptation of laboratory bioassay protocols
tofieldexposure.Hence,
measurementswitheither
organisms (Soares et al., 2005; Galloway et al., 2006; Barata
et al., 2007; Krell et al., 2011), and protocols for in situ
measurements with caged organisms (Maltby and Crane,
1994; Dedourge-Geffard et al., 2009) are developed for physi-
ological rate and life-history trait measurements. Among the
individual responses which can be monitored, feeding inhi-
bition is of great interest for multi-scale assessment of water
quality. On one hand, it is an ecological concern because it can
be related to alteration in life-history traits (Maltby, 1999;
Baird et al., 2007; Barata et al., 2007) and because it can be
correlated with ecosystem processes (Forrow and Maltby,
2000; Maltby et al., 2002). On the other hand, its interpreta-
tion can be linked with the modulation of molecular
biomarkers of specific modes of action (Barata et al., 2007;
Xuereb et al., 2009b). In aquatic invertebrates, feeding inhibi-
tion is in most cases one of the first observed responses to
environmental pollution (Gerhardt, 1995; Macedo-Sousa et al.,
2007; Alonso et al., 2009; Mouneyrac et al., 2010). Since the
1990s, several laboratory studies have shown that the feeding
rate (FR) of amphipods (in particular freshwater gammarids)
can be inhibited by a large rangeof chemical stressors(metals,
insecticides, fungicides, herbicides, drugs, organic com-
pounds. see Suppl. Table 1A). Gammarus pulex (Linnaeus) and
Gammarus fossarum (Koch) are highly relevant as sentinel
species to study feeding inhibition in streams. They are
widespread in European ecosystems, where they play a key
role in nutrient cycles as decomposers of coarse organic
matter. By performing a short review of the literature since
1990, we noted that several studies showed in situ feeding
inhibitions in gammarids in various contamination profiles
(industrialwastes,acidmine
acomplementary approach
rarely usedbecausethe
protocols
indigenous
forpost-exposure
transplantedor
drainage, agricultural
catchments. see Suppl. Table 1B). Consequently, FR assess-
ment that can be easily measured in situ with caged gam-
marids (mainly by leaf-mass feeding assays), has been
proposed as an ecologically relevant in situ indicator of water
quality (Maltby et al., 2002).
The main limitation for the use of individual responses in
monitoring programs is the difficulty to define baseline values
due to spatial and seasonal variability related to the effects of
biotic and abiotic factors (Hagger et al., 2008; Hanson et al.,
2010). Such biotic and non-toxic environmental influences
could lead to the misinterpretation of individual markers in
water chemical quality assessment during in situ or post-
exposure assays with caged organisms (Maltby et al., 2002;
Moreira et al., 2006; Kater et al., 2001; Krell et al., 2011).
Indeed, the inflated variability of responses in controls results
in a decreased statistical power explaining a low sensitivity of
bioassays (i.e. high rate of false negatives). In addition, con-
founding effects could give rise to false-positive cases, when
deviation from controls is causedby a difference in the level of
anon-toxicinfluentialfactor
measurement in gammarids can be affected by many biotic
and abiotic factors. Biotic factors include source population
(Maltby and Crane, 1994; Veerasingham and Crane, 1992;
Crane et al., 1995), parasite load (McCahon et al., 1988;
Pascoe et al., 1995; Fielding et al., 2003; Lettini and Sukhdeo,
2010), or body size (Nilsson, 1974). With the aim to reduce
the variability related to these biotic factors, the use of
transplanted standard organisms is proposed for water
quality assessment (Liber et al., 2007) because it allows to play
down the impact of biotic factors (one population source,
same physiological parameters such as size, sex, reproductive
and energetic status).
The confounding effect of abiotic factors, which can not be
controlled during in situ exposure, has limited the application
of bioassays with transplanted organisms to paired compari-
sons between stations upstream/downstream from identified
point-source pollutions. In this specific context, the assess-
ment of chemical water quality strongly relies on a question-
able experimentaldesign
physicochemical conditions are similar between stations,
excepted for levels of bioavailable toxic compounds (Liber
et al., 2007). As an alternative, modelling the influence of
confounding factors can make measurements comparable in
space and time (Maltby et al., 2002; Moreira et al., 2006; Krell
et al., 2011). This could allow to benefit from robust refer-
ence conditions defined at larger scales of space and time. For
instance, through an empirical analysis of the influence of
environmental conditions (temperature, alkalinity,.) on FRs
in Gammarus, Maltby et al. (1990b, 2002) underlined that taking
into consideration the most influential environmental condi-
tions in order to define reference values of biological activities
could improve the in situ approach for site-specific studies.
Furthermore, such a methodological advance could permit
the application of FR in situ bioassays to large scale and long-
term biomonitoring programs.
The present study illustrates how modelling the influence
of confounding factors allows to improve the interpretation of
in situ feeding assays with the widespread keystone species G.
fossarum as an indicator of water quality. We proceeded in
four steps: (i) we quantified the influence of important
(i.e.lowspecificity).FR
which assumes that
water research 45 (2011) 6417e6429
6418

Page 3

confounding factors in laboratory conditions; (ii) based on
these laboratory results, we computed a feeding inhibition
index (FI ), which proved to be robust to environmental
conditions and allowed us to define a reference statistical
distribution of feeding activity values through the data
compilation of 24 in situ assays among diverse reference
stations at different seasons; (iii) we tested the sensitivity of
the feeding assay using this statistical framework by per-
forming 41 in situ deployments in contaminated stations
presenting a large range of contaminant profiles; and (iv) we
illustrated how the proposed methodology improved water
quality diagnosis in two site-specific studies of impacted
watersheds previously reported in the literature, which were
focused on the development of biomarkers (Dedourge-Geffard
et al., 2009; Lacaze et al., 2011).
2. Material and methods
2.1.
fossarum
Sampling and maintenance of transplanted G.
Organisms were collected by kick sampling at La Tour du Pin,
upstream of the Bourbre River (France). This station displayed
good water quality according to RNB data records (French
Watershed Biomonitoring Network), and a high density of
gammarids was found. The organisms were kept during a 15
days acclimatisation period in 30 L tanks under constant
aeration. They were continuously supplied with groundwater
mixed with osmosed water at constant conductivity, 200 or
600 mS cm?1, depending on the conductivity level of the
subsequent experimental environment (water hardness: 88.2
or 223.0 mg L?1of CaCO3, respectively). A 10/14 h light/dark
photoperiod was maintained and the temperature was kept at
12 (?1)?C. Organisms were fed ad libitum with alder leaves
(Alnus glutinosa), previously conditioned for at least 6 (?1) d in
groundwater. Twice a week, freeze-dried Tubifex sp. worms
were added as a dietary supplement.
2.2.
FR assays
Analyses of water physicochemical parameters were per-
formed for each experiment by a French accredited chemical
analysis laboratory (Laboratoire d’analyses physicochimiques
des milieux aquatiques, Cemagref, UR Milieux Aquatiques,
Ecologieet Pollutions). Temperature
measured using Tinytag temperature logger Aquatic 2?.
wascontinuously
2.2.1.
Because we chose to perform in situ FR assays through the
transplantation of standard organisms, we tried to improve
this methodology by analyzing, first, the influence of size of
selected organisms on the FR, and second, the influence of
main potential confounding environmental factors reported
in the literature (temperature and conductivity). The body size
of transplanted organisms was thereafter fixed for the in situ
bioassay protocol used for field experiments (Section 2.2.2).
During experiments 1 and 2, conductivity, temperature, pH,
and dissolved oxygen were monitored daily.
Laboratory exposure
Experiment 1: Influence of body size on FR. Three size classes of
gammarids - 7.3 (?0.5), 10.6 (?0.7) and 12.8 (?0.9) mm - were
considered for the experiment (water temperature: 12.1
(?0.01)
88.2 mg L?1of CaCO3). Here the body size corresponded to the
dorsal length between the start of the prosoma (at the base of
the antenna) and the end of the metasoma (thus excluding
urosoma and telson). For the first class, we selected juvenile
gammarids, and for the two others, adult male gammarids
were selected in order to exclude impacts of sex on the FR.
Four replicates of 20 gammarids were studied for each
condition. We used a flow-through system which consisted of
0.5 L glass beakers filled with continuously renewed water
(four renewals per day), a continuous pumping system, and
a 10/14 h light/dark photoperiod. 20 alder leaf discs (20 mm in
diameter, without major veins) were supplied in each beaker.
Two beakers, containing only leaf material, were deployed to
control leaf weight gain or loss unrelated to gammarid feeding
activity. After 7 days of exposure, gammarids were counted
(for survival rate assessment), alder leaf discs were collected,
and a new batch of leaf discs was placed in each beaker for the
second period of 7 days. At the end of the experiment, gam-
marids were counted again. The methodology for FR compu-
tation is described in Section 2.3.
Experiment 2: Influence of water temperature and conductivity
on FR. The influence of temperature and conductivity on FR
wasstudied byexposing
(10.6 ? 0.7 mm) to three temperatures: 6.9 (?0.05), 12.1 (?0.01)
and 16.4 (?0.4)
600 mS cm?1in a fully factorial design.These levels corre-
sponded to the range of physicochemical characteristics
usually encountered in the streams in Rho ˆne-Alpes region.
24 h before initiating the experiment, gammarids were accli-
matised to each water temperature and conductivity treat-
ment. We used the same experimental design as for
experiment 1 to conduct the exposure under the six
treatments.
?C; conductivity: 600 mS cm?1; water hardness:
adultmalegammarids
?C and two conductivity levels: 200 and
2.2.2.
In situ feeding assays were adapted according to the method
described by Maltby et al. (1990a,b). We deployed four
replicates of 20 adult male gammarids with homogenous
body size (10e11 mm) in stations presented in Tables 1AeD.
As in experiment 2, we used a size class, which corresponds
in mean to the first class of adults in experiment 1. This is
because it is the more numerous in collected samples from
the source population and thus it makes easier to constitute
homogenous replicates. Organisms were placed in poly-
propylene cylinders (diameter 5 cm, length 10 cm) capped at
their ends with pieces of net (mesh size: 1 mm). 20 alder leaf
discs (20 mm in diameter, without major veins) were
supplied in each container. Two containers with only leaf
material, were deployed at each station as a control. After 7
days of exposure, the gammarids were counted (for survival
rate assessment), and the alder leaf discs were collected.
The methodology for FR computation is described in
Section 2.3.
Experiment 3: In situ characterisation of FR variability among
reference stations. 24 deployments in reference stations
(detailedinTable 1A)were
In situ deployments
implementedduring two
water research 45 (2011) 6417e6429
6419

Page 4

campaigns in October 2009 (R1 to R12) and June 2010 (R13 to
R24). These stations were chosen among the national refer-
encenetwork (WFDimplementation) in collaboration with the
regional public water agency, which built this network by
expert judgement using data on land use, chemical moni-
toring (including micropollutants), and ecological diagnosis
(http://sierm.eaurmc.fr/eaux-superficielles). For our study,
stations were selected on rivers in Rho ˆne-Alpes region,
seeking to cover a large range of physicochemical character-
istics and geographical locations (w20000 km2) (Suppl.
Fig. 1A). Mean weekly water temperature ranged from 9.2?C
to 15.1?C, and conductivity from 110 mS cm?1to 420 mS cm?1
between the 24 deployments.
Experiment 4 : In situ exposure at contaminated stations. 41
deployments (detailed in Table 1B) were performed during the
same two campaigns in the same region (Suppl. Fig. 1A) as
experiment 3 in October 2009 (P1 to P15) and June 2010 (P16 to
P41). These stations were chosen among the national control
network (WFD implementation) in collaboration with the
regional water agency. They displayed depreciated water
chemical quality and faunistic indices and typology of
micropollutant contamination (Table 1B) was previously
established by the water agency (http://sierm.eaurmc.fr/eaux-
superficielles). For our study, stations were selected in order to
supply diverse contamination profiles (pesticides, metals,
urban). Mean weeklywater temperature varied between 11.8?C
and 20.7
868 mS cm?1.
?C, and conductivity between 80 mS cm?1and
Experiment 5: The Lot watershed. Four campaigns of in situ
caging were performed from November 2009 to June 2010 on
the Lot watershed in the Decazeville area. This river system
has been intensively studied during different scientific
programs (e.g. ANR 08-CES-014 RESYST) because of its poly-
metalic contamination due to former open-cast coal mining
and zinc ore treatment. Experiments were performed in
autumn, winter, spring and summer. Four stations were used
(Suppl. Fig. 1B), for which chemical characterisation, survival
and biomarker responses in caged G. fossarum were provided
in a previous report (Lacaze et al., 2011): i) one station
upstream of the Lot-Riou Mort confluence (Upstream Lot),
considered as a reference site for the studied water system
because of low metal concentrations in the water column; ii)
one station downstream of the Lot-Riou Mort confluence
(Downstream Lot); iii) one station on the Riou Mort river
upstream of the polymetallic contamination (Decazeville)
which was another metal-free site but located in an urban
area, iv) the fourth station (Riou Viou) is on the Riou Viou river,
a tributary of the Riou Mort river which presented significant
metal concentrations in the water column. Mean water
temperature varied between these 20 deployments from 6.2 to
17.4?C, and mean conductivity from 135 to 1552 mS cm?1
(detailed in Table 1C).
Experiment 6: The Amous watershed. In situ caging was
performed in March 2008 in four stations on the Amous
watershed (Suppl. Fig. 1C), a French river known to be highly
contaminated by heavy metals originating mainly from acid
Table 1A e Detailed information on in situ deployments for the different feeding assays considered in experiment 3: 24
deployments in reference stations in Rho ˆne-Alpes region.
Station GPS coordinates National grid
reference
Deployment
date (no.)
Water
temperature
(?C)
Conductivity
(mS cm?1)
Hardness
(mg L?1
of CaCO3)
River, location, (coding
letter on Suppl. Fig. 1)
Doux, Labatie d’Andaure (a)04?29041.5" E
45?01023.6" N
04?30011.9" E
45?10039.5" N
04?30036.4" E
45?26036.3" N
05?11020.0" E
45?48027.5" N
05?32031.8" E
45?57022.8" N
05?28023.7" E
45?58036.1" N
04?58052.0" E
45?26015.3" N
05?07050.3" E
45?15026.5" N
05?27055.5" E
45?12011.6" N
05?45017.4" E
45?21042.2" N
04?23039.3" E
45?51052.7" N
04?31015.9" E
46?11011.8" N
04?26045.5" E
46?08021.2" N
610556810/2009 (R1)
06/2010 (R13)
10/2009 (R2)
14.8
13.4
13.4
64
55
78
20
<14
21
Cance, Saint Julien Vocance (b)
6101905
Gier, La Valla en Gier (c)
682013810/2009 (R3)13.1 7821
Ain, Saint Maurice
de Gourdans (d)
Albarine, Chaley (e)
6092000 10/2009 (R4)15.0394 266
630000110/2009 (R5)11.9440316
Mandorne, Oncieux (f)
606965010/2009 (R6)
06/2010 (R14, R20)
10/2009 (R7)
06/2010 (R15, R21)
10/2009 (R8)
06/2010 (R16, R22)
10/2009 (R9)
11.9
10.9, 11.9
13.9
14.0, 15.0
13.5
13.8, 15.1
13.9
380
415
411
325
292
130
323
245
238
288
168
260
55
242
Vareze, Cours et Buis (g)
6820073
Galavayson, Saint Clair
sur Galaure (h)
Drevenne, Rovon (i)
6104900
6147220
Guiers Mort, Saint
Laurent du Pont (j)
Boussuivre, Saint
Marcel l’Eclaire ´ (k)
Ardie `res, Les Ardillats (l)
6078200 10/2009 (R10)
06/2010 (R17, R23)
06/2010 (R18, R24)
10.7
9.2, 10.1
12.9, 13.6
300
290
245
266
156
726580673
605137510/2009 (R11)12.8295 60
Ergues, Poule les
Echarmeaux (m)
605383010/2009 (R12)
06/2010 (R19)
12.6
14.2
150
110
75
28
water research 45 (2011) 6417e6429
6420

Page 5

Table 1B e Detailed information on in situ deployments for the different feeding assays considered in experiment 4: 41 deployments in contaminated stations in Rho ˆne-
Alpes region. Typology of contamination (D) established by the regional water agency for national control network (WFD implementation).
StationGPS
coordinates
National grid
reference
Deployment
date (no.)
MetalsPesticidesOther
contaminants
Water temperature
(?C)
Conductivity
(mS cm?1)
Hardness
(mg L?1of CaCO3)
River, location, (coding
letter on Suppl. Fig. 1)
Doux, Saint Jean de Muzols (n) 04?49039.5" E
45?04040.2" N
04?47047.6" E
45?11030.9" N
05?26001.8" E
45?56032.1" N
05?11048.4" E
46?06058.7" N
05?10031.3" E
46?07037.9" N
05?33005.3" E
46?10012.3" N
05?42004.3" E
45?11036.6" N
04?36009.1" E
45?50015.5" N
04?34021.4" E
45?54024.6" N
04?43033.1" E
45?54041.5" N
04?45042.3" E
45?35015.4" N
04?47003.4" E
45?35036.4" N
05?10029.9" E
45?04000.3" N
04?49057.3" E
45?47049.4" N
04?44000.9" E
46?07018.4" N
610603010/2009 (P1)
06/2010 (P16, P29)
10/2009 (P2)
06/2010 (P17, P30)
10/2009 (P3)
+++17.1
18.0, 18.7
14.8
16.9, 17.3
12.7
123
80
613
160
386
47
20
173
36
243
Cance, Sarras (o)
6103500+++
Albarine, Saint Rambert (p)
6300001+
Veyle, Lent (q)
6048570 10/2009 (P4)
06/2010 (P18, P31)
10/2009 (P5)
06/2010 (P19, P32)
10/2009 (P6)
++12.6
15.6, 16.0
14.0
17.3, 17.9
12.7
473
345
447
330
602
364
167
360
154
289
Veyle, Servas (r)
6049550+++
Ange, Brion (s)
6086100+++++
Drac, Fontaine (t)
614650010/2009 (P7)
06/2010 (P20, P33)
10/2009 (P8)
06/2010 (P21, P34)
10/2009 (P9)
06/2010 (P22, P35)
10/2009 (P10)
06/2010 (P23, P36)
10/2009 (P11)
06/2010 (P24, P37)
10/2009 (P12)
06/2010 (P25, P38)
10/2009 (P13)
06/2010 (P26, P39)
10/2009 (P14)
06/2010 (P27, P40)
10/2009 (P15)
06/2010 (P28, P41)
++++++ +++15.1
11.8, 12.1
13.6
17.2, 17.2
13.2
15.3, 15.3
15.8
18.2, 17.6
16.1
17.8, 17.0
19.1
17.4, 17.9
15.2
16.6, 16.7
18.8
20.7, 20.3
14.3
17.0, 16.6
317
240
868
465
307
265
664
430
377
240
434
390
720
650
702
490
270
160
191
118
242
116
187
99
364
151
275
75
280
180
348
331
156
218
153
53
Turdine, Arbresle (u)
6057200++
Azergues, Legny (v)
6800009 ++++
Azergues, Lucenay (w)
6057700+++ ++++
Gier, Givors (x)
6097000 +++ ++
Rho ˆne, Givors (y)
na +++ ++
Bourbre, Pont de Cheruy (z)
na+++
Sao ˆne, Ile Barbe (aa)
6059500++++++
Ardie `res, Saint Jean (ab)
6051550+++++++
water research 45 (2011) 6417e6429
6421

Page 6

mine drainage from the former lead and zinc mine at Car-
noule `s (Dedourge-Geffard et al., 2009). Four stations were
studied: three stations along the Amous river: Upstream
?1500 m, Downstream þ1200 m and Downstream þ3500 m,
withdifferentlevelsofmetallic
a maximum in Downstream þ1200 m); and a fourth refer-
ence station on a tributary from the same river system
(Tributary) that is not impacted by metal-loaded mine
leachates (see Dedourge-Geffard et al., 2009 for chemical
characterisation data). Mean water temperature varied
betweenstations from9.5
water conductivity from 520 to 600 mS cm?1(detailed in
Table 1D).
contamination(with
to 11.8
?C and mean
2.3.
FR computation
Leaf discs were numerically scanned using an Epson perfec-
tion 3490 PHOTO ? scanner after 7 d of exposure. The surfaces
of the discs were measured using SigmaScan ? Pro v5.0
imaging software(Systat Software).FR, expressed as
a consumed surface per day per gammarid (mm2.d?1.
organism?1), was calculated for each replicate as follows:
FRi¼
ðScontrol? SiÞ
??li;0þ li;t
where FRiis the feeding rate of replicate i; Scontrolthe total
surface of leaf discs present at the end of experiment in the
control without gammarids; Sithe total surface of leaf discs
present at the end of the experiment in replicate i; t is the
duration in days of the experiment (here t ¼ 7 days for all
assays); li,0and li,tare the number of living gammarids at the
start and at the end of experiment (here li,0¼ 20). We reported
on Suppl. Fig. 2 the relationship between surface and mass of
leaf discs after different levels of consumption to allow the
comparison with FR from the literature expressed as mg
consumed per day and per gammarid. In order to propose
a sufficiently reliable measure of FR for all the experiments (in
the laboratory or in situ), we decided to only consider situa-
tions when gammarid survival remained higher than 75%.
??2?? t
(1)
Table 1C e Detailed information on in situ deployments for the different feeding assays considered in experiment 5: Lot
watershed (2009/2010).
SeasonStation GPS
coordinates
Water
temperature (?C)
Conductivity
(mS cm?1)
Hardness
(mg L?1of CaCO3)
AutumnUpstream Lot02?14030.1" E
44?35053.5" N
02?11051.5" E
44?34049.7" N
02?14014.9" E
44?33038.1" N
02?12044.6" E
44?33010.1" N
10.7 14568
Downstream Lot
10.6158 77
Decazeville
11.11180 485
Riou Viou
9.2270120
WinterUpstream Lot
Downstream Lot
Decazeville
Riou Viou
Upstream Lot
Downstream Lot
Decazeville
Riou Viou
Upstream Lot
Downstream Lot
Decazeville
Riou Viou
6.4
6.2
8.2
7.7
12.1
12.2
15.8
12.1
14.9
15.2
17.4
15.6
162
180
717
215
135
177
1072
255
135
155
1530
207.5
75
76
256
88
38
67
556
79
55
67
894
85
Spring
Summer
Table 1D e Detailed information on in situ deployments forthe different feedingassays considered in experiment 6: Amous
watershed (March 2008).
StationGPS.coordinatesWater temperature (?C)Conductivity (mS cm?1) Hardness (mg L?1of CaCO3)
Tributary 03?59051.9" E
44?06013.1" N
03?59017.2" E
44?06035.5" N
03?59043.2" E
44?05057.7" N
03?58057.8" E
44?04030.6" N
9.45520308
Upstream-1500 m
11.80 525311
Downstream þ1200 m
10.25590 347
Downstream þ3500 m
11.65600351
water research 45 (2011) 6417e6429
6422

Page 7

2.4.Statistical analyses
Statistical procedures were carried out with the R software (R
Development Core Team, 2008). Normality and homosce-
dasticity were checked using ShapiroeWilk test and Bartlett
test, respectively. The influence of body size and exposure
week on FRs in experiment 1, and the influence of tempera-
ture, conductivity and exposure week on FRs in experiment 2
were tested using linear modelling (ANCOVA procedure)
including the interaction terms in full models. Among
explanatory variables, body size and temperature were
included as continuous, conductivity and exposure week as
categorical. In experiment 3, the between-deployment vari-
ability was also tested by linear modelling (ANOVA test). To
compare FR measurement between deployments during
experiments 4, 5 and 6, non-parametric testing was used
because of heteroscedasticity in FR distributions observed
between deployments in contaminated contexts. Further-
more, as in biomarker studies reported in Xuereb et al. (2009a,
2011), we built a reference distribution of in situ feeding levels
based on the measurements in the reference stations
(experiment 3). This was possible after removing the
between-deployment variability in FR measurements in
reference stations, by means of the computation of a feeding
inhibition index (FI ) (see Section 3.2). A normal distribution
was fitted to FI values from reference stations thanks to the
fitdistr function in the MASS R-package. Such a distribution
allowed us: (i) to compute a 5% confidence interval of FI
values expected in reference conditions; and (ii) to calculate
the likelihood that the replication of FI values measured in
a given in situ deployment could be observed under reference
conditions. A p-value associated to this likelihood was calcu-
lated using an empirical null distribution built by computing
the likelihoods of replications of FI values in 105theoretical
deployments simulated from the reference distribution.
3.Results
3.1.
on FRs in the laboratory
Influence of body size, temperature and conductivity
In experiment 1 (Fig. 1A), we did not observe any significant
differences in FR between the two successive weeks of
experiments (ANCOVA test: interaction term p ¼ 0.53, week
effect p ¼ 0.68). We noted that the feeding activity of gam-
marids increases with body size (ANOVA test, p < 10?14). This
influence was strong considering for instance that a deviation
from 10 to 11 mm in the mean body size of the 20 selected
organisms would give rise to a relative increase of 20% in FR
level.
In experiment 2 (Fig. 1B), no effect of conductivity and
exposure week on FR was detected (ANCOVA test: interaction
termsp>0.10,weekeffectp¼0.99,conductivityeffectp¼0.80).
We noted that the FR increases with temperature (ANOVA test,
p < 10?15). Therefore, for male gammarids measuring
10.6 ? 0.7 mm, we related FR, expressed as a consumed surface
pergammaridperday(mm2.day?1.organism?1),totemperature
T (?C) with a linear regression model (r2¼ 0.79, n ¼ 48):
FR ¼ 1:85 ? T þ 3:14
From this equation, it appeared for instance that an
increase in mean temperature from 12?C to 13?C would result
in an augmentation of 7.3% in FR.
(2)
3.2.
measurements among reference stations
Modelling the in situ variability of FR
Important
between the
temperature (from 9.2 to 15.1?C) or conductivity (from 110 to
420 mS cm?1) (details in Table 1A). We observed significant
differences for FR measurements between deployments
(ANOVA test, p ¼ 0.04) (Fig. 2A). Considering the findings from
laboratory experiment 2, we assumed that temperature was
the main determinant of this variability. This was also
confirmed a posteriori by the analysis of FR variance among the
24 deployments fitting conductivity and temperature as
physicochemical
24deployments,
differences
concerning
wereobserved
meanwater
Fig. 1 e Effects of body size, temperature and conductivity
on feeding rates (FR) in laboratory conditions: (A) FRs for
three body size classes of gammarids (7.3 (±0.5), 10.6 (±0.7)
and 12.8 (±0.9) mm); (B) FRs for three water temperatures
(6.9 (±0.05), 12.1 (±0.01) and 16.4 (±0.4)?C) and two levels of
conductivity (200 mS cmL1in black and 600 mS cmL1in
grey). Circle and triangle symbols correspond to first and
second weeks of experiment, respectively.
water research 45 (2011) 6417e6429
6423

Page 8

explaining factors (ANOVA test: conductivity effect p ¼ 0.49;
temperature effect p < 10?7). In order to take into account the
influence of temperature on FR, we calculated a feeding inhi-
bition index (FI) which was calculated as:
FI ¼
?FRpred ? FRobs
?
FRpred
? 100(3)
where FRobsis the feeding rate measured during the in situ
experiment, and FRpredis the feeding rate predicted with Eq.
(2) at the mean temperature during the in situ exposure. With
this index, we no longer detected significant differences in
feeding activity between the different deployments in refer-
ence stations (ANOVA test, p ¼ 0.12) (Fig. 2B). Fitting a normal
distribution to the calculated FI values led to a mean value of
FI, which is not significantly different from zero (Student test,
p ¼ 0.52). This demonstrates congruence between the in situ
measurements and the predicted values from the laboratory
experiments. This finding validated the assumption that
temperature was the main determinant of FR variability
between deployments. In addition, the robustness of FR
measurement was reinforced since we observed that not only
the mean value of FR was unchanged between laboratory and
in situ measurements, but also that the variability of FR was
not affected (see Suppl. Fig. 3). Using the normal distribution
of FI in the reference stations, we confirmed the good repli-
cability between the stations and the seasons. Indeed data
invariably fitted the reference distribution when the likeli-
hood of observed replicated FI values for each deployment
( p > 0.5 for all deployments, Fig. 2B) was tested. From these
findings, FI can be interpreted as a feeding inhibition index,
because it contrasts a given observed FR in one assay with the
expected value of feeding activity in a non-contaminated
context for the same temperature.
3.3.
at contaminated stations
Sensitivity of the FR assay during in situ exposure
Important
between the 41 deployments : mean temperature ranged from
11.8 to 20.7?C, conductivity from 80 to 868 mS cm?1(details in
Table 1B). We observed significant differences between the FR
(Fig. 2A, Kruskal & Wallis rank sum test, p < 10?10) and greater
variability compared to reference stations (experiment 3). We
assumed that this increase in the range of of observed FRs was
due on one hand to the possible influence of higher temper-
atures, and on the other hand, to possible inhibitions. In order
to remove temperature-induced variability in feeding activity,
notably for the comparison of contaminated and reference
stations (experiment 3), we calculated FI for all deployments
(Fig.2B).Thisrevealedsignificant
( p < 0.05) for 14 deployments (P7, P20, P21, P24, P25, P28, P29,
P30, P32, P33, P34, P37 and P41). We observed feeding
physicochemicaldifferences wereobserved
feeding inhibitions
A
B
Fig. 2 e In situ feeding assays in reference and contaminated stations in October 2009 (deployments R1-R12 and P1-P15) and
June 2010 (deployments R13-R24 and P16-P41): (A) Raw measurement of feeding rates (FR) (black points: mean for each
deployment; grey points: four replicates per deployment; segments: min-max range); (B) Temperature correction via the
computation of the feeding inhibition index (FI ) (see Eq. (3) in the text). The solid and dotted lines represent the mean and
the 95% confidence interval of the distribution of FI values assessed in deployments within reference stations. The
deployments with a significant deviation of FI values from this reference distribution are marked with stars (see Section 2.4
for the calculation of likelihood and p-value).
water research 45 (2011) 6417e6429
6424

Page 9

inhibition in 7% of the deployments in October 2009, and 54%
in June 2010. In addition, FI values did not exceed the upper
limit of the reference confidence interval, implying that the
higher values of the FR were mainly explained by higher
exposure temperatures.
3.4.Case studies
The Lot watershed. We first performed a paired comparison
(within each season) between FR measurements in contami-
nated stations and the reference station Upstream Lot
(Fig. 3A). Significantly lower FR values were observed from the
Decazeville station in winter, spring and summer, from the
Riou Viou station in autumn and from the Downstream lot
station in spring (unilateral Wilcoxon rank sum tests,
p < 0.05). The Riou Viou station also presented higher FR
values than the reference station in winter (unilateral Wil-
coxon rank sum tests, p < 0.05). The computation of FI values
(Fig. 3B) confirmed the significant feeding inhibition at the
Decazeville station in winter, spring and summer, and at the
Downstream lot station in spring when compared to the
reference distribution from experiment 3 ( p < 0.05). In
A
B
Fig. 3 e Site-specific surveys of Lot (four seasons) and Amous (spring) watersheds (reference stations are in bold): (A) Raw
measurement of feeding rates (FR) (same conventions as Fig. 2). Crosses represent stations with significant differences in
FRs in comparison with the reference station; (B) Temperature correction via the computation of the feeding inhibition index
(FI ) (see Eq. (3) in the text). The solid and dotted lines represent the mean and the 95% confidence interval of the distribution
of FI values assessed in deployments within reference stations from Fig. 2. The deployments with significant deviation of FI
values from this reference distribution are marked with stars (see Section 2.4 for the calculation of likelihood and p-value).
water research 45 (2011) 6417e6429
6425

Page 10

addition, FI values revealed significant feeding inhibitions at
the Decazeville station in autumn, as well as at the Down-
stream Lot station in winter. This constituted two false-
negative cases, i.e. inhibitions not detected with the FR.
These false-negatives would not have been uncovered if we
had simply compared FI to values from the only reference
station (Upstream Lot) for the same season (unilateral Wil-
coxon rank sum tests, p > 0.05). The computation of FI values
also revealed false-positive cases: the decrease of FR values
from the Riou Viou station in autumn and the increase in
winter were masked once temperature heterogeneity influ-
ence was taken into account with FIs. These false-positives
were eliminated either by the comparison with the reference
station or by the comparison with the reference distribution
from experiment 3. In addition, we observed that the FI
values from the reference Upstream Lot station corresponded
well with the reference distribution, and that the significant
seasonal variability in FR values for this station (Kruskal
Wallis rank sum test, p < 0.05) was negated once the FIs which
integrated temperature influence were considered (Kruskal
Wallis rank sum test, p ¼ 0.68).
The Amous watershed. Considering FR values (Fig. 3A), we
observed significantly lower feeding activities from the two
stations Upstream ?1500 m and Downstream þ1200 m in
comparison to FR measurements from the reference station
(Tributary) (unilateral Wilcoxon rank sum test, p < 0.05). Using
FI values (i.e. temperature corrected) (Fig. 3B), we still detected
significant feeding inhibition but only at the Downstream
þ1200 m station (unilateral Wilcoxon rank sum test, p < 0.05).
This pattern was supported when comparing the FIs with
either the measurements from the reference station (Tribu-
tary) or with the reference distribution of FI values (experi-
ment 3).
4.Discussion
4.1.
fossarum
Identification of influential factors on FR in caged G.
Despite their recognized importance, the influence of biotic
(including body size, source population or parasite load) and
abiotic (including dissolved oxygen concentration, alkalinity,
temperature or pH) factors on feeding activity of gammarids
has rarely been scrutinized. The influence of parasitism is one
of the most described biotic factor (McCahon et al., 1988;
Pascoe et al., 1995; Fielding et al., 2003; Lettini and Sukhdeo,
2010). For other factors, quantitative studies are scarcer.
Here, we showed a significant linear positive relationship
between FR values and organism size (Fig. 1A), agreeing with
the study of Nilsson (1974) on G. pulex. Blockwell et al. (1998),
on the contrary, did not find any significant differences in
feeding activity of G. pulex between juveniles of 5.1 and
7.0 mm. However, they measured feeding activity through the
consumption of Artemia salina eggs, which makes the
comparison with our results difficult, and their methodology
may have been less sensitive than leaf consumption methods.
The linear relationship reported on Fig. 1A contradicts the
theoretical prediction, which states that FR should be
proportional to length squared due to allometric constrains
(Kooijman, 2000). Such a parabolic relationship may be con-
cealed because we tested a too narrow range of body sizes.
Nevertheless, considering that the weight of organisms is
proportional to the length cubed, the positive relationship
between FR and length is consistent with negative correla-
tions reportedforamphipods
consumption and body weight (Nilsson, 1974; Sutcliffe et al.,
1981; Lozano et al., 2003). Furthermore, in Diporeia, the coef-
ficient of this exponential decrease has been quantified as
?0.84 ? 0.08 (Lozano et al., 2003) which is closer to our finding
(?2/3 is expected) than to a parabolic pattern (?1/3 is ex-
pected). The possible interaction between the effect of body
size on FR and abiotic factors such as temperature is reported
in some studies (Nilsson, 1974) but not in others (Lozano et al.,
2003). In ourstudy,we did nottestsuch aninteractionbecause
we chose to control the variability induced by biotic factors
thanks to the transplantation of standard organisms, using
male gammarids from a unique population, with no visible
parasites, with homogenous body size, and acclimatized in
the laboratory before in situ transplantation.
Regarding the influence of environmental conditions
(abiotic factors) with caged gammarids, Maltby et al. (2002)
showed that temperature heterogeneity explained 76% of
the between-deployment FR variation in reference stations.
Taking into account additional physicochemical variability
did not strongly increase the amount of explained variance
(less than 8%). Our results confirmed these findings. We
described a significant linear influence of temperature on
feeding activity (Fig. 1B), with a 50% reduction of the FR at 7?C
compared to 16?C. This was consistentwith results on G. pulex
(Nilsson, 1974) where the FR was dropped by 90% at 2
compared to 15?C under laboratory conditions. According to
the theoretical model of Arrhenius, we could expect an
exponentialrelationship between
because it is a physiological rate (Kooijman, 2000). This
pattern could have been described more precisely if we had
extended the range and the number of tested temperatures.
Nevertheless, considering the residual variability of FR values
(Fig. 1B), the minor difference which could be detected
between linear and exponential model should not improve
our prediction of temperature effect within the range of the
temperatures considered in our study, which already consti-
tutes a large range of conditions for in situ biotests with G.
fossarum. We did not detect any influence of conductivity on
FRs in the laboratory. This pattern of a major influence of
temperature on the FR was supported by in situ results in
different rivers, and during different seasons, with highly
contrasting physicochemical characteristics and geographical
locations. Indeed, removing temperature effects, erased
spatial and seasonal differences in measurements under
reference conditions (during experiment 3 and for the two
case studies, Figs. 2 and 3) and we observed an important
decrease in the variability of feeding activity in contaminated
stations (Fig. 2).
betweenweight-specific
?C
temperatureand FR,
4.2.
situ FR assay
Definition of reference values and accuracy of in
In order to limit false positives induced by confounding
environmental factors, one approach - implemented for some
water research 45 (2011) 6417e6429
6426

Page 11

biomarkers (Xuereb et al., 2009a; Hanson et al., 2010) - would
consist in defining a range of referencevalues by including the
whole annual and spatial variability observed in reference
stations (Hagger et al., 2008). This appears problematic for
individual responses such as FR, because of their high natural
variability (induced by temperature) which would lead to
a lack of statistical power and thus to difficulties in discrimi-
nating feeding inhibitions related to contaminations (Fig. 2A).
A current practical solution consists in using only local/
seasonal controls. As exemplified in our two case-studies
(Fig. 3), this is questionable since it does not fully prevent
confounding effects of temperature occurring even at small
spatial scales. In the Lot case-study, the deviation of the FR
between the Riou Viou and the reference station was entirely
explained by temperature heterogeneity within the water-
shed. In the Amous case-study, the same pattern was
observed between the Upstream ?1500 m and Tributary
stations. In addition, it appears that the interpretation of
bioassays according to local controls does not solve the
problematic lack of statistical power, not because of signifi-
cant variability in controls but due to the reduced number of
available control measurements (Hanson et al., 2010). For
example, in the Lot case-study, FI values measured from
Decazeville station in autumn and from the Downstream Lot
station in winter showed inhibition of feeding activity (as
during other seasons), but this was not revealed by the simple
comparison to the reference Upstream Lot station (Fig. 3).
As an alternative, modelling the influence of confounding
factors permits to correct observed FR and supplies compa-
rable feeding activities even in variable environmental
conditions (Maltby et al., 2002; Moreira et al., 2006; Krell et al.,
2011). We show here that this allows to take advantage of
robust reference conditions defined at large scale of space and
time. This permits a gain in the statistical power for the
detection of impacted feeding activities because of the
reduced variability of FI in controls, and because of the large
set of reference measurements which could be compiled to
definethedistributionofreferencevalues(Hansonetal.,2010).
Following the same approach as Moreira et al. (2006) and Krell
et al. (2011), we chose to model the influence of temperature
and conductivity in preliminary experimentsunder controlled
laboratory conditions. Maltby et al. (2002) modelled the influ-
ence of temperature and other environmental factors on
gammarids FR through the statistical analysis of results from
asetofinsitubioassaysinreferenceconditions.Inourcase,we
observed similar results with the two approaches (Suppl.
Fig. 3). Nevertheless, when the measurements of individual
responses are comparable between laboratory and in situ
conditions, the first approach would be preferred because it
should be more predictive (between regional contexts for
instance). Indeed, the latter data-driven approach may
confoundtheinfluencesofenvironmentalfactorswhichcould
be correlated only in the specific reference dataset.
For the assessment of the impact of an identified point-
source pollution, the definition of baseline values robust to
environmental variability also permits to qualify the value of
a local reference. For instance, in the Lot study, the Decaze-
ville station was first considered as a reference location to
evaluate the impact of the past mining activity because of its
metal-free water chemistry. But, it appears that urban
effluents highly impact FR in this station, consistent with the
alteration of water chemical quality revealed by the induction
of biomarkers of genotoxicity in this station (Lacaze et al.,
2011). Thus, as already proposed with laboratory controls for
post-exposure FR bioassays in Daphnia (Barata et al., 2007), in
such comparative approaches (e.g. upstream/downstream),
the use of independent reference benchmarks for feeding
activity could help to accurately assess any alteration of
chemical water quality.
4.3.
fossarum
Relevance of the in situ FR bioassay with G.
With the aim to integrate the complexity of field exposure to
contaminants in water quality assessment, bioassays with
gammarids appear as promising tools because leaf-mass
consumption methodology permits in situ measurements of
FR, while the majority of protocols with invertebrates are only
achievable by post-exposure measurement in laboratory
conditions: bivalves (Hagger et al., 2008), gastropods (Krell
et al., 2011), daphnids (Barata et al., 2007; Dama ´sio et al.,
2008), decapods (Moreira et al., 2006), chironomids and anne-
lids (Soares et al., 2005). In addition, these assays are not less
influenced by environmental conditions during exposure than
in situ FR measurements. The feeding assay with Gammarus
appeared sensitive to contaminants in multiple contexts
(agricultural, industrial, mining,.): in experiment 4, 37% of
the contaminated stations showed significant feeding inhibi-
tions. These results supported findings from previous labo-
ratory or field studies (Suppl. Table 1) that demonstrated
feeding inhibition with gammarids in response to a large
variety of environmental contaminants. The FI index allowed
us to compare impacts of contamination at different times.
We observed important seasonal variations of in situ impacts
on feeding activity. For instance in experiment 4, only 7% of
the stations showed significant feeding inhibitions in autumn,
whereas 54% of the stations were impacted in summer. Such
a seasonal variation is also reported by Maltby et al. (2002),
who explained this pattern by variation in waterflow.
Complementary hypotheses are variation in run-off and more
specifically seasonal treatments (pesticides) in agricultural
contexts. Because we used standard organisms, such varia-
tion in individual responses could not be understood by vari-
ation in susceptibility due to changes in the biological status
of tested organisms, as in the proposed interpretation of
bioassays performed with indigenous organisms (Hagger
et al., 2008). Thus, our approach, which makes short-term
tests comparable in time would facilitate the inclusion of
the seasonality of in situ toxicity for water quality assessment.
The transplantation of standard organisms only supplies
partial information to understand long-term chronic impacts
of contamination on higher integration levels of biological
organization (Liber et al., 2007). Nevertheless, feeding inhibi-
tion is of great interest for such multi-scale assessment. On
one hand FR is recognized as an ecologically relevant endpoint
because it can be related to alteration in life-history traits, and
mechanistic modelling is proposed to fill the gap between
feeding inhibition and drop in population dynamics in
particular with Gammarus (Baird et al., 2007). Because they are
keystone species in freshwater ecosystems, alteration in
water research 45 (2011) 6417e6429
6427

Page 12

population dynamics of Gammarus sp could be interpreted in
terms of alteration of ecosystem functioning (e.g. leaf
decomposition). On the other hand, feeding inhibition has
been correlated with the impact of contaminants with diverse
modes of actions traced by the modulation of specific molec-
ular biomarkers (Barata et al., 2007; Xuereb et al., 2009b). This
position of FR between biomarkers and fitness traits offers the
opportunity to describe adverse outcome pathways in multi-
scale assessment schemes (Kramer et al., 2011), which could
reinforce weight of evidence approaches for the diagnostic of
contaminant impacts on ecosystems (Dama ´sio et al., 2008). It
is illusory to think that all sources of uncertainty (standard vs
indigenousorganisms,within-population,
ulations, or between-species variability, intra and interspecies
interactions, food availability, .) could be taken into account
mechanistically in such a scheme which is based on infor-
mation obtained with standard organisms in specific experi-
mental conditions. Yet, it has already been shown that
reduction of feeding activities of transplanted Gammarus can
be correlated to reduction in leaf decomposition efficiency in
streams, regardless of the presence of Gammarus in indige-
nous communities (Forrow and Maltby, 2000). This underlines
the strong potential of FR in situ bioassay with Gammarus as an
ecologically relevant indicator of water quality.
between-pop-
5.Conclusions
We proposed an innovative protocol for an in situ feeding
assay based on the standardisation of FR measurements
through the combination of experimental and computational
methodologies (caging and statistical modelling). As it cor-
rected the confounding influence of temperature, which
appeared as the main environmental influence on in situ FR
values, our protocol permitted a more accurate assessment of
the alteration of feeding activity when between-station
comparisons in space and time were performed. A reliable
interpretation of our bioassay results was made feasible via
the comparison to a distribution of reference values. Such
a methodology increased the specificity and the sensitivity of
theassayinmultiplecontaminant,
seasonal contexts. It also enhanced the relevance of toxicity
assessment in site-specific studies by validating reference
station measurements. Lastly it could offer the possibility to
assess water quality in isolated stations as part of large scale
surveys, notably in non-point-source pollution contexts.
Further research will focus on the influence of source
population for FR measurements (within and between-species
variability), which could limit the development of such
bioassays for large scale (national, continental) biomonitoring
programs. Modelling approaches will also be developed with
G. fossarum to extrapolate such assessments of feeding inhi-
bitions to the potential impacts on population dynamics.
geographical, and
Acknowledgements
RC received financial grants from the Cluster Environnement
Re ´gion Rho ˆne-Alpes. The present work was partially funded
by the programs ANR 08-CES-014 RESYST, ANR ECCO-
ECODYN convention # 06CV050, and the French national
agency for water and aquatic ecosystems (ONEMA). We are
grateful to T. Pelte (Regional water agency Rho ˆne-Me ´di-
terrane ´e-Corse) and A. Tilghman (Cemagref) for providing
information from national river reference/control networks
for the selection of deployment stations. The authors thank A.
Tilghman for her critical reading of the English of the MS.
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.09.035.
r e f e r e n c e s
Alonso, A., De Lange, H.J., Peeters, E.T.H.M., 2009. Development of
a feeding behavioural bioassay using the freshwater
amphipod Gammarus pulex and the multispecies freshwater
biomonitor. Chemosphere 75 (3), 341e346.
Baird, D.J., Brown, S.S., Lagadic, L., Liess, M., Maltby, L., Moreira-
Santos, M., Schulz, R., Scott, G.I., 2007. In situ-based effects
measures: determining the ecological relevance of measured
responses. Integrated Environmental Assessment and
Management 3 (2), 259e267.
Barata, C., Damasio, J., Lopez, M.A., Kuster, M., de Alda, M.L.,
Barcelo, D., Riva, M.C., Raldua, D., 2007. Combined use of
biomarkers and in situ bioassays in daphnia magna to monitor
environmental hazards of pesticides in the field.
Environmental Toxicology and Chemistry 26 (2), 370e379.
Blockwell, S.J., Taylor, E.J., Jones, I., Pascoe, D., 1998. The influence
of fresh water pollutants and interaction with Asellus aquaticus
(L.) on the feeding activity of Gammarus pulex (L.). Archives of
Environmental Contamination and Toxicology 34 (1), 41e47.
Chapman, P.M., 2007. Determining when contamination is
pollution-weight of evidence determinations for sediments
and effluents. Environment International 33 (4), 492e501.
Crane, M., Delaney, P., Mainstone, C., Clarke, S., 1995.
Measurement by in situ bioassay of water quality in an
agricultural catchment. Water Research 29 (11), 2441e2448.
Dagnino, A., Sforzini, S., Dondero, F., Fenoglio, S., Bona, E.,
Jensen, J., Viarengo, A., 2008. A "Weight-of-Evidence"
approach for the integration of environmental "Triad" data to
assess ecological risk and biological vulnerability. Integrated
Environmental Assessment and Management 4 (3), 314e326.
Dama ´sio, J., Tauler, R., Teixido ´, E., Rieradevall, M., Prat, N.,
Riva, M.C., Soares, A.M.V.M., Barata, C., 2008. Combined use of
Daphnia magna in situ bioassays, biomarkers and biological
indices to diagnose and identify environmental pressures on
invertebrate communities in two Mediterranean urbanized
and industrialized rivers (NE Spain). Aquatic Toxicology 87 (4),
310e320.
Dedourge-Geffard, O., Palais, F., Biagianti-Risbourg, S., Geffard, O.,
Geffard, A., 2009. Effects of metals on feeding rate and
digestive enzymes in Gammarus fossarum: an in situ
experiment. Chemosphere 77 (11), 1569e1576.
Fielding, N.J., MacNeil, C., Dick, J.T.A., Elwood, R.W., Riddell, G.E.,
Dunn, A.M., 2003. Effects of the acanthocephalan parasite
Echinorhynchus truttae on the feeding ecology of Gammarus
pulex (Crustacea: Amphipoda). Journal of Zoology 261 (3),
321e325.
Forrow, D.M., Maltby, L., 2000. Toward a mechanistic
understanding of contaminant-induced changes in detritus
water research 45 (2011) 6417e6429
6428

Page 13

processing in streams: Direct and indirect effects on
detritivore feeding. Environmental Toxicology and Chemistry
19 (8), 2100e2106.
Galloway, T.S., Brown, R.J., Browne, M.A., Dissanayake, A.,
Lowe, D., Depledge, M.H., Jones, M.B., 2006. The ECOMAN
project: a novel approach to defining sustainable ecosystem
function. Marine Pollution Bulletin 53 (1e4), 186e194.
Gerhardt, A., 1995. Effect of Metals on Stream Invertebrates. Lund
University.
Hagger, J.A., Jones, M.B., Lowe, D., Leonard, D.R.P., Owen, R.,
Galloway,T.S.,2008.Applicationofbiomarkersforimprovingrisk
assessments of chemicals under the water framework directive:
a case study. Marine Pollution Bulletin 56 (6), 1111e1118.
Hanson, N., Fo ¨rlin, L., Larsson, A˚, 2010. Spatial and annual
variation to define the normal range of biological endpoints:
an example with biomarkers in perch. Environmental
Toxicology and Chemistry 29 (11), 2616e2624.
Kater, B.J., Postma, J.F., Dubbeldam, M., Prins, J.T.H.J., 2001.
Comparison of laboratory and in situ sediment bioassays using
Corophium volutator. Environmental Toxicology and
Chemistry 20 (6), 1291e1295.
Kooijman, S.A.L.M., 2000. Dynamic Energy and Mass Budgets in
Biological Systems. Cambridge University Press, Cambridge.
Kramer, V.J., Etterson, M.A., Hecker, M., Murphy, C.A.,
Roesijadi, G., Spade, D.J., Spromberg, J.A., Wang, M., Ankley, G.
T., 2011. Adverse outcome pathways and ecological risk
assessment: Bridging to population-level effects.
Environmental Toxicology and Chemistry 30 (1), 64e76.
Krell, B., Moreira-Santos, M., Ribeiro, R., 2011. An estuarine
mudsnail in situ toxicity assay based on post-exposure
feeding. Environmental Toxicology and Chemistry 30 (8),
1935e1942.
Lacaze, E., Devaux, A., Mons, R., Bony, S., Garric, J., Geffard, A.,
Geffard, O., 2011. DNA damage in caged amphipods Gammarus
fossarum: a tool for freshwater quality assessment.
Environmental Pollution 159 (6), 1682e1691.
Lettini, S.E., Sukhdeo, M.V.K., 2010. The energetic cost of
parasitism in isopods. Ecoscience 17 (1), 1e8.
Liber, K., Goodfellow, W., den Besten, P., Clements, W.,
Galloway, T., Gerhardt, A., Green, A., Simpson, S., 2007. In situ-
based effects measures: considerations for improving
methods and approaches. Integrated Environmental
Assessment and Management 3 (2), 246e258.
Lozano, S.J., Gedeon, M.L., Landrum, P.F., 2003. The effects of
temperature and organism size on the feeding rate and
modeled chemical accumulation in Diporeia spp. for Lake
Michigan sediments. Journal of Great Lakes Research 29 (1),
79e88.
Macedo-Sousa, J.A., Pestana, J.L.T., Gerhardt, A., Nogueira, A.J.A.,
Soares, A.M.V.M., 2007. Behavioural and feeding responses of
Echinogammarus meridionalis (Crustacea, Amphipoda) to acid
mine drainage. Chemosphere 67 (8), 1663e1670.
Maltby, L., 1999. Studying stress: the importance of organism-
level responses. Ecological Applications 9 (2), 431e440.
Maltby, L., Clayton, S.A., Wood, R.M., McLoughlin, N., 2002.
Evaluation of the Gammarus pulex in situ feeding assay as
a biomonitor of water quality: robustness, responsiveness,
and relevance. Environmental Toxicology and Chemistry 21
(2), 361e368.
Maltby, L., Crane, M., 1994. Responses of Gammarus pulex
(Amphipoda, crustacea) to metalliferous effluents:
identification of toxic components and the importance of
interpopulation variation. Environmental Pollution 84 (1),
45e52.
Maltby, L., Naylor, C., Calow, P., 1990a. Effect of stress on
a freshwater benthic detritivore: scope for growth in
Gammarus pulex. Ecotoxicology and Environmental Safety 19
(3), 285e291.
Maltby, L., Naylor, C., Calow, P., 1990b. Field deployment of
a scope for growth assay involving Gammarus pulex,
a freshwater benthic invertebrate. Ecotoxicology and
Environmental Safety 19 (3), 292e300.
McCahon, C.P., Brown, A.F., Pascoe, D., 1988. The effect of the
acanthocephalan Pomphorhynchus laevis (Muller 1776) on the
acute toxicity of cadmium to its intermediate host, the
amphipod Gammarus pulex (L.). Archives of Environmental
Contamination and Toxicology 17 (2), 239e243.
Moreira, S.M., Moreira-Santos, M., Guilhermino, L., Ribeiro, R.,
2006. An in situ postexposure feeding assay with Carcinus
maenas for estuarine sediment-overlying water toxicity
evaluations. Environmental Pollution 139 (2), 318e329.
Mouneyrac, C., Perrein-Ettajani, H., Amiard-Triquet, C., 2010.
Influence of anthropogenic stress on fitness and behaviour of
a key-species of estuarine ecosystems, the ragworm Nereis
diversicolor. Environmental Pollution 158 (1), 121e128.
Nilsson, L.M., 1974. Energy budget of a laboratory population of
Gammarus pulex (Amphipoda). Oikos 25 (1), 35e42.
Pascoe, D., Kedwards, T.J., Blockwell, S.J., Taylor, E.J., 1995.
Gammarus pulex (L.) feeding bioassay - Effects of parasitism.
Bulletin of Environmental Contamination and Toxicology 55
(4), 629e632.
R Development Core Team, 2008. R: a Language and Environment
for Statistical Computing. R Foundation for Statistical
Computing.
Soares, S., Moreira, S., Andrade, S., Guilhermino, L., Ribeiro, R.,
2005. Freshwater-saltwater interface and estuarine sediment
in situ assays based on post-exposure feeding of chironomids
and polychaetes. Estuaries 28 (2), 314e319.
Sutcliffe, D.W., Carrick, T.R., Willoughby, L.G., 1981. Effects of
diet, body size, age and temperature on growth rates in the
amphipod Gammarus pulex. Freshwater Biology 11 (2),
183e214.
Veerasingham, M., Crane, M., 1992. Impact of farm waste on
freshwater invertebrate abundance and the feeding rate of
Gammarus pulex L. Chemosphere 25 (6), 869e874.
Xuereb, B., Bezin, L., Chaumot, A., Budzinski, H., Tutundjian, R.,
Garric, J., Geffard, O., 2011. Vitellogenin-like gene expression
in freshwater amphipod Gammarus fossarum (Koch, 1835):
Characterization and potential of use as endocrine disruption
biomarker in males. Ecotoxicology 20 (6), 1286e1299.
Xuereb, B., Chaumot, A., Mons, R., Garric, J., Geffard, O., 2009a.
Acetylcholinesterase activity in Gammarus fossarum (Crustacea
Amphipoda). Intrinsic variability, reference levels, and a liable
tool for field surveys. Aquatic Toxicology 93 (4), 225e233.
Xuereb, B., Lefe `vre, E., Garric, J., Geffard, O., 2009b.
Acetylcholinesterase activity in Gammarus fossarum (Crustacea
Amphipoda): Linking AChE inhibition and behavioural
alteration. Aquatic Toxicology 94 (2), 114e122.
water research 45 (2011) 6417e6429
6429