Cryptic species diversity: an overlooked factor in environmental management?



  1. Molecular genetic methods continuously uncover cryptic lineages harboured by various species. However, from an applied perspective, it remains unclear whether and to which extent such a genetic diversity affects biological traits (e.g. ecological, behavioural and physiological characteristics) and environmental management.

  2. We assessed potential deviations regarding the trait ‘environmental stress tolerance’ using individuals from five field populations of each of two cryptic lineages (called A and B) comprised under the nominal species Gammarus fossarum. We used ammonia as a chemical stressor while assessing the feeding rate on leaf discs as a measure of sublethal response. In this context, we established a restriction fragment length polymorphism assay to allow a rapid identification of the lineages.

  3. We observed a biologically meaningful and statistically significant twofold higher overall tolerance of one cryptic lineage, lineage B, over the other. Confounding factors that may have the potential to influence the test results, such as life stage, sex, season of collection, parasitism, physiological status of organisms and upstream land-use patterns of the river catchments, were either controlled for or displayed only minor deviations between lineages.

  4. Synthesis and applications. The trait differences observed in the present study seem to be mainly explained by the considerable genetic differentiation between cryptic lineages of one nominal species. Although traits other than tolerance have been minimally investigated in this context, this study indicates implications in the reliability and quality of environmental monitoring and management if cryptic lineage complexes are ignored.


Individuals of a species are considered to be morphologically and genetically similar according to the morphological and biological species concepts, respectively (cf. Bickford et al. 2007). However, these assumptions seem to contradict two phenomena. First, when species with high levels of phenotypic plasticity are of great genetic similarity (Agrawal, Laforsch & Tollrian 1999) and secondly, where a species lack morphological differentiation despite considerable genetic divergence (Müller 2000). The latter describes the case of cryptic lineage complexes, incorrectly considered as nominal species, which are widespread throughout the biosphere and continuously reported for diverse taxonomic groups and biomes (Pfenninger & Schwenk 2007). The genetic divergence observed within cryptic lineage complexes may be the result of genetic drift due to reproductive isolation among lineages with manifestations of differences in biological rather than morphological traits (i.e. ecological, behavioural and physiological characteristics; cf. Westram 2011). Such groups of organisms may feature unique adaptations and evolutionary potential due to their specific evolutionary histories (Westram et al. 2011b). These evolved biological traits among lineages, like intraspecific competition and predator avoidance (cf. Cothran et al. 2013), make generalizations regarding their response to environmental change for a whole cryptic lineage complex (nominal species) inappropriate on the basis of one single lineage. This seems particularly alarming if different cryptic lineages are treated as one nominal species, thus ignoring their potential deviation in traits, in environmental management in a larger sense, inter alia ecological risk assessment, biomonitoring and conservation biology (Bálint et al. 2011; Feckler, Schulz & Bundschuh 2013). Identification of cryptic lineage complexes and disclosure of lineage-specific differences in biological traits are fundamental for the development of appropriate guidelines ensuring a sound and reliable environmental management (cf. Sattler et al. 2007).

The benthic freshwater amphipod Gammarus fossarum Koch (Crustacea: Amphipoda), a common model organism in the field of eco(toxico)logy (e.g. Dehedin, Piscart & Marmonier 2013) and water quality assessment (e.g. Stucki 2010), represents a cryptic lineage complex that harbours at least three lineages (Müller 2000). The diverse application of G. fossarum is mainly driven by (i) its widespread distribution (Karaman & Pinkster 1977), (ii) its ecological importance as predator as well as prey (Dahl 1998; Felten et al. 2008) and (iii) its key role in the ecosystem function of leaf litter breakdown (Dangles et al. 2004). Recent studies reported differences in biological traits between two of the three lineages, referred to as A and B. While Westram et al. (2011a) displayed higher infection rates by acanthocephalan parasites in lineage B, Feckler et al. (2012) uncovered a higher tolerance for lineage B towards two currently used plant protection products.

Since Feckler et al. (2012) lacked an assessment of individuals from multiple populations representing both lineages, we aimed at providing a more general and comprehensive picture regarding the hypothesized potential differences in the biological trait of ‘environmental stress tolerance’ within cryptic complexes, using G. fossarum as a model organism. Specifically, we (i) developed a polymerase chain reaction restriction fragment length polymorphism (RFLP) assay to allow rapid and cost-effective identification of cryptic G. fossarum lineages, (ii) exposed individuals from five populations exhibiting pure stands of each cryptic G. fossarum lineage (lineage A and B) to various concentrations of unionized ammonia (NH3), a known environmental stressor, whereas the alteration in the individual feeding rate on leaf material was used as a measure of sublethal response and (iii) jointly assessed differences in the relative feeding rate between each NH3 treatment and the respective stressor-free control mean – that is, effect sizes (ES) – using meta-analysis tools and checked for deviations between lineages. To judge their impact on the results of the present study, we additionally quantified potentially confounding factors, such as physiological status of gammarids (employing lipid reserves as proxy; Prato & Biandolino 2009) and upstream land-use patterns of the river catchments of the sampling sites (Adam et al. 2010).

Materials and methods

Identification of Gammarus fossarum lineages using RFLPs of the 16S ribosomal RNA gene: establishment and validation

We aligned reference sequences of the 16S ribosomal RNA (rRNA) gene for cryptic G. fossarum lineages (Accession Nos. AJ269587AJ269625 and JN900486JN900488) and Gammarus pulex (Accession Nos. AJ269626, AJ269627 and EF582877) using the ClustalW algorithm (Higgins, Thompson & Gibson 1996) implemented in geneious basic 5.5.6 (Kearse et al. 2012). We included G. pulex into the method development because of the potentially arising confusion based on their morphological similarity (Karaman & Pinkster 1977) and co-occurrence with G. fossarum (e.g. Englert et al. 2013). We identified restriction sites for endonucleases (BcuI, Eco147I, and MunI; Table 1; see Appendix S1, Supporting Information) that are specific for taxon discrimination for each cryptic lineage of G. fossarum included in the present study (lineages A and B; lineage C was excluded since no samples were available) and G. pulex using the restriction analysis tool available in chromaspro version 1.5 (Technelysium Pty Ltd 2009). For the validation of the RFLP assay, a total of 298 G. fossarum and 15 G. pulex from streams located in Central and Western Europe were collected according to their distribution range and preserved in pure ethanol (purity >99%; Fig. 1; Appendix S2, Table S1, Supporting information). After we identified individuals of both species to the species level by their morphological characteristics using taxonomic literature (Eggers & Martens 2001), we extracted total DNA using a modified salt-extraction method (Appendix S3, Supporting information). Using a NanoDrop 1000 (NanoDrop products, Wilmington, Delaware, USA), we measured the amount of extracted DNA and prepared working stocks with a final concentration of 5–10 ng DNA μL−1. We amplified a 429- to 447-base-pair-long fragment of the 16S rRNA gene as described in Feckler et al. (2012) with one modification: we set annealing temperature to 51 °C instead of 49 °C to improve amplification specificity. The subsequent RFLP reaction contained a total volume of 15 μL, including 5 μL PCR amplicon, 1 unit endonuclease (either BcuI, Eco147I, or MunI), 1·5 μL buffer (endonucleases and buffers both Fermentas GmbH, St. Leon-Rot, Germany) and 8·4 μL sterile ddH2O. We let the digestion of PCR amplicons run for 4 h at 37 °C. To exclude false-positive (=non-specific restriction digestions) and false-negative results (=failure of restriction digestions at the specific site) that may lead to an incorrect assignment of specimens to taxa, we tested each endonuclease for all individuals. Afterwards, we mixed 10 μL of restricted fragments with 5 μL of a loading buffer (0·25% bromophenol blue and 40% sucrose; both w : v) and separated them on 2% agarose gels (w : v; using 1× TBE buffer: 10·8 g tris base, 5·5 g boric acid and 4 mL 0·5 M EDTA pH 8·0 L−1) at 100 V for c. 1 h. We simultaneously separated a commercial size marker (100 bp-DNA-ladder extended) to compare fragment lengths. Subsequently, we stained gels in a 0·01% ethidium bromide – 1× TBE solution (v : v) for c. 30 min, rinsed them briefly in clean 1× TBE buffer and visualized DNA bands using UV light (c. 302 nm). We additionally validated the established RFLP assay by sequencing the same fragment for a subsample of animals used for restriction analysis (= 2–3 depending on the availability of individuals; Appendix S2, Table S1, Supporting information) of all analysed populations. For this purpose, PCR amplicons were commercially sequenced by SeqIT (Kaiserslautern, Germany) on a 3730 DNA Analyzer eight capillary sequencer (Applied Biosystems, MA, USA) in the forward direction. We manually proofread the obtained sequences using geneious basic 5.5.6 (Kearse et al. 2012) and subjected them to similarity analysis using the nucleotide Basic Local Alignment Search Tool algorithm at the National Center for Biotechnology Information afterwards. If not further specified, we obtained chemicals either from Roth (Karlsruhe, Germany) or Sigma-Aldrich (Seelze, Germany).

Table 1. Diagnostic RFLP patterns of the assessed 16S rRNA gene using the endonucleases BcuI, Eco147I and MunI. Approximate restricted fragment sizes are given in base pairs. Dashes are shown if no restriction was observed. For the sample size of each taxon, see Table S1 in Appendix S2 of the Supporting Information
Gammarus fossarum lineage Ac. 131 + 152 + 153c. 73 + 363
G. fossarum lineage Bc. 87 + 349
Gammarus pulex c. 103 + 326c. 83 + 364
Figure 1.

Sampling sites for cryptic Gammarus fossarum lineages A (open circles) and B (filled circles) and Gammarus pulex (open squares) in Europe.

Quantification of the stressor

We chose NH3 as model substance since it is common in streams as a result of anthropogenic pollution by ammonium (math formula), in terms of sewage effluents, land-use of habitats surrounding the stream and farm waste discharge (cf. Alonso & Camargo 2004). math formula is converted to NH3 at higher pH levels (Berenzen, Schulz & Liess 2001). Its toxicity is thereby mainly mediated by an increased haemolymph pH of aquatic invertebrates and damage of their respiratory surfaces (Colt & Armstrong 1981). We prepared test solutions with specific math formula concentrations by diluting ammonium chloride (NH4Cl) in artificial test medium (=SAM-5S; Borgmann et al. 1998) immediately prior to the start of each feeding trial to achieve nominal test concentrations of 1·5, 15·0 and 37·5 mg math formula L−1. These concentrations correspond to nominal concentrations of 0·06, 0·6 and 1·5 mg NH3 L−1 under the test conditions of 20 °C and a pH of c. 8 (cf. for conversion BMU 1996). Furthermore, we included a stressor-free control (=pure SAM-5S). math formula concentrations in the field range from 0·3 to 30 mg L−1 (Berenzen, Schulz & Liess 2001), corresponding to NH3 concentrations of c. 0·008–0·858 mg L−1 for pH values and temperatures reported for European lowland streams (cf. Schäfer et al. 2007; Englert et al. 2013; van Vliet et al. 2013). Therefore, at least the two lowest nominal test concentrations selected in this study (0·06 and 0·6 mg L−1) are considered as environmentally relevant.

Following DIN 38406 E5-1 (1983), we performed chemical analysis of NH4+-N, with modifications for the use of a microwell plate reader (Tecan Infinite® M200, Tecan Group, Männedorf, Switzerland). Briefly, we took samples of all treatments at the start and the termination of each feeding trial and stored them at −20 °C until further use. After thawing at 20 °C, we measured pH and temperature of all samples as the most important factors influencing the aqueous NH3 equilibrium (Emerson et al. 1975). Afterwards, we added 0·4 mL of a salicylate citrate solution (130 g sodium salicylate, 130 g trisodium citrate dehydrate and 0·97 g disodium pentacyanonitrosylferrate (II) dehydrate per litre) to 4 mL of each sample and briefly vortexted them. Subsequently, we added 0·4 mL of a reagent solution (32 g sodium hydroxide and 2 g sodium dichloroisocyanurate per litre) and again mixed samples by vortexing. We transferred a volume of 80 μL of each sample to a 96-well microwell plate (TC MicroWell 96F SI W/Lid Nunclon D, Nunc, Wiesbaden, Germany) and incubated them for 1 h at 25 °C in the plate reader. After incubation, we measured the absorbance at 655 nm. We directly quantified mass concentrations of NH4+-N in the samples by using a standard curve prepared with an NH4Cl stock solution ranging from 0·1 to 0·8 mg N L−1 and calculated the corresponding math formula concentrations in the samples (BMU 1996). If nominal concentrations were higher than the maximum concentration used for calibration, we diluted the sample accordingly. We converted the outcome into corresponding NH3 concentrations under given test conditions and calculated time-weighted mean concentrations (i.e. 0, 0·072, 0·493 and 1·292 mg NH3 L−1; see Appendix S4, Table S2, Supporting information), which we refer to throughout the present study.

Origin and maintenance of tested organisms

We collected G. fossarum to be used during feeding trials by kick sampling from c. 200-m-long stretches at ten sites in Southern Germany (Fig. 1; Appendix S2, Table S1, Supporting information) on a rotating basis for the two lineages during autumn and winter 2011/2012. We conducted sampling at least 1 week prior to the start of the respective feeding trial to allow animals to acclimate to laboratory conditions and the test medium, and to limit possible differences in their response towards the stressor due to distinct acclimation to characteristics of the sampling sites. In this context, we additionally compared the water quality parameters among sampling sites, uncovering statistically non-significant differences (Appendix S5, Tables S3 and S4, Supporting information).

To control for the confounding effects of gammarids' life stage and parasitism by acanthocephalan parasites on the outcome of the present study (Williams, Green & Pascoe 1984; Pascoe et al. 1995), we assorted organisms immediately after sampling. At first, we used a passive underwater separation technique (Franke 1977) to divide gammarids into different size classes. We exclusively used adult animals with a cephalothorax length of c. 1·2–1·6 mm for the feeding trials. Afterwards, we removed individuals either evidently infected by acanthocephalan parasites or obviously ovigerous since they may display lower tolerances towards chemical stress (McCahon & Pascoe 1988; Pascoe et al. 1995), although this procedure may have affected the sex ratio in our experiment. We kept the test organisms at 20 ± 1 °C and fed them ad libitum with pre-conditioned black alder leaves (Alnus glutinosa Gaertn.) until the start of the respective feeding trial. Since the physiological status of organisms may affect their tolerance towards toxicants (Prato & Biandolino 2009), we measured the initial lipid content of gammarids. This procedure avoided a possible bias for the measurement of the physiological status by an exposure towards NH3. Therefore, five randomly selected individuals per field population were frozen in liquid nitrogen prior to the start of each feeding trial and stored at −75 °C until further processing.

Conditioning of leaf discs

We conditioned leaf discs as described in Bundschuh, Zubrod and Schulz (2011). Briefly, we collected leaves of A. glutinosa from a group of trees near Landau, Germany (N 49°12′01″; E 8°05′40″), in autumn 2010 shortly before abscission and stored them at −20 °C until further use. We cut leaf discs with a diameter of 20 mm using a cork borer after thawing and subsequently conditioned these leaf discs for 10 days in aerated nutrient medium (Dang, Chauvet & Gessner 2005), together with black alder leaves previously exposed in the Rodenbach, Germany (N 49°33′59″; E 8°02′33″), to establish a near-natural microbial community inter alia consisting of fungi and bacteria. These micro-organisms alter the leaves (physically and chemically) and, thereby, modify their palatability and nutritional value for leaf-shredding organisms (Bärlocher 1985). After conditioning, we dried the leaf discs at 60 °C and weighed them to the nearest 0·01 mg to ensure an accurate measurement of the amphipods' feeding rate (Maltby et al. 2002). We resoaked pre-weighed leaf discs in SAM-5S for 48 h prior to the start of the respective feeding trial to avoid floating at the surface of the test medium.

Feeding trials

Altogether we performed ten independent feeding trials (one per sampling site; = 5 per cryptic lineage; = 15 per concentration). For each replicate, we placed three randomly selected gammarids and four pre-weighed, resoaked leaf discs in a 250-mL glass beaker containing 200 mL SAM-5S and the appropriate concentration of NH3 for 5 days. We used both sexes of G. fossarum in the present study, since differences in the tolerance towards chemical stress between male and non-ovigerous female gammarids are unlikely (Malbouisson, Young & Bark 1995). We randomly assigned fifteen replicates for each treatment (i.e. 0, 0·072, 0·493 and 1·292 mg NH3 L−1) to positions in a climate-controlled chamber at 20 ± 1 °C in complete darkness. Since G. fossarum requires a high level of oxygen (Meijering 1991), we aerated replicates to avoid additional stress. We considered leaf mass loss driven by microbial activity and abiotic factors by five additional replicates per treatment without gammarids that we treated likewise. After the test duration, we removed surviving gammarids, remaining leaf material and any leaf tissue shredded off, dried them to constant weight at 60 °C and weighed them to the nearest 0·01 mg. In case replicates contained dead or fewer than three animals as a consequence of cannibalism, we discarded those from the subsequent calculation of gammarids' feeding rate (for detailed information on mortalities see Appendix S8, Table S7, Supporting information).

Quantification of energy reserves

We quantified energy reserves by examining lipid content as detailed in Zubrod et al. (2011), whereas we used only peroxymonosulphuric acid and 1 : 1 chloroform : methanol solution (v : v) washed glassware to avoid a bias due to lipid residues. Briefly, we transferred freeze-dried and weighed gammarids to 0·5 mL of a 1 : 1 chloroform : methanol (v : v) solution for 72 h. Afterwards, we homogenated organisms, centrifuged the homogenate in the short spin mode and transferred the supernatant into a sterile culture tube. After the solvent evaporated at 95 °C, we added 0·2 mL of sulphuric acid (≥ 95%), boiled samples at 95 °C for 10 min and afterwards cooled them down to room temperature. Subsequently, we added 5 mL of a vanillin–phosphoric acid reagent, incubated samples for 5 min and transferred 80 μL of each sample to a 96-well microwell plate. We measured the absorbance at 490 nm in a microwell plate reader and determined the lipid content per gammarid directly from a calibration curve, which we generated by using commercial soya bean oil (Sojola Soja-Öl, Vendemoortele, Dresden, Germany) at appropriate concentrations. Finally, we normalized the lipid content to gammarids' dry weight and expressed it as μg lipids per mg dry weight of G. fossarum to exclude size effects.

Upstream land-use of river catchments

For the characterization of the upstream land-use of river catchments, we used digital images from Google Earth (Google Inc. 2012) which we georeferenced and digitalized using ESRI® ArcMap™ 9.3 (ESRI Inc. 2008). For both riparian zones, we modelled a 500-m-wide buffer zone for a length of 1000 m upstream of each sampling site. Afterwards, we quantified percentage proportions of the different land-usages (agriculture, forestry, stagnant water bodies and urban land-use) within the buffer zone using the implemented clip function. We combined the land-usage types present into a single factor per sampling site, referred to as land-usage index (LUI; compare derivation of habitat degradation scores in von der Ohe & Goedkoop 2013). Therefore, we ranked each type of land-usage based on a score system, which relies on its percentage proportion: not present = 0, 0·1–5·0% = 1, 5·1–12·5% = 2, 12·6–31·3% = 3, 31·4–78·1% = 4 and ≥78·2% = 5. Thereby, we weighted land-usage types expected not to influence the populations' tolerance towards stressors (forestry and stagnant water bodies) positive, while those hypothesized to affect the tolerance (agriculture and urban land-use) were weighted negative (Liess et al. 2001; Zhao & Newman 2006; Jansen et al. 2011). Finally, we summed up single land-usage type scores to calculate LUIs:

LUI = forestry + stagnant water bodies − agriculture − urban land-use.

Data analysis

Given the shortcomings of conventional null hypothesis significance testing (Nakagawa & Cuthill 2007), we applied unpaired, two-sided 95% confidence intervals (CIs) as an alternative for statistical significance testing (Altman et al. 2000). If the 95% CI of a difference between means or medians did not include zero (the value reflecting no effect), we judged the test outcome as statistically significant.

At first, we calculated absolute feeding rates of gammarids according to Maltby et al. (2002), expressed as mg consumed dry leaf material per mg dry weight of G. fossarum per day. Afterwards, we calculated the relative feeding rate of each treatment expressed in per cent relative to the absolute feeding rate in the corresponding control, while we set the latter to 100%.

To give an overview of each treatment's effect on the assessed cryptic lineages, we conducted a meta-analysis based on a fixed effect model (Borenstein et al. 2009). We preferred meta-analysis over conventional statistical significance testing, as it has the advantage of combining the outcome of all independent experiments based on the observed difference in relative feeding between each NH3 treatment and the respective stressor-free control mean (=effect size; ES) and hence does not conflict with the above-mentioned criticism on null hypothesis significance testing (Nakagawa & Cuthill 2007). Therefore, we calculated raw (unstandardized) mean differences in relative feeding rates between the stressor-free control and respective NH3 treatments separately for each sampling site. In order to rescale this original data, we divided raw mean differences by the associated within-groups' standard deviation to obtain Cohen's d as a measure of ES since it can cope with differences in the response variable and associated variability (cf. Borenstein et al. 2009). Finally, we combined these standardized mean ES for each combination of lineage and NH3 concentration to provide combined ES (‘overall effects’) together with their corresponding 95% CIs (Borenstein et al. 2009). To uncover statistically significant differences in tolerance, we compared combined ES and associated standard errors between both cryptic G. fossarum lineages at each NH3 concentration separately using the ‘comped’-function (‘drc’-package). Again, we judged the test outcome to be statistically significant if the 95% CIs of combined ES and median differences among them, respectively, did not include zero.

We checked the impact of land-usage by using Pearson's product moment correlation coefficient (ρ). For this purpose, we calculated site-specific ES for each sampling site separately using meta-analysis that combines the results of all treatments obtained with gammarids from the respective sampling site. Afterwards, we correlated these site-specific ES against the respective LUIs, both combined for all sampling sites and separately for sampling sites of each cryptic lineage. For all statistical analyses and figures, we used r for Windows version 2.13.1 (R Development Core Team 2013), supplemented by the extension package ‘drc’ version 2.2-9 (Ritz & Streibig 2005).


Identification of Gammarus fossarum lineages using RFLPs of the 16S rRNA gene

Restriction digestion of PCR amplicons resulted in the predicted RFLP profiles as each endonuclease showed the expected band pattern for each taxon (Table 1; Appendix S1, Fig. S1, Supporting information). This was consistent both for cryptic G. fossarum lineages and G. pulex.

Feeding trials

Results of the meta-analysis showed that feeding by individuals of cryptic lineage A decreased with increasing NH3 concentrations. At 0·493 mg NH3 L−1 (combined ES: 0·56; 95% CI 0·18 to 0·94; = 5, while n in each of the comparisons represents up to 15 independent replicates; Fig. 2a) and 1·292 mg NH3 L−1 (combined ES: 0·87; 95% CI 0·47 to 1·28; = 5; Fig. 2a), feeding rates were impaired. This was shown to be statistically significant compared to the corresponding controls. However, at 0·072 mg NH3 L−1, only a statistically non-significant tendency for a decline in feeding was found (combined ES: 0·33; 95% CI −0·07 to 0·72; = 5; Fig. 2a). For cryptic lineage B, a statistically significant impairment in the feeding rate was detected at the highest test concentration of 1·292 mg NH3 L−1 (combined ES: 1·21; 95% CI 0·79 to 1·62; = 5; Fig. 2a), whereas at the lower test concentrations, slightly increased feeding rates were observed that were not statistically significantly different from the control (0·072 mg NH3 L−1: combined ES: −0·30; 95% CI −0·66 to 0·07; = 5; 0·493 mg NH3 L−1: combined ES: −0·09; 95% CI −0·46 to 0·27; = 5; Fig. 2a).

Figure 2.

(a) Combined effect sizes (ES; as Cohen's d; ±95% CI) of Gammarus fossarum separated by their affiliation to cryptic lineage A (open circles) and lineage B (filled circles) at the respective ammonia concentration (each = 5, while each n represents up to 15 independent replicates). A positive combined ES displays an impaired feeding at the respective ammonia concentration compared with the corresponding control. (b) Median differences (±95% CI) in combined ES of the two assessed cryptic lineages at each test concentration. Positive median differences indicate a higher tolerance of cryptic lineage B compared to lineage A. Deviations in combined ES between treatments and the corresponding controls and median differences between them, respectively, are considered to be statistically significant when zero is not included within the 95% CI, which is denoted by asterisks.

The comparison of combined effect sizes at each NH3 level between the two cryptic lineages displayed statistically significant deviations at the two lowest test concentrations, namely 0·072 mg NH3 L−1 (median difference of combined ES: 62·5; 95% CI 8·6 to 116·4; Fig. 2b) and 0·493 mg NH3 L−1 (median difference of combined ES: 65·5; 95% CI 12·9 to 118·1; Fig. 2b). However, no statistically significant deviation was observed at 1·292 mg NH3 L−1 (median difference of combined ES: −33·5; 95% CI −91·7 to 24·7; Fig. 2b). Finally for completeness, we performed complementary analyses (for results see Appendix S6, Table S5, Supporting information), which did confirm the statistically significant deviation in tolerance between cryptic G. fossarum lineages A and B to be rather a ‘lineage effect’ than a ‘population effect’.

Quantification of energy reserves

Prior to the start of the feeding trials, the two cryptic lineages sampled from five field populations exhibited, with a difference of c. 5%, no statistically significant deviation in their lipid content (absolute difference of means: 11 μg mg gammarid−1; 95% CI −12 to 34; = 25 with five individuals per sampling site; Appendix S7, Fig. S3, Supporting information).

Upstream land-use of river catchments

The analysis of the land-use patterns upstream of the sites where gammarids were collected for the feeding trials were dominated by agriculture and forestry usage. These data were independent from the lineage considered (Table 2). LUI did not correlate significantly with the site-specific ES, calculated for both cryptic lineages separately (Pearson's ρ = 0·49; 95% CI −0·69 to 0·96 for lineage A; Pearson's ρ = 0·1; 95% CI −0·86 to 0·90 for lineage B) or combined (Pearson's ρ = 0·11; 95% CI −0·56 to 0·69).

Table 2. Means (±95% CIs; = 5) of percentage proportions of land-use types and the respective ranges of land-usage indices (LUIs) at the sampling sites of Gammarus fossarum lineages A and B, respectively
Type of land-useLineage ALineage B
Agriculture54 (±37)38 (±42)
Forestry38 (±44)52 (±53)
Stagnant water body0·02 (±0·35)0·04 (±0·07)
Urban land-use8 (±9)10 (±14)
LUI−5 to +2−6 to +4


Identification of Gammarus fossarum lineages using RFLPs of the 16S rRNA gene

Since identification of lineage(s) within the G. fossarum complex is commonly based on sequencing methods (Müller 2000; Westram et al. 2011b) that are rather cost-intensive and time-consuming (cf. Russell et al. 2000), we established a RFLP assay as a relatively fast and cost-efficient alternative. Results confirmed the RFLP assay to be a rapid and reliable tool for routine identification of cryptic lineage(s) within the G. fossarum complex since RFLP patterns were identical for individuals of the same cryptic lineage originating from different sampling sites (cf. Grünig et al. 2004; Alam et al. 2007). We additionally demonstrated its applicability for discriminating between the two species G. fossarum and G. pulex. We further supported the method's robustness by comparing DNA sequences of a subsample of each population (Appendix S2, Table S1, Supporting information) with reference sequences deposited at the National Center for Biotechnology Information. These similarity analyses revealed that the RFLP assays correctly assigned individuals to one of the cryptic lineages.

Differences in tolerance between cryptic lineages

According to former studies assessing effects of NH3 on Gammarus (e.g. Dehedin, Piscart & Marmonier 2013), we expected an impact of NH3 on the feeding rate of both assessed cryptic G. fossarum lineages. However, based on the results of a previous study (Feckler et al. 2012), we furthermore hypothesized a divergence in lineages' tolerances. In effect, our results supported the pattern of lineage-specific tolerance with cryptic lineage B being more tolerant compared to lineage A. Moreover, our complementary analyses (namely the ratio of squared mean differences in relative feeding of populations between and within lineages; see Appendix S6, Supporting information) suggest the observations are rather a ‘lineage effect’ than a ‘population effect’, although there was – particularly pronounced for lineage B – variability among populations in each lineage. More precisely, while individuals of lineage A showed a statistically significant decline in the feeding rate already at the second lowest test concentration of 0·493 mg NH3 L−1, we observed impairment in the feeding rate of lineage B only in the highest treatment (Fig. 2a). Additionally, in contrast to a constant decline in feeding of lineage A, lineage B showed a relative increase in feeding at the two lower test concentration (Fig. 2a; Table S6, Supporting information). An increased shredding activity of Gammarus at lower levels of stress was already observed in former studies (e.g. Zubrod, Bundschuh & Schulz 2010) and may be attributed to a compensatory behaviour, aiming at optimizing the energy input to maintain basic physiological processes (cf. Rasmussen et al. 2012). The lack of such a compensatory behaviour in lineage A may partly explain the difference between lineages observed in the present study, while the relevance of this mechanism needs to be assessed in further detail.

Nonetheless, we discovered a factor of >2 between tolerances of the two cryptic lineages with individuals of G. fossarum lineage B being more tolerant compared to cryptic lineage A at the environmentally relevant test concentrations (cf. Berenzen, Schulz & Liess 2001; Fig. 2b). These data were derived by comparison of relative ES at respective concentrations (cf. Appendix S8, Table S6, Supporting information).

However, at the highest (not environmentally relevant) test concentration, the deviation in tolerance appeared to be minor and statistically non-significant (Fig. 2b). This is in accordance with previous results from another study (Feckler et al. 2012) where plant protection products, namely the insecticide thiacloprid and the fungicide tebuconazole, were used as stressors. Feckler et al. (2012) found the tolerance to vary up to a factor of 6 between lineages, with statistically significant deviations prevalent at the lower test concentrations. The non-existence of statistically significant deviations between both lineages at higher concentrations in Feckler et al. (2012) as well as the present study may be explained by the substantial effect of the chemical stressors assessed (cf. Appendix S8, Supporting information), superposing any difference in lineage-specific tolerance.

The consideration of only one population for each cryptic lineage in the previous study by Feckler et al. (2012), however, questioned the general applicability of these findings. As we assessed the tolerance of five populations for both lineages towards three concentrations of NH3 (each = 15) during the present study, which we combined in the final meta-analysis, the outcomes we presented here allow for a more general interpretation: Selection and genetic drift most likely caused the considerable genetic differentiation observed between the examined cryptic lineages (uncorrected genetic p-distance of 13%; Feckler et al. 2012), which is only marginally lower than between described species. This may be seen as the driving factor for the differences in tolerance towards ammonia.

This hypothesis is supported both by a study of Sturmbauer et al. (1999), reporting lineage-specific deviations of up to a factor of eight among cryptic lineages of the Tubifex tubifex complex regarding cadmium toxicity, and by the complementary analyses of the present study: we either controlled for potentially confounding factors that may affect organisms' tolerance towards chemical stress prior to or during the experiments or investigated them additionally. We controlled for implications of life stage, ovigerous females and parasitism, which may all influence gammarids' tolerance (Williams, Green & Pascoe 1984; McCahon & Pascoe 1988; Pascoe et al. 1995), either by actively selecting only individuals of a certain size class or excluding those specimens evidently ovigerous or parasitized. Moreover, we conducted specimens' sampling (and hence testing) on a rotating basis for the two lineages during autumn and winter, reducing the conceivable impact of seasonality on the outcome. Also the physiological status of the lineages, indicated by the lipid content, did not differ (neither statistically significantly nor substantially) at the beginning of the respective feeding trials (Appendix S7, Fig. S3, Supporting information). Furthermore, water quality parameters at the sampling sites only showed marginal differences that were not statistically significant (Appendix S5, Tables S3 and S4, Supporting information). In addition to the above, these three factors, the seasonal variation in tolerance, physiological status and/or pre-exposure towards math formula (Dehedin, Piscart & Marmonier 2013), suggest to be of minor importance for the present study's outcome. Finally, anthropogenic land-use of upstream river catchments and associated release of pollutants into the surface water body is assumed to alter the tolerance of organisms (Liess et al. 2001; Zhao & Newman 2006; Jansen et al. 2011). However, both the descriptive comparison of respective types of land-use (Table 2) and the statistically non-significant correlation between site-specific ES and respective LUIs suggested also this factor to be unimportant for the outcome of the present study.

Conclusions and management implications

Although, we solely focused on differences with regards to one trait during the present study, namely tolerance towards one single stressor, cryptic lineages may show deviations regarding further biological traits as well (cf. Cothran et al. 2013) with meaningful consequences on environmental management in a larger sense:

  1. Lineages' capabilities to cope with alterations in environmental conditions, such as those predicted for global climate change (Intergovernmental Panel on Climate Change 2007), may be considerably affected. The distribution pattern of species harbouring cryptic lineage complexes may be altered in a way that a more tolerant lineage replaces the less tolerant lineage and act as an invasive species exhibiting high morphological similarity to the native one. However, given the lack of knowledge regarding the differentiation in further behavioural endpoints frequently involved in the displacement of local species, such as intraguild predation, competition, dispersal ability and cannibalism (cf. Dick, Montgomery & Elwood 1993; Westram 2011), a firm conclusion on ecosystem implications of potentially invasive cryptic lineages cannot yet be drawn. Nevertheless, environmental monitoring data, which are usually based on morphological characteristics of the species, should be carefully interpreted (e.g. in terms of the invasion status of a water body). Additionally, biomonitoring used for water quality assessment, commonly based on invertebrate metrics like taxa richness and density, should proactively consider cryptic species diversity as the occurrence of cryptic lineages displaying a high tolerance towards pollution may possibly mask low water quality and lead to deceptive interpretations of the ecological status of ecosystems.
  2. Regarding conservation strategies for species known to harbour cryptic lineage complexes, the concept of evolutionary significant units may be preferred since it prioritizes to define conservation units below the species level (Fraser & Bernatchez 2001). This acknowledges cryptic lineages as genetically distinct but morphologically alike subunits of nominal species with specific requirements for conservation guidelines. Otherwise, cryptic lineages may be lost due to inadequate conservation strategies resulting in reduced evolutionary potential of nominal species that eliminates ongoing diversification processes with impact on future biodiversity (as reviewed in Bálint et al. 2011). This is of particular concern for endangered species. However, this implies further taxonomic and ecological research as only a sound knowledge on lineages' status and capabilities to cope with environmental changes will facilitate the development of successful conservation strategies (Sattler et al. 2007).
  3. The distinct impacts of chemical stressors observed within cryptic lineage complexes contradict the common assumption that species' tolerance only varies around one mean. Cryptic lineages hence pose a deficiency for ecological risk assessment of chemicals, depending on the tolerance of the assessed lineage(s) (cf. Feckler, Schulz & Bundschuh 2013). The protectiveness of ecological risk assessment may thereby be influenced in a way that safety factors, commonly applied to accommodate uncertainty in data (Chapman, Fairbrother & Brown 1998), may be diminished (up to a factor of 6) due to deviations in tolerance between cryptic lineages, which were demonstrated at field relevant levels of chemical stressors. It may hence be questionable whether the uncertainties intended to be addressed by the safety factors (e.g. interspecies extrapolations and laboratory-to-field extrapolations; cf. Chapman, Fairbrother & Brown 1998) are still adequately covered. Thus, further taxonomic and ecotoxicological knowledge needs to be generated, quantifying the extent of tolerance deviations within cryptic complexes to reliably judge their importance for regulatory purposes.


We thank the Fix-Stiftung, Landau, for financial support of research infrastructure as well as F. Altermatt, J. Arce Funck, J. Augusiak, B. Ganser, W. Graf, R. B. Schäfer, A. Schmidt-Kloiber and P. van den Brink for kindly providing Gammarus specimens for the establishment of the RFLP assay. R. B. Schäfer is acknowledged for statistical advice, T. Bürgi, D. Englert, M. Köster, B. Lösch, R. R. Rosenfeld and F. Seitz for laboratory assistance. At last, grateful acknowledgement goes to E. R. Bennett for his valuable comments during the peer review stage of this manuscript. J. P. Zubrod received funding through a scholarship of the German Federal Environmental Foundation (Deutsche Bundesstiftung Umwelt).