Novel insights into the diet of the Pyrenean desman (Galemys pyrenaicus) using next-generation sequencing molecular analyses

Abstract

The Pyrenean desman, a threatened, semiaquatic mammal, is considered a specialist predator feeding on aquatic benthic invertebrates. This categorization comes from visual identification of prey in scat or gut contents, often based on a limited number of samples and locations. We combined diet analyses using next-generation sequencing methods with an extensive survey to explore the summer diet of Pyrenean desmans across the French Pyrenees. This study thus provides an unprecedented level of detail on the trophic ecology of Pyrenean desmans. Our results revealed a diverse diet containing a high proportion of rare prey and substantial consumption of terrestrial prey, which suggests a more generalist diet than previously understood. Three diet groups were identified, with significant differences in prey composition. These differences were not related to geographic location, but rather to local environmental variables. The spatial variation in diet was likely induced by local abiotic parameters that affect prey availability or use of foraging habitats.

foraging habitat, scat analysis, semiaquatic mammal, spatial structure

Knowledge of how species interact with their environment and with other organisms is required to evaluate species vulnerability (Rodrigues et al. 2006). Biotic interactions such as competition, predation, and trophic resource availability are known to play a central role in the distribution of species and their abundance (Boulangeat et al. 2012; Kissling et al. 2012) and to influence their responses to changing environments at different scales (Pearson and Dawson 2003; Araújo and Luoto 2007; Wisz et al. 2013; Belmaker et al. 2015). For instance, the decline of the Iberian lynx, Lynx pardinus, the world’s most threatened felid, was reported to be caused by the decline of its staple prey, the European rabbit, Oryctolagus cuniculus (Real et al. 2009), due to changes in environment and diseases (e.g., myxomatosis). In studies focused on single species, biotic interactions have often been overlooked due to sparse data and the difficulty of quantifying and incorporating complex relationships between organisms (Soberón and Peterson 2005). This is particularly true when the species of interest is rare and elusive, and thus difficult to study.

The Pyrenean desman, Galemys pyrenaicus (E. Geoffroy Saint-Hilaire, 1811), is a small, semiaquatic mammal (Talpidae) endemic to the Pyrenees Mountains (France, Spain, and Andorra) and the Iberian Peninsula (northern and central Spain, northern Portugal). The species is listed as vulnerable by the IUCN (Fernandes et al. 2008) and is legally protected in the 4 countries encompassing its distributional area. For a long time, the Pyrenean desman remained relatively unstudied, while also suffering a substantial decline across its range (Fernandes et al. 2008; Gisbert and García-Perea 2014; Charbonnel et al. 2016). Although recent studies have improved general knowledge about this species (e.g., habitat—Charbonnel et al. 2015 and Biffi et al. 2016; space use—Melero et al. 2012, 2014; population genetics—Gillet et al. 2015, 2016), a lack of information on biotic interactions, both with predators and trophic resources, remains to be supplemented.

To study the diet of small mammals, morphological identification of prey remains has been widely used (e.g., Castién and Gosálbez 1999; Churchfield and Rychlik 2006). With these methods, the Pyrenean desman was considered a specialist predator that fed on benthic invertebrates in streams (mainly Trichoptera, Ephemeroptera, Plecoptera, or Diptera—e.g., Santamarina and Guitian 1988; Santamarina 1993; Bertrand 1994; Castién and Gosálbez 1995).

Even though traditional methods enable the identification of a wide variety of prey, they are becoming less and less popular because 1) analyses of gut contents often involve the destruction of animals, 2) degraded prey remains cannot be identified with precision (rarely beyond the family level), 3) they underestimate the consumption of prey that are thoroughly masticated (e.g., arthropods consumed by bats) and soft-bodied prey (e.g., molluscs and earthworms) that leave no hard remains after digestion, and 4) they are time-consuming and require expert knowledge of the range of potential prey, and the ability to identify prey taxa based on small fragments of bone, hair, feathers, scales, exoskeletons, or other nondigested parts, particularly in analyses of scats (feces). The recent development of molecular analyses based on barcoding of fecal DNA fragments and next-generation sequencing has overcome these difficulties (see a review by Pompanon et al. 2012).

In that context, the first aim of this study was to improve knowledge of the trophic ecology of the Pyrenean desman by combining next-generation molecular analyses and extensive sampling covering the species range throughout the French Pyrenees. Gillet (2015) recently identified 3 genetic populations of Pyrenean desmans using different habitats (Biffi et al. 2016) in the 3 main hydrographic regions of the area. The second aim of the study was to compare the diet of Pyrenean desmans in those 3 regions. We hypothesized that the diet would vary according to the 3 hydrographic regions of the French Pyrenees, which differ in terms of catchment surface, climate, elevation, and type of land cover.

Materials and Methods

Study area

The Pyrenees Mountains are located in southwestern Europe and are a natural barrier between France and Spain. This study focused on the French part of the Pyrenees (W1°400–E3°100, N43°080–N42°230; Fig. 1), which extends for about 400 km from the Bay of Biscay to the Mediterranean Sea and has a maximum elevation of 3,298 m. The stream network can be divided into 3 main hydrographic regions: 1) the western streams that flow mostly westward to the Atlantic Ocean (coastal streams and Adour catchment; 9,412 km² in the Pyrenees), 2) the central streams that flow mainly northward and form the upstream part of the Garonne river catchment (7,702 km² in the Pyrenees), and 3) the eastern streams that flow to the Mediterranean Sea (Aude, Tech, and Têt catchments; 6,773 km² in the Pyrenees).

Fig. 1.

Location of the study area in France. The 3 hydrographic regions of interest are shown in light gray (western region: coastal streams and Adour catchment), medium gray (central region: Garonne river catchment), and dark gray (eastern region: Aude, Tech, and Têt catchments). A) Location of all sampling sites (dark dots). B) Location of sampling sites according to the 3 summer diet groups for the Pyrenean desman (Galemys pyrenaicus).

Feces collection

A national survey was conducted from 2011 to 2013 within the framework of the Conservation Action Plan for the Pyrenean desman (Némoz et al. 2011). Given the elusive behavior of this species, searches for indirect signs (i.e., feces) were conducted in 1,330 sites covering the entire French Pyrenees (Fig. 1A). Skilled observers meticulously inspected each emergent item (i.e., rock, tree root, and branch) along 500-m riverbed transects (Castién and Gosálbez 1992; Bertrand 1994; Aymerich and Gosálbez 2002; Charbonnel et al. 2014, 2015; Biffi et al. 2016). From the 1,330 original sites, 989 feces suspected of being left by a Pyrenean desman, based on their color, size, smell, and position, were collected in 958 sites and preserved in absolute ethanol for further molecular analyses.

Molecular analyses

DNA was extracted from feces using the Stool Mini Kit (Qiagen Inc., Hilden, Germany), following the manufacturer’s instructions. PCR amplifications and Ion Torrent PGM sequencing (Life Technologies, Carlsbad, California) were duplicated on the 989 fecal samples following the tagging and multiplexing method developed by Galan et al. (2012). Briefly, a 133 bp mini-barcode of the cytochrome oxidase I gene (COI) was amplified for each sample using a modified forward primer LepF1 (Hebert et al. 2004): 5′-CCATCTCATCCCTGCGTGTC TCCGACTCAGNNNNNNNATTCHACDAAYCAY AARGAYATYGG-3′, and a modified reverse primer EPT-long-univR (Hajibabaei et al. 2011): 5′-CCTCTCTATGGGCAG TCGGTGATNNNNNNNACTATAAAARAAAATYT DAYAAADGCRTG-3′. The 5′ parts of the primers were modified by the addition of individual-specific MIDs (Multiplex IDentifiers NNNNNNN), consisting of a short 7 bp sequence and adaptors required for emPCR and Ion Torrent sequencing. By using a combination of different forward and reverse MIDs sequences, several hundred samples can be multiplexed in the same sequencing run, and the sequences can be recognized after sequencing, when all the PCR products from the different samples are mixed together (Gillet et al. 2015).

PCRs were carried out in a 10-µl reaction volume using 5 µl of 2× QIAGEN Multiplex Kit (Qiagen), 0.5 µM of each primer (LepF1 and EPT-long-univR, concentrated at 10 µM), and 1 µl of DNA extract. The PCR conditions consisted of an initial denaturation step at 95°C for 15 min, followed by 40 cycles of denaturation at 94°C for 30 s, annealing at 45°C for 45 s, and extension at 72°C for 30 s, followed by a final extension step at 72°C for 10 min. After pooling all PCR products at 5 ng/µl, the amplicon libraries were sequenced by the company Genes Diffusion (Douai, France) on an Ion Torrent PGM system using an Ion 316 Chip version 2 (Life Technologies).

In addition, a customized database of COI sequences was built from 84 typical invertebrate species from the study area collected in some Pyrenean rivers and identified by local entomologist experts from the Conservatoire d’Espaces Naturels Midi-Pyrénées (CEN MP) and the EcoLab laboratory of Toulouse. A 710 bp fragment of COI was amplified in these samples with universal primers LCO1490 and HCO2198 (Folmer et al. 1994), following the PCR conditions reported in Folmer et al. (1994).

The sequences were sorted using SESAME barcode software (SEquence Sorter and AMplicon Explorer—Piry et al. 2012). These sequences were compared with sequences available in the customized and BOLD databases (Ratnasingham and Hebert 2007). Sequences that had a unique best-hit with an identity score greater than or equal to 98% were considered to be positive matches and allowed identification of the predator producing the feces as well as the prey contained in them.

Taxa were validated as possibly occurring in France and in the Pyrenees region using the French National Inventory of Natural Heritage (Muséum National d’Histoire Naturelle 2003–2017), the French Office for Insects and their Environment (OPIE-Benthos 2017) online databases, and the help of local experts. Taxa identified as endemic to other parts of the world were kept in the analysis and are designated by an asterisk hereafter (*), as they correspond more likely to a genetically similar taxon present in the Pyrenees but not available in the reference databases.

The frequency of occurrence (i.e., the number of feces containing the taxon divided by the total number of Pyrenean desman feces) in the diet of Pyrenean desmans was then calculated for each order, family, and genus, and for the different modes of life of prey taxa (i.e., exclusively aquatic, exclusively terrestrial, or with aquatic and terrestrial stages).

Summer diet of Pyrenean desmans

To investigate potential spatial structure in the diet of Pyrenean desmans, 287 Pyrenean desman feces collected at 115 sampling sites were kept for statistical analysis. This strong reduction in the number of feces compared to the initial pool of feces (n = 989) was due to drastic selection following 3 criteria: 1) the predator producing the feces was identified by molecular analyses, 2) the prey contained in the feces were identified by molecular analyses, and 3) the feces were collected during summer (i.e., June to September) in order to prevent potential strong seasonal variation. The summer season was chosen as most samplings were conducted during this low-flow period between 2011 and 2013. Prey occurrences were not different between the 3 summers of sampling, as revealed by a nonparametric permutation-based multivariate analysis of variance (PERMANOVA—Anderson 2001), which was not significant (P = 0.12). To correct for potential sampling bias (i.e., more than 1 feces collected per site), all the feces collected in a single site were pooled. Consequently, a taxon was assumed to be present in the diet of Pyrenean desmans at a site if it was found in at least 50% of the feces collected for this site. All prey taxa were kept for subsequent analysis.

Cluster analysis and identification of summer diet groups

—The binary matrix (i.e., presence–absence of prey genera at each site) was converted to a distance matrix calculated with the Sǿrensen similarity coefficient, which is the equivalent of Bray-Curtis distance but for presence–absence instead of abundance data (Borcard et al. 2011). Then, a hierarchical ascendant clustering was computed with Ward’s algorithm to build a dendrogram representing the distance between each pair of sites according to their similarity in the summer diet of Pyrenean desmans (i.e., prey taxa genetically identified in feces). The number of diet groups was chosen according to the dendrogram, so that it increases the variation between groups and decreases the variation within groups while keeping a relatively balanced number of sites within groups.

The mean composition of prey communities within sites was compared among the 3 diet groups using a PERMANOVA (Anderson 2001). As PERMANOVA may be sensitive to within-group effects, PERMDISP (Anderson et al. 2006), a permutation dispersion analysis, was then used to test for differences in within-group dissimilarity (i.e., the variability of diet composition within the groups as the mean distance of diet compositions to their group centroid). When the PERMDISP test was significant, a pairwise Tukey HSD test was run to examine which of the diet groups had higher dispersion. The Sǿrensen matrix was used as the measure of similarity in both PERMANOVA and PERMDISP.

To identify the prey taxa specific to each diet group, an indicator value (IndVal) analysis was performed (Dufrêne and Legendre 1997; Cáceres and Legendre 2009) using the occurrence matrix of the summer prey taxa and the classification of sites in diet groups obtained through the hierarchical clustering. The IndVal is the product of the specificity (i.e., the probability that the survey site belongs to the target site group, given the fact that the species has been found) and the fidelity (i.e., the probability of finding the species in sites belonging to the site group) of species for each cluster. A taxon was only able to be a potential indicator of 1 diet group, as group combinations were not allowed. The IndVal statistical significance was tested using a permutation test (9,999 permutations) for each prey taxon.

Finally, a chi-squared test was performed to test whether the assignment of sites to the 3 diet groups according to prey community composition found in feces was dependent on their location in the 3 hydrographic regions of the French Pyrenees.

Comparison of environmental parameters among diet groups

—To investigate potential local constraints that could influence the diet of Pyrenean desmans during the summer, a linear discriminant analysis (LDA) was performed on a set of 12 environmental variables that described local habitat conditions at the site (Biffi et al. 2016) and reach scales (i.e., stream reaches of approximately 1 km long—Charbonnel et al. 2016) in each diet group (see Table 1 for a complete description). These variables were expected to influence most invertebrate communities in mountain streams (Tachet et al. 2000; Usseglio-Polatera et al. 2000). This multivariate statistical method searches for linear combinations of quantitative variables (i.e., environmental variables) that maximize intergroup variance (i.e., diet groups). Significance of the LDA was tested using a Monte-Carlo test (9,999 permutations).

Table 1.

Environmental variables (Biffi et al. 2016; Charbonnel et al. 2016) expected to influence the prey composition of diet groups of Pyrenean desmans (Galemys pyrenaicus) in the French Pyrenees.

Scale	Category	Code	Description
Site	Riverbed	SHELTER	Proportion (%) of the heterogeneity at the site comprised of shelters and substrate (i.e., presence of emerging items and cavities, diversity of substrate types, and banks components) as a proxy for riverbed heterogeneity
		CLOG	Proportion (%) of the surface of streambed rocks covered by fine sediment (silt and mud)
		SHEET	Proportion (%) of the site with nonturbulent fast water units of shallow water that flows uniformly over smooth bedrock
		RUN	Proportion (%) of the site with nonturbulent fast water units of shallow gradient with relatively fine substrata ranging from sand to cobble
		RIFFLE	Proportion (%) of the site with turbulent fast water units with rapid and shallow flow with steep water surface gradient
	Riverbanks	BANKS	Shannon’s diversity index derived from the proportion (%) of banks type components along the site (DIRT, ROOTS, ROCKS)
		WOOD	Proportion (%) of bankside with shrubby-woody vegetation
	Floodplain	HUM_IMP	Proportion (%) of urbanized land within the site floodplain as a proxy for human impacts
Reach	Climate	TEM	Atmospheric mean annual temperature (°C)
	Hydrology	FLO	Mean monthly stream flow (m3/s)
	Hydrography	SLO	Mean slope of the river section (%)
		TRI	Number of tributaries in the focal section and in its adjacent upstream and downstream sections

All statistical analyses were conducted in R 3.1.1 (R Development Core Team 2014) using the ade4, vegan, and indicspecies packages.

Results

Molecular identification of the predator producing the feces and contents.

After the 2 PCR amplifications, a total of 9,489,679 reads were obtained, among which 2,962,289 were correctly assigned to 560 of the 989 analyzed samples (57%), after removing singleton sequences that were likely to be PCR or sequencing errors from the dataset. Among them, 390 samples (2,695,260 reads; 90%) belonged to the Pyrenean desman, whereas the remaining 170 samples (267,029 reads) were assigned to 25 other host species, including 7 birds (e.g., Turdus) and 14 mammals (e.g., Neomys sp., Glis glis). Among the 390 samples assigned to Pyrenean desmans, 383 samples also provided information on prey content (representing 37% of the reads), including 287 desman feces collected during the summer.

The proportion of items obtained twice in the PCR duplicates for the same DNA extract was computed as the percentage of repeatability. Repeatability of the results reached 99% when only host species were considered, 83% when only prey species were considered, and 87% when all species were considered.

Overall diversity of prey contained in the feces of Pyrenean desmans.

The 383 Pyrenean desman feces collected between 2011 and 2013 contained prey belonging to 11 classes, 30 orders (Table 2), 91 families, and 156 genera (Supplementary Data SD1). A mean of 5.8 ± 2.0 genera were present per feces (5.1 ± 1.9 and 3.7 ± 1.2 for family and order, respectively). Among the 156 genera, 100 were confirmed to be present in the Pyrenees, 31 in France, and 21 were endemic to other parts of the World (e.g., New Zealand, Australia) and were thus misidentified by genetic databases. The presence in the Pyrenees of the 4 additional genera remains unknown. For 1 taxon, databases were not able to discriminate between 2 macroinvertebrate genera and families (designated hereafter as “Perlodes_Epeorus*” at the genus level).

Table 2.

Prey taxa identified with positive matches (≥ 98 %) from 383 feces of Pyrenean desmans (Galemys pyrenaicus) collected in the French Pyrenees. Frequencies of occurrence (FO: % of feces containing the prey) are displayed. See Supplementary Data SD1 for the complete list of taxa at a finer taxonomic resolution (family and genus).

Class	FO (%)	Order	FO (%)
Insecta	99.7	Ephemeroptera	86.7
		Plecoptera	84.9
		Trichoptera	64
		Diptera	45.7
		Plecoptera_ Ephemeroptera*	30.8
		Coleoptera	14.1
		Orthoptera	6.5
		Neuroptera	1.8
		Hemiptera	1.3
		Lepidoptera	1.0
		Odonata	0.5
		Hymenoptera	0.3
Malacostraca	18	Amphipoda	16.7
		Isopoda	1.6
		Decapoda	0.3
Diplopoda	3.7	Glomerida	2.4
		Polydesmida	0.8
		Julida	0.5
		Chordeumatida	0.3
Lissamphibia	2.9	Urodeles	2.9
Arachnida	2.4	Araneae	0.8
		Opiliones	0.5
		Sarcoptiformes	0.5
		Trombidiformes	0.5
Clitellata	0.5	Haplotaxida	0.5
Collembola	0.3	Entomobryomorpha	0.3
Eurotatoria	0.3	Bdelloida	0.3
Eutardigrada	0.3	Parachela	0.3
Gastropoda	0.3	Littorinimorpha	0.3
Unidentified fish			0.5

*Taxa misidentified by genetic databases.

The diet of Pyrenean desmans was mainly composed of Insecta and Malacostraca, which were present in 99.7% and 18.0% of the 383 feces, respectively (Table 2). Other classes were present in less than 4% of the Pyrenean desman feces (e.g., Diplopoda, Lissamphibia, Arachnida, Clitellata, Gastropoda). Only 7 orders (23.3%), 14 families (15.4%), and 14 genera (9.0%) had a frequency of occurrence in the feces above 10%, meaning that the majority of prey taxa were found in less than 10% of the feces. The most frequent order was Ephemeroptera (86.7%), which included the most frequent family, Heptageniidae (59.0%), and the most frequent genus, Baetis (Baetidae, 56.4%; Supplementary Data SD1).

Ephemeroptera, Plecoptera, and Trichoptera (EPT) together represented on average 79.1% of the identified genera in each feces. However, when considering the total number of identified taxa (TIT) in the overall diet of Pyrenean desmans, EPT orders represented only 28.2% of identified prey. Exclusively aquatic items represented 4.1% of prey (versus 7.7% of TIT), whereas exclusively terrestrial prey represented 7.7% of prey (versus 36.5% TIT). Taxa with at least 1 aquatic stage and 1 terrestrial stage represented 88.3% of prey (versus 55.8% TIT).

Variation of summer diet across the French Pyrenees.

When considering the summer months (i.e., June to September), analyses were limited to 287 feces collected from 115 sites. From these feces, 91 prey genera were kept to study possible spatial variation in the summer diet of Pyrenean desmans. Hierarchical ascendant clustering made it possible to identify 3 distinct diet groups of sites (with 39, 48, and 28 sites, respectively) based on prey assemblage (Fig. 2). There was no significant difference (χ²₂ = 4.05, P = 0.13) between the number of samples collected per site and its classification by cluster analysis (i.e., diet group), excluding potential bias due to diet profiles merging. Prey composition was closer between Groups 1 and 2 than with Group 3. Within-site prey composition differed significantly among groups (PERMANOVA: F_2,112 = 7.19, P < 0.01; Fig. 3). The dissimilarity of prey assemblages among sites similarly differed by diet group (PERMDISP: F_2,112 = 3.28, P < 0.05) and was driven by slightly lower among-site variability in Group 3 than in Group 2 (Tukey HSD: P = 0.045). There was no significant difference in among-site variability between Groups 1 and 2, and Groups 1 and 3 (Tukey HSD: P > 0.05).

Fig. 2.

Hierarchical clustering dendrogram of Sǿrensen distances between the sampling sites according to their similarity in the summer diet (i.e., 91 prey taxa genetically identified in feces) of Pyrenean desmans (Galemys pyrenaicus). Three diet groups are identified.

Fig. 3.

Prey composition of the 3 groups of sampling sites based on summer diet composition (% of sites with prey) of Pyrenean desmans (Galemys pyrenaicus). The diet groups were determined from a hierarchical ascendant clustering based on a Sǿrensen distance matrix of presence–absence of prey genera per site (see Fig. 1). Prey are sorted by decreasing proportion in the first group of sites. *Prey taxa that were misidentified by the genetic databases. Indicator taxa for diet groups are identified in bold and underlined characters (see Table 3 for details). Habitat type of taxa is given (black triangles: exclusively aquatic; black rectangles: exclusively terrestrial; no symbol: with aquatic and terrestrial stages).

Table 3.

Significant indicator prey taxa (P < 0.05) of the 3 summer diet groups of Pyrenean desmans (Galemys pyrenaicus) in the French Pyrenees identified by an IndVal analysis.

Groups	Indicator taxa		IndVal index	P-value
	Order	Genus
1	Trichoptera	Hydropsyche	0.72	0.0001
	Plecoptera	Perlodes	0.62	0.0001
	Diptera	Allotrichoma	0.36	0.0035
	Plecoptera	Halticoperla*	0.35	0.0289
	Ephemeroptera	Epeorus	0.32	0.0488
2	Ephemeroptera	Ecdyonurus	0.73	0.0001
	Ephemeroptera	Rhithrogena	0.58	0.0008
	Ephemeroptera	Ephemerella	0.54	0.02
3	Amphipoda	Gammarus	0.62	0.0001
	Plecoptera	Protonemura	0.61	0.0003
	Diptera	Simulium	0.45	0.0023

*Taxa misidentified by genetic databases.

Diet Group 3 exhibited the least diverse prey assemblage, with 45 different invertebrate genera consumed by Pyrenean desmans, versus 51 in Group 1 and 57 in Group 2. Only 21 genera were consumed in all 3 groups. Eleven significant indicator taxa were identified for the 3 summer diet groups of Pyrenean desmans (P < 0.05; Table 3; Fig. 3). Group 2 was characterized by 3 Ephemeroptera taxa, whereas the indicator taxa of Group 1 and Group 3 were more diverse (Trichoptera, Plecoptera, Diptera, and Ephemeroptera for Group 1 and Amphipoda, Plecoptera, and Diptera for Group 3; Table 3). The differences in diet composition were not significantly related to the geographical location of sites within the 3 hydrographic regions of the French Pyrenees (χ²₄ = 5.93, P = 0.20; Fig. 1B).

Linear discriminant analysis computed 2 significant functions, F1 and F2 (Monte-Carlo test: P = 0.001), accounting for 60.8% and 39.2% of the variability between the 3 diet groups, respectively (Fig. 4). The first function, F1, separated Group 3 from the 2 other groups, whereas the second function, F2, separated Group 2 from Groups 1 and 3. The environmental variables discriminating the 3 diet groups were 1) SHELTER (i.e., proxy for riverbed heterogeneity) and SLO (i.e., slope of river section), which were positively correlated with F1; 2) WOOD (i.e., proportion of bankside with shrubby-woody vegetation), which was negatively correlated with F1; 3) HUM_IMP (i.e., proxy for human impacts), which was positively correlated with F2; and 4) SHEET (i.e., proportion of the stretch with nonturbulent fast water units of shallow water that flows uniformly over smooth bedrock) and TRI (i.e., number of tributaries), which were negatively correlated with F2 (Fig. 4). This suggests that on average, the sites in Group 3 exhibited lower slopes, less heterogeneous riverbeds, and wooded riverbanks (Table 4). Group 2 clustered sites with low human impact along the near floodplain. The sites in Group 1 showed intermediate local environmental conditions (Table 4).

Fig. 4.

Output of the linear discriminant analysis using environmental variables to distinguish local habitat conditions in each diet group. The projection onto the first 2 linear discriminants is displayed: A) 12 environmental variables; B) sites clustered by diet groups. See Table 1 for description of variables.

Table 4.

Mean and SD of values for the 12 environmental variables used to discriminate the 3 summer diet groups of Pyrenean desmans (Galemys pyrenaicus) in the French Pyrenees. SHELTER through HUM_IMP (except BANKS) are given as proportions (%); BANKS is given as values of Shannon’s diversity index; TEM is given in °C; FLO is given in m³/s; SLO is given in percentage of slope; TRI is given in number of stream tributaries. See Table 1 for description of variables and their measurement.

Environmental variables	Groups
	1	2	3
SHELTER	74.42 ± 18.33	67.79 ± 17.16	59.42 ± 21.76
CLOG	11.89 ± 12.61	15.43 ± 16.79	19.71 ± 16.23
SHEET	4.81 ± 6.20	6.60 ± 11.08	4.64 ± 5.12
RUN	23.39 ± 16.04	20.75 ± 11.91	21.24 ± 12.77
RIFFLE	35.77 ± 23.52	37.26 ± 18.82	44.43 ± 22.60
BANKS	0.35 ± 0.25	0.47 ± 0.23	0.52 ± 0.28
WOOD	50.50 ± 23.03	44.51 ± 17.63	53.47 ± 17.74
HUM_IMP	28.79 ± 29.69	9.29 ± 17.49	19.79 ± 20.64
TEM	9.70 ± 1.93	8.94 ± 1.73	10.34 ± 1.48
FLO	1.32 ± 1.87	0.68 ± 0.83	1.27 ± 1.64
SLO	8.78 ± 8.35	10.79 ± 8.91	8.07 ± 10.54
TRI	3.31 ± 2.13	3.42 ± 1.93	2.68 ± 1.19

Discussion

Next-generation sequencing as a powerful tool for studying the diet of Pyrenean desmans.

The amplification of a COI mini-barcode successfully identified the Pyrenean desman and its prey species, as well as several other host species, in 57% of fecal samples. This confirms the relevance of using such a genetic marker with next-generation sequencing methods in diet assessments (Pompanon et al. 2012; Piñol et al. 2014; Gillet et al. 2015) without the need for predator-specific blocking probes. The remaining 43% of the samples could not be correctly assigned. According to McInnes et al. (2017), results are highly dependent on freshness and size of fecal samples. Small feces such as those of the Pyrenean desman (10–15 mm long) contain small amounts of DNA, and the limited amount of DNA also affects the reproducibility of the extraction step, since the entire feces has to be used. Moreover, DNA is rapidly degraded by contact with water and UV radiation (Lindahl 1993).

In spite of these shortcomings, molecular analysis allowed the identification of 156 different invertebrate genera across the French Pyrenees. These genera belonged to 30 orders and 91 families, of which almost 70 were identified as prey of Pyrenean desmans for the first time. Using traditional methods of fecal analysis, Bertrand (1994) identified only 20 families from an extensive sampling of 521 feces collected in 2 small French catchments. In Northern Spain, Castién and Gosálbez (1995) trapped 49 desmans throughout the year and were able to identify only 11 orders as the lowest taxonomic level. The present study is therefore the most extensive ever done on the diet of this species in terms of number of samples, geographic coverage, and taxonomic resolution altogether. The high number of newly identified prey emphasizes the efficiency of molecular analysis in detecting taxa that are difficult to identify in feces through morphological analysis and highlights the Pyrenean desman’s capacity to adapt to its trophic environment. Nevertheless, some taxa identified, such as Collembola, Eurotatoria, and Eutardigrada, are unlikely to be direct prey of the Pyrenean desman, as they are part of the soil microfauna or aquatic zooplankton. Other taxa, such as Sargus and small Coleoptera, may develop at the larval stage or feed on scat and may have thus been collected with the feces. The high sensitivity of next-generation sequencing methods means that secondary predation can sometimes be included in the results (Sheppard et al. 2005). Molecular diet analyses also are limited by deficiencies in the reference databases: a taxon can be identified at the species (or genus) level only if it has already been sequenced (Pompanon et al. 2012). In our study, 21 prey species were endemic to other parts of the world and were thus misidentified. Inventories of genetic biodiversity will become increasingly valuable as molecular analyses find new applications in studies of ecology and conservation. Recently, the use of COI markers led to some concerns in metabarcoding studies (see Deagle et al. 2014) using environmental DNA or bulk biodiversity samples. However, in this study, we used only samples that belonged to previously identified species. Some taxa could still not have been discovered but as we specifically designed the primers and, in sight of the number of genera found among prey, we are confident that the number of missing taxa is very low.

Diversity of prey in the diet of Pyrenean desmans.

The Pyrenean desman has been described as a specialist predator targeting prey in aquatic environments (e.g., Bertrand 1994; Castién and Gosálbez 1995). In this study, we confirmed that dietary preferences seem to be directed toward Ephemeroptera, Plecoptera, and Trichoptera. However, the wide variety of prey identified suggests a more generalist diet for Pyrenean desmans. First, 91% of taxa were identified in less than 10% of the sites, representing a very high proportion of infrequent prey. Second, the most frequent prey (e.g., Baetis, Protonemura, and Rhithrogena) are among the most abundant aquatic macroinvertebrates in the Pyrenees (e.g., Brown et al. 2006; Finn et al. 2013). Third, strictly terrestrial prey represent about 8% of the prey identified in each feces and more than 35% of all identified prey taxa. This substantial percentage of terrestrial invertebrate consumption may result from 1) active hunting for terrestrial prey, 2) opportunistic feeding on terrestrial prey while moving on the banks, or 3) consumption of drowned terrestrial invertebrates. Other aquatic (e.g., brown trout, Salmo trutta) and semiaquatic (e.g., European otter, Lutra lutra) species are known to rely to some extent on a pool of alternative prey including terrestrial subsidies to fulfil their energetic needs (Clavero et al. 2003; Evangelista et al. 2014; Milardi et al. 2016). This diversification of diet may be linked to aquatic stressors (e.g., pollution) or seasonal effects (e.g., variation of climatic conditions and water flow) that limit or modify in situ aquatic communities (Clavero et al. 2003; Kraus et al. 2016; Milardi et al. 2016). The highly diverse summer diet of Pyrenean desmans could thus be considered here, as for the otter, a response to summer drought conditions (Ruiz-Olmo et al. 2001). Indeed, during the summer period, most aquatic insects have already emerged in mountain streams (Füreder et al. 2004) or occur in small-sized life stages, especially Ephemeroptera, Plecoptera, and Trichoptera, which induces a potential diversification of the prey of Pyrenean desmans towards an increasing number of alternative aquatic and terrestrial food items. The diet of desmans would be less diverse, with less abundant terrestrial prey, during other seasons. Confirmation of this prediction requires additional seasonal surveys. Such temporal surveys should be combined with standardized sampling of both terrestrial and aquatic potential prey on river banks and in streams to get a quantitative assessment of the trophic resources available for Pyrenean desmans and therefore better understand its feeding behavior.

Spatial variation in the diet of Pyrenean desmans and the influence of the environment.

While accounting only for the most common prey consumed during the summer by Pyrenean desmans, 3 different diets were identified in the French Pyrenees according to prey composition and indicator prey taxa. There was no major regional influence from the 3 main hydrographic regions of the French Pyrenees, despite differing environmental conditions from the western wet Atlantic area to the eastern dry Mediterranean. This result does not support our assumption that the diet of the Pyrenean desman would differ according to the genetic populations identified by Gillet (2015) or to the different habitat use reported by Biffi et al. (2016).

However, sites grouped within the 3 Pyrenean desman diets exhibited differences in environmental variables, suggesting some influence at the site scale. Group 2 diet sites exhibited conditions typical of upstream parts of river basins with higher reach slope and lower impact from human activities (i.e., urbanization). At the opposite end, the Group 3 diet clustered lower-altitude sites (e.g., low slope of river stretches) with quite high levels of human impact on the near floodplain. These sites also included more homogeneous riverbeds.

First, the difference in prey composition among groups could be explained by the local availability of prey, which is dependent on fine-scale environmental conditions. The abundance and richness of aquatic macroinvertebrates are directly dependent on local habitat conditions, such as 1) human-induced pollution (e.g., near agricultural or urban areas), which impacts water quality (e.g., Johnson et al. 2013; Pallottini et al. 2017); 2) heterogeneity of substrate types and emerging items, which provides different types of microhabitat in streams (Reid et al. 2010); 3) water current and oxygenation (Tachet et al. 2000; Usseglio-Polatera et al. 2000), which are influenced by the slope of river reaches; and 4) small-scale climate variables, which can modify the period of invertebrate emergence and their availability in streams (e.g., Füreder et al. 2004). Such diet adaptation to site-scale differences in resource availability would suggest an opportunistic and flexible foraging strategy for the Pyrenean desman.

Secondly, the availability of foraging habitats for Pyrenean desmans could be constrained by local physical features. The species would thus feed on the fauna that is present in the microhabitats of streams it can access. For instance, in sites of diet Group 3 where riverbeds are the most homogeneous, the Pyrenean desman may forage on the littoral margins, whereas it may have access to a more diverse choice of microhabitats and prey in sites of diet Groups 1 and 2, where riverbeds are more heterogeneous. This assumption is corroborated by the indicator species identified for each diet. Indicator species of Group 3, such as Gammarus and Protonemura, can be found in zones of plant and organic debris accumulation (i.e., litter) and dense root hairs, which are typical of low-slope areas along riverbanks (Tachet et al. 2000; Usseglio-Polatera et al. 2000). Indicator species of Group 2 are Ephemeroptera taxa sensitive to water quality. This is consistent with the apparently less-disturbed environmental conditions of these sites, which also provide higher water velocity (i.e., higher slopes) in more open areas (i.e., higher altitude). These local conditions make possible the development of biofilms, which are important food resources for scraper taxa such as Ecdyonurus. The intermediate environmental conditions of Group 1 likely induced a higher diversity of indicator taxa, which may reflect a more heterogeneous set of habitats.

Conclusion and Perspectives.

This study reveals that Pyrenean desmans adopt a more generalist foraging strategy than previously reported. Even if a wider dietary niche increases the chance of adaptation to altered environments (Murgatroyd et al. 2016), food availability remains an important issue for species conservation. Many disturbances to freshwater environments result in a decline of abundance and richness in aquatic invertebrate communities (Paul and Meyer 2001) and thus have detrimental consequences for Pyrenean desmans. Other threats affecting its nesting (on river banks) and foraging (in river beds) habitats as well as population viability, such as predation and mortality induced by human activities, should be limited as much as possible to protect Pyrenean desmans. Further studies are needed to 1) investigate the extent to which the diet and spatial distribution of Pyrenean desmans are influenced by prey availability and other biotic interactions and 2) get a thorough knowledge of its foraging areas within streams to determine how habitat heterogeneity affects diet.

Supplementary Data

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.— Complete list of taxa identified as prey of Pyrenean desmans (Galemys pyrenaicus) by molecular analysis of 383 feces collected in the French Pyrenees. Frequencies of occurrence of each prey (FO: % of feces with taxa) are displayed. *Misidentified taxa whose distribution areas exclude the Pyrenees and France. Habitat type of taxa is given (aq: exclusively aquatic; aq/te: with aquatic and terrestrial stages; te: exclusively terrestrial).

Acknowledgments

We thank all the partners for their help in conducting the extensive fieldwork at the scale of the French Pyrenees: Laboratoire d’Ecologie Fonctionnelle et Environnement (EcoLab – CNRS/UPS/INPT), Fédération Aude Claire, Fédération des Réserves Naturelles Catalanes, Office National des Forêts, Groupe de Recherche et d’Etudes pour la Gestion de l’Environnement, Office National de la Chasse et de la Faune Sauvage, Parc National des Pyrénées, Conservatoire d’Espaces Naturels Ariège, Conservatoire d’Espaces Naturels Aquitaine, Conservatoire d’Espaces Naturels Midi-Pyrénées (CEN MP). We are also grateful to G. Grenouillet for statistical support and to F. D’Amico. This study was funded by ANRT (Cifre n° 2011/1018 et n° 2011/1571), the European Union (FEDER and LIFE+ Nature), EDF (Electricité de France), Agence de l’eau Adour-Garonne, DREAL (Direction Régionale pour l’Environnement, l’Aménagementet le Logement) Aquitaine, Midi-Pyrénées, and Languedoc-Roussillon, Conseil Régional Aquitaine, Midi-Pyrénées and Languedoc-Roussillon, Conseil Départemental Pyrénées-Atlantiques, Aude, and Pyrénées-Orientales, SHEM (Société Hydroélectrique du Midi), and Patagonia. It is part of the French Conservation Action Plan for the Pyrenean desman (2010–2015) supervised by DREAL de Midi-Pyrenees and coordinated by the CEN MP.

Author notes

†

Co-first authors

Associate Editor was Tereza Jezkova.