DNA metabarcoding unveils multiscale trophic variation in a widespread coastal opportunist

Abstract A thorough understanding of ecological networks relies on comprehensive information on trophic relationships among species. Since unpicking the diet of many organisms is unattainable using traditional morphology‐based approaches, the application of high‐throughput sequencing methods represents a rapid and powerful way forward. Here, we assessed the application of DNA metabarcoding with nearly universal primers for the mitochondrial marker cytochrome c oxidase I in defining the trophic ecology of adult brown shrimp, Crangon crangon, in six European estuaries. The exact trophic role of this abundant and widespread coastal benthic species is somewhat controversial, while information on geographical variation remains scant. Results revealed a highly opportunistic behaviour. Shrimp stomach contents contained hundreds of taxa (>1,000 molecular operational taxonomic units), of which 291 were identified as distinct species, belonging to 35 phyla. Only twenty ascertained species had a mean relative abundance of more than 0.5%. Predominant species included other abundant coastal and estuarine taxa, including the shore crab Carcinus maenas and the amphipod Corophium volutator. Jacobs’ selectivity index estimates based on DNA extracted from both shrimp stomachs and sediment samples were used to assess the shrimp's trophic niche indicating a generalist diet, dominated by crustaceans, polychaetes and fish. Spatial variation in diet composition, at regional and local scales, confirmed the highly flexible nature of this trophic opportunist. Furthermore, the detection of a prevalent, possibly endoparasitic fungus (Purpureocillium lilacinum) in the shrimp's stomach demonstrates the wide range of questions that can be addressed using metabarcoding, towards a more robust reconstruction of ecological networks.


| INTRODUC TI ON
Trophic interactions provide important insights on a wide range of ecological dynamics, ranging from individual to ecosystem levels, which include animal behaviour, predator-prey interactions, food web structure and community ecology (e.g., Leray, Meyer, & Mills, 2015;Pinol, San Andres, Clare, Mir, & Symondson, 2014). The feeding strategy of key consumers can have pronounced influences on ecosystem dynamics (Hanski, Hansson, & Henttonen, 1991;Holling, 1965) and their stomach contents can reveal essential information on food item distribution and prey assemblage structure (Lasley-Rasher, Brady, Smith, & Jumars, 2015). Crustaceans are a key component in marine/estuarine soft-bottom habitats (Evans, 1983(Evans, , 1984, and evaluating their diet is very challenging due to the complexity of direct observations on predation rates and the limitations associated with the identification of partially digested food items (Asahida, Yamashita, & Kobayashi, 1997;Feller, 2006;Symondson, 2002).
The recent application of high-throughput sequencing (HTS) tools, such as metabarcoding, promises to revolutionize the way prey diversity and composition are estimated from gut contents or faeces of consumers (Kartzinel & Pringle, 2015;. Metabarcoding refers to the identification of multiple taxa based on the screening of bulk DNA extracted from community or environmental samples (i.e., water, soil, faeces; Barnes & Turner, 2016), by means of massive parallel sequencing of PCR amplicons (Barnes & Turner, 2016;Taberlet, Coissac, Pompanon, Brochmann, & Willerslev, 2012a). Metabarcoding has proven to be highly effective for the identification of prey remains with improved taxonomic resolution, accuracy and speed of analysis, compared to traditional morphological methods (Berry et al., 2015;Casper, Jarman, Deagle, Gales, & Hindell, 2007;Symondson, 2002). Yet, some challenges remain, such as fragmentation of partially digested DNA, variability in taxon-specific digestion rates, secondary predation and, typically, the presence of high proportion of DNA from the study organisms itself, which may reduce sequencing depth and render cannibalism undetectable (Barnes & Turner, 2016;Berry et al., 2015;Pinol et al., 2014). Furthermore, due to the sensitivity of these methods, in some cases, it might be difficult to discriminate between contaminant DNA and target DNA.
The brown shrimp, Crangon crangon (L.), is a key crustacean species in European coastal waters. Its wide distribution (i.e., from the White Sea to Morocco), year-round occurrence and high abundance (>100 ind./m 2 ; van der Veer, Feller, Weber, & Witte, 1998) make it an essential part of the coastal benthic food web (Ansell, Comely, & Robb, 1999;Campos & van der Veer, 2008;Evans, 1984), a major prey item for birds and fish (Evans, 1984;Walter & Becker, 1997), and an important target for fisheries, with recorded catches in 2011 up to 35,000 tons and more than 500 fishing vessels employed in the North Sea (Aviat, Diamantis, Neudecker, Berkenhagen, & Müller, 2011;Campos & van der Veer, 2008). The trophic position of C. crangon is still being discussed, being described as trophic generalist (Evans, 1983), carnivorous opportunist (Pihl & Rosenberg, 1984) omnivorous (Ansell et al., 1999;Raffaelli, Conacher, McLachlan, & Emes, 1989;Tiews, 1970) and probable scavenger (Ansell et al., 1999). As a juvenile, it relies mostly on the consumption of meiofaunal prey items while it switches to larger demersal organisms as an adult, including conspecifics and juvenile stages of several commercially important teleosts and bivalves (Evans, 1984;Oh, Richard, & Richard, 2001;Pihl & Rosenberg, 1984;van der Veer & Bergman, 1987;van der Veer et al., 1998). Previous studies showed considerable variation in prey item consumption, partly due to the brown shrimp's inherent trophic flexibility and niche breadth, but also because studies have relied on microscopic identification of prey remains (e.g., Boddeke, Driessen, Doesburg, & Ramaekers, 1986;Oh et al., 2001; but see also Nordström, Aarnio, & Bonsdorff, 2009); yet, prey items are usually macerated to a fine degree by C. crangon, and a high proportion of its stomach content is, consequentially, impossible to identify through morphological examination (Asahida et al., 1997;Wilcox & Jeffries, 1974). Furthermore, most studies on C. crangon's diet to date have focused on a limited number of locations and relatively small spatial scales (e.g., Evans, 1984;Oh et al., 2001;Pihl & Rosenberg, 1984) while large-scale studies are required to assess geographical variation in the shrimp's diet and to understand the relative importance of different prey items.
The degree to which food items are actively selected or passively ingested by consumers is an essential consideration in assessing the relative importance of different prey categories and understanding the trophic niche of consumers. Traditionally, indices are used to infer the predator's preference for prey based on the relative abundance of prey in the predator's diet and the prey's relative abundance in the environment (e.g., Peterson & Ausubel, 1984). Examples of commonly used indices are the Ivlev's electivity index (Ivlev, 1961) and the Jacob's index of selectivity (Jacobs, 1974), which also corrects for item depletion (Jacobs, 1974). Although some attempts have been made to link diet metabarcoding data with food availability in managed forests (Kowalczyk et al., 2011) and artificial mesocosms (Ray et al., 2016), no examples exist for wild marine animal trophic studies contrasted with whole-community environmental DNA (eDNA) data.
Here, we report on a large-scale analysis of the trophic ecology of C. crangon, which reveals its ecological role in estuarine systems and provides a key for the reconstruction of ecological networks of European coastal marine communities. By using nearly universal primers for mitochondrial cytochrome c oxidase I, we used metabarcoding to describe the diet of the shrimp, alongside the soft-bottom communities on which they feed, over a European scale. We were expecting a wide variety of prey items, reflecting variation in environmental conditions and prey availability across European coasts. More specifically, we tested whether metabarcoding can (a) provide a detailed overview of C. crangon's diet, including prey selectivity, using DNA extracted from stomach and environmental samples; (b) identify geographical patterns in its trophic ecology, at both local and regional scales; and (c) assess consistent and general trophic patterns in order to better define the ecological role of this widespread species.

| Sample collection and processing
Brown shrimp and sediment samples were collected from 24 sites distributed over six estuaries in the Netherlands, Portugal and the United Kingdom (Figure 1). Adult shrimp (>20 mm total length, TL; tip of the rostrum to tip of the telson) were captured in the intertidal zone (0-1 m depth) by push-net at low tide (±3 hr). Shrimp (30-50 per site) were placed on ice and transported to the laboratory to be stored at −20°C. Sediment was collected for the extraction of eDNA to characterize the biological community present at each site. Sediment was sampled from the upper 2-cm surface layer, which represents the most recent DNA deposits and the habitat where the shrimp live and feed (Pinn & Ansell, 1993;Turner, Uy, & Everhart, 2015), with a PVC corer (3.2 mm Ø). Per site, three sediment subsamples were collected at several metres distance from each other and combined to reduce the influence of local heterogeneity (Taberlet et al., 2012b). The sediment was stored in 96% ethanol, transported on ice and kept at −20°C. At each site, temperature, salinity (Fisher Scientific Traceable Salinity Meter), pH (Hanna HI 98129), dissolved oxygen (OxyGuard Handy Mk I) and turbidity (Eutech TN-100) were measured in triplicates. Extra sediment was collected, in triplicates, from each site for granulometric analyses (Horiba LA-950 Particle size analyser) and total organic matter (TOM) determination by means of ashing (550°C, 6 hr). One site (Mersey 3) was not included for analysis because HTS of its stomach samples did not result in sufficient read depth (<1,000 reads; see Section 3).

| DNA extraction
Overall, 1,025 shrimp (20-50 mm TL) were caught and 494 full stomachs (visual determination) were dissected using flame-sterilized tools to avoid cross-contamination. Stomachs were pooled in batches F I G U R E 1 Overview of sample locations, illustrating (a) the overall western European scale; (b) the Dutch estuaries, Western Scheldt (WS) and Eastern Scheldt (ES); (c) the British estuaries, Mersey (Me) and Kent (Ke); (d) the Aveiro Ria (Av) and (e) the Minho estuary (Mi) in Portugal. Small dots within estuaries represent individual collection points for shrimp and sediment samples. *Site removed prior to molecular analysis. Source map: OpenStreetMap of eight (from shrimp collected at the same site) prior to DNA extraction (Ray et al., 2016). Though the pooling of samples increases the number of stomachs analysed per sequencing run, pooling might reduce the detected molecular taxonomic unit (MOTU) richness since large, recently ingested diet items in single stomachs may obscure the stomach contents of other individuals. Nevertheless, this potentially negative effect of pooling was deemed to be negligible, due to the pool replication conducted within location, and the population emphasis of the study. Three replicate pools were extracted per site.
However, due to a high percentage of empty stomachs in natural populations (20%-60%;Feller, 2006;Oh et al., 2001;Pihl & Rosenberg, 1984), some sites contained only two replicates and some replicates contained less than eight full stomachs (see Supporting Information

| DNA amplification and highthroughput sequencing
Amplification of DNA, for both stomach and sediment samples, The Leray fragment has already been successfully applied for both the characterization of marine communities and marine fish gut contents (Leray & Knowlton, 2015;Leray et al., 2013. Eightbase oligo-tags (Coissac, Riaz, & Puillandre, 2012) attached to the metabarcoding primers were added to the amplicons during a single PCR step, in order to label different samples in a multiplexed library.
Also, a variable number (2, 3 or 4) of fully degenerate positions (Ns) was added at the beginning of each primer, in order to increase variability of the amplicon sequences (Guardiola et al., 2015). The PCR mix recipe included 10 μl AmpliTaq gold 360 Master mix (Applied Biosystems), 3.2 μg bovine serum albumin (Thermo Scientific), 1 μl of each of the 5 μM forward and reverse tagged primers, 5.

| Bioinformatic and data analyses
Bioinformatic analyses were performed using the obitools metabarcoding software suite (Boyer et al., 2016). Read quality assessment was performed with FastQC and paired-end read alignment using illuminapairedend, retaining reads with an alignment quality score >40. Demultiplexing and primer removal were achieved using ngsfilter with the default options.
Obigrep was applied to select all aligned reads with a length between 303 and 323 bp and free of ambiguous bases. Obiuniq was used to dereplicate the reads, and the uchime-denovo al- taxonomic assignment of the representative sequences for each MOTU was performed using the ecotag algorithm (Boyer et al., 2016), using a local reference database (Wangensteen et al., 2018) containing filtered COI sequences retrieved from the bold database (Ratnasingham & Hebert, 2007) and the EMBL repository (Kulikova et al., 2004). This algorithm uses a phylogenetic approach to assign sequences to the most reliable monophyletic unit, based on the density of the reference database. The data were refined by clustering MOTUs assigned to the same species, abundance renormalization (to remove false positives due to tag-switching; Wangensteen & Turon, 2017) and by removing bacterial reads and contaminations of human or terrestrial origin.
MOTUs with a maximum of four or less reads per sample were removed on a sample by sample basis to avoid false positives and low-frequency noise (De Barba et al., 2014;Wangensteen et al., 2018). All MOTUs for which the abundance in the PCR-negative controls was higher than 10% of the total reads of that MOTU were removed (Wangensteen & Turon, 2017  stomach samples ( Figure 2) and 24 sediment samples (Figure 3). A total of 14 pooled stomach samples were removed, prior to further analyses, because they contained less than 1,000 diet-related reads (Supporting Information Table S4). One sediment sample (

| Description of Crangon crangon diet
Analysis of C. crangon stomach contents showed large variation in relative MOTU abundances between samples (Figure 2

| Selectivity in Crangon crangon diet
MOTU diversity within phyla was generally higher in the sediment than in the stomach samples, except for Arthropoda, Annelida, Mollusca and Chordata (Figure 4a). The proportion of MOTUs that could not be assigned at the phylum level was higher in the sediment (73%) than in the stomach samples (58%), and many abundant taxa in the sediment could not be identified at lower taxonomic ranks (Figure 3). Data combined per sample type (sediment/stomach) and MOTUs pooled at the phylum level showed that sediment samples contained high relative read abundances of Bacillariophyta (20 ± 3%), Discosea (10 ± 2%), Dinoflagellata (6 ± 2%) and Arthropoda (5 ± 1%) while C. crangon stomach samples contained a high mean (±SE) relative read abundance (%) for Arthropoda (53 ± 5%), Annelida (12 ± 3%) and Chordata ( read abundances between the sediment and stomach samples (Supporting Information Table S5). Visualization of the importance of the phyla detected in the stomach samples based on the mean relative abundance (%), presence (%) and Jacobs' selectivity index (D) is shown in Figure 4c.

| Variation between estuaries
Multivariate analysis on the stomach contents (MOTUS ≥0.5% abundance) showed significant differences between estuaries (Figure 6b; PERMANOVA: pseudo-F 5,29 = 2.7, p < 0.001) and sites nested within Stepwise model selection (both forward and reverse) and CCA (Figure 6c) showed significant influences of salinity (p < 0.01), median grain size (p < 0.01) and TOM (p < 0.05; see Supporting Information Table S7 for means per estuary) on MOTU composition in C. crangon stomach samples (≥0.5% abundant MOTUs). The environmental variables (constrained CCA axes) explained 34% of the variance in the data set.
Temperature, turbidity and oxygen saturation did not have a significant influence on the model, and pH was strongly correlated with salinity (r 2 = 0.75, p < 0.001, N = 23). These factors were, therefore, not included in the final model. MOTU richness (rarefied to 1,000 reads) in C. crangon stomach contents also showed differences between estuaries, with the Aveiro and Eastern Scheldt estuaries showing a higher number of MOTUs than the others (Figure 7). The slopes of the MOTU accumulation curves, however, did not approach an asymptote, offering a glimpse of the vast amount of marine biodiversity yet to be uncovered.

| Evaluation of Crangon crangon diet
This study provides a detailed overview of the brown shrimp's trophic ecology, focusing on dietary variations at multiple geographical scales. Adult brown shrimp were caught in a variety of sandy estuarine intertidal habitats. Mainly females were captured, probably due to the spatial sex-specific segregation of C. crangon during the summer-autumn period (Bamber & Henderson, 1994;Henderson & Holmes, 1987). The results confirm C. crangon as a generalist consumer feeding on a broad variety of food items but preferring larger mobile epifaunal prey items such as crustaceans, annelids and fish. The present investigation uncovered a great diet contribution of decapods and teleosts, while these were usually not considered to be important contributors to the shrimp diet in previous studies (e.g., Ansell et al., 1999;Oh et al., 2001;Plagmann, 1939;Raffaelli et al., 1989). Although comparisons of prey contribution should be made cautiously due to diversity of quantification methods used, the observed trend could be partly explained by scavenging behaviour on large organisms, previously not recorded in crangonid shrimps.
Crangonid shrimps generally macerate and eat the soft body parts of larger preys (Asahida et al., 1997;Gibson, Yin, & Robb, 1995;Seikai, Kinoshita, & Tanaka, 1993;Wilcox & Jeffries, 1974). Smaller food items, on the other hand, are often ingested as a whole, including their hard body parts (Tiews, 1970), and are thus more easily identified by morphological methods. This discrepancy in detectability might possibly have played a role in studies that have detected low amount of fish and decapods but considerable amounts of unidentified soft tissue (e.g., Oh et al., 2001;Raffaelli et al., 1989).
Metabarcoding methods can detect and taxonomically identify such soft tissues, thus highlighting the enhanced suitability of molecular approaches to present a more realistic picture of trophic ecology in marine invertebrates.
The diet of C. crangon showed a high MOTU richness, including previously described food items ( Table 1). The number of COI MOTUs (2,342) detected in the shrimp's stomachs may be an overestimation of the total number of real species (e.g., due to detection of pseudogenes; Tang et al., 2012;Vamos, Elbrecht, & Leese, 2017) and includes protists and microalgae (Wangensteen et al., 2018), which are unlikely to be prey items of C. crangon. Nonetheless, even just the 306 ascertained or 20 most abundant (>0.5% relative abundance) species in the shrimp's diet was remarkably higher than the number found in previous studies based on morphological identification (see Table 1). Furthermore, twenty taxa showed a high abundance (>5%) in any given sample, probably representing important prey items at some locations or times.
Two species were predominant in our study: the shore crab Carcinus maenas across the overall geographic distribution, and the amphipod Corophium volutator in UK localities (characterized by muddy sediments and high organic matter content). Both species are well-known prey of C. crangon (Evans, 1984;Moksnes, Pihl, & van Montfrans, 1998;Pihl & Rosenberg, 1984) and can occur at high densities in softbottom estuarine habitats (Meadows & Reid, 1966;Moksnes, 2002).
This is also the first study showing a high occurrence of

| The application of metabarcoding in crustacean trophic studies
Metabarcoding using universal primers is generally considered as a simple, rapid and relatively inexpensive method to define in detail the feeding ecology of organisms (Berry et al., 2015;Kartzinel & Pringle, 2015;Pinol et al., 2014). The fraction of the brown shrimp DNA detected in its own gut was low allowing for the detection of prey items without using predator-specific blocking primers (average: 28%; compared to, e.g., Olmos-Péerez, Roura, Pierce, Boyer, & González, 2017;Pinol et al., 2014). Metabarcoding has several clear advantages over traditional trophic methods including the better detection of soft-bodied, small and cryptic taxa, higher speed of analysis (Berry et al., 2015;Casper et al., 2007;Chariton et al., 2015;Symondson, 2002), Both traditional morphological examination and DNA metabarcoding of food items suffer from limitations in providing quantitative descriptions of the diet of consumers (Casper et al., 2007). For metabarcoding, errors can occur due to technical artefacts specific to DNA amplification and sequencing (Barnes & Turner, 2016;Pompanon et al., 2012), and biological limitations such as speciesspecific digestion and DNA degradation rates (Deagle, Chiaradia, McInnes, & Jarman, 2010;Murray et al., 2011;Pinol et al., 2014;Sakaguchi et al., 2017). Furthermore, some of the DNA detected might come from secondary predation (taxa present in the stomach of preyed organisms; Berry et al., 2015;Kartzinel & Pringle, 2015).
Cannibalism also imposes a specific problem in trophic molecular studies since it cannot be identified by means of metabarcoding (Berry et al., 2015;Ray et al., 2016). Large brown shrimps are known to be cannibalistic (Evans, 1984;Pihl & Rosenberg, 1984) but the removal of C. crangon sequence reads from our data set makes it impossible to gauge insights into the extent of cannibalism in this species. Due to the restrictions in the quantification of consumed prey volume, many trophic studies only use presence/absence data (e.g., Deagle et al., 2010;Harms-Tuohy et al., 2016;Pinol, Mir, Gomez-Polo, & Agusti, 2015). This might, however, result in an overestimation of small taxa that are abundant in the sediment, but with low trophic relevance, as they could, in the case of C. crangon, be passively acquired when shrimp ingest sediment to crush food in their stomach (Ansell et al., 1999;Deagle et al., 2018;Tiews, 1970).
Multiple stomachs were pooled prior to analysis and data were subjected to rigorous filtering to allow for a semiquantitative estimation of proportions of prey DNA (Deagle et al., 2005;Lejzerowicz et (Deagle et al., 2005). This study provides a significant addition to a growing body of studies in showing the applicability of semiquantitative estimations in molecular trophic ecology (e.g., Albaina et al., 2016;Deagle et al., 2018;Ray et al., 2016;Sakaguchi et al., 2017;Soininen et al., 2013).
Finally, the results presented draw a close link between prey distribution in estuarine habitats and ingested prey item abundance. The use of eDNA from sediments to assess community composition and to relate this to the shrimp's diet is a novel contribution to the fields of molecular trophic analysis and eDNA, which goes beyond the taxon studied. It should be noted, however, that a correct assessment of the predator's trophic niche by means of prey selectivity determination relies on a correct assessment of prey abundance, both in the stomach and in the environment. Issues with incorrect abundance estimations, for example due to species-specific detection rates (e.g., due to different rates of DNA sequestering in the environment; Barnes & Turner, 2016), are not specific to molecular studies (Strauss, 1979) and the constant work on improving the reliability of relative abundance estimations from eDNA (Deagle et al., 2018;Thomas et al., 2016;Ushio et al., 2018) should substantially enhance the applicability of selectivity indices in molecular research.

| Geographic variation in C. crangon trophic ecology
This study also assesses for the first time a large geographical variation in the brown shrimp's trophic ecology at multiple spatial scales.
Previous studies have shown local variability in C. crangon diet (Evans, 1984;Oh et al., 2001;Pihl & Rosenberg, 1984) but no studies have been performed across multiple European estuaries. The results indicate that the consumed prey community can vary at local (within estuary, as discussed above) and regional (between estuaries) scales.
The seasonal and tidal migratory behaviour of C. crangon (Al-Adhub & Naylor, 1975;Henderson & Holmes, 1987) (2); fish 0-year-juveniles (3); 4 mysids (4); (pico) phytoplankton (5); fish carcasses (6); decapod carcasses (7); amphipods (8); chironomid, mollusc and barnacle larvae (9); meiofauna (10). Letters define method of ingestion: Secondary predation (SP); ambush predation (AP); gulping predation (GU); passive ingestion (PI); scavenging (SC). Images not to scale F I G U R E 7 MOTU accumulation curves showing MOTU richness (based on all MOTUs detected) in Crangon crangon pooled stomach samples in several European estuaries. Each sample has been rarefied to 1,000 reads prior to the construction of the accumulation curves. Stomach samples consisted of a pool of up to eight stomachs , 1972). The Mersey estuary also showed a relatively low species richness detected in the stomach contents of C. crangon, probably related to its history of anthropogenic stress (Jones, 2000). Overall, trophic variation in C. crangon depends on patterns in the local abundance and distribution of its prey (in line with: Oh et al., 2001;Pihl, 1985;Pihl & Rosenberg, 1984). In order to evaluate this variation more exhaustively, knowledge on the ecology and seasonality of the local macrozoobenthic community is required.

| Crangon crangon's ecological role
Based on the results of this study, C. crangon can best be described as a highly opportunistic carnivore and scavenger. Despite its broad dietary range (Figure 8), it shows a prominent level of selectivity for larger mobile epifaunal prey items. This high level of flexibility in its trophic ecology might contribute to its very wide distribution on European coasts (Campos et al., 2009). In order to feed on diverse prey taxa, adult C. crangon are capable of employing a variety of methods ( Figure 8) including (camouflage-assisted) ambush predation (Gibson et al., 1995;Pinn & Ansell, 1993;Siegenthaler, Mastin, Dufaut, Mondal, & Benvenuto, 2018), gulping behaviour (Tiews, 1970) and scavenging ( Figure 8; Ansell et al., 1999;Price, 1962). Since meiofaunal and protist phyla were not selected as prey items based on Jacobs' selectivity index (present but not abundant; in line with : Evans, 1983;Feller, 2006), it is possible that these taxa were passively consumed during the ingestion of sand to aid digestion (Ansell et al., 1999;Tiews, 1970) or through secondary predation. Several studies classify C. crangon as an omnivore (Ansell et al., 1999;Raffaelli et al., 1989;Tiews, 1970), but we cannot confirm this classification, because the primers used during this study have a very low affinity for chlorophytes resulting in many algal taxa not being detected (Wangensteen et al., 2018).
Nevertheless, the algal phyla that can be detected with these primers (e.g., Rhodophyta, Phaeophyta and Bacillariophyta) had a low selectivity, indicating a negligible trophic importance for C. crangon. More research is required with plant-specific primers to assess the actual contribution of herbivory to the diet of the brown shrimp. Overall, the results of this study yield a level understanding of the trophic ecology of this species that would not have been possible through traditional morphological analysis, and which is key to providing essential insights into coastal community interactions.

ACK N OWLED G EM ENTS
We would like to express our thanks to C. Adiele, A. Bio, T.

DATA ACCE SS I B I LIT Y
The data set, including sequences, taxonomic assignment and abundances for all MOTUs in every sample, has been deposited in Dryad (https://doi.org/10.5061/dryad.sk2155m). Custom R scripts are publicly available from https://github.com/Andjin/Crangon-dietanalysis and https://github.com/metabarpark for scripts related to the bioinformatics pipeline.