Characterizing niche differentiation among marine consumers with amino acid δ13C fingerprinting

Abstract Marine food webs are highly compartmentalized, and characterizing the trophic niches among consumers is important for predicting how impact from human activities affects the structuring and functioning of marine food webs. Biomarkers such as bulk stable isotopes have proven to be powerful tools to elucidate trophic niches, but they may lack in resolution, particularly when spatiotemporal variability in a system is high. To close this gap, we investigated whether carbon isotope (δ13C) patterns of essential amino acids (EAAs), also termed δ13CAA fingerprints, can characterize niche differentiation in a highly dynamic marine system. Specifically, we tested the ability of δ13CAA fingerprints to differentiate trophic niches among six functional groups and ten individual species in the Baltic Sea. We also tested whether fingerprints of the common zooplanktivorous fishes, herring and sprat, differ among four Baltic Sea regions with different biochemical conditions and phytoplankton assemblages. Additionally, we investigated how these results compared to bulk C and N isotope data for the same sample set. We found significantly different δ13CAA fingerprints among all six functional groups. Species differentiation was in comparison less distinct, due to partial convergence of the species' fingerprints within functional groups. Herring and sprat displayed region‐specific δ13CAA fingerprints indicating that this approach could be used as a migratory marker. Niche metrics analyses showed that bulk isotope data had a lower power to differentiate between trophic niches than δ13CAA fingerprinting. We conclude that δ13CAA fingerprinting has a strong potential to advance our understanding of ecological niches, and trophic linkages from producers to higher trophic levels in dynamic marine systems. Given how management practices of marine resources and habitats are reshaping the structure and function of marine food webs, implementing new and powerful tracer methods are urgently needed to improve the knowledge base for policy makers.


| INTRODUC TI ON
Direct pressures on marine systems such as increasing temperatures, eutrophication, introduction of nonindigenous species, and overfishing are affecting the performance of individual species and the structure of entire systems. Examples of these consequences include the malnutrition of ecologically and commercially important fish species (Eero et al., 2015), niche shifts following the introduction of nonindigenous species (Ojaveer et al., 2017), and evidence for system-wide shifts in many regions (Alheit et al., 2005). In this context, identifying organic matter sources at the base of the food web is key for understanding resource partitioning and trophic niche differentiation across time and space.
Resource partitioning among marine species and trophic groups is often poorly understood due to the complexity of marine food webs (Lynam et al., 2017) and methodological constraints (Nielsen, Clare, Hayden, Brett, & Kratina, 2018). Diet identification has traditionally relied on visual taxonomic assessment of stomach and fecal contents (Hyslop, 1980), but visual assessments are now increasingly complemented with DNA metabarcoding (Bowser, Diamond, & Addison, 2013). While the taxonomic resolution of these methods can be high, they only provide instant snapshots of ingested diets provided that the identifiable fragments or DNA sequences are intact. Obtaining intact sequences can be logistically challenging when assessing multiple species over space and time. In comparison, it is possible to integrate dietary histories with stable isotope ratios since the diet-derived building blocks for animal tissues are sourced over time. Stable isotopes of elements can be informative of diet sources because lighter stable isotopes enter reactions and physical processes at faster rates than heavier stable isotopes, resulting in different isotope ratios among different organic pools. The rate by which elements shifts their isotopic ratios during trophic transfer differs greatly: elements such as carbon and sulfur are used as source tracers because they discriminate (Mittermayr, Hansen, & Sommer, 2014) less compared to nitrogen, which is used as a marker of trophic position (Vander Zanden & Rasmussen, 1999). However, isotope ratios of whole tissues (bulk SIA) often lack source specificity because of variable, and at times, unpredictable isotope discriminates processes and isotope baseline values for different systems (Fry, 2006;Post, 2002). To overcome these limitations, ecologists are increasingly using compound-specific isotope analyses (CSIA), in which stable isotope ratios are determined for individual compounds, as a complementary approach (Whiteman, Elliott Smith, Besser, & Newsome, 2019).
CSIA of protein amino acids has emerged as one of the most promising approaches to trace the origins and fate of food sources (McClelland & Montoya, 2002;O'Brien, Fogel, & Boggs, 2002).
Amino acids (AAs) are among the major conduits of organic carbon in food webs and well suited as a source tracer because metazoans cannot synthesize the carbon backbones of about half of the 20 protein AAs de novo. Instead, metazoans depend on essential amino acids (EAAs) from food sources (McMahon, Fogel, Elsdon, & Thorrold, 2010) or more rarely bacterial symbionts (Larsen, Ventura, et al., 2016). EAAs are powerful source tracers because δ 13 C EAA values remain largely conserved through trophic transfer and because the producers of these EAA, algae, bacteria, fungi, and vascular plants each generate unique δ 13 C EAA patterns or fingerprints (Larsen, Taylor, Leigh, & O'Brien, 2009;Larsen et al., 2013;Scott et al., 2006) (see Box 1 for an illustration). Thus, by analyzing δ 13 C EAA ecologists can circumvent the problem of variable and unknown isotopic fractionation during trophic transfer, but the ability of fingerprints to resolve primary production sources is still unclear. Larsen et al. (2013) compared two dozen species of laboratory cultures comprising of diatoms, cyanobacteria, chrysophytes, chlorophytes, and haptophytes to macroalgae, seagrass, fungi, bacteria, and terrestrial vascular plants and found that of all these groups, phytoplankton displayed the largest intragroup variability in δ 13 C EAA patterns across species and taxonomic groups. Despite some unresolved questions for applying δ 13 C EAA fingerprints in marine environments, they have been applied successfully to track habitat use of fishes with distinct ontogenetic migration patterns (Vane, Larsen, Scholz-Böttcher, Kopke, & Ekau, 2018), resource and habitat use in marine systems (McMahon, Berumen, & Thorrold, 2012), and proportional contributions of primary production sources to marine consumers (Elliott Smith, Harrod, & Newsome, 2018;Rowe et al., 2019;Vokhshoori, Larsen, & McCarthy, 2014). A recent study on mesozooplankton in the Baltic Sea showed promise in distinguishing between interannual algal assemblages (Eglite et al., 2019). Taken together, these results indicate that δ 13 C EAA fingerprints may be able to provide detailed insights into ecological niches of consumers to a much larger extend than previously realized.
Exploring further use of CSIA to elucidate changes in basal resources and ecological niches is particularly pertinent for regional seas because of their rapidly warming sea surface temperatures and increasing stressors from anthropogenic activities such as eutrophication and overfishing, with corresponding changes in food webs (Reusch et al., 2018). In this study, we selected the western and central Baltic Sea as a study area because it is a brackish inland sea characterized by strong spatial differences in phytoplankton composition (Eglite et al., 2019;Gasiūnaitė et al., 2005;Wasmund, Dutz, Pollehne, Siegel, & Zettler, 2017) driven by a gradient in hydrographic-hydrochemical conditions (Naumann et al., 2017). In this sea, food web-related processes have been identified as driver of changes in ecosystem composition (Möllmann et al., 2009) and declines of key commercial species (Casini et al., 2016;Reusch et al., 2018).
Compared to euhaline systems, this brackish water system is characterized by a relatively low diversity (Ojaveer et al., 2010) and a tight coupling of benthic and pelagic food webs (Griffiths et al., 2017;Kiljunen et al., 2020). Across the gradient, the small pelagic fish species herring (Clupea harengus) and sprat (Sprattus sprattus) are the dominant zooplanktivores, and of large commercial value (Ojaveer, Lankov, Raid, Põllumäe, & Klais, 2018). As zooplanktivores, these species are also natural integrators of pelagic planktonic production.
To test the power of CSIA to identify niche differentiation among marine consumers in the spatially variable Baltic Sea, we obtained δ 13 C AA values for 10 species. These species encompass both fishes and invertebrates across six different functional groups: suspension feeders, planktivores, benthic predators, benthic flatfishes, and scavengers. Furthermore, to assess the power of the method to identify differences across larger spatial scales, we obtained δ 13 C AA values for herring and sprat from four locations along the Baltic Sea gradient (Figure 1). We first assessed the power of δ 13 C EAA fingerprints to identify (a) trophic niche differentiation among functional groups and among species, and (b) the presence of spatial patterns among planktivorous fishes, positing that different δ 13 C EAA profiles of phytoplankton assemblages would propagate via mesozooplankton to zooplanktivore fishes. Finally, to assess the relative performance of CSIA versus bulk SIA in differentiating functional groups, we obtained bulk isotope (δ 13 C and δ 15 N) values for a subset of the samples assessed with CSIA and compared the niche separation among functional groups with niche metrics analysis (Jackson, Inger, Parnell, & Bearhop, 2011).

Box 1 Carbon isotope fingerprinting of essential amino acids (EAAs)
This conceptional model depicts δ 13 C EAA values of consumers feeding in both estuarine and marine habitats. The consumers and their potential food sources mirror δ 13 C baseline values along this salinity gradient, and the δ 3 C EAA intramolecular variability is from Larsen et al. (2015). The two plots in the left pane (a and c) are based on baseline δ 13 C EAA values, and the two plots in the right pane (b and d) are based on δ 13 C EAA values centered to the δ 13 C mean across all EAAs of a given sample. (a) Varying biogeochemical conditions across the estuarine-marine gradient cause highly variable δ 13 C EAA values. (b) This variability is greatly reduced within each food source when centring the δ 13 C EAA values of each sample to the mean of all five EAAs. (c) To find out which combination of variables explain most of the variability among the three food sources, we applied principal component analysis (PCA), an unsupervised dimensionality reduction method. Prior to the PCA, we omitted lysine because it is the least informative EAA for separating the three food groups. Since the PCA is based on baseline δ 13 C values, the PCA factor scores (PC1 and PC2 coordinates) are influenced by both baseline and intermolecular δ 13 C variability. (d) By using mean-centered data in the PCA, we have generated a δ 13 C EAA fingerprint where the resulting factor score variability within each group is reduced substantially. By factoring out δ 13 C baseline variability and instead using the source diagnostic power of δ 13 C EAA fingerprinting, it is now evident that regardless of habitat use all three consumers derive most of their dietary EAAs from Food-III. Abbreviations used on the x-axes in a and b: Ile = isoleucine, Leu = leucine, Lys = lysine, Phe = phenylalanine, and Val = valine.

| Study system
The Baltic Sea is a shallow (mean depth 58 m) temperate regional sea, which displays a strong salinity gradient from marine salinity (30 g/kg) at the connection to the North Sea in the west to near freshwater (2 g/kg) in the northeastern inner part (Meier, 2007). The Baltic environmental situation entails strong fluctuations in temperature and light availability, a horizontal salinity gradient and strong vertical stratification, low oxygen conditions in the deep parts of the basins (Carstensen, Andersen, Gustafsson, & Conley, 2014), and an F I G U R E 1 Sampling stations in the Baltic Sea for the AL476 cruise (fauna; filled red circles) and black filled squares for the IOW stations (phytoplankton monitoring; Wasmund et al., 2017). The color gradient on the map shows showing surface concentration of the chlorophyll-a on 15 April 2016 observed by satellite and supplemented by the results of the ecohydrodynamic model EcoSat (http:// satba ltyk.iopan.gda.pl). The four pie charts present the relative biomass fraction of major taxonomic algal groups integrating monitoring results from three cruises from January to May 2016 (Wasmund et al., 2017). "Het." is an abbreviation for heterotrophic Note: CSIA indicates the number of specimens analyzed for compound-specific stable isotope analysis and BSIA the number of specimens analyzed for bulk stable isotope analysis. abundant nutrient supply due to eutrophication (Gustafsson et al. 2012), with seasonal minima when nutrients are taken up during phytoplankton blooms. Due to an accumulation of anthropogenic pressures on a level that is expected for other coastal seas, the system has been coined a "time machine for the future coastal oceans" (Reusch et al., 2018).

| Fauna sampling
Sampling for this study took place during research cruise AL476 with research vessel ALKOR in April 2016 (see sampling stations in Figure 1). All specimens were measured (total length or diameter to the nearest mm, mass to the nearest g), and ca. 0.5 cm 3 of muscle tissue was taken for isotope analysis and immediately frozen at −20°C on board of the vessel for further analyses. Our sampling was designed with our two main research questions in mind: (a) can δ 13 C AA fingerprints differentiate feeding niches at functional group and species levels, and (b) Table 1 for a summary and Supplementary S1 for detailed information.

| Phytoplankton assemblages
Information of phytoplankton communities during the study period was obtained from publicly available plankton monitoring data

| Stable isotope analysis
Isotope data are expressed in delta (δ) notation: For the element E, the ratio of heavy (i) to light (j) isotope is measured in both sample and references (Coplen & Shrestha, 2016). To express the isotopic data as per mil (‰), they are multiplied by 1,000.
The isotope ratios are expressed relative to international standards; Vienna Pee Dee Belemnite (VPDB) for carbon and atmospheric air for nitrogen.

| Statistical analyses
All statistical analyses were performed in R version 3.5.1 (R-Development-Core-Team, 2018). To assess whether the EAAs in consumers originate from bacteria, fungi or marine phytoplankton, we applied linear discriminant function analysis (LDA) (R: MASS) using δ 13 C EAA training data from Larsen et al. (2013). To assess the power of differentiating among functional groups and among species with δ 13 C EAA data, we applied principal component analysis  (Jackson et al., 2011). We also plotted the 95% confidence interval of the groups' bivariate means, which is commonly referred to as the standard ellipse area (SEA). The niche width distribution of the entire community was defined according to community-level Layman metrics, that is, the total area of the convex hull encompassing the group means (TA c ) (Layman, Arrington, Montana, & Post, 2007). To make the community niche space comparable for the two isotope methods, we took the ratio between TA c and the combined SEAs. A higher number signifies a greater niche width separation.

| Biosynthetic origins of the essential amino acids
According to our LDA using training data of broad phylogenetic groups, phytoplankton were the primary EAA source for all consum-

| δ 13 C EAA fingerprints across Baltic regions
The δ 13 C EAA fingerprints of clupeids from the four Baltic Sea regions show region-specific clustering of most herring and sprat specimens (Figure 5a,b). The separation is stronger for herring (Pillai's trace = 1.05, F 6,32 = 5.9; p < .001) than for sprat (Pillai's trace = 0.90, F 6,32 = 4.4; p < .01) due to larger principal component variability of the Arkona Basin specimens. The region-specific separation becomes weaker when joining the two clupeid species (see Figure S1).

| Niche differentiation with compound and bulk isotope methods
The differentiation of the functional groups' feeding niches, visualized based on biplots of δ 13 C EAA discriminant scores and bulk δ 13 C and δ 15 N isotope data, is shown in Figure 6a,b. The comparison of the groups' niche areas shows that benthic flatfishes and planktivores overlap in both the δ 13 C EAA (38.3%) and bulk (34.6%) biplots.
In contrast, while there are no other group overlaps in δ 13 C EAA biplot, there are several group overlaps in the bulk biplot: benthic predators and suspension feeders (22.4%), benthic flatfishes and benthic predators (6.7%), and benthic flatfishes and suspension feeders (9.2%).
Moreover, the greater TA c to SEA ratio for δ 13 C EAA (1.68) than bulk isotope (0.72) indicates a higher level of niche separation for the latter than the former.
To investigate niche differentiation, we pooled Kiel Bight and Arkona Basin specimens because on a species level, baseline δ 13 C and δ 15 N differences between the two locations are inconsistent.
For example, European flounder is significantly more 13 C enriched in

| D ISCUSS I ON
With accelerating global and regional environmental changes, improved understanding of food web structures is essential to project the corresponding changes of biological systems (Barth, Walter, Robbins, & Pasulka, 2020;Kortsch, Primicerio, Fossheim, Dolgov, & Aschan, 2015). Here, we provide a systematic assessment of the potential of CSIA to provide insights into resource partitioning and trophic niche differentiation among marine consumers in the Baltic Sea; a rapidly changing sea with a strong spatial environmental gradient.

| Understanding niche differentiation and resource partitioning with CSIA
Our results show that the δ 13 C EAA fingerprinting method holds considerable potential for identifying feeding differences in marine habitats. In our two westernmost Baltic locations, the Kiel Bay and the Arkona Basin, we were able to identify niche differentiation among all putative functional groups, as well as most species. This differentiation is in agreement with previous knowledge based on traditional methods like stomach content analysis, for example, Hislop et al. (1997). Species with similar modes of feeding clustered closely. It is surprising, however, that sea stars clustered very differently than bivalves, considering that blue mussels are considered a major prey (Sommer, Meusel, & Stielau, 1999).
We posit that such mismatches do not pertain to limitations of the fingerprinting method, but rather limited sampling and analysis of relevant endmembers because sea stars also feed on other F I G U R E 4 Principal component analysis for species using δ 13 C EAA values centered to the EAA mean of consumers from Kiel Bay (a) and Arkona Basin (b), respectively. Values in parentheses are the percentage variations accounted by each axis. In (a and b), the first two axes account for 84% and 83% of the variations, respectively. The ellipses signify 95% confidence boundaries for each group. Amino acid abbreviations: isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), phenylalanine (Phe), threonine (Thr), and valine (Val) invertebrates such as sponges, snails, and isopods (Anger, Rogal, Schriever, & Valentin, 1977). Similarly, the closer clustering of benthic flatfishes with planktivorous fishes than other benthic piscivores may be related to similar phytoplankton sources fueling their respective compartments of the food web. However, a more systematic sampling approach would be required to more fully characterize this benthic-pelagic coupling. Taken together, our results highlight the potential of δ 13 C EAA fingerprinting to elucidate the dietary niches of marine consumers, and how fluxes of carbon and nutrients from primary producers to detritus and consumers structure marine ecosystems (Cebrian, 1999;Lartigue & Cebrian, 2012).
The highly dynamic and complex nature of marine food webs can make it challenging to assess trophic relationships between consumers and producers, particularly on a taxon-specific level (Armengol, Calbet, Franchy, Rodríguez-Santos, & Hernández-León, 2019;Woodward, Speirs, Hildrew, & Hal, 2005). The clear spatial and trophic group differences observed in our study underscore the potential of δ 13 C EAA fingerprinting to determine the trophic basis of production, that is, how particular production sources are linked to consumers. At the same time, it is important bearing in mind that consumer fingerprints will lag behind primary producer fingerprints and that lower level consumers will integrate more recent photosynthates in their tissue than higher level consumers. Hence, frequent sampling would be needed to establish a more holistic picture of trophic connectivity and niche differentiation. Likewise, it will be critical for future studies to establish a reference phytoplankton library based on well-characterized in situ algal assemblages and single-species cultures. Increased application of this method to identify the taxonomic groups fueling production on higher trophic levels could improve our understanding of trophic links in many marine food webs and reduce the current bias toward larger prominent species feeding on clearly identifiable food items.
The fingerprinting method is well suited for quantifying inconspicuous sources because laboratory cultures of bacteria, phytoplankton, and other potential endmembers can be used as a proxy for in situ samples (Arthur, Kelez, Larsen, Choy, & Popp, 2014;Larsen et al., 2013;Rowe et al., 2019). For example, the fingerprinting method has yielded invaluable insight into detritus-based energy channels in soil food webs (Larsen, Pollierer, et al., 2016;Pollierer et al., 2019). Our study did not examine detrital feeders, but bacterial and fungal EAA contributions were undetectable even among benthic and suspension feeders.
Since dead phytoplankton biomass usually undergoes a distinct F I G U R E 5 Principal component analysis (PCA) with δ 13 C EAA values centered to the EAA mean of herring (a) and sprat (b), respectively. The convex hulls represent the maximum range in PC1 and PC2 scores for each of the four sampling locations. The most important EAAs for variations among locations are displayed in two first ordination components. Values in parentheses are the percentage variations accounted by each axis. In (a and b), the first two axes account for 95% and 92% of the variations, respectively. For a PCA with both species, see Figure S1. Amino acid abbreviations: isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), phenylalanine (Phe), threonine (Thr), tyrosine (Tyr) and valine (Val) succession of biotic activity and chemical decomposition (Azam & Malfatti, 2007;Biddanda, 1988), we attribute the lack of bacterial fingerprints to two independent processes. First, bacteria lack certain sterols and fatty acids essential for most metazoans (Phillips, 1984), which may explain why consumers such as the ocean quahog feed on recent primary production sources rather than organic matter from surface sediments (Erlenkeuser, 1976;Larsen, Yokoyama, & Fernandes, 2018). Second, the rate by which bacteria rework phytoplankton-derived EAAs appears to be very slow possibly because microbes assimilate them directly into their tissue rather than synthesizing them de novo (Hannides, Popp, Choy, & Drazen, 2013;Larsen et al., 2015). Our results confirm that tracing detrital-based energy channels in marine food webs is challenging and may require additional tracer techniques such as bacterial fatty acid biomarkers (Hayakawa, Handa, Kawanobe, & Wong, 1996;Taipale et al., 2015) and DNA metabarcoding of gut content (Fernández-Álvarez, Machordom, García-Jiménez, Salinas-Zavala, & Villanueva, 2018). While the latter two methods are unsuited for quantifying relative nutritional contributions, they provide important information on how detrital processes enter and alter marine energy channels.

| Assessing spatial differences in marine consumers and food webs with CSIA
Spatial isotope differences of marine consumers can inform about underlying differences in the organic matter at the base of food webs, as well as migration patterns of individuals (Hansson et al., 1997;McMahon et al., 2012;Torniainen et al., 2017). The geographically distinct δ 13 C EAA fingerprints of herring and sprat observed in our study would be consistent with limited mixing among schools from the different locations, that is, spatial population structuring, in combination with the presence of different phytoplankton assemblages or isotopic baselines among locations. This corresponds well with monitoring studies of phytoplankton highlighting the change in assemblages along the environmental gradient in the Baltic Sea, and with different baselines linked to spatially variable terrestrial organic matter inputs (Rolff & Elmgren, 2000;Wasmund et al., 2017) ( Figure 2). The additional observation of substantial variability within the same locations for both sprat and herring could be related to sizerelated differences in feeding (Casini, Cardinale, & Arrhenius, 2004;Kleppel, 1993;Last, 1989) as well as differences in migrations, both between areas (Aro, 1989;Gröhsler, Oeberst, Schaber, Larson, F I G U R E 6 Niche spaces of Kiel Bay and Arkona Basin functional groups based on multivariate δ 13 C EAA (a) and bivariate bulk isotope (b; δ 13 C lipid corrected and δ 15 N) values. To represent the δ 13 C EAA data in a bivariate space (see Figure 3 for the independent variables), we used the first two linear discriminant scores encompassing 97.8% of the variability. The groups depicted here contain ≥10 specimens. The niche spaces are visualized by 95% confidence interval around the bivariate means, also called the standard ellipse area (SEA; inner ellipses with broken lines) and 95% prediction ellipses (outer ellipses with full lines), respectively. The convex hulls encompassing the group means are denoted TA c for the total area of each community. The community niche space is generally more separated for δ 13 C EAA than bulk isotopes; for example, there is only one overlap of the prediction ellipses between functional groups in the δ 13 C EAA biplot, but four overlaps in the bulk biplot & Kornilovs, 2013;Jørgensen, Hansen, Bekkevold, Ruzzante, & Loeschcke, 2005) and in the case of herring between coastal and offshore areas during spawning runs (Šaškov, Šiaulys, Bučas, & Daunys, 2014). Our finding suggests that with further development, δ 13 C EAA fingerprinting have the potential to complement telemetric (Chittenden, Ådlandsvik, Pedersen, Righton, & Rikardsen, 2013;Pincock, Welch, McKinley, & Jackson, 2010) and bulk isoscape (Soto, Wassenaar, & Hobson, 2013;St. John Glew, 2019;Torniainen et al., 2014Torniainen et al., , 2017 approaches to track migration of single species in offshore systems, as well as novel migration trackers such as δ 15 N AA (Matsubayashi et al., 2020) and bulk radiocarbon analysis . It could also provide further and much needed insight into dietary response to changing physiochemical conditions (Casini et al., 2004;Kulke, 2018), When we pooled the two clupeid species, the spatial differentiation of the δ 13 C EAA fingerprints weakened considerably. Although the two clupeid species have substantial dietary overlap as corroborated by our results (Figures 3 and 4), it is important to note that both species differ in their dietary preferences. For example, adult herring are not strictly zooplanktivorous; they can opportunistically shift from pelagic to benthic prey by feeding on nektobenthos, that is, consumers such as mysids and amphipods that tend to migrate daily in the water column (Casini, Bartolino, Molinero, & Kornilovs, 2010;Kiljunen et al., 2020). In contrast, all size classes of sprat are strictly zooplanktivorous (Casini et al., 2004). By pooling the two species, we therefore increased variability within each locations. The resulting loss in spatial differentiation is in line with a previous δ 13 C EAA fingerprinting study in the southern Baltic Sea that found poor spatial differentiation after pooling multiple zooplankton species with different dietary preferences (Eglite et al., 2019). These results underline that to leverage the full power of the fingerprinting approach to track migratory patterns, it is important to focus on single species.

| Dietary niche differentiation with compoundspecific and bulk isotope approaches
Our ability to answer research questions in trophic ecology and food web studies depends on methodological approaches (Nielsen et al., 2018). EAAs are among the most powerful carbon tracers because EAA carbon backbones are usually passed through multiple trophic levels with minimal modifications in contrast to bulk carbon.
The substantially higher differentiation among functional groups and species with the δ 13 C EAA than the bulk isotope approach confirms the usefulness of EAA as high fidelity source tracers. At the same time, it is important to bear in mind that the EAA and bulk isotope approaches are not directly comparable. As demonstrated by our results, the multidimensional δ 13 C EAA niche space is powerful to delineate among primary producers at the base of the food chain, consumers supported by these different sources, and spatial differences among the same organisms from areas with different baselines. This advantage is highlighted by the differentiation between diet sources that are nearly indistinguishable in terms of δ 13 C baselines in past studies, such as marine phytoplankton and kelp (Vokhshoori et al., 2014). In comparison, the isotopic niches identified with the bulk method were less distinguishable, but may hold more easily interpretable information regarding trophic position and terrestrial versus marine or benthic versus pelagic production.
Moreover, the lower costs per analyzed sample and the larger repository of reference data (e.g. Pethybridge et al., 2018;de la Vega, Jeffreys, Tuerena, Ganeshram, & Mahaffey, 2019) compared to EAA can be practical considerations in particular for temporal comparisons or studies requiring large sample numbers (e.g., high spatialtemporal resolution). Ultimately, whether EAA or bulk SIA is the best approach will therefore strongly depend on the study question at hand; complementary use of both methods may in many cases be the optimum solution.

| Perspectives
Our study highlights the applicability of δ 13 C EAA fingerprinting in a regional sea with strong salinity and temperature gradients by differentiating among the trophic niches of both functional groups and species at an unprecedented resolution, and by identifying spatial fingerprinting differences of widely distributed species. These differences are likely driven by regional differences in basal resources, that is, algal composition, and the strength of trophic links between various phytoplankton producers and consumers. Our study also highlights how CSIA can provide new insights into food web structuring in spatially and temporally dynamic systems, and thus complement traditional tools in trophic ecology, including insights that are complementary to those from the "traditional" bulk stable isotope analysis.
Current marine food webs are predicted to be fragile and susceptible to structural changes with consequent alterations in the functioning of the ecosystem (Marina et al., 2018). As environmental changes are accelerating, it is crucial to understand whether and how quickly marine food webs can adapt to changes in phytoplankton assemblages (Barth et al., 2020) and top predator abundances (Kortsch et al., 2015). For this reason, it is key identifying and quantifying feeding interactions across trophic levels, from phytoplankton to zooplankton to higher trophic levels, but many of these interactions remain crucial knowledge gaps (Griffiths et al., 2017). The combination of δ 13 C EAA and the more affordable bulk stable isotope analysis holds considerable promise to address these gaps in the future.

ACK N OWLED G M ENTS
We thank Dr. Nils Andersen, Karsten Gramenz, and Robert Priester for technical assistance at the Leibniz Laboratory for Isotope Research (University of Kiel), and the scientific and permanent crew of RV Alkor cruise AL476 for their support during fieldwork. The study was supported by the Cluster of Excellence 80 "The Future Ocean," which is a framework within the Excellence Initiative by the Deutsche Forschungsgemeinschaft (DFG). Sampling on board of RV Alkor took place in the framework of the BONUS BIO-C3 project. TL was supported by the Germany's Federal Ministry of Education and Research (BMBF) via LOMVIA (03V01459) and JD was in part supported by the BONUS XWEBS project, both supported by BONUS (Art 185), funded jointly by the EU and the German BMBF.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data associated with this paper are available in the Supplementary Information and DRYAD: https://doi.org/10.5061/dryad.crjdf n321.