Amino acid δ15N underestimation of cetacean trophic positions highlights limited understanding of isotopic fractionation in higher marine consumers

Abstract Compound‐specific stable isotope analysis (CSIA) of amino acids (AAs) has been rapidly incorporated in ecological studies to resolve consumer trophic position (TP). Differential 15N fractionation of “trophic” AAs, which undergo trophic 15N enrichment, and “source” AAs, which undergo minimal trophic 15N enrichment and serve as a proxy for primary producer δ15N values, allows for internal calibration of TP. Recent studies, however, have shown the difference between source and trophic AA δ15N values in higher marine consumers is less than predicted from empirical studies of invertebrates and fish. To evaluate CSIA‐AA for estimating TP of cetaceans, we compared source and trophic AA δ15N values of multiple tissues (skin, baleen, and dentine collagen) from five species representing a range of TPs: bowhead whales, beluga whales, short‐beaked common dolphins, sperm whales, and fish‐eating (FE) and marine mammal‐eating (MME) killer whale ecotypes. TP estimates (TPCSIA) using several empirically derived equations and trophic discrimination factors (TDFs) were 1–2.5 trophic steps lower than stomach content‐derived estimates (TPSC) for all species. Although TPCSIA estimates using dual TDF equations were in better agreement with TPSC estimates, our data do not support the application of universal or currently available dual TDFs to estimate cetacean TPs. Discrepancies were not simply due to inaccurate TDFs, however, because the difference between consumer glutamic acid/glutamine (Glx) and phenylalanine (Phe) δ15N values (δ15NGlx‐Phe) did not follow expected TP order. In contrast to pioneering studies on invertebrates and fish, our data suggest trophic 15N enrichment of Phe is not negligible and should be examined among the potential mechanisms driving “compressed” and variable δ15NGlx‐Phe values at high TPs. We emphasize the need for controlled diet studies to understand mechanisms driving AA‐specific isotopic fractionation before widespread application of CSIA‐AA in ecological studies of cetaceans and other marine consumers.


| INTRODUC TI ON
Trophic connections among producers and consumers contribute to ecosystem structure, function, and stability (e.g., Polis & Strong, 1996;Post, 2002;Worm & Duffy, 2003). Indirect characterization of marine food webs using isotopic analysis of bulk tissues (Rau, 1982) has become routine, particularly for estimating the trophic position (TP) of cetaceans and other marine mammals (Lesage, Hammill, & Kovacs, 2001;Newsome, Clementz, & Koch, 2010). Nitrogen (N) isotope ratios differ by several per mil between food and consumer, and therefore serve as a proxy for TP (DeNiro & Epstein, 1981;McCutchan, Lewis, Kendall, & McGrath, 2003). However, in addition to the suite of dietary (e.g., protein content) and physiological factors (e.g., nutritional state and growth rate) that also influence 15 N discrimination (Gorokhova, 2018;Nuche-Pascual, Lazo, Ruiz-Cooley, & Herzka, 2018;Robbins, Felicetti, & Sponheimer, 2005;Trueman, McGill, & Guyard, 2005), underlying biogeochemical processes impart distinct baseline δ 15 N values to entire food webs (e.g., McClelland, Holl, & Montoya, 2003;Ruiz-Cooley, Koch, Fiedler, & McCarthy, 2014). Variation in isotopic baselines across a range of spatiotemporal scales can equal or vastly exceed typical trophic 15 N enrichment (Hannides, Popp, Landry, & Graham, 2009;McMahon, Hamady, & Thorrold, 2013;Rolff, 2000), often leading to the question "do (bulk) nitrogen isotope differences among consumers reflect diet differences or foraging within isotopically distinct food webs?" The confounding influences of trophic and baseline isotopic variation on bulk tissue SI values can be resolved with concurrent measurement of the isotopic composition of primary producers or available prey. This approach, however, has inherent challenges. The fast growth rates and nutrient uptake of phytoplankton, for example, lead to short-term isotopic variation that is mismatched with the longer integration periods in consumer tissues (Hannides et al., 2009). Baseline characterization is challenging to resolve for marine mammals that occupy large geographic ranges, and is especially problematic for migratory species whose movements span pronounced regional and seasonally variable isotope gradients. The increasingly popular approach of reconstructing long-term diets from isotopic profiles of baleen and teeth (Matthews & Ferguson, 2014Newsome, Etnier, Monson, & Fogel, 2009;Pomerleau et al., 2018) also introduces temporal baseline SI variation over the period of tissue growth as an additional confounding factor. In such retrospective studies, baseline or prey SI databases over matching temporal scales, which often exceed decades, are typically nonexistent.
Compound-specific stable isotope analysis (CSIA) of individual amino acids (AAs) offers a means to tease apart trophic and baseline contributions to bulk tissue isotopic variation that circumvents these challenges. Amino acids designated as trophic AAs enter metabolic pathways involving transamination and deamination reactions, during which isotopic discrimination causes 15 N enrichment of the AA pool (Chikaraishi, Kashiyama, Ogawa, Kitazato, & Ohkouchi, 2007). Source AAs, on the other hand, predominantly enter metabolic pathways in which the amine bonds remain intact, such that primary producer δ 15 N values are conserved with minimal 15 N enrichment throughout the food web (Chikaraishi et al., 2007(Chikaraishi et al., , 2009McClelland & Montoya, 2002). The relative difference between trophic and source AA δ 15 N values of consumer tissues therefore allows for internal calibration of TP while accounting for baseline isotopic variation, making CSIA-AA ideal for both diet and distribution studies.
Pioneering empirical studies of green algae, zooplankton, and fish larvae measured the relative 15 N enrichment of multiple AAs with trophic transfer (Chikaraishi et al., 2007(Chikaraishi et al., , 2009McClelland & Montoya, 2002). Among source AAs, phenylalanine (Phe) δ 15 N values (δ 15 N Phe ) were the most conservative, with only a slight increase of ~0.4‰ with each trophic transfer. Glutamic acid + glutamine (Glx; see Methods), on the other hand, exhibited consistent, high 15 N enrichment of ~8‰ with each trophic transfer (Chikaraishi et al., 2009). It was therefore proposed that consumer TP be calculated as: where δ 15 N Glx and δ 15 N Phe are the consumer δ 15 N values of those AAs, β Glx-Phe is the difference between primary producer δ 15 N Glx and δ 15 N Phe values (3.4 ± 0.9‰ in marine cyanobacteria and algae), and TDF Glx-Phe is the trophic discrimination factor, or the difference in fractionation of Glx (Δ 15 N Glx ) and Phe (Δ 15 N Phe ) with each trophic step (7.6 ± 1.2‰; Chikaraishi et al., 2009). Note that the original publication, along with most other cited ecology studies, uses the abbreviation Glu; here, we use Glx to specify the AAs that are actually measured (see "Methods").

| Specimen collection
Bowhead whale skin biopsies (n = 10 whales) were collected from free-ranging animals at Disko Bay, Greenland, using a crossbow.
Baleen was collected from subsistence hunted whales (n = 2 different whales, from the same population) in the eastern Canadian Arctic. Dolphin skin was collected from animals (n = 9) killed incidentally in gillnet fisheries in the Southern California Bight, California, USA. Sperm whale skin was biopsied from free-ranging and stranded animals (n = 13) in the upper California Current. Sperm whale teeth were collected from commercially harvested whales (n = 6) off the coast of Peru (Clarke, Paliza, & Aguayo, 1988). Beluga skin (n = 4 whales) and teeth (n = 9 different whales from the same population) were collected from subsistence hunted animals in the eastern Canadian Arctic. Finally, teeth were collected from genetically assigned FE (n = 3) and MME (n = 4) killer whale ecotypes stranded around Vancouver Island, Canada (G. Hanke, Royal British Columbia Museum, Pers. Comm.). Tissues were frozen at −20°C with no preservative (bowhead baleen, common dolphin skin, and beluga skin and teeth), frozen at −20°C in 20% dimethyl sulfoxide (DMSO; bowhead and sperm whale skin), or stored dry at room temperature (killer whale teeth).

| Sample preparation
Most of the isotope data presented here have been compiled from previously published studies where detailed sample preparation and analysis procedures can be found (Matthews & Ferguson, 2014Pomerleau et al., 2017;Ruiz-Cooley et al., 2014 Zupcic-Moore, Ruiz-Cooley, Paliza, Koch, & McCarthy, 2017).
Briefly, baleen samples were drilled from the proximal end of each plate where the most recent growth corresponds to foraging on the summer grounds (Matthews & Ferguson, 2015), and no further sample preparation was carried out prior to isotope analysis. Bowhead whale skin samples were rinsed of DMSO using deionized water and were not lipid-extracted prior to analysis. Sperm whale skin samples were also rinsed of DMSO using deionized water and then lipid-extracted using a 2:1 chloroform:ethanol mixture (Lesage et al., 2010;Ruiz-Cooley, Engelhaupt, & Ortega-Ortiz, 2012). Dolphin skin was thawed and lipid-extracted with petroleum ether. Annual dentine growth layers of sperm whale teeth were sampled using a micromill and later combined, while a handheld rotary tool was used to collectively sample multiple dentine growth layers of beluga and killer whale teeth. All dentine was demineralized using repeated washes of 0.25 N HCl for 12-hr periods, and the remaining collagen was rinsed with distilled H 2 O. All samples except baleen were freeze-dried and finely homogenized. The suite of AAs that can be accurately quantified depends on the derivatization agent  and tissue AA content; we include here the nine AAs that were measured in all samples: Glx, aspartic acid (Asx), alanine (Ala), isoleucine (Ile), leucine (Leu), proline (Pro), valine (Val), glycine (Gly), and Phe. We note that acid hydrolysis converts glutamine (Gln) and asparagine (Asn) to glutamic acid (Glu) and aspartic acid (Asp), respectively. Glx (Glu + Gln)

| Compound-specific stable isotope analysis
and Asx (Asp + Asn) are the IUPAC-recognized abbreviations for the resultant AA combinations that are measured.

| Trophic position estimates and data analysis
TP CSIA estimates were calculated using four published equations that use the trophic-source AA pair Glx and Phe (Table 1), with the exception of equation (2) Equations (1) and (2), based on the work of McClelland and Montoya (2002) and Chikaraishi et al. (2007Chikaraishi et al. ( , 2009, apply a single TDF Glx-Phe of 7.6‰ to all trophic transfers, while equations (3) and (4) apply a dual TDF Glx-Phe to account for variation among trophic transfers. Equation (3) (1) and (2) apply a blanket trophic discrimination factor (TDF) to all trophic transfers, while equations (3) and (4) incorporate a dual TDF to account for TDF variation in higher consumers. Note that the original publications use the abbreviation Glu instead of Glx   (Pauly et al., 1998). in sperm whale dentine to 14.31‰ in MME killer whales; Table 2).

| RE SULTS
Notably, Phe δ 15 N values were ~4‰ higher in MME (14.31‰) than FE (10.48‰) killer whales. δ 15 N values of Phe and Gly were similar between bowhead whale baleen and skin, but differed between skin and dentine collagen of beluga and sperm whales. Gly in particular differed by more than 10‰ (Table 2 and Figure 1).
TP CSIA estimates using all equations were lower than published TP SC (Table 3). Equations (1) and (2) produced TP CSIA estimates that were generally 1 to 2.5 positions lower than TP SC for all species (Table 3). Equations (3) and (4), which apply dual TDFs to account for lower TDF in higher consumers, produced TP CSIA estimates that were generally comparable to TP SC for bowhead whales, belugas, and FE killer whales, but 0.5-1.5 positions lower than TP SC for dolphins and sperm whales, and 2-2.5 lower for MME killer whales (Table 3). Tissue-specific TP CSIA estimates were similar within species (Table 3).
Mean δ 15 N Glx-Phe values did not follow TP SC order. The lowest values were in bowhead whales (12.57 and 13.50‰ in baleen and skin, respectively) and, notably, MME killer whales (13.46‰;

| D ISCUSS I ON
This study presents the most comprehensive compilation of AA δ 15 N values and TP CSIA estimates for cetaceans, which have been underrepresented in recent meta-analyses of marine consumer CSIA-AA due to lack of published data, and for which no published controlled diet studies exist. Our findings are consistent with general patterns across a range of other taxa from diverse marine ecosystems that have shown variable AA isotopic fractionation (Bradley et al., 2015;McMahon & McCarthy, 2016;Nielsen et al., 2015) and low TP CSIA estimates for higher TP consumers (Dale et al., 2011;Lorrain et al., 2009Lorrain et al., , 2015Matthews & Ferguson, 2014;Ruiz-Cooley et al., 2013).
We anticipated lower TP CSIA estimates assuming a uniform TDF across all trophic transfers (Equations 1 and 2), given previous studies have shown that TDF Glx-Phe values in other high TP consumers like harbor seals (Germain et al., 2013) and penguins (McMahon, Polito, et al., 2015) were considerably lower than those measured in invertebrates and fish (Chikaraishi et al., 2007(Chikaraishi et al., , 2009McClelland & Montoya, 2002). Equation (1) assumes that δ 15 N Glx values are the highest among trophic AAs, which was not the case in any cetacean tissue. Other studies have also found slightly lower TDF Glx values relative to other trophic AAs such as Pro (e.g., Bradley et al., 2015). Higher δ 15 N Pro values in many of our samples might reflect its role in formation of collagen (Germain et al., 2013), a prominent protein in skin and dentine.
Averaging across the δ 15 N values of all trophic AAs using equation (2), which is intended to minimize the impact of such δ 15 N variation in any single AA, did not improve TP estimates, possibly because TP CSIA estimates using the dual TDF approach, which has been advocated to account for variation in trophic 15 N enrichment across TPs, were still often more than one TP lower than TP SC . The low TP CSIA estimates for all species using equation (3), which applies a smaller TDF of 4.3‰ to just the final trophic transfer (Germain et al., 2013), could possibly reflect differences in AA metabolism between the captive seals and wild cetaceans (driven, e.g., by rate and amount of food intake, protein content, metabolic processing, etc.). However, Germain et al.'s (2013) own TP CSIA estimate of 2.8 for harbor seals from which the equation was derived is also unrealistically low for seals fed wild-caught herring. Herring is a secondary consumer of zooplankton that itself occupies at a TP ~ 3 (Pauly & Christensen, 1995), thus putting the seals at an expected TP of ~4.

F I G U R E 1
Mean (solid diamonds) and individual (hollow circles) δ 15 N measurements of nine amino acids (AAs) in tissues (baleen, skin, and/or dentine) of five cetacean species TA B L E 3 Amino acid δ 15 N-derived trophic position (TP) estimates for five cetacean species compared against those established from stomach contents (TP SC ) from Pauly et al. (1998). TP estimates (1-4) were calculated using published equations ( Abbreviations: FE, fish-eating; MME, marine mammal-eating. a Pauly et al. (1998) calculated one value for killer whales (4.5), assuming approximately equal proportions of miscellaneous fishes and higher vertebrates, and lesser amounts of squids and pelagic fishes. Following their methodology, we calculated a FE killer whale TP assuming diet comprised 100% salmon (Ford & Ellis, 2006), which Pauly et al. (1998) assigned a TP of 3.3, and MME killer whale TP assuming diet comprised 100% higher vertebrates, which Pauly et al. (1998) assigned a value of 4.0. The mean trophic positions for each prey type are originally from Pauly and Christensen (1995).
TA B L E 3 (Continued) (Bradley et al., 2015), and may therefore be more appropriate for TP reconstructions (Fuller & Petzke, 2017). Unfortunately, Thr was not measured across all samples in this study and therefore cannot be evaluated here.
Baleen, skin, and dentine are composed of different proteins, and their unique AA compositions and metabolic rates may impart tissue-specific 15 N fractionation as each tissue draws differentially on AA pools during formation (see Schmidt et al., 2004). Sampled tissues were not from the same animals, preventing direct comparison of tissue-specific δ 15 N AA values. Relative patterns of variation for trophic AAs were nevertheless largely similar between tissues and across species, suggesting that the metabolic processes driving their isotopic fractionation follow similar biochemical pathways in cetaceans. A notable exception was Gly, which had considerably lower δ 15 N values (~10‰) relative to Phe in beluga and dolphin skin, but not in bowhead and sperm whale skin. Skin from by-caught dolphins and hunted belugas that were potentially molting may have been subjected to some degree of bacterial degradation, as opposed to freshly biopsied bowhead and sperm whale skin. Calleja, Batista, Peacock, Kudela, and McCarthy (2013), however, showed bacterially degraded organic nitrogen had approximately 15‰ higher δ 15 N Gly values than fresh material, which is inconsistent with the lower values we observed in skin. Baleen, skin, and teeth are routinely sampled from cetaceans during field research programs F I G U R E 2 δ 15 N Glx-Phe values in five cetacean species plotted against their estimated trophic position from stomach contents (Pauly et al., 1998). The lack of relationship between δ 15 N Glx-Phe and trophic position (predicted to be positively correlated) indicates δ 15 N Glx-Phe is not a reliable proxy for relative trophic position in these species F I G U R E 3 δ 15 N Glx-Phe (left panel) and δ 15 N AveTrop-Phe (right) values in five cetacean species ordered by diet type (zooplankton eating bowhead whales, squid-eating sperm whales, fish/ invertebrate eating dolphin and beluga whales, fish-eating killer whales, and marine mammal-eating killer whales) and necropsies, and tissue-specific AA-specific δ 15 N variation merits further study (preferably using multiple tissues sampled from the same individuals) to understand how tissue selection might impact TP CSIA estimation.
Some degree of discrepancy between TP CSIA and TP SC estimates could reflect inaccurate diet assumptions, as stomach contents may be biased toward recent diet and items with differential digestion rates (Bowen & Iverson, 2013). However, we have included species whose diets have been well characterized through meta-analysis of numerous studies (Pauly et al., 1998). Population-specific and individual diet differences can be considerable, but not to the degree required to make sense of TP CSIA estimates that are off by 1-2 TPs.
Variation in ß Glx-Phe (see Vander Zanden et al., 2013) would also lead to erroneous TP CSIA estimates using equations that assume a constant ß across different marine food webs. Back-calculated estimates of ß from regression analyses of hundreds of marine consumer δ 15 N AA values (Bradley et al., 2015;Nielsen et al., 2015), however, are consistent with the value used in our calculations (Chikaraishi et al., 2009). ß values of sea grasses and terrestrial C3 plants can be more than 10‰ lower than marine algae (Chikaraishi et al., 2009;Vander Zanden et al., 2013), which could explain the low TP CSIA estimates of dolphins feeding in coastal food webs with sea grass or allochthonous terrestrial inputs (see Barros, Ostrom, Stricker, & Wells, 2010). While we know of no measured ß values in ice algae, which contribute considerably to the Arctic food webs of bowhead whales and belugas (Brown et al., 2017), TP CSIA estimates were no worse for these species than the others. δ 15 N values (Matthews & Ferguson, 2014;Yarnes & Herszage, 2017).
We therefore conclude that our low TP CSIA estimates are not an artifact of derivatization method, which is supported by the similar results for sperm whale samples derivatized using trifluoroacyl-isopropyl esterification. This argument, however, will be unavoidable until international standards are used for interlaboratory CSIA-AA calibration, as is routinely done with bulk SIA.
In the absence of accurate TDFs and estimating equations, we had anticipated that δ 15 N Glx-Phe would at least serve as an index of relative TP; that is, the relative difference in δ 15 N Glx-Phe among consumers would follow TP SC order. Nonsystematic differences in δ 15 N Glx-Phe with respect to TP SC are nevertheless consistent with recent meta-analyses that, despite finding significant overall positive correlations with TP, report considerable variation in δ 15 N Glx-Phe spanning 10‰ in higher consumers (Bradley et al., 2015;Nielsen et al., 2015). McMahon and McCarthy (2016) found TDF Glx-Phe values across 70 species varied predominantly with mode of N excretion and diet quality. Because our analysis focused on a single infraorder, mode of N excretion cannot account for the observed δ 15 N Glx-Phe variation. Reorganizing species based broadly on diet type (Figure 3), however, shows δ 15 N Glx-Phe differences may reflect diet quality differences. Zooplanktivorous bowhead whales had the lowest δ 15 N Glx-Phe , followed by offshore, primarily mesopelagic squid-eating sperm whales, and then fish-eating belugas and FE killer whales, a pattern that is consistent with the increasing deviation of TP CSIA with mean TP of feeding guild (invertivores vs. piscivores) of carnivorous fishes (Bradley et al., 2015). Incorporation of N derived from the foregut fermentation of chitin exoskeletons of crustacean prey (Herwig, Staley, Nerini, & Braham, 1984;Sanders et al., 2015) may have contributed to the δ 15 N Glx-Phe of bowhead whales, the only mysticete in our sample, although the relative importance of this potential influence is unknown. The lower (and considerably more variable) δ 15 N Glx-Phe values of common dolphins, which feed near the coast on small sized fish (mainly myctophids) and cephalopods, could reflect different primary producer inputs (see above), or prey assemblages from different length food chains related to temporal variation in environmental conditions (see Ruiz-Cooley et al., 2017).
The most surprising divergence from any apparent relationship between δ 15 N Glx-Phe and diet type was the unexpectedly low δ 15 N Glx-Phe of MME killer whales (Figure 3). The fish diets of belugas and FE killer whales and marine mammal diets of MME killer whales would presumably both be high in protein content, and Beach et al. (1943) report similar proportions of 10 AAs in muscle of fish and mammals, suggesting they would be of similar quality (i.e., AA composition).
However, it is possible that dietary AA imbalances (e.g., amounts present in the diet vs. those required for growth and metabolism) might be larger for whales feeding on fish than those feeding on other marine mammals, thereby leading to differences in isotopic Like previous studies attributing "compressed" TDF Glx-Phe in higher TP consumers to mechanisms affecting trophic 15 N enrichment of Glu (Chikaraishi et al., 2015;Germain et al., 2013;McMahon & McCarthy, 2016;, both of these studies linked variation in TDF Glx-Phe to variation in 15 N fractionation of Glu, since δ 15 N Phe values were essentially constant among treatments. Few studies have suggested 15 N enrichment of Phe as a contributing factor to variation in δ 15 N Glx-Phe with TP, since Phe is assumed to undergo negligible trophic 15 N enrichment (Bradley et al., 2015;Chikaraishi et al., 2009;McMahon & McCarthy, 2016;Nielsen et al., 2015). However, Nuche-Pascual et al. (2018)  Pacific. Finally, δ 15 N Phe values in skin of belugas (9.14 ± 1.3‰) were considerably higher than those of bowhead whale baleen grown in summer (mean 6.17‰), when both species share a similar distribution. Similarly, δ 15 N Phe values of common dolphins (9.97 ± 2.1‰) and sperm whales (10.74 ± 1.5‰) off the California coast were considerably higher than those of lanternfish (Myctophidae) from the same region (~4 to 6‰; Choy et al., 2012).
The large geographic ranges of these species introduce potential for integration of spatially and seasonally variable baseline SI values that precludes any rigorous conclusions regarding trophic 15 N enrichment of Phe. The only published controlled diet study on marine mammals, harbor seals, does not support this hypothesis (Germain et al., 2013). However, controlled diet studies of rats (Fuller & Petzke, 2017)

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

AUTH O R S' CO NTR I B UTI O N S
CJDM, RIR-C, CP, and SHF conceived the ideas and designed methodology. CJDM, RIR-C, and CP collected and analyzed the data.
CJDM and RIR-C wrote the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data analyzed in this paper have been uploaded to the Dryad Digital Repository (https://doi.org/10.5061/dryad.9kd51 c5d3).