Evaluating the importance of mesopelagic prey to three top teleost predators in the northwest Atlantic Ocean

Abstract

The ocean’s twilight zone is a vast area of the global ocean that lies between the sunlit surface waters and perpetually dark midnight zones, covering depths from ∼200 to 1000 m. Recent work in the twilight (or mesopelagic) zone has revealed unexpected biomass and diversity that may not only challenge scientific understanding of marine systems but also provide a new and largely untapped resource for fisheries harvest. A key knowledge gap in our understanding of the mesopelagic is how its food webs support foraging activity by commercially valuable, highly migratory top predators. Here, we use compound-specific stable isotope analyses to trace the flow of carbon through pelagic ecosystems in the northwest Atlantic to three predators: bigeye tuna (Thunnus obesus), swordfish (Xiphias gladius), and yellowfin tuna (Thunnus albacares). Temperate mesopelagic-associated carbon was estimated as both a direct and an indirect source of predator carbon, alongside temperate epipelagic and mixed epi-mesopelagic tropical carbon, via Bayesian mixing models. The contribution of temperate mesopelagic carbon to individual predators ranged from 5% to 94%, with means of 62%, 46%, and 28% for bigeye tuna, yellowfin tuna, and swordfish, respectively. We also found that carbon sources of predators shifted seasonally as they moved between temperate and tropical waters by contrasting tissues (liver, muscle) and season of sampling (summer, fall). These results inform our understanding of the adaptive value of deep diving behaviors in large marine predators and provide key estimates of food web linkages to inform multi-species fisheries management of both mesopelagic prey and migratory predators.

Introduction

The ocean twilight zone, also called the mesopelagic zone, lies between ∼200 and 1000 m water depth and is home to a remarkable diversity and abundance of life. Recent studies have estimated ∼10 billion tons of fish (Irigoien et al. 2014, Proud et al. 2019) and half of the ocean’s zooplankton biomass (Hernández-León et al. 2020) reside in the mesopelagic zone. Mesopelagic animals have numerous morphological and behavioral adaptations to life in the deep ocean, including diel vertical migration between the surface waters at night to feed and the mesopelagic by day to escape visual predators (Haddock and Choy 2024). The active transport of carbon by these diel vertical migrators is also likely a substantial component of the global biological carbon pump (Robinson et al. 2010, Aksnes et al. 2023, McMonagle et al. 2023). Small mesopelagic fishes have largely avoided commercial exploitation to date, with only short-lived and unsuccessful small-scale or exploratory fisheries since the ∼1970s (FAO 1997, Hoagland et al. 2019, Roberts et al. 2020, Standal and Grimaldo 2020, Pauly et al. 2021). Growing demand for fishmeal and fish oil (Tacon and Metian 2015, FAO 2022) has, however, increased interest in the exploitation of mesopelagic fish stocks. Yet the potential impacts of industrial-scale fishing on the ecosystem services provided by mesopelagic communities remain uncertain despite the high value of these services (St. John et al. 2016). For instance, a recent study found that carbon sequestration from surface waters to the deep ocean by the biological carbon pump was worth over |${\$}$|200 billion in market-value carbon prices (Hoagland et al. 2019).

The ecosystem services provided by mesopelagic fauna across the global ocean likely extend to providing important foraging resources for commercially important fishes and marine mammals of conservation concern. Numerous large marine predators make extensive movements from surface waters into the mesopelagic zone and beyond (reviewed in Braun et al. 2022). Deep-diving behavior has been observed in economically valuable fish species, phocid seals, and toothed whales, among others. Potential motivations for marine predators to make deep dives include thermoregulation (Teo et al. 2007), navigation (Willis et al. 2009, Klimley et al. 2017), and refuge from their own predators (Lam et al. 2020, Beltran et al. 2021). But for those predators that expend the necessary energy to make repeated dives to the mesopelagic over periods of hours to days, foraging is likely the primary motivation for most species and behavior modes (Braun et al. 2022). The list of large fish predators displaying deep-diving behavior includes globally distributed tunas and swordfish that, together with other billfishes, represent ten percent of the global landed value of marine-capture fisheries (FAO 2022). The value of economic services provided by mesopelagic biomass as forage for these fish stocks is, therefore, likely to be significant but is currently unknown.

Direct observations of mesopelagic habitat use by large, often highly migratory predators have proved challenging due to the difficulties of tracking individuals in dark and remote environments [but see, e.g. Yoshino et al. (2020)]. Gut content analyses provide a snapshot of the animal’s diet at a single moment in time and have been used to identify mesopelagic-associated prey in predator stomachs (e.g. Iglesias et al. 2023). However, gut content analysis has several shortcomings, including lack of recognition or identification of species without distinctive hard structures (e.g. bones) and the possibility for contents to be regurgitated upon capture (Hyslop 1980, Bowman 1986). Biochemical analyses of predators’ tissue, including stable isotope analysis (SIA) and fatty acid analyses, provide time-integrated estimates of trophic interactions that often complement tag or gut content-based results (Young et al. 2015).

Compound-specific SIA, usually focusing on individual amino acids, has been shown to be particularly effective at tracing carbon from basal resource channels such as primary producers, bacteria, and fungi through to species at top trophic levels (Howland et al. 2003, Larsen et al. 2013). Essential amino acids (EAAs) are transferred between trophic levels with negligible fractionation of carbon isotopes because consumers cannot synthesize the carbon skeletons of EAAs. Moreover, differences in biosynthesis pathways across broad phylogenies for building EAAs result in unique differences between EAA δ¹³C values among basal ocean food web members. These differences result in separation in multivariate space that propagates up from primary producers to top predators (Larsen et al. 2013). By establishing the unique isotopic carbon signatures of mesopelagic vs. other open ocean food webs, we can potentially quantify the relative contributions of various carbon sources in predator tissues and thus better resolve the reliance of predators on mesopelagic forage. The EAA analyses provide time-averaged estimates of source carbon at each trophic level, with the specific time interval determined by factors including the individual’s size and metabolism (MacAvoy et al. 2006) and the carbon turnover rates of the specific tissue sampled (Thomas and Crowther 2015).

Here, we investigated the sources of carbon contributing to the biomass of three large marine predators with differing spatial use of the mesopelagic: bigeye tuna (Thunnus obesus), yellowfin tuna (Thunnus albacares), and swordfish (Xiphias gladius). All three species consume mesopelagic prey, including squids, lanternfish, and pomfret (e.g. Duffy et al. 2017, Logan et al. 2021); however, it is unclear when during the diel cycle these trophic interactions occur. Swordfish mirror the diel vertical migration of mesopelagic zooplankton and nekton, spending nights in the surface mixed layer and descending to >500 m at dawn (e.g. Dewar et al. 2011). Bigeye tuna typically make a series of short, periodic daytime dives to mesopelagic depths of ∼400 m (e.g. Lam et al. 2014). Yellowfin also make short dives to the mesopelagic, but to shallower depths (200–300 m) and at a lower frequency than bigeye (e.g. Weng et al. 2009). Thus, these three species all may interact with mesopelagic prey in the near-surface at night and, due to their varying vertical movement strategies, may do so to different extents at mesopelagic depths during the day.

We evaluated the carbon isotope signatures of these three predators using SIA of EAAs and characterized the mesopelagic and epipelagic communities in which they feed using a sample collection from the northwest Atlantic Ocean. Furthermore, we compared carbon isotope signatures of reference temperate and tropical food webs to better resolve the influence of seasonal horizontal migrations on the dietary sources of these highly migratory predator species as they move between the northern and central Atlantic (Lam et al. 2014, Braun et al. 2019, ICCAT 2019). Finally, we described the relative carbon contributions of both these reference communities and of general pelagic end members at the base of ocean food webs (e.g. phytoplankton, bacteria) to predators to determine the reliance of our study species on mesopelagic food webs. Filling this knowledge gap of predators’ reliance on mesopelagic resources is necessary to manage current fisheries on tuna and swordfish in light of the growing interest for mesopelagic commercial fisheries to provide fishmeal and nutraceuticals (St. John et al. 2016).

Methods

Sample collection

To investigate sources of carbon assimilated by top pelagic predators in the northwest Atlantic, we sampled predators along with lower trophic-level mesopelagic and epipelagic fish species from temperate and tropical waters. Predator samples for carbon isotope analysis were collected from fish caught on commercial longline gear by the F/V Monica in waters seaward and adjacent to the shelf edge of the eastern USA from Massachusetts to New Jersey in 2019, 2020, and 2022 (Fig. 1; Table 1). We collected samples from bigeye tuna (Thunnus obesus, n = 49), yellowfin tuna (Thunnus albacares, n = 51), and swordfish (Xiphias gladius, n = 23). Samples were collected in summer (July 2019, August 2020 and 2022) and autumn (November 2019 and 2020). Liver tissue samples were collected from all fish (n = 123); white muscle samples were collected from a subset of individuals (n = 48). Both liver and muscle samples were collected to compare shifting carbon sources across the seasons, as they have different isotopic turnover times (Tieszen et al. 1983).

Four reference groups were used as proxies for representative carbon signatures at the base of food webs supporting tunas and swordfish. First, we used carbon isotope data from two species of myctophid (Hygophum hygomii and Lobianchia gemellarii) and one species of hatchetfish (Argyropelecus aculeatus) collected in the temperate northwest Atlantic Ocean (Fig. 1) on the R/V Bigelow in July 2018 (K. Gardner unpublished data) via the Multiple Opening/Closing Net and Environmental Sensing System (Wiebe et al. 1985) and processed for compound-specific SIA using the same methods as this study. All three species are documented to make diel vertical migrations from mesopelagic depths during the day to surface waters at night (Roe 1984, Linkowski 1996, Sassa et al. 2002). Second, we collected muscle samples from common dolphinfish (Coryphaena hippurus) and flying fish (Exocoetus obtusirostris) caught by pole and line and dip nets, respectively, on the F/V Monica during August 2022 to characterize epipelagic food webs in the same general temperate study area. Third, we obtained mesopelagic fishes from the stomach contents of swordfish caught by the F/V Trouble offshore of Miami, Florida, USA using deep-set buoy gear in March 2024 to characterize mesopelagic food webs in the western tropical Atlantic. The mesopelagic species included myctophids (Lampanyctus nobilis, Bolinichthys photothorax, and Diaphus dumerlii), snake mackerel (Gempylus serpens, Nesiarchus nasutus), dragonfish (Heterophotus ophistoma), fanfish (Pterycombus brama), and sawtooth eel (Serrivomer beanii). Visual identifications of these species were confirmed by DNA barcoding (see next section). Finally, we purchased common dolphinfish from a fishing charter (Captain Cook) in St. Croix, US Virgin Islands (Fig. 1), in April 2023 to characterize tropical epipelagic food webs. All common dolphinfish used in reference groups (both temperate and tropical) were ∼1 year or less of age, based on length (temperate samples) or weight (tropical samples) curves at age for Atlantic common dolphinfish (Oxenford 1999, Schwenke and Buckel 2008). Furthermore, the dolphinfish from the two collections were likely from separate stocks with distinct migration circuits (Oxenford and Hunte 1986). The temperate common dolphinfish were presumably foraging in the Gulf Stream region offshore from the USA, and the tropical common dolphinfish from the northeast coast of South America up to the Virgin Islands in the months prior to capture. Overall, these four groups represented the temperate mesopelagic (n = 62), temperate epipelagic (n = 16), tropical mesopelagic (n = 18), and tropical epipelagic (n = 4) food webs in which the predators feed.

The carbon signature of a northwest Atlantic ommastrephid squid Illex illecebrosus was also examined given its large proportional abundance in observations of stomach contents from the same field collections of bigeye and yellowfin tuna that we analyzed (Austin 2022). We used a collection of I. illecebrosus (n = 5) collected on the July 2018 research cruise for comparison to the reference groups and predator samples.

We used generic end members at the base of the food web to place the predator and reference animal samples in a larger carbon source context. Consumers sourcing their carbon from the surface ocean were assumed to have a carbon signature reflective of the particular phytoplankton community contributing to primary productivity at that location (e.g. Stahl et al. 2023). In contrast, consumers assimilating carbon from deeper in the water column were predicted to have a carbon signature resembling isotope values of heterotrophic bacteria, as microbially reworked sinking particles are an important basal food source in the deep ocean (Gloeckler et al. 2018, Boeuf et al. 2019, Hannides et al. 2020, Shea et al. 2023). We used four isotopically distinct pelagic end members (Table 2): diatoms (Stahl et al. 2023), dinoflagellates (Stahl et al. 2023), bacteria (Larsen et al. 2009, 2013), and cyanobacteria (K. Gardner unpublished data) for these analyses. Cyanobacteria samples were collected from lab cultures onto polycarbonate 0.2 µm filters and then processed for compound-specific SIA using the same methods as this study. Additional Synechococcus and Prochlorococcus culturing details can be found in Waterbury et al. (1986) and Moore et al. (2007), respectively. We note the limitation of the bacteria end member collection from Larsen et al. being largely terrestrial in origin; however, it remains the best available amino acid carbon stable isotope bacterial data and has been used for similar purpose in other marine studies (e.g. McMahon et al. 2015a).

DNA barcoding

The visual identifications of fish prey from tropical swordfish stomach contents were confirmed via DNA barcoding. Genomic DNA from prey tissue was extracted with DNEasy Extraction Kits (Qiagen, Germantown, MD, USA) following the manufacturer’s protocol, and the COI barcode marker was amplified with PCR as described in Quigley et al. (2023) and Govindarajan et al. (2023). The Fish F1/R1 primer set (Ward et al. 2005) was used successfully on most specimens; for fish identified as L. nobilis, B. photothorax, and N. nasutus, Fish F2/R2 (Ward et al. 2005) yielded high-quality amplicons. The PCR consisted of 95°C for 3 min, 35 cycles of 95°C for 30 s, 48°C for 30 s, 72°C for 1 min, and 72°C for 5 min. PCR success was assessed using 1.2% agarose gels stained with GelRed (Biotium, Hayward, CA, USA). Amplicons were purified with Qiaquick PCR purification kits (Qiagen, Germantown, MD, USA) and quantified with a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and amplicons were sequenced in both directions at Eurofins Genomics. Consensus sequences were generated from the raw reads in Geneious v.9.0.5 (Biomatters, Inc.) and were compared to sequences in GenBank using BLAST. Matches with >98% identity over at least 90% of the sequence length were considered positive species identifications. DNA sequences were deposited in GenBank (accession numbers PQ344747-PQ344760).

Sample preparation and analysis

Sample preparation and analysis for compound-specific SIA of amino acids generally followed the methods of Walsh et al. (2014). All fish tissue samples were frozen at sea or rapidly upon return to shore. In the lab, samples were freeze-dried overnight. Approximately 25 mg of freeze-dried tissue was sectioned, then hydrolyzed in 6 N hydrochloric acid under an atmosphere of nitrogen (N²), tightly capped, and placed on a heating plate at 110°C for 16–18 h. The hydrolyzed sample was dried under a stream of nitrogen. 0.01 M hydrochloric acid was added to reconstitute the sample prior to loading on a Dowex column 50WX8 50-100 (H). Samples loaded on the Dowex column were allowed to elute freely, with 2 M ammonium hydroxide added to the Dowex column to elute the amino acids. Samples were then dried to completeness under a stream of nitrogen at 110°C. A volume of 100 µl of 0.1 N HCl was added to each sample, and 50 µl of sample was pipetted into a V-shaped vial. A volume of 35 µl of methanol was added, followed by 30 µl of pyridine and 15 µl of methyl chloroformate. A volume of 100 µl of chloroform was added, and samples were thoroughly mixed. The chloroform layer containing the derivatized amino acids was transferred to an autosample vial for analysis via gas chromatograph isotope ratio monitoring mass spectroscopy (GC-irm-MS) at Woods Hole Oceanographic Institution (Woods Hole, MA, USA).

Carbon isotope analysis of amino acids was performed on a Finnigan MAT253 mass spectrometer using a high polarity VF-23MS GC column (30 m by 0.25 µm ID, 0.25 µm film thickness; Agilent Technologies, Santa Clara, CA, USA) with an injector temperature of 250°C. The amino acids were combusted after passing through a reactor containing two platinum and two nickel wires with the furnace temp set to 1030°C. Data were collected and analysed with the Isodat software package (Thermo Electron).

Samples were run in duplicate along with an amino acid standard of known isotopic composition and an internal lab standard (Atlantic cod white muscle). The amino acid standard was made from a mixture of highly purified individual AAs and was previously analysed by several outside labs to verify isotope ratios. These standards were used to account for kinetic fractionation and the introduction of carbon during derivatization (Silfer et al. 1991, Walsh et al. 2014). Methods are further described in McMahon et al. (2011). Five essential amino acids (EAAs: threonine, isoleucine, valine, phenylalanine, and leucine) and five non-EAAs (alanine, glycine, proline, aspartic acid, and glutamic acid) of sufficient peak height and separation were routinely measured in the samples. Long-term precision (RSD) of amino acid isotope measurements, calculated from repeated measurement of the internal lab standard, was 0.5.

Data analysis

All isotope data were initially mean normalized to account for baseline inorganic carbon differences among geographic locations and years by subtracting each sample's EAA mean from its individual EAA values. Any differences in baseline carbon values were therefore removed from the resulting analyses. Furthermore, mean normalization of values minimized any potential systematic offset in measured δ¹³C values from different lab facilities, as end member sample data from the literature were processed in several different labs.

We used principal component analysis (PCA) to visualize and interpret the multivariate signatures of EAA δ13C values. First, we generated a PCA using reference and predator samples only. We calculated a metric (ΔPC2_M-L) for comparing muscle and liver samples collected from the same fish by subtracting PC2 values for paired muscle and liver samples from the same individual. A positive value for this metric suggests that the isotopic composition of the muscle tissue is more similar to the tropical reference samples, while a negative value suggests muscle samples are closer to the temperate references. Second, we generated a new PCA parameterized with generic pelagic end members at the base of the food web (diatoms, dinoflagellates, bacteria, cyanobacteria). Reference and predator samples were then projected onto this end member PCA. This second PCA examined the broader food web context to examine the basal differences among the reference groups’ carbon signatures and to confirm patterns of similarity between reference groups and predator species.

Stable isotope mixing models (Stock et al. 2018) were used to assess the relative contribution of the food webs (represented by reference groups) to predator carbon at the individual level via the MixSIAR R package (Stock and Semmens 2016). Mixing models of EAAs can quantify the relative contribution of various putative carbon sources to a consumer. The models were run with EAA δ¹³C values from the reference groups or end members as source inputs. We assumed mean ± SD trophic discrimination factors of 0.1 ± 0.1 based on minimal modification of EAA δ¹³C values through the food chain (Howland et al. 2003, McMahon et al. 2015b). The models were run at length “very long” (1 000 000 iterations with an initial discard of the first 500 000 iterations as burn-in), at which point Gelman-Rubin and Gweke diagnostics indicated that the models had converged (see Supplemental Information for details). The two tropical reference groups (epipelagic and mesopelagic) were combined into one general “tropical” source group, as they fully overlapped in the PCA. Differences in mesopelagic carbon input with intraspecific animal size or year of collection were examined via statistically preliminary (due to low sample sizes) linear regression and ANOVA, respectively.

To examine changes in carbon source contributions calculated by the mixing model over horizontal migrations, we compared liver samples collected in the summer vs. fall and white muscle vs. liver samples from the same individual. Muscle and liver samples from the same individual were compared to examine individuals’ shifting carbon sources as they move seasonally due to longer tissue turnover rates in muscle compared to liver (Tieszen et al. 1983), e.g. ∼162 days for liver and 255 days for white muscle in southern bluefin tuna (Madigan et al. 2012). We hypothesized that liver samples from fall should primarily represent feeding over the temperate geographic range represented in the mesopelagic and epipelagic reference groups, while liver samples from the summer and muscle samples from all time points would have greater tropical contribution.

All statistical analyses described above and plotting were conducted in R (v 4.1.1, R Core Team 2023) with plots using the package ggplot2 (Wickham and Chang 2015).

Results

We quantified δ¹³C values of five EAAs (threonine, isoleucine, valine, phenylalanine, and leucine) from a total of 171 predator tissue samples and 100 reference group samples. Results of predator liver and temperate reference group samples are described first, then these are placed into the larger contexts of the tropical reference groups, predator muscle, and end member data.

Principal component analysis

Data for our initial PCA were comprised of the two temperate reference groups (epipelagic and mesopelagic) and liver samples from the three predator species. The results showed both a clear distinction between epipelagic and mesopelagic reference groups along PC1 (Fig. 2a) and overlap of liver samples among the three predator species (Fig. 2b). The carbon isotope signature of the squid I. illecebrosus was similar to that of epipelagic fishes. However, it was apparent that the predator samples extended beyond the reference samples along the second principal component, suggesting that the reference data set was missing an end member.

The addition of the tropical epi- and mesopelagic samples to our reference data set addressed this “missing end member” issue, as they were clearly separated from the two temperate reference groups on PC2 (Fig. 2a). Moreover, the three reference groups together approximately bounded the PCA carbon space of the predators (Fig. 2b). However, in contrast to the temperate reference groups, the tropical reference groups did not show a distinction between the signatures of epipelagic and mesopelagic samples and were combined into a single group for further analyses. No noteworthy differences in PCA results were observed among years of sampling for any of the three predator species (Fig. S1 and S2).

We also analyzed white muscle samples from a subset of the predators. Carbon in muscle tissue has a longer turnover time than liver and therefore may reflect feeding conditions at earlier points in time during the seasonal migration of the three predator species. Muscle samples were more similar to the tropical reference group than were the liver samples (Fig. 3a and b). Only 5 of a total of 48 muscle-liver comparisons had ΔPC2_M-L values < 0.

Mixing models

We used a Bayesian mixing model to quantify the relative contribution of carbon assimilated from our three reference prey groups (temperate epipelagic, temperate mesopelagic, tropical) in tuna and swordfish (Fig. 4, Table 3). Model results suggested that all three species sourced a substantial portion of their carbon from temperate mesopelagic communities, but with significant variability among individuals within a species. Posterior probability distributions were unimodal for all individuals (Fig. S3), and Gelman-Rubin and Gweke diagnostics suggested a satisfactory model fit (SI). Estimates of carbon from temperate mesopelagic sources in liver samples ranged from 22% to 93% in bigeye tuna, from 9% to 94% in yellowfin tuna, and from 5% to 70% in swordfish. Bigeye tuna had the greatest proportion of temperate mesopelagic carbon in their livers (mean ± sd: 62% ± 15%), followed by yellowfin tuna (46% ± 19%) and swordfish (28% ± 18%). The proportion of temperate mesopelagic carbon in muscle samples was lower than in the liver across all three species and ranged from 14% to 71% in bigeye tuna, from 5% to 87% in yellowfin tuna, and from 5% to 24% in swordfish. Muscle samples from yellowfin tuna had the greatest proportion of temperate mesopelagic carbon (mean ± sd: 32% ± 24%), followed by bigeye tuna (27% ± 20%) and swordfish (16% ± 7%). There were no meaningful observed differences in mesopelagic carbon input with intraspecific animal size or year of collection. To coarsely estimate the temperate-only division of diet sources between epi- and mesopelagic food webs, we subtracted out the tropical samples from the above liver model results. After this subtraction, our estimates of (temperate-only) mesopelagic carbon sourcing showed a small increase for bigeye (77% ± 12%) and yellowfin (67% ± 22%) tunas, and a more significant increase for swordfish (49% ± 18%).

Carbon assimilation from the different prey sources also varied by season (Table 3). For example, from summer to fall, the mean percent contribution of temperate mesopelagic carbon to swordfish livers increased by 30%, and to yellowfin tuna livers by 18%, with nearly equivalent summer-to-fall decreases in contribution from tropical sources. In contrast, the carbon sources of bigeye tuna livers changed little with season (seasonal change within 5% across all three reference groups). Muscle tissue composition across seasons could only be calculated for swordfish. Here, there were minimal shifts in carbon source with season (0%–6% by reference group), emphasizing the longer turnover times of muscle compared to liver tissue. No noteworthy differences in mixing model results were observed among years of sampling (Fig. S4) or fish length (Fig. S5) for any of the three predator species.

Pelagic end members

Putative pelagic end members (diatoms, dinoflagellates, bacteria, and cyanobacteria) were used to identify the ultimate sources of carbon at the base of the food that was assimilated by the consumers we sampled. Both reference fishes and predators were centrally located when projected into the end member PCA (Fig. 5). Our epipelagic reference group was located centrally to the phytoplankton groups (i.e. diatoms, dinoflagellates, and cyanobacteria; Fig. 5a), while the mesopelagic group primarily overlapped with the bacteria end member. The tropical group that contained both epipelagic and mesopelagic species aligned with cyanobacteria. The predator samples occupied a relatively wide area with a mean value across all predator samples that was close to the overall centroid of combined end member values (Fig. 5b).

Discussion

Identifying the sources of carbon contributing to migratory marine predators is important both for sustainable management of pelagic fisheries and timely given recent studies that have suggested significantly more biomass at mesopelagic depths in the open ocean than previously thought (Irigoien et al. 2014, St. John et al. 2016). Our study analysed carbon isotopic signatures of EAAs in white muscle and liver sampled from three large pelagic predators (swordfish, bigeye tuna, yellowfin tuna) with different patterns of mesopelagic habitat use. We then compared the carbon isotope signatures with members of three reference prey groups (temperate mesopelagic, temperate epipelagic, and a combined epi/mesopelagic tropical), and four pelagic end members (diatoms, dinoflagellates, cyanobacteria, and bacteria) at the base of food webs in the northwest Atlantic. We found that mesopelagic food webs contributed substantially to carbon assimilated by all three predator species.

We used stable carbon isotope signatures in EAAs to estimate relative contributions of different basal carbon end members to higher trophic levels in ocean food webs. The approach requires each of the end members to be constrained in multivariate isotope space. Earlier studies in ocean environments used results from Larsen et al. (2009, 2013) to constrain different algal and bacterial groups (e.g. McMahon et al. 2015). More recently, two studies have added greater taxonomic specificity by analyzing specific marine phytoplankton groups in the northwest Atlantic, including diatoms, dinoflagellates, and cyanobacteria (Stahl et al. 2023, K. Gardner unpublished data). However, we relied on the isotope data from terrestrial heterotrophic bacteria reported by Larson and co-workers, as δ¹³C EAA values from marine bacteria are not available as far as we are aware. The δ¹³C values of specific amino acids are generally thought to be phylogenetically constrained due to conserved amino biosynthetic pathways. Nonetheless, future food web studies in ocean environments would clearly benefit from a better understanding of taxonomical variability in δ¹³C EAA signatures within marine heterotrophic bacteria.

Swordfish, bigeye tuna, and yellowfin tuna are considered generalist predators with varied diets across their near-global ranges (Duffy et al. 2017, Logan et al. 2021). The three species broadly overlap in their primary teleost and cephalopod prey taxa, with particular shared emphasis on epipelagic and mesopelagic ommastrephid squids, including I. illecebrosus in the northwest Atlantic (Toll and Hess 1981, Logan and Lutcavage 2013, Duffy et al. 2017, Austin 2022). Other important prey identified from stomach contents include mesopelagic paralepidid, gempylid, alepisaurid, and myctophid fishes for swordfish and bigeye, and epipelagic scombrid and mesopelagic paralepidid fishes for yellowfin tuna (Duffy et al. 2017, Lin et al. 2020, Logan et al. 2021, Austin 2022). These globally reported findings align well with the fish prey identified in our tropical swordfish stomachs (collected off Miami, FL). Swordfish typically consume larger prey items than bigeye tuna (Logan and Lutcavage 2013), which in turn have consistently been reported to feed on larger prey than yellowfin tuna (Duffy et al. 2017, Albuquerque et al. 2019, da Silva et al. 2019, Austin 2022). Yellowfin tuna also consume more small crustaceans than bigeye tuna, contributing to yellowfin’s overall broader diet. Based on historical diet preferences and data reported by Austin (2022) on tuna from the same field collections used here, we predicted that yellowfin stable isotope signatures would show higher levels of epipelagic carbon than either bigeye or swordfish. Moreover, Austin (2022) reported a relatively large proportion of epipelagic fish in yellowfin stomach contents, while bigeye and swordfish stomachs were dominated by ommastrephid squid. We found similar (∼50%–60%) mean contributions of temperate mesopelagic-associated carbon in yellowfin and bigeye livers, with lower (c. 30%) mean contributions to swordfish. This unexpectedly low mesopelagic carbon contribution to swordfish may have been driven by a particular preference for ommastrephids based on mean proportion by weight (Duffy et al. 2017, Logan et al. 2021). Local I. illecebrosus are abundant at a wide range of epi- and mesopelagic depths. However, in the summer in our study region, they are most abundant in the lower epipelagic at depths of 100 to 200 m (Hendrickson and Holmes 2004) and have a SIA-EAA carbon signature that strongly overlaps with our temperate epipelagic reference group. A preference for these squid may thus underlie the differences we found in swordfish carbon sourcing compared to the more fish-based diets of the tunas. The two tunas were then differentiated by the broader and more opportunistic diet of yellowfin compared to bigeye, as observed both in stomach contents (Austin 2022) and by yellowfin’s more variable carbon sources among individuals and seasons, compared to the very consistent carbon sourcing we observed among bigeye tuna.

Access to mesopelagic prey by the three focal predators is facilitated by species-specific movement strategies. Swordfish approximately match the diel vertical migrations of the mesopelagic community, spending nights in surface waters and days at ∼600 m (e.g. Braun et al. 2019). Bigeye tuna also occupy the mesopelagic during the daytime, but via a series of short, repeated dives from the lower epipelagic to ∼400 m (e.g. Lam et al. 2014), presumably due to physiological limitations preventing prolonged cold exposure at depth (Holland et al. 1992). Daytime depths are estimated to, on average, place swordfish slightly below and bigeye tuna just above, the primary deep scattering layer (Braun et al. 2023). Bigeye tuna vertical movements have also been observed to match the dawn vertical migration of the deep scattering layer (Josse et al. 1998). These behaviors provide both species with access to vertically migrating mesopelagic prey, suggesting that predator vertical movements are attuned over evolutionary time to match their migratory prey. In contrast, yellowfin tuna are predominantly distributed in epipelagic waters (e.g. Weng et al. 2009) and may not be feeding during the dark nighttime periods when their vertical habitat overlaps with mesopelagic migrants (Buckley and Miller 1994, Wright et al. 2021, Machful et al. 2024). Their substantial consumption of mesopelagic prey is likely facilitated by crepuscular foraging, when prey are at vulnerable transition depths, and there is sufficient light for the tuna to hunt (Reintjes and King 1953, Roger and Grandperrin 1976, Buckley and Miller 1994). Our results suggest that yellowfin are highly successful at this crepuscular predation on the migratory layer, to the point of sourcing the majority of their carbon from mesopelagic food webs. Overall, our results are consistent with the hypothesis that large pelagic predators focus their foraging effort on highly concentrated layers of mesopelagic prey in the “thin soup” (low average concentration) prey field of the open ocean (Bakun 1996, Arostegui et al. 2020, 2023, Benoit-Bird 2024).

Turnover times of biological tissues can influence the interpretation of stable isotope results (Thomas and Crowther 2015, Vander Zanden et al. 2015). Here, we compared results from liver (relatively fast turnover time) and muscle (slower) for individuals of our three study species. Tissue-specific carbon turnover rates for swordfish, bigeye, or yellowfin tuna have not been reported. We initially assumed them to be similar to turnover times observed for Pacific bluefin tuna (Thunnus orientalis): liver 162 days (95% confidence interval: 90–850 days), white muscle 255 days (95% CI: 168–532 days) (Madigan et al. 2012), as they are of similar size and taxonomy. This study’s three focal species occupy temperate waters from approximately June to November. Our summertime samples were therefore collected after the animals had been in temperate waters for ∼60–100 days, and our fall samples after ∼155–185 days. Thus, if our species had mean turnover times equivalent to those of Pacific bluefin tuna, there would be ∼50% tropical-sourced carbon in the liver of animals captured in the fall. We found fall liver samples to have ∼25% tropical-sourced carbon across all three species, suggesting similar but slower liver turnover times than bluefin (not accounting for other factors like temperature, isotopic composition of diet sources for wild vs. experimental animals, etc.). Comparing putative muscle turnover times is more difficult, given the likely turnover time is greater than the temperate residence time and that this study was not able to collect tuna muscle in the fall. However, we assume the muscle turnover times for swordfish to be similar to those of Pacific bluefin tuna, given that both summer and fall swordfish muscle samples were estimated to be 66% tropical carbon sourced (i.e. no seasonal shift was observed). In contrast, tropical carbon contribution to swordfish liver decreased by ∼30% from summer to fall, aligning with our expectation of shorter liver turnover times and thus the associated temperate influence on EAA signatures. Future research in this area would be beneficial to resolve taxonomic, dietary, and tissue-based differences in tissue turnover times among tuna and billfish species.

Movements and foraging patterns of mesopelagic fishes may contribute substantially to the biological carbon pump by moving organic carbon from surface to deeper waters (McMonagle et al. 2023). Current estimates of the contribution of fish to the biological carbon pump range widely (McMonagle et al. 2023). Yet the contribution of large fishes or other top predators to the biological carbon pump is frequently overlooked. However, an increasing number of studies across ecosystems demonstrate that predators have critical roles in carbon cycling (Schmitz et al. 2010) and potentially biosequestration of carbon (Wilmers et al. 2012, Atwood et al. 2013, 2015, Heithaus et al. 2014), including in the context of high trophic level pelagic fish (Stafford et al. 2022) and fisheries (Andersen et al. 2024). Thus, our focal predators may make meaningful contributions to the marine carbon cycle via their own vertical movements, fecal pellet flux, and eventual mortality (Falciani et al. 2022, Mouillot et al. 2023) as well as their impacts on mesopelagic biomass. However, potential impacts as they relate to mesopelagic foraging of large fish predators remain unclear. Future work incorporating large fishes in ocean carbon models is needed to resolve the implications of their diverse movement and feeding strategies.

The extent to which mesopelagic food webs support top predator biomass is a key knowledge gap essential to sustainable management of emerging mesopelagic fisheries alongside existing fisheries for predators such as swordfish and tuna (St. John et al. 2016). The estimated impacts of fishing forage populations on their predators vary from largely unaffected (Hilborn et al. 2017, but see Pikitch et al. 2018) to highly impacted, especially when the forage populations are mesopelagic fishes (Smith et al. 2011). We argue that the swordfish and tunas in this study are likely to be highly impacted by large-scale exploitation of mesopelagic fishes, as they source the majority of their carbon from mesopelagic food webs. Furthermore, if mesopelagic stocks are depleted, these predators may have limited ability to diet switch and replace lost mesopelagic forage with epipelagic prey, as many epipelagic fish resources are fully or overexploited (FAO 2022). We further emphasize that the “mesopelagic” contributions reported here are minimum values, referring to temperate mesopelagic carbon only, as we were not able to distinguish between tropical epipelagic and mesopelagic sources. The true global mesopelagic carbon contribution to these predators is thus likely to be even greater. Further work refining these trophic ecology results and integrating them into fisheries models and management will be critical to ecosystem-based management of both predator and prey stocks.

In conclusion, we found that mesopelagic carbon sources were important to supporting the biomass of three large, highly migratory marine predators: bigeye tuna, swordfish, and yellowfin tuna. Temperate mesopelagic-associated carbon was the dominant source of carbon to bigeye and yellowfin tunas, and a major source of carbon to swordfish caught in the northwest Atlantic. These results were supported by investigating the contributions of pelagic end members to all consumer groups, including mesopelagic reference animals and predators. Comparisons of predator tissues (muscle, liver) and season of sampling demonstrated seasonally shifting carbon sources as predators move between temperate and tropical waters. Our description of carbon flows through pelagic food webs to predators informs how human activities like emerging mesopelagic fisheries may influence the trophic ecology and abundances of these important animals.

Acknowledgments

We thank Kelton McMahon and Angela Stahl for sharing SIA end member data; Danny Mears and the crew of FV Monica for temperate sample collection; Captain Cook Fishing Charters for their assistance in securing tropical dolphinfish samples; Jackson Coates and the crew of FV Trouble for collecting tropical swordfish stomachs; and the Scanlan Family Foundation for providing the Pelagic Fisheries Lab at the University of Maine support to conduct components of this research.

Author contributions

Conceptualization: CW and SRT. Methodology: CW, LH, and AFG. Investigation: CW, LH, and AFG. Resources: KGG, CDB, MCA, WG, JKL. Data Curation: CW and AFG. Visualization: CW. Supervision: SRT. Writing – Original Draft: CW and SRT. Writing – Review & Editing: All authors.

Conflict of interest

None declared.

Funding

This work was funded in part by the WHOI Ocean Venture Fund (to C.W.), a Natural Sciences and Engineering Research Council of Canada Postgraduate Scholarship—Doctoral (to C.W.), a Massachusetts Institute of Technology Martin Family Society of Fellows for Sustainability Fellowship (to C.W.), and the WHOI President’s Innovation Fund (to M.C.A.). This work was part of the Woods Hole Oceanographic Institution’s Ocean Twilight Zone Project, funded as part of the Audacious Project housed at TED.

Data availability

DNA sequences from swordfish stomach contents are available in the GenBank Nucleotide Database and can be accessed under accession numbers PQ344747-PQ344760. Stable isotope data are available upon request.

References