Introduction

Deep-water corals and sponges support complex habitats which provide shelter, substrate and nutrients for myriad benthic fauna and commercially important fish stocks1,2,3. Corals and sponges are generally thought to feed on particulate (POM) and/or dissolved (DOM) organic matter reservoirs, yet we still know little about their specific source diets4. Moreover, climate-driven changes in marine primary productivity5 and biogeochemical cycles6,7 may alter food quantity and quality in deep-water habitats8,9. A more comprehensive understanding of benthic community trophodynamics is therefore needed to help manage and conserve deep-water coral and sponge habitats.

Deep-water corals are passive suspension feeders and may consume both surface-water derived and deep suspended materials, potentially including phytodetritus, zooplankton and their faecal pellets10,11,12. Early work suggested that many deep-water corals are opportunistic feeders13,14. More recent studies suggest that coral feeding behaviour may be more selective than previously thought15,16, and might include seasonal variability and/or dormancy11,17. Sponges actively filter large volumes of water to extract POM and DOM18,19,20. Recent studies propose a wide range of possible trophic positions spanning primary consumers to top predators21,22,23, which collectively highlight the need for better tools to comprehend their complicated microbial symbioses. Diet overlap has been inferred between deep-water corals and sponges24, but few studies of specimens cohabitating small areas exist to support this interpretation. In addition to niche overlap, feeding experiments demonstrate that corals are capable of consuming recycled sponge-derived material25 and that sponges can assimilate mucus secreted from corals21,26. The difficulty of accessing deep-water habitats and the resulting scarcity of observational feeding data for cohabiting corals and sponges severely limit our ability to expand and generalize on these interpretations.

In 2022, we visited the upper continental slope off Nova Scotia (Eastern Canada) to investigate deep-water coral and sponge habitats using a remotely operated vehicle (ROV; Fig. 1). In this region, incised canyons host rich assemblages of corals and sponges along steep to vertical ridge walls27,28. We collected specimens from a vertical wall within Shortland Canyon at a depth of 438–454 m over a 35 m wide transect (Fig. 2; supplemental videos). Octocorals of the order Scleralcyonacea were observed to exhibit a conspicuous vertical zonation, with dense stands of Paragorgia arborea located along ridge crests, and Primnoa resedaeformis located directly below. More sparsely distributed Keratoisis grayi were located toward the ridge bases. Sampling also included Malacalcyonacea corals (Duva florida), sea anemones (Hormathiidae), sponges (Weberella bursa, and an unidentified Hexactinellid), and ascidians (Didemnidae), which were distributed with less apparent vertical zonation. This setting provided an unparalleled opportunity to investigate trophic resource partitioning among primary and secondary consumers where food availability and local hydrodynamics should be relatively equal.

Fig. 1
figure 1

Location of Shortland Canyon on the continental slope off Nova Scotia, Canada. (a) Inset map of North America. (b) Multi-beam imagery of Shortland Canyon showing sample location and contour lines at 200 m intervals. (c) Nova Scotia shelf and slope highlighting location of Shortland Canyon. Black dots in Gully Marine Protected Area (MPA) and Haddock Channel mark DIC/POM sampling locations. Contour lines at 500 m intervals. Multibeam imagery of Shortland Canyon from the Geological Survey of Canada29 and modified using QGIS version 3.40.1.

Full size image
Fig. 2
figure 2

Distribution of faunal samples collected along a 35 m wide vertical wall transect within Shortland Canyon. Lower corners of the plot show latitude and longitude. Symbols and colours used to represent taxa are consistent throughout the rest of this article. Manipulator hand is ~ 20 cm long. See supplemental videos for more detailed distribution patterns.

Full size image

We use bulk and compound-specific stable isotopes to test whether resource partitioning exists between species of corals, sponges, ascidians, and anemones living metres apart. Bulk stable carbon (δ13Cbulk) and nitrogen (δ15Nbulk) isotopes were used to assess differences in resource utilization. Compound-specific nitrogen isotope analysis of amino acids (δ15NAA) was used to further differentiate taxon-specific diets by baseline sources, trophic position, and microbial resynthesis. We also compare radiocarbon (Δ14C) of tissue samples with a new dissolved inorganic carbon (DIC) depth profile and mixed layer suspended POM (POMsusp) from the Scotian Slope to elucidate food provenance.

Results and discussion

Bulk stable isotopes reveal significant trophic niche partitioning

When plotted in δ13Cbulk vs. δ15Nbulk space, the POMsusp and taxa formed a positive linear relationship (Fig. 3). POMsusp had the lowest isotopic values (δ13Cbulk = −26.4 – −24.1 ‰; δ15Nbulk = 2.7 – 4.1 ‰). Corals P. arborea and P. resedaeformis had the next lowest isotopic values (−23.3 – −20.2 ‰; 7.3 – 9.1 ‰). Corals K. grayi and D. florida, anemones and ascidians had intermediate values (−22.6 – −20.0 ‰; 7.6 – 12.5 ‰). Sponges (Hexactinellid and W. bursa) had the most enriched values (−21.2 – −19.7 ‰; 13.6 – 17.6 ‰).

Fig. 3
figure 3

Distribution of δ13Cbulk vs. δ15Nbulk data for specimens collected at Shortland Canyon wall transect and POMsusp (Gully MPA and Haddock Channel, Fig. 1).

Full size image

Conventional interpretations of the bulk isotope data in Fig. 3 would ascribe a mostly pelagic, low trophic position to taxa with lower δ13Cbulk and δ15Nbulk (P. resedaeformis, P. arborea) and a more benthic, high trophic position to taxa with higher δ13Cbulk and δ15Nbulk (sponges)30,31. Strikingly, the total range of values in our assumed primary and secondary consumers (δ13Cbulk = −23.3 – −19.7 ‰; δ15Nbulk= 7.3 – 17.6 ‰) overlaps with reported ranges across virtually all trophic guilds of offshore marine ecosystems, from zooplankton to piscivorous fish, seabirds, and top predators31,33,33. In fact, the exceptionally high δ15N values in W. bursa approach those of polar bears when variations in baseline nitrogen are accounted for32. Traditional δ15Nbulk-based calculations of trophic position (TP) using POMsusp for the baseline δ15N and a trophic enrichment factor of 3.431 would suggest a range of TPs from 3.4 (P. resedaeformis, P. arborea) to 6.0 (W. bursa), which is unrealistic. Nevertheless, given that all taxa were contemporaneously sampled from the same cliff face, their separation in δ13Cbulk vs. δ15Nbulk space likely signifies actual differences in resource utilization34, which we assessed further through compound-specific isotope analysis of amino acids.

Compound-specific isotope analysis of amino acids (CSIA-AA)

Analysis of δ15NAA was used to deconvolve factors influencing the observed wide range of δ15Nbulk. A strong positive correlation (linear regression: r2 = 0.68) between δ15Nbulk and the δ15N of total hydrolysable amino acids (δ15NTHAA) confirmed that δ15Nbulk represents the isotopic mass balance of individual amino acids. The overall distribution of δ15NAA data (Fig. 4) was typical of heterotrophic consumers, with relatively lower values for “source” amino acids (Src-AAs; e.g., phenylalanine; Phe) and higher values for “trophic” amino acids (Tr-AAs; e.g., glutamic acid; Glx)35. Src-AAs undergo minimal isotopic fractionation in animal metabolism and therefore retain baseline δ15N signatures of primary producer biomass through successive trophic levels. Tr-AAs undergo rate-limiting transamination reactions in animal metabolism, leading to stepwise enrichment of δ15NAA values by as much as 7 ‰ with each trophic transfer36,38,38. In addition, microbial metabolisms may alter the distribution of δ15NAA values through various pathways35. One such pathway is AA-selective heterotrophic resynthesis, which results in a scattering in the δ15N of Tr-AAs39; another pathway is extracellular hydrolysis, which leads to uniform enrichment of all AAs40,41.

Fig. 4
figure 4

δ15NAA results divided into trophic, intermediate, source and metabolic groupings based on established fractionation patterns35,36,38. Abbreviations: Total hydrolysable amino acids (THAA), glutamic acid (Glx), aspartic acid (Asx), alanine (Ala), leucine (Leu), isoleucine (Ile), proline (Pro), valine (Val), glycine (Gly), serine (Ser), phenylalanine (Phe), lysine (Lys), threonine (Thr). Connecting lines are shown to help visualize different patterns among species.

Full size image

The δ15NPhe proxy for baseline δ15N clearly separates corals and anemones from sponges and ascidians. For corals and anemones, δ15NPhe clustered tightly around a value of 5.9 ± 0.4 ‰. This value is slightly higher than the δ15N of mixed layer POMsusp collected during August 2022 (Table 1), but within range of spatio-temporally averaged POMsusp and nitricline nitrate in the study region42. This is strong evidence that the diets for these taxa originate ultimately from mixed layer primary production. In sponges, δ15N was substantially enriched across all AAs, including Phe (13.4 ± 0.2 ‰). This pattern of enrichment indicates that the diets of sponges may be influenced by extracellular hydrolysis40,41 and implies that they exploit an entirely different basal food web from corals and anemones. Ascidians showed intermediate levels of δ15N enrichment across most AAs, implying a different or mixed diet between that of corals and sponges.

Table 1 Summary table of radiocarbon, bulk carbon and nitrogen isotopes and CSIA-measured parameters of collected specimens and POMsusp. POM taken from a depth of 20 m. Further detail on radiocarbon data in SI.
Full size table

The difference in δ15N between Glx and Phe provides an internally normalised index of metazoan trophic position (TPmet37; see Methods). Calculated TPmet among all taxa ranged from 2.3 to 3.2, consistent with feeding at primary to secondary consumer levels (Table 1). An alternative formulation for TP based on the δ15N difference between alanine (Ala) and Phe accounts for additional trophic steps induced by protozoan heterotrophy (TPprot43; Methods). For all taxa except W. bursa, calculated TPprot was higher than TPmet by 0.6–1.1 units (Table 1), indicating that protozoan heterotrophy added about one trophic level in the diets of most taxa. In W. bursa, TPprot was lower than TPmet by 0.7 units. In addition, the δ15N of threonine (Thr) normalized to δ15N of Phe became more negative with increasing TP for corals and anemones (Fig. 4). This pattern is consistent with inverse trophic fractionation of Thr38 and adds further support for separation of taxa by trophic level.

In addition to the effects of baseline δ15N and heterotrophic fractionation, the distributions of δ15NAA exhibit differences relating to AA-selective heterotrophic microbial resynthesis, which is quantified by the ∑V parameter (Methods). ∑V provides a relative index of diet “freshness”. In corals and anemones, ∑V clustered around 1.4 ± 0.1 (Table 1). These values are typical of fresh algae and zooplankton biomass39,44. In sponges and ascidians, ∑V clustered around 2.6 ± 0.3. Such high values of ∑V are characteristic of highly degraded materials including meso- and bathy-pelagic POMsusp39,45 and DOM41. These distinct groups of ∑V values further support the inference of at least two distinct basal food webs.

Δ14C values of tissue samples differentiate DOM from POM in diet composition

Food provenance was also investigated by comparing Δ14C from tissue samples to a new Δ14CDIC depth profile collected from the upper slope region off Nova Scotia (Fig. 5). The Δ14CDIC profile, which peaks at + 20 ‰ from 150 to 200 m depth and decreases to −10 ‰ below that, reflects the uptake and subsequent mixing and dilution of atmospheric nuclear bomb-produced 14C46. POMsusp was also sampled at the depth of maximum primary productivity (0.6 ± 5.2 ‰) and was lower than that of DIC, which may be explained by temporal offsets in the integration of Δ14C between these two pools at the time of sampling. The Δ14C value of deeper POMsusp and dissolved organic material (DOM) was not measured in the present study, but in northwest Atlantic slope waters these pools typically have far more negative Δ14C (POMsusp: < −50 ‰; DOM: < −350 ‰46. All the taxa except for Hexactinellid sponges had positive Δ14C values (Table 1), confirming that their diets originate from the surface mixed layer10 (Fig. 5). Potential routes for transporting these mixed layer Δ14C signals to depth include passive sinking of phytodetritus and other particulates; vertically migrating zooplankton; or downward tidal pumping47,48. Δ14C values of corals and anemones (avg. +13.3 ‰) were slightly higher than that of ascidians and W. bursa (avg. 7.7 ‰), possibly reflecting zonation of food sources in the upper vs. lower part of the mixed layer, respectively49. Alternatively, ascidians and W. bursa may record a mixed diet comprising both mixed layer and deeper 14C-depleted food sources. Hexactinellid sponges had a remarkably low Δ14C value of −75.5 ‰ (14C age = 560 ± 12 years) more consistent with the Δ14C of deep POMsusp or DOM.

Fig. 5
figure 5

Comparison between ∆14C from seawater DIC and POM, and from faunal tissues. Values for DIC are from three stations in each of the Gully MPA and the Haddock Channel; black dashed line represents a composite depth profile fitted through all points. Values for POM are from the same stations, collected at the depth of peak primary productivity (~ 20 m) based on in situ O2 profiles measured during CTD downcasts. Taxa were collected at depths of 438 to 454 m, but their tissue ∆14C values are mapped to the corresponding depths on the composite profile of DIC ∆14C to infer the depth at which their food sources originated.

Full size image

To summarize, the δ15NAA data demonstrate that: (1) Based on TPmet values, all taxa feed at the level of primary to secondary consumers; (2) Higher relative values of TPprot indicate that protozoan heterotrophy is an important component of the underlying food web in all diets, except for W. bursa; (3) Based on differences in isotopic baseline (δ15NPhe) and microbial resynthesis (∑V), corals and anemones rely on relatively fresh materials derived from mixed layer primary productivity, while sponges and ascidians exploit highly microbially degraded food sources; (4) Radiocarbon data confirm that the diets of all taxa, except Hexactinellid sponges, ultimately originate from the surface mixed layer.

Taxon-specific trophic niches

To further explore potential resource partitioning, isotopic metrics (except Δ14C because only one sample per taxon was measured) were analysed by principal component analysis (PCA; Fig. 6). Unlike conventional interpretations of isotopic niches in two-dimensional δ15Nbulk vs. δ13Cbulk space, the separation of multiple isotopic metrics in PCA space deconvolves factors influencing δ15Nbulk (i.e., baseline isotopic variability, trophic processing, and microbial degradation pathways) independently of each other. Because of this, the relative positions of taxa in PCA space are notably different than in Fig. 3. The first PC separated the taxa mainly along continua of δ13Cbulk, δ15Nbulk, δ15NPhe, and ∑V. Lower values of each loading are more closely representative of diets derived from surface primary productivity, and higher values represent a decoupling via microbial processing; therefore, this axis represents an index of benthic-pelagic coupling. The second PC separates taxa mainly according to trophic level. Within this framework, corals and anemones are separated from sponges and ascidians according to benthic-pelagic coupling, while the taxa within each of these groups are separated according to trophic level. Accordingly, and with important caveats noted below, the separation of taxa in multi-dimensional PCA space strongly supports significant resource partitioning among the taxa, likely reflecting actual differences in trophic niches.

Fig. 6
figure 6

Taxon-specific niche separation identified through principal component analysis (PCA) of bulk carbon (δ13Cbulk) and nitrogen (δ15Nbulk) isotopes and CSIA-based parameters including δ15N of metazoan (TPmet) and protistan (TPprot) trophic positions, baseline δ15N (δ15NPhe) and microbial resynthesis (∑V). The two axes explain 86.1 % of the variance within the dataset.

Full size image

Based on the distribution of data in PCA space along with the surface-derived Δ14C signatures, P. arborea and P. resedaeformis occupy a similar trophic niche, with tight benthic-pelagic coupling (mean δ15NPhe = 5.8 ± 0.7; ∑V = 1.5 ± 0.2) and low trophic position (TPmet = 2.3 ± 0.2; TPprot = 3.0 ± 0.3). The most probable food sources for both taxa include recently exported phytodetritus and protozoa. Fan-shaped colonies of P. arborea sat atop ridge crests oriented perpendicular to fast moving currents (based on ROV footage), which may optimise capture of fresher, non- or weakly motile particles27,50. This positioning is consistent with in situ time lapse photography showing that feeding cycles in P. arborea are tightly coupled to periods of maximum flow of the diurnal tidal cycle directly above the colonies51. Bushier colonies of P. resedaeformis were located directly below P. arborea, protruding a shorter distance from the cliff face, in what presumably would be a marginally slower hydrodynamic regime. Therefore, the different vertical positions and morphologies of P. arborea and P. resedaeformis may reflect preferential colonization of areas where the hydrodynamic regime suits the morphology of the coral to optimize nutrient acquisition52. Alternatively, their positioning may reflect differences in larval settlement53 rather than dietary selectivity.

Keratoisis grayi occupied a distinct niche with the highest TP of all the taxa (TPmet = 3.2 ± 0.1; TPprot = 4.3 ± 0.1) while maintaining a high degree of benthic-pelagic coupling as evidenced by the low ∑V value (1.3 ± 0.1) and similar baseline (δ15NPhe and δ13Cbulk) characteristics as P. arborea and P. resedaeformis. These results support previous interpretations that K. grayi consumes zooplankton16,49. Keratoisis grayi has a spindly morphology and was observed in the lower, more sparsely populated and hydrodynamically protected part of cliff faces. It is unclear whether and how this vertical positioning might facilitate capture of motile zooplankton prey unless K. grayi feeding is optimized at slower current speeds52. Collectively, our results mirror feeding experiments where corals efficiently capture phytoplankton at higher current speeds, but preferentially prey on zooplankton at lower speeds54. Thus, even in such a confined sampling area, meter-scale variability in hydrodynamic patterns appears to be tied to complementary morphological traits, supporting that feeding behaviour of corals is inextricably linked to morphology and flow at metre-scale resolution52. These results shed light on the drivers of localized species distribution segregation commonly observed in deep-water benthic vertical wall communities55,56.

Malacalcyonacean soft corals (D. florida) and anemones (Hormathiidae) occupied a region in PCA space between that of P. arborea/P. resedaeformis and K. grayi, based on intermediate values of TPmet (avg. 2.8) while maintaining a high degree of benthic-pelagic coupling (Fig. 6; Table 1). Low ∑V values in D. florida (1.2) and Hormathiidae (1.4) may support a high proportion of live prey consumption (Table 1). In contrast to larger octocorals, the seemingly more random spatial distribution of soft corals and hemisessile nature of anemones (Fig. 2) may decouple their feeding strategy from local hydrodynamics. These characteristics imply opportunistic feeding of relatively fresh food sources for soft corals and anemones, consistent with observed feeding in shallow and deep-water counterparts57. At Shortland Canyon, their diets likely comprise smaller zooplankton prey such as foraminifera, amphipods, and invertebrate larvae58. Our results support previous observations that Malacalcyonacean corals are niche generalists in the northwest Atlantic59. For anemones, feeding based on CSIA-AA proxies appears less carnivorous than generally recognized60.

Sponges and ascidians exploit a different basal food web characterised by a high degree of microbial reprocessing. The high values of ∑V (mean = 2.6 ± 0.2) and baseline δ15NPhe (12.4 ± 0.4) suggest that both selective heterotrophic resynthesis and extracellular hydrolysis may play important resource utilisation roles for these taxa. The δ15NAA enrichment patterns in sponges and ascidians resembles the few existing measurements of δ15NAA in mesopelagic suspended POM40,41. For Hexactinellid sponges, the additional constraint of pre-bomb Δ14C strongly indicates that these sponges feed on a fraction of pre-aged, hydrolytically degraded, suspended POM and DOM. Weberella bursa and ascidians, however, had distinct post-bomb Δ14C signatures, implying that neither deep POMsusp or DOM are proximate food sources for these taxa. Rather, the enzymatic hydrolysis responsible for the elevated δ15NAA patterns in these taxa may happen internally within their holobionts22,61,62. Despite these uncertainties, the data clearly support additional trophic niche partitioning even among morphologically similar sponges and ascidians.

Combined, bulk stable carbon (δ13Cbulk) and nitrogen (δ15Nbulk), δ15NAA and 14C data capture a wide separation in resource utilisation among deep-water benthic fauna living only metres apart. Our multifaceted approach explains anomalously elevated δ15Nbulk values commonly observed in sponges22,23,30,63,64 as due to baseline nitrogen and microbial reprocessing. The spatially constrained sampling location standardised many environmental aspects, but factors such as tissue turnover rates, isotopic fractionation associated with starvation, fasting or reproduction, recruitment strategies and microbiome-specific metabolic processes are challenging to constrain through isotopic proxies. In addition, the potential impact of ecological interactions and specific anatomic peculiarities could not be assessed. Isotopic proxies of tissue samples record time-averaged feeding preferences, but seasonal changes such as winter dormancy as observed in P. arborea17, may preserve a bias towards high productivity months. While the parameters we measured do not fully encompass individual trophic niches, our results reveal significant differences in resource utilisation among functionally similar taxa, which collectively increase community level nutrient acquisition and support niche complementarity65 rather than competition.

Implications for broader species distribution patterns of coral and sponge habitats

Vertical zonation patterns at the sampling location were also seen at other cliff faces in Shortland Canyon and the Gully Marine Protected Area (MPA) during the 2022 and earlier expeditions28. Interestingly, these meter-scale zonation patterns reflect broader bathymetric distributions long recognized by fishermen66 and documented in bycatch records and targeted surveys67,69,69. For example, along the eastern Scotian slope, P. resedaeformis and P. arborea are concentrated within a relatively shallow and narrow range of water depths (230–500 m), K. grayi occurs deeper (350–900 m), while Polymastiidae (including W. bursa) and Hexactinellid sponges have wider depth tolerances (190–915 m)70. Because environmental factors such as temperature, salinity, seawater chemistry, slope (as a proxy for substrate) and current regime tend to co-vary with depth, it has remained unclear which factors are limiting to taxon distributions70,71. In the present study, all these environmental factors except for current regime should be constant along the cliff face. Currents on the other hand would be expected to vary on the cliff face with vertical position, aspect, and physical obstacles. Thus, we hypothesize that current regime and its influence on food availability and utilization must play a first order constraint on overall taxon-specific depth distributions. If true, then the taxa studied here should exhibit similar patterns of resource utilization over their broader depth distributions.

Nutrient recycling pathways in deep-water coral and sponge habitats

Global shifts in plankton biomass and community composition5,72 pose poorly constrained threats to deep-water coral and sponge habitats9. Where POM flux is negatively impacted, sponges may add resilience to their habitats by consuming DOM and supplying POM to neighbouring fauna via predation or detritovory (i.e. the sponge loop)73,74. Our results support that sponges and ascidians do not demonstrate strong benthic-pelagic coupling, and exploit recalcitrant, microbially degraded food sources. In particular, Hexactinellids are completely decoupled from surface primary productivity because they feed on DOM and deep POMsusp. While our study does not include predators, similar food web studies using CSIA parameters observe that taxa one trophic guild above sponges (i.e. sea stars, sea urchins, and anemones) have similarly high ∑V values22. This may hint that sponge-derived food sources are commonly used by neighbouring organisms. These results support the use of CSIA-AA for recognizing specific nutrient recycling pathways in natural environments. The ability of deep-marine benthic communities to recycle recalcitrant sources of organic matter will be critical in regions experiencing long-term food shortages due to decreased nutrient export from the surface ocean.

Methods

Collection of biological samples

Sampling was conducted during an expedition aboard the M/V Atlantic Condor with a remotely operated vehicle (ROV) in August 2022. Biological specimens were collected from a single cliff face at Shortland Canyon (44°11’9.64"N, 58°25’56.64"W) at depths of 438–454 m. Five specimens each of corals (P. arborea, P. resedaeformis, K. grayi, D. florida) and demosponges (W. bursa), three anemones (Hormathiidae), two unidentified hexactinellid sponges (same morphotype), and three ascidians (Didemnidae) were collected. The challenges of maintaining ROV position in strong currents prevented the collection of larger numbers of replicates. Upon recovery, the specimens were photographed and stored at −20°C.

Collection of seawater POM and DIC samples

Mixed layer suspended particulate organic matter (POMsusp) and dissolved inorganic carbon (DIC) in seawater was collected from three stations at each of the Gully MPA and Haddock Channel, which spatially bracket Shortland Canyon (Fig. 1). Samples were collected using a CTD/Rosette equipped with temperature, salinity and oxygen sensors, and 20 L Niskin bottles. POMsusp was collected by filtering seawater collected at the depth of maximum dissolved oxygen (~ 20 m) onto pre-combusted 70 mm QMA filters using custom stainless-steel manifolds and immediately frozen. Depth profiles of DIC were collected at nominal depths of 2, 20, 80, 125, 250, 400, 700 and 900 m, then transferred into 250 mL bottles, poisoned with 50 µL saturated HgCl2 and stored in darkness under room temperature.

Bulk stable isotope analysis of biological tissues

Frozen tissues from each specimen were freeze-dried and homogenized with a mortar and pestle. Aliquots for δ13Cbulk analysis were decarbonated in 10 % HCl, neutralized with de-ionized water, and re-dried at 50°C. Aliquots for δ15Nbulk analysis were not acid treated. Analysis was conducted by Elemental Analyzer - Isotope Ratio Mass Spectrometry (EA-IRMS) at Dalhousie University. δ13Cbulk and δ15Nbulk are reported in δ-notation: δX = [(Rsample/Rstandard) − 1]*103 where X is 13C or 15N and R is 13C/12C or 15N/14N, respectively. Values are presented in per mil (‰) units referenced to the standards Vienna Pee Dee Belemnite (VPDB) or atmospheric N2 for carbon and nitrogen, respectively. Repeatability of sample replicates averaged 0.2 ‰ for both δ13Cbulk and δ15Nbulk.

Compound-specific isotope analysis of biological tissues

Compound-specific isotope analysis was conducted at Dalhousie University following procedures in Chen et al.44 and references therein. Aliquots of homogenized tissues were hydrolysed in 6 N HCl at 110°C for 20 h, then spiked with a norleucine internal standard and purified through cation exchange resin. Amino acids were converted to isopropyl-trifluoroacetic anhydride derivatives and analysed with co-derivatized calibration standards on a Thermo Trace 1310 GC coupled with a Delta V IRMS. Reproducibility of individual δ15NAA values averaged 0.5 ‰. “Glx” refers to both glutamic acid (Glu) and the glutamine converted to Glu during hydrolysis. Similarly, “Asx” refers to aspartic acid (Asp) and asparagine. Trophic position (TP = (δ15NGlx or Ala - δ15NPhe - β)/TEF) + 1 was calculated following Chikaraishi et al.37 with trophic enrichment factor (TEF) and β values from Decima and Landry43. For TPmet, βGlx = 3.4 and TEFGlx = 6.1; for TPprot, βAla = 3.2 and TEFAla = 4.5. The ∑V parameter for microbial resynthesis was calculated as 1/n∑Abs(xi), where xi is the difference in δ15N between each of the Tr-AAs (Glx, Asx, Ala, Leu, Ile, Pro, Val) and the average of those same Tr-AAs39.

Preparation and analysis of seawater POM and DIC samples

DIC samples were prepared for isotopic analysis using Gas Bench IRMS methods of Torres et al.75 for δ13C measurements and headspace extraction methods of Gao et al.76. Slight modifications to 14C methods used are described in Zeidan et al.77. POMsusp samples were prepared following methods of Fox and Walker78. Briefly, they were dried at 50°C/8 h in pre-combusted (540°C/2 h) petri dishes and subsequently 12 N HCl vapour phase acidified for 4 h to remove inorganic carbon. POM samples were dried again at 50°C/8 h to remove residual HCl and H2O vapour. Dried POM filter fractions were either pelletized for analysis of δ13C and δ15N via EA-IRMS at UC Santa Cruz, or closed-tube combusted for 14C analysis.

Accelerator Mass Spectrometry 14C analysis

One sample of decarbonated tissue from each of the biological taxa and seawater DIC and POC were analysed for natural abundance 14C dating either at the André E. Lalonde National Facility using a Mini Carbon Dating System (MICADAS) or a compact 500MV NEC at the UC Irvine Keck Carbon Cycle Laboratory. CO2 evolved from all samples were reduced to graphite via H2 reduction or the sealed-tube Zn method79,80. Results are reported in accordance with the conventions in Stuiver and Polach81. DIC Δ14C data were corrected for year of sample collection with a total analytical reproducibility better than ± 2 ‰. All other Δ14C measurements were corrected for year of measurement. Isotopic Δ14C results are reported as Fraction Modern (Fm), Δ14C, and conventional radiocarbon age (ybp).