Lipid transfers within the lower food web of western Arctic seas

Copepods are key organisms in the Arctic Ocean, where they provide the crucial link between phytoplankton and the upper food web, including pelagic animals that Inuit consume (Falk-Petersen et al., 1990; Dalsgaard et al., 2003). By grazing on phytoplankton, primary consumers such as Calanus glacialis and Calanus hyperboreus accumulate energy and essential molecules that are subsequently passed on to other consumers in the food web. The intake of lipids is crucial in polar waters because it represents the most condensed form of energy that organisms can store during the short growing season (Sargent and Falk-Petersen, 1988; Kattner and Graeve, 1991; Falk-Petersen et al., 2000; Parrish, 2009; Leu et al., 2015). Among the lipids, essential fatty acids (EFA) are necessary to ensure the survival and healthy function of organisms, including growth and reproduction. Because animals cannot synthesize EFA de novo, their acquisition via zooplankton grazing on phytoplankton is vital for the entire pelagic food web. The synthesis, accumulation, and trophic transfer of different lipids are therefore crucial determinants of the sustainability of Arctic marine ecosystems.

Fatty acids (FA) are characterized by the length of their carbon chains and their degree of unsaturation caused by the insertion of a double bond and resulting in a spatial reconfiguration that increases the potential energy released when the carbon chain is catabolized. The location of the unsaturation also determines the fatty acid isomer; of general acceptance is that animals cannot insert double bonds between the ninth carbon and the terminal methyl group. In contrast, primary producers have desaturase enzymes that allow them to insert a double bond on the first, third, fourth, and sixth carbon, thereby producing ω1-FA, ω3-FA, ω4-FA, and ω6-FA, respectively (Dalsgaard et al., 2003).

Numerous studies have shown that the fatty acid profiles of phytoplankton are diverse and depend primarily on their taxonomic composition and, to a lesser extent, ambient growth conditions (Viso and Marty, 1993; Napolitano, 1999; Dalsgaard et al., 2003; Galloway and Winder, 2015; Jónasdóttir, 2019; Marmillot et al., 2020). This variability in the nutritional value of particulate organic matter (POM) can be expected to be reflected in the fatty acid composition of zooplankton grazers, as experimental studies have shown that dietary fatty acid structures are conserved during incorporation into storage lipids. In the laboratory, the fatty acid profiles of Calanus finmarchicus and C. hyperboreus tracked those of their food source when switched between single cultures of dinoflagellates (rich in 18:4ω3) and diatoms (rich in 16:1ω7), highlighting the potential of specific FA to be used as trophic markers of the POM assimilated by copepods (Graeve et al., 1994; Brett et al., 2009). Because a seasonal succession of phytoplankton occurs in nature, the lipid biomass of copepods is expected to reflect both their current and past diets depending on the turnover rate of stored FA and the persistence of phytoplankton assemblages. Using ¹³C labeling, Graeve et al. (2005) observed that the lipid pool of C. hyperboreus was renewed in approximately 11 days, while fatty acid turnover rates in C. glacialis and C. finmarchicus were relatively slow, with 45% and 22% of their lipid pool exchanged over 14 days, respectively.

The fatty acid profiles of copepod assemblages can also be affected by the environment, either directly through physiological responses of the animals or indirectly due to taxonomic shifts in the species dominating the biomass. Homeoviscous responses in poikilotherms, including copepods, have been amply demonstrated in the laboratory (e.g., Farkas, 1979; Farkas et al., 1984; Hazel, 1995; Brett et al., 2009). For example, some studies report a negative correlation between temperature and the content of polyunsaturated fatty acids (PUFA) in phospholipids, which probably serves to maintain membrane fluidity. In the copepod Parastenhelia sp., the proportion of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) declined with decreasing pH (Jayalakshmi et al., 2016). With regard to taxonomy, Falk-Petersen et al. (2009) showed that Calanus hyperboreus dominates assemblages in deep-water areas, whereas C. glacialis occurs mostly on shelves. The gross lipid composition of these two species differs (Sargent and Falk-Petersen, 1988) and is also modulated by their developmental stage. For instance, copepodite stages IV (CIV) and V (CV) are characterized by enhanced wax ester synthesis, presumably as a means to increase buoyancy and acquire enough energy for diapause, sexual maturation, and reproduction.

Prior studies of lipids in Arctic copepods have focused on the inflow shelves of the Atlantic sector. The oceanographic context is very different on the outflow shelves of the Canadian Arctic, where relatively fresh and strongly stratified waters that originate from the Pacific Ocean and, to a lesser extent, rivers dominate (Tremblay et al., 2015). This setting is conducive to relatively short-lived surface blooms and the subsequent establishment of productive subsurface chlorophyll maxima (SCM) that persist throughout summer and fall in the rich nutricline (Martin et al., 2010; Martin et al., 2013). Given the importance of large calanoid copepods for carbon cycling and energy flux at high northern latitudes (Falk-Petersen et al., 2009; Rubao et al., 2012), the present study aimed to assess the relative influence of taxonomic composition, phytoplankton, and climate-sensitive physicochemical water properties on the bulk lipid and fatty acid composition of assemblages in the Canadian Arctic. Given the results of the experimental studies reported above (Graeve et al., 1994; Graeve et al., 2005; Brett et al., 2009), we hypothesized that the proportions of specific FA in copepods and POM in nature would be positively related at the time of sampling and that the coefficient of determination would be higher where Calanus hyperboreus dominates zooplankton biomass. Given the vital role of EFA for animals (e.g., EPA, DHA), we hypothesized that homeostasis would keep their content relatively stable when compared to non-essential FA.

Sampling was conducted with the Canadian icebreaker CCGS Amundsen from July 29 to September 28, 2016 (Figure 1). The 35 stations were spread across Baffin Bay (July 29 to August 3 and September 26–28), Nares Strait (August 6–16), the Northwest Passage (August 4, August 17–24, and September 18–25) and the Beaufort Sea (August 28 to September 5). The overall region is influenced by different water masses according to the general ocean circulation (Rudels, 2015). The western part of Baffin Bay is exposed to cold and relatively fresh waters descending from the high Arctic with the Baffin Island Current and the Labrador Current (Tremblay et al., 2018). These two currents carry predominantly Pacific-derived waters that previously transited across the Beaufort Sea and enter the Canadian Archipelago via the Nares Strait and the Northwest Passage (predominantly through Barrow Strait and Lancaster Sound). By contrast, Atlantic waters enter the survey area around the southern tip of Greenland and propagate northward with the West Greenland Current along the eastern edges of Baffin Island. This water eventually crosses to the west to feed the offshore branches of the Baffin Island and Labrador currents. Most of the sampling stations were located in relatively shallow waters, but regional differences in average bottom depths (± standard deviation, SD) were present, with shelf stations of the Northwest Passage (204 ± 26 m, n = 15) being shallower than stations in the Beaufort Sea (352 ± 70 m, n = 5), the Nares Strait (476 ± 31 m, n = 10) and Baffin Bay (489 ± 42 m, n = 5).

At each station, sensors mounted on a Rosette sampler provided detailed vertical profiles of temperature and salinity (Sea-Bird SBE-911 CTD), nitrate (In situ Ultraviolet Spectrometer, ISUS, Satlantic), dissolved oxygen (Seabird SBE-43, calibrated onboard against Winkler titrations; Martin et al., 2010), photosynthetically available radiation (400–700 nm; Biospherical Instruments QDP2300), and chlorophyll fluorescence (Seapoint). The latter was used to determine the depth of the SCM for sampling purposes. Water was collected at the surface and at the SCM with 12 L Niskin bottles mounted on the Rosette.

The pH of seawater was measured at 25°C with a spectrophotometer using phenol red (433 and 558 nm) and cresol purple (434 and 578 nm) according to Robert-Baldo et al. (1985), Clayton and Byrne (1993), and Millero et al. (2009). Total alkalinity was measured by potentiometric titration combined with pH electrodes and dilute HCl (0.02 M) as a titrant. We used CO₂sys (Lewis et al., 1998) to convert pH measurements into in situ pH and pCO₂ by using alkalinity, temperature, and salinity data; pH measurements were not performed in the Beaufort Sea. Sea-ice concentrations provided by AMSR-E/AMSR2 and SSMI (Earth Observing System Data and Information System, EOSDIS; Brodzik and Stewart, 2016; Meier et al., 2018) were used to estimate the date when sea ice had declined by 50% and to calculate the number of days elapsed between this date and sampling time for each station (hereafter abbreviated as SID for sea-ice decline).

For both sampling depths at each station, a 200 ml aliquot of seawater was preserved with acidic Lugol’s and stored at 4°C in the dark for taxonomic analysis of phytoplankton phyla, class, and species. As described by Marmillot et al. (2020), the carbon content of different phytoplankton groups was estimated from visual counting and sizing of cells under visible light microscopy (Lund et al., 1958; Utermöhl, 1958; Menden-Deuer and Lessard, 2000). POM was collected by filtering 3 L of seawater through 47 mm GF/C filters with 1.2 µm pore size for lipid analysis. The filters were stored in pre-burned aluminum foil at −80°C until analysis in the home laboratory.

Zooplankton was collected with a 1 m² aperture, 4.5 m long, conical-square plankton net with 200 µm mesh size, mounted on a 2 × 2 m metal frame carrying other plankton nets. This frame was hauled vertically from 10 m above the bottom to the surface at a speed of 24 m min⁻¹. The samples were then subdivided using a Motoda splitting box. As described by Darnis and Fortier (2012), the first halves were stored with 4% formaldehyde seawater solution for further laboratory analyses where zooplankton organisms were counted and identified to the lowest possible taxonomic level using a Borogov counting chamber and microscope. In the second halves, macrozooplankton was removed (e.g., Chaetognathas, euphausiids, large amphipods, and medusae) and subdivided by size fractioning with a 750 µm mesh in order to select the largest Calanus copepods, mainly composed of late-stage copepodites (CV and CIV) and adult females, and then pooled, as this study aims to elucidate the fatty acid content and nutritional quality of complex natural communities that potentially feed higher trophic levels of food chains. As higher trophic level organisms can feed on biomass-rich copepod communities without selecting specific species, our results therefore intended to provide information on the gross lipid quality of copepods that would be available to subsequent animals in food chains in a macroecological context of trophic transfers in a multi-stressor/driver environment. The fraction was then subdivided using the Motoda splitting box until reaching the desired quantity of copepods (i.e., approximately 1000 mg), which was deposited on a pre-burned 47 mm GF/C filter (1.2 µm pore size) and stored in pre-burned aluminum foil at −80°C until lipid analysis. The carbon content of copepod species was estimated based on the length-mass relationships established by Forest et al. (2011).

Each filter collected during fieldwork was put in lipid clean vials with 4 ml of chloroform, flushed with nitrogen and sealed with Teflon tape before storage at −20°C. Lipids were extracted with chloroform:methanol:water (1:2:1) following Folch et al. (1957) as modified by Parrish (1999). Filters were ground using a rod, sonicated, and centrifuged three times in order to extract lipid layers (lower) with two Pasteur pipettes (double pipetting). Lipid extracts were stored at −20°C in 2 ml vials capped under N₂ and sealed with Teflon tape until further analyses.

Extracts were put in lipid clean vials where a mixture of H₂SO₄–MeOH was added prior to heating at 100°C for 1 hour. Then 1.5 ml of hexane was added and the upper layer containing lipids was removed. Samples were dried under N₂ prior to resuspension and storage at −20°C until fatty acid methyl ester analyses using an HP 6890 GC FID equipped with a 7683 autosampler (Budge and Parrish, 1998, 1999). Retention times and 37 component standards were used to identify peaks with (Varian) Galaxie chromatography software (Guerra et al., 2023). Although lipid class analysis could have refined the interpretation of some results, we decided to focus on FA in this study, while the interpretation of lipid classes will be addressed in the future.

As the total number of phytoplankton cells was 1.8 times higher at the SCM than surface (1.9 times higher in terms of carbon content) and herbivorous copepods potentially feed throughout the upper water column, we used the discrete phytoplankton lipid data obtained at both depths to calculate a weighted average that could then be compared with the zooplankton data obtained from integrated vertical tows. The weighting was done using the carbon content of POM at each depth as obtained by the combustion of POM samples collected on GF/F filters and analyzed using an elemental analyzer (ECS 4010, Costech Analytical Technologies Inc.) coupled to a mass spectrometer (Delta V Advantage, Thermo-Finnigan). We then created POM coefficients for each depth (fraction of surface + SCM) and multiplied each coefficient by its corresponding phytoplankton lipid value. The resulting values for surface and SCM were then summed to obtain a single weighted phytoplankton lipid value for each station. For the assemblage composition of phytoplankton, the sum of cell numbers measured at surface and SCM was converted into carbon equivalents in order to obtain a specific proportion of carbon biomass for the different phytoplankton groups. As described by Marmillot et al. (2020), the biovolume of single cells was multiplied by the number of cells to obtain total biovolume which was then converted into carbon quantity using conversion factors from Menden-Deuer and Lessard (2000). However, in an attempt to eliminate the potential biases introduced by those calculations and to evaluate whether copepods feed in preference at the SCM, we also performed a subset of analyses considering phytoplankton data from this depth only. In seeking possible relationships between the lipid composition of POM and physicochemical water properties, we considered averages of surface and SCM data for the latter.

In the Results section, deviations from mean values are reported as ± one standard error. To simplify the text, use of the words “abundance” or “carbon” after a taxonomic group (e.g., diatom carbon) hereafter designates the relative contribution of this group to total phytoplankton or copepod counts or to carbon. ANOVA and multivariate PERMANOVA were used to test for differences among oceanographic regions (i.e., the Beaufort Sea, the Northwest Passage, Nares Strait, and Baffin Bay), where p_perm refers to the p-value of PERMANOVA.

Given the complexity of the data set obtained, different statistical approaches were employed to explore relationships between physicochemical variables, taxonomic composition, and fatty acid profiles. Although we recognized covariations between abiotic factors and assemblage composition (Figures S1 and S2), both of which affect the lipid profiles of copepod communities, we explored the best variables explaining fatty acid proportions of total FA using the Akaike information criterion in order to rank different possible models and test for interactions between variables. While highly useful, this analysis excludes stations with incomplete datasets (e.g., pH in the Beaufort Sea) and possibly overlooks significant relationships for variables with the greatest availability. The scores from this analysis (not shown) were therefore used as initial guidance in the selection of generalized linear models (where the slopes of regression models were defined by β) and generalized linear mixed models with region as a random effect as needed. Linear regression models were used to address the effects of environmental and biological variables on copepod lipid composition, while polynomial models allowed us address the storage or use of FA based on need and resource availability. Pearson correlation tests and variance inflation factor (VIF) of the R package “car” were used to assess the co-linearity of variables in multiple regression models (VIF < 2 was set as a threshold for including variables). Grubbs test (R package Outliers; Komsta, 2011) was used to identify statistical outliers in the dataset.

As we aimed to provide community-level insights into the dynamics of the fatty acid content in natural copepod assemblages, the fatty acid analyses that follow are based on unsorted samples that include all the species and stages comprising the large size fraction at each station. This approach recognizes that the aggregate nutritive quality of the assemblage impacts diverse consumers higher up the food web and that most predators of copepods may feed indiscriminately and/or opportunistically on the assemblages they come across (Karnovsky et al., 2008; Pomerleau et al., 2014; Buckley and Whitehouse, 2017), although Calanus glacialis has been shown to be the preferred prey of polar cod during early stages of development (Bouchard and Fortier, 2020). Since the lipid composition of copepods is known to be affected by species and life stage (Falk-Petersen et al., 2008), we used this information to describe the assemblages and interpret variability in their bulk fatty acid and lipid composition.

On average and regardless of developmental stage, more than 95% of the carbon in the large size fraction was contributed by Calanus hyperboreus (54.9% ± 3.9%), Calanus glacialis (32.5% ± 4%), and Metridia longa (9.47% ± 1.8%). Paraeucheata species accounted for only 2.1% of the total, with even lower contributions (<1%) from other copepod species (Figure 2). C. hyperboreus dominated in all regions when considering the average of all stations and, while the PERMANOVA indicates no significant differences among regions, dominance tended to be higher in the North (mean of 68.9% for Nares Strait) than in Baffin Bay, the Beaufort Sea, and the Northwest Passage (mean = 55.7%, 52.4%, and 47.5%, respectively; Figure 2). A quadratic regression was found between the proportion of C. hyperboreus and latitude + latitude² (r² = 0.24, p < 0.01: β0 = −5673, β1 = 152.9, β2 = −1.01), suggesting a latitudinal optimum during our sampling period at approximately 75°N for this species (in orange in Figure 3). Overall, the proportions of C. hyperboreus and C. glacialis (in purple in Figure 3) were consistent with their known preferences for deep-water and shallow-water habitats, respectively (Falk-Petersen et al., 2009). C. hyperboreus proportions related positively to bottom depth (r² = 0.32, p < 0.001), decreased with temperature (r² = 0.10, p < 0.05), and increased with pH (r² = 0.46, p < 0.001) or salinity (r² = 0.49, p < 0.001; Figure 3); the latter two variables were also strongly related to bottom depth (r² = 0.43 and 0.45, respectively p < 0.001). By contrast, the mean proportions of C. glacialis differed among regions (p_perm < 0.05), with higher values in the western part of the sampling area (43.1% in the Northwest Passage, 41.9% in the Beaufort Sea) than in the East (19.4% in Nares Strait, 15.1% in Baffin Bay). C. glacialis proportions decreased northward (r² = 0.16, p < 0.05) and related negatively to bottom depth (r² = 0.41, p < 0.001). We also found that C. glacialis proportions increased with temperature (r² = 0.20, p < 0.01) and decreased with pH (r² = 0.61, p < 0.001) or salinity (r² = 0.67, p < 0.001). Furthermore, ANOVA results revealed a higher proportion (p < 0.05) of Metridia longa in Baffin Bay (13.1%) and Nares Strait (9.6%) than in the Northwest Passage (6.1%) and Beaufort Sea (3.7%).

When the 3 dominant species (i.e., Calanus hyperboreus, C. glacialis, and Metridia longa) were pooled together, the distribution of stages in terms of carbon biomass was approximately one-third each for CV (37.6% ± 2.5%), adult females (34.2% ± 3%), and CIV (23.6% ± 3.4%; Figure 2). While this roughly even distribution reduces the potential biases introduced by developmental stage in the interpretation of species effects, we nevertheless investigated stage distributions within species as they are known to affect fatty acid composition in copepods. Among the dominant species, C. hyperboreus was composed predominantly of adult females (34.1% ± 3.3%), CV (31.9% ± 2.5%), and CIV (28.3% ± 3.7%), with differences among regions for CV only (p_perm < 0.05; Figure 2). The proportion of CV in Baffin Bay (44.2% ± 3.6%) was higher than in the Northwest Passage, Nares Strait, and the Beaufort Sea (34.8% ± 3.9%, 23.3% ± 3.5%, and 22.2% ± 3.9%, respectively).

In addition to these regional differences, the estimated time separating sea-ice melt from sampling date affected the proportion of copepodite stages present. For Calanus hyperboreus, this SID related positively with the proportion of CIV (r² = 0.41, p < 0.001) and negatively with the proportion of females (r² = 0.23, p < 0.01), suggesting that the latter fed primarily on the surface phytoplankton bloom early in the season while smaller stages were more dependent on prolonged production at the SCM. The C. glacialis population was composed predominantly of CV (57.2% ± 3.5%), adult females (28.0% ± 3.4%), and CIV (13.9% ± 3.1%). Here again, inter-regional differences emerged for the CV and CIV stages (p_perm < 0.01), with a higher proportion of CIV in the Beaufort Sea (48.2% ± 12.1%) than in other regions (Nares Strait = 35.7% ± 8%, Northwest Passage = 23.0% ± 4.5%, and Baffin Bay = 18.1% ± 8.2%) and a high proportion of CV in Baffin Bay and the Northwest Passage (66.8% ± 5.6% and 66.4% ± 2.4%, respectively) relative to Nares Strait and the Beaufort Sea (48.8% ± 9.1% and 25.6% ± 5.6%, respectively). As for C. hyperboreus, the SID affected stage proportion positively for CIV (r² = 0.26, p < 0.01) and negatively for females (r² = 0.11, p < 0.05) in C. glacialis. Finally, adult females were dominant within M. longa populations (60.6% ± 3.7%), followed by CV (18.4% ± 2%) and adult males (14.3% ± 2%), with no significant differences among regions.

Among the significant relationships observed between fatty acid and taxonomic composition, we found, considering the area as a random effect, that ω6-FA increased with Calanus glacialis proportion (r² = 0.26, p < 0.01, n = 31), while specific EFA such as EPA increased with C. hyperboreus proportion (r² = 0.18, p < 0.05, n = 31). Moreover, PUFA and ω3-FA contents increased with C. hyperboreus (r² = 0.17 and 0.21, respectively, p < 0.05, n = 31; Figure 4). While Kattner and Hagen (2009) concluded there was a lack of clear difference in the proportion of PUFA or ω3-FA between C. hyperboreus or C. glacialis, increases in the two fatty acid groups with C. hyperboreus here possibly result from an effect of developmental stage, as previously reported by Falk-Petersen et al. (2009). Indeed, considering only the CIV stage of C. hyperboreus when establishing its numerical contribution improved the above relationships (r² = 0.31 for PUFA and 0.28 for ω3-FA, p < 0.01, n = 29) and led to a new positive one with EPA (r² = 0.22, p < 0.05, n = 29). Likewise, a positive relationship between PUFA and CV stage was observed in C. glacialis (r² = 0.15, p < 0.05, n = 29). These results are consistent with the literature, as the highest proportion of PUFA and ω3-FA rich wax esters are found in the CIV and CV stages for the two species (Sargent and Falk-Petersen, 1988).

The assemblage composition of copepods strongly affected content of saturated fatty acids (SFA). The proportion of SFA increased as Calanus hyperboreus declined (r² = 0.42, p < 0.001, n = 31) and C. glacialis increased (r² = 0.82, p < 0.001, n = 31; Figure 4). The particularly strong relationship with C. glacialis might be explained by the relatively high variability of physicochemical water properties in coastal areas and the need for membrane acclimation in response to osmotic stress. As we already showed a strong inverse regression between C. glacialis and salinity or pH, the variation of SFA in relation to the assemblage composition suggests physiological adjustments of C. glacialis in low-salinity or low-pH environments.

Only a few studies have explored direct relationships between copepod lipid content and acidification. Of these studies, Jayalakshmi et al. (2016) reported a lower proportion of EFA, such as EPA and DHA, in adults of Parastenhelia sp. collected from tropical regions and exposed for 30 days to reduced pH. This study used very low pH (<6) compared to other ones in which the reductions were relatively modest. Garzke et al. (2016) found that the fatty acid content of adult Paracalanus sp. from the Baltic Sea was affected significantly by an interaction between temperature (i.e., 9°C versus 15°C) and pCO₂ (560 µatm versus 1400 µatm) but that the latter as a single variable had a very limited effect compared with temperature. Furthermore, Mayor et al. (2015) concluded from experiments that short-term exposure (i.e., 5 days) to pCO₂ conditions expected by the end of 2100 should not affect the fatty acid content of the CV of temperate copepods species such as Calanus helgolandicus and C. finmarchicus. This lack of response to realistic and moderate pH reductions possibly extends to planktonic crustaceans in general, as Ericson et al. (2019) reached the same conclusion for adult Antarctic krill. In this context, our results suggest that the effects of moderate pH reductions, possibly in combination with a salinity decrease, may be unique to large Arctic copepods and are stage-specific, because the relationships observed here were significant when considering CIV or CV proportions (p < 0.05, n = 29) but not the proportion of adult females.

Notwithstanding the above relationships between assemblage composition and physicochemical seawater properties or geographic location, direct and sometimes unique relationships were observed between the latter and the bulk lipid composition of copepod assemblages. Indeed, the proportion of SFA in copepods correlated negatively with salinity (r² = 0.57, p < 0.001, n = 31) and seawater pH (r² = 0.66, p < 0.001, n = 25). As pH and salinity are positively correlated (p < 0.001), these results are consistent with the osmotic response and membrane adjustment suggested above and the experiment of Lee et al. (2017), which reported a decreased in SFA when the estuarine copepod Paracyclopina nana was exposed to high salinities (25 ppt and 30 ppt) relative to control conditions (i.e., 15 ppt). Given the relationships observed above between the lipid composition of copepods and their assemblage composition, physicochemical seawater properties such as pH and salinity can be regarded as additional drivers of fatty acid variability (Figure S1). By contrast, only a non-significant tendency was found between the proportion of SFA of total FA and temperature contrary to other studies that reported negative correlations between this variable and phospholipid unsaturation levels as a means to increase membrane fluidity (Harwood, 1998; Guschina and Harwood, 2009). However, the impact of this homeoviscous response on copepod lipid profiles is expected to be less than the effects of changes in assemblage composition (Gladyshev et al., 2011) and our results suggest that other water properties may supersede the effects of temperature as secondary drivers in a cold setting where the temperature range is relatively narrow.

Among the physicochemical and environmental variables considered above, pH had the largest effect on copepod fatty acid composition based on the number of significant relationships involving this variable. While latitude, salinity, or SID also had apparent effects, these variables all correlated with pH (Figure S1), and the latter generally exhibited the highest coefficients of determination with fatty acid proportions of total FA, especially for ω6-FA or SFA (Table S1).

Finally, we found that ω6-FA, linoleic acid (LA; accounting for 50% of ω6-FA), and SFA were lower (p < 0.01) in Nares Strait compared to the other regions (Figure 5), whereas the PERMANOVA test revealed that PUFA, ω3-FA, and specific EFA, such as EPA, DHA, or arachidonic acid (ARA), did not differ among regions. These results emphasize the requirement of EFA in copepods and support the hypothesis that these FA are less susceptible to variability in environmental conditions than non-essential FA. Furthermore, the lower proportion of SFA in northern areas is consistent with higher unsaturation levels in colder environments, while the lower proportion of ω6-FA could be caused by a decreasing contribution of ω6-FA-rich flagellates with increasing latitude (Marmillot et al., 2020).

In order to evaluate the extent to which the fatty acid composition of copepods reflects that of the phytoplankton, this section first compares the overall lipid profiles of copepods and phytoplankton, focusing on significant relationships and variability for specific FA, such as EFA. As many environmental and biological factors can affect the lipid composition of copepods, we then sought to identify the best predictors of copepod fatty acid profiles among several independent variables such as physicochemical seawater properties, phytoplankton assemblages and developmental stage of Calanus species. All sampling was done around the time of maximum temperature in the water column so that the metabolic rate would be higher than for most of the rest of the year; during this period we can expect FA in zooplankton to be most representative of those in the diet. Because the temporal scales of lipid synthesis by phytoplankton and utilization by copepods differ and may influence the apparent degree of trophic coupling based on snapshot sampling, we devised an indirect means to assess fatty acid turnover and the origin of copepod fatty acid biomass using a fatty-acid marker of diatoms (FADM).

Only 9 of the 56 FA that phytoplankton and copepods had in common co-varied between the two groups of organisms (Table S2). EPA proportions of total FA were positively related (r² = 0.22, p < 0.05, n = 28; Figure 6), with an intercept (β0 = 12.6) and a slope (β1 = 1.09) that suggest both an accumulation and a minimal level of 12.6% of this fatty acid in copepods. The strongest relationship was observed for precursors of EFA, such as α-linolenic acid (ALA; r² = 0.30, p < 0.001) for which the regression slope (β0 = 0.06) and intercept (β1 = 0.22) suggest a differential loss (Figure 6). Because an experimental study of Calanus species found no evidence that the animals are able to elongate ALA to EPA and DHA (Parrish et al., 2012), the low proportion of ALA in copepods (0.42%) compared to phytoplankton (1.46%) is likely attributable to catabolism in the former.

Simultaneous samples of copepods and phytoplankton had different fatty acid profiles overall (Figure 7). A Mantel test showed no significant relationship between phytoplankton and copepod FA in a distance matrix (significance = 0.34). The proportion of EFA, such as EPA or DHA, was generally highest in copepods (Tables S2 and S3), reflecting the accumulation of these essential drivers of organismal health and function in animals (Parrish, 2009; Parrish et al., 2012). Copepods had relatively low proportions of dietary FA, such as 14:0, 16:0, and 18:1ω9 known as the potential precursors of 20:1ω9 and 22:1ω11, with the corresponding fatty alcohols accounting for a significant portion of wax esters (Graeve et al., 2005; Brett et al., 2009). Because 20:1ω9 and 22:1ω11 are among the major FA in northern species of copepods (Kattner and Hagen, 2009) but are almost absent in phytoplankton, the overall proportion of monunsaturated fatty acids (MUFA) differed greatly between the two groups of organisms. The MUFA represent 24.1% of total lipids in phytoplankton but 47.2% in copepods. Of this 47.2% of MUFA, nearly half was comprised of the dietary fatty acid and diatom marker 16:1ω7, while a quarter was comprised of the wax ester components 20:1ω9 (15%) and 22:1ω11 (12%) that are typically synthesized de novo by the animals and thus represent only a small proportion of FA in phytoplankton (Sargent and Henderson, 1986; Falk-Petersen et al., 2002). The results emphasize the importance of wax ester synthesis in copepods and account for most of the difference in their lipid profiles relative to those of their phytoplankton food sources.

We then compared the proportion of specific FA in copepods and phytoplankton and found that different patterns emerged with respect to the biosynthesis capacities and metabolic requirements of producers versus consumers. Indeed, t-tests and paired t-tests revealed significant differences (p < 0.001) between the most studied FA, such as EPA, DHA, ALA, ω3-FA, ARA, LA, and ω6-FA (Figure 8). Opposite trends were observed depending on the fatty acid, as the proportions of ω3-FA, EPA, and DHA of total FA were higher in zooplankton while the ALA precursor was higher in phytoplankton as mentioned above. The same patterns were observed for the ω6-FA, as the LA precursor was higher in phytoplankton while the ARA proportion was higher in copepods, although the overall proportion of ω6-FA was lower. The latter might be explained by the high proportion of LA, which accounted for more than 43% of the overall ω6-FA. These general trends showing a higher proportion of specific EFA, such as EPA, DHA, or ARA, in copepods but lower proportions of precursors, such as ALA or LA, likely reflect a differential trophic accumulation of EFA, but they could also suggest catabolism or even some ability to reconfigure these precursors. Moreover, a comparison of the standard deviation (SD) of specific fatty acid proportions of total FA in copepods and phytoplankton also revealed different dispersion patterns for EPA, ALA, LA, and ω6-FA (Fisher p < 0.001) but not for DHA, ω3-FA, or ARA (Figure 8). The SD of EPA was highest in copepods, and no significant difference appeared for DHA and ARA. This pattern might result from a homeostatic response for EPA content in copepods; it also suggests that EPA is assimilated to a larger extent than DHA or has a different turnover time, which is discussed below. We emphasize that using fatty acid proportions (% total fatty acids) indicates the nature of the fatty acids transferred, but not the quantity. A further refinement would thus be to consider mass ratios (e.g., proportion of dry mass, organic mass, mass of carbon) and/or concentrations (e.g., mass per liter), which would also permit an estimate of total fatty acid energy transferred.

Considering the numerous factors affecting copepod lipid profiles previously shown, we sought to determine the best models that could explain fatty acid content of copepods using the biological and abiotic variables measured here. We found that the proportion of CIV among the 3 dominant copepod species (i.e., Calanus hyperboreus, C. glacialis, and Metridia longa) + FADM in phytoplankton (see below) + SID explained the EPA content of copepods fairly well (r² = 0.58, p < 0.001, n = 28) and that the best model included only CIV + diatom carbon (r² = 0.66, p < 0.001; Figure 9). These 2 models are also significant when considering phytoplankton data from the SCM only (r² = 0.54 and 0.44, p < 0.001, n = 25, respectively). Using the same approach, we found that the DHA proportion of total FA in copepods was explained mainly by flagellate carbon + latitude + SID (r² = 0.45, p < 0.001, n = 26; Figure 9) and, again, that this model remained significant when using only SCM data (r² = 0.35, p < 0.001). The significant effects of the above independent variables are consistent with the determining influences of diet, assemblage composition, and developmental stage reported in the literature (Graeve et al., 2005; Galloway and Winder, 2015), but the models obtained here provide additional insights in suggesting that EPA proportions of total FA are mostly affected by taxonomy and ontogeny whereas DHA proportions are more closely linked to phytoplankton assemblage composition. These models also highlight the effect of seasonal progression and SCM development, which are closely linked.

Because EPA and DHA, which are generally taken up from the diet, can be considered both structural FA and a major component of wax esters (Sargent and Falk-Petersen, 1988; Graeve et al., 2005; Brett et al., 2009), our results also suggest recent high feeding and storage of EFA. In addition, the specific CIV stage in the best model explaining the variations in EPA could be related to the maximal formation of wax ester during this developmental stage (as well as CV) that would be used as an energy source during overwintering and for gonadal development (Sargent and Falk-Petersen, 1988).

While EPA and DHA are often considered the major EFA in terms of proportion, quantity, or benefit to animals, physicochemical seawater properties do not figure among the best models explaining variability in these EFA. However, some studies have clearly shown variations in fatty acid profiles of copepods as a function of temperature (Farkas et al., 1984; Hazel, 1995; Schlechtriem et al., 2006; Brett et al., 2009), with a higher proportion of PUFA, such as DHA and EPA, in colder environments being consistent with the homeoviscous adaptation theory. The contrast may be due to the fact that water temperatures across our study area generally remain near or below zero throughout the upper 200 m (including the SCM), with warmer conditions occurring only in the top few meters of the water column at a few stations.

Experimental studies based on ¹³C-labeled phytoplankton show that lipid assimilation and turnover rates differ greatly among FA, species, and developmental stages of copepods (Graeve et al., 2005; Boissonot et al., 2016; Graeve et al., 2020). However, the interpretation of fatty acid assimilation and turnover is complicated in natural phytoplankton and copepod communities where the assemblage and developmental stage composition is complex. Because the turnover time of lipids in the environment cannot be assessed directly, we sought to evaluate the extent to which the observed lipid profiles of copepods were consistent with a diet based on the phytoplankton present at the time of sampling or indicative of feeding on a prior stage of algal succession, although we recognize that this approach contains various uncertainties, as discussed below. As our sampling occurred in late summer and early fall, surface phytoplankton communities had already shifted from spring-bloom diatoms to flagellate-dominated communities, whereas SCM communities continued to be largely dominated by diatoms, with some stations showing a transition toward flagellates (Marmillot et al., 2020).

Firstly, we used the data presented in Marmillot et al. (2020) to examine which specific fatty acid or combination of FA in POM provided the best signature for diatom carbon. This analysis indicated that 16:1ω7 was the single fatty acid that best related to diatom carbon (r² = 0.47, p < 0.001, n = 27). The relationship improved a little (r² = 0.51, p < 0.001, n = 27) when different diatom markers were combined, as suggested by Pepin et al. (2011). Specifically, 16:1ω7 + EPA + 16:4ω1 emerged as the best FADM.

Because other FA are typical of flagellates, the second most abundant phytoplankton group in our study area (Marmillot et al., 2020), we also examined correlations between ratios of FADM to 18:4ω3, DHA, LA, C18-PUFA, or ω6-FA and the contribution of diatoms to the sum of diatom and flagellate carbon (hereafter “diatom proportion”):

In the end, the most robust regression (r² = 0.53, p < 0.001, n =27) was found between FADM in phytoplankton and diatom proportion. A linear extrapolation of this relationship (Figures 10 and S3) implies that FADM would reach 26.1% on average in a theoretical phytoplankton community composed only of diatoms (“diatom-only”) while the average FADM proportion would be only 8.74% in a phytoplankton community composed exclusively of flagellates (“flagellate-only”). Although informative, we recognize that these values are based on a regression where the phytoplankton FADM explains just over 50% of the copepod FADM and that the theoretical community is not necessarily realized in nature.

FADM in copepods, when estimated from the same constituent FA, is expected to track the FADM of their phytoplankton diet after accounting for offsets resulting from the differential accumulation of specific FA (see above). This expectation can be validated using prior experimental studies. The mean proportion of FADM in copepods here (46.5% ± 1%, n = 31) was similar, albeit somewhat elevated with respect to the experimental study of Graeve et al. (2005), where the cumulative proportion of FA that constitutes FADM averaged 34.3% in Calanus glacialis (female) and C. hyperboreus (CV and females combined) but reached 39.1% when considering only CV for the latter. The agreement was especially good when comparing 16:1ω7 only, which reached a maximum proportion of 23.1% in our study compared to 25% for C. hyperboreus at the end of the feeding experiment of Graeve et al. (2005). The small differences could easily be caused by a disparity in growth conditions or in the fatty acid profiles of the diatom food (single culture versus natural assemblages). This analysis implies that FADM, which accounts for nearly half of the FA in copepods (i.e., 46.5%), provides a useful means to assess the herbivorous diet of copepods in our study area, especially because the FA used in the calculation of FADM are synthesized mainly by diatoms. One caveat to this approach is that the fatty acid constituents of FADM in copepods can originate to a small extent from other phytoplankton groups (Jónasdóttir, 2019).

The proportions of FADM in copepods and phytoplankton were well related at the time of sampling (r² = 0.51 with a quadratic effect: β0 = 29.6, β1 = 1.32, β2 = −0.02 for the quadratic term, p < 0.001, n = 28; grey line in Figure 11). The positive intercept of 37.1% FADM in copepods is consistent with a prior accumulation of FADM and matches very closely the value found by Graeve et al. (2005) for copepods feeding exclusively on diatoms. Furthermore, Figure 11 shows a smaller FADM range in copepods (minimum = 31.5%, maximum = 56.7%) than in phytoplankton (minimum = 7.52%, maximum = 37.1%) while the average proportion of FADM was much lower in phytoplankton (18.8% ± 1.2%) than in copepods (46.5% ± 1.0%), highlighting here again a prior accumulation of the fatty acid constituents of FADM (Figures 6 and 11). Thus, the proportions of FADM in phytoplankton and copepods are comparable when the FADM content in phytoplankton is high. However, our results suggest that copepods store FADM when the proportions of diatoms decrease, suggesting a regulation in the use or turnover of FA that constitute this marker depending on ambient food availability.

Using the values for a diatom-only community and flagellate-only community derived from Figure 10 and based on the relationship shown in Figure 11, we estimated that the proportion of FADM in copepods feeding exclusively on diatoms should reach 50.7% on average and decrease to 41.6% for animals feeding only on flagellates (i.e., corresponding to the flagellate-only tie-line in Figure S3). While the three FA that constitute FADM together account for 46.5% of the fatty acid content in copepods and given that the flagellate-only tie-line of the regression in Figure 11 and Figure S3 is 41.6%, these results suggest that, in a flagellate-dominated community, a minimum of 19.3% of copepod FA (i.e., 46.5% of 41.6%) may be from past feeding activities. This suggestion is consistent with copepods feeding on a prior diatom bloom at the surface, but also suggests that most copepod FA are from recent feeding activity, highlighting the predominant role of SCM in these regions.

The indirect evidence for a similar FADM turnover in Calanus glacialis and C. hyperboreus in our study is potentially an artefact of the methodology, as the distribution of developmental stages may have been uneven across species. Indeed, interesting trends emerged with regard to developmental stages when we compared the parameters of regressions between FADM in copepods and phytoplankton depending on the proportion of CIV, CV, and females among the 3 dominant copepod species (i.e., C. hyperboreus, C. glacialis, and Metridia longa). To make this comparison, a subset of our dataset was created keeping only the stations where the proportion of CIV, CV, or females was higher than 40%. From this analysis we observed that the regression slopes increased with developmental stage progression. The highest regression slopes (β1 = 0.63, r² = 0.74, p < 0.01, n = 7) were observed when females were dominant (i.e., >40%), followed by CV (β1 = 0.52, r² = 0.47, p < 0.05, n = 28) and CIV, although for CIV the relationship was not significant (β1 = 0.43, r² = 0.19, p = 0.21, n = 6; Figure S4). The same trends appeared when we considered only the proportions of CIV, CV, or females above 50% instead of 40%, but the significance decreased because the number of observations was smaller. These results suggest an increase in turnover rates with respect to developmental stages but could also point to a progressive transition toward feeding on diatom-rich SCM layers in late developmental stages or selective feeding by younger stages, although our dataset did not explore potential trends in early copepodite stages (CI, CII, or CIII).

Finally, we observed that neither the proportion of Calanus hyperboreus nor proxies for the seasonal maturity of the system (i.e., SID or phosphate) seem to affect the relationship between FADM in phytoplankton and copepods (Figure 11). This observation suggests that the turnover of FA was relatively rapid and similar in both Calanus species, which departs from previous studies indicating that C. hyperboreus exhibits rapid turnover (nearly complete exchange of lipids in 11 days; Graeve et al., 2005) relative to smaller species (3 weeks for exchanging half the lipids in Pseudocalanus minutus; Boissonnot et al., 2016; 45% and 22% of lipids exchanged over 14 days for C. glacialis and C. finmarchicus, respectively; Graeve et al., 2005).

Our study examined the lipid composition of large copepods across wide latitudinal and longitudinal gradients where physicochemical water properties and the assemblage composition of phytoplankton varied greatly. As hypothesized, the fatty acid profiles of POM affected the lipid composition of copepods in their natural environment despite the different turnover of primary producers and grazers. However, among the FA that copepods and phytoplankton had in common, only one-fifth co-varied, which is consistent with their different life traits and the fact that copepods possess high storage capacities. Significant relationships were also found between EFA and the proportion of dominant copepod species. The proportion of ω6-FA in copepod communities was positively related with Calanus glacialis, while the proportion of ω3-FA and PUFA in general increased with C. hyperboreus. While a prior laboratory study found that lipid accumulation and turnover were particularly high in C. hyperboreus, we found no evidence that relationships between copepods and phytoplankton fatty acid profiles were strongest where C. hyperboreus was dominant. This lack of evidence could be explained in part by an uneven distribution of developmental stages among the species that dominated the assemblages. Indeed, improvements in coefficients of determination resulting from considering only the CIV stage indicate that developmental stage is a key driver, as previously shown in the Atlantic sector of the Arctic Ocean (Falk-Petersen et al., 2009).

We hypothesized that homeostasis would make EFA less sensitive to variability in biological and environmental parameters, which was partially verified. Based on model comparisons, physicochemical water properties mainly affected non-essential FA. However, EFA, such as EPA and DHA, were strongly affected by developmental stage and the assemblage composition of copepods and their phytoplankton diet at the time of sampling. While co-variations appeared between copepod assemblage composition, bathymetry and physicochemical water properties, distinct relationships emerged between the latter and the fatty acid content of assemblages. The proportion of SFA was strongly related to salinity and pH, which could reflect differences in membrane plasticity between species such as Calanus glacialis, which thrives in coastal areas where environmental fluctuation can be high compared to the offshore environment where species such as C. hyperboreus typically dominate. Furthermore, among the most studied EFA, we found that EPA was affected mainly by the copepod developmental stage and the relative contribution of diatoms to total phytoplankton carbon, while DHA was affected mainly by the proportion of flagellates, latitude, and SID, suggesting that changes in plankton assemblages directly affect the nutritive quality of copepods that feed subsequent trophic levels in the food web.

Beyond the environmental and taxonomic effects on copepod fatty acid profiles, this study also provides new insights into the accumulation of FA that may account for more than half of the copepod biomass. Although a high accumulation of diatom fatty acid markers in copepods has been observed previously during the spring bloom, suggesting strong feeding activities at that time (Falk-Petersen et al., 2008), our estimations suggest that most FA in copepods come from recent feeding activity occurring weeks or even months after the surface spring bloom. Daase et al. (2008) also underscored sustained feeding activities of Calanus species on phytoplankton-rich layers in late summer and early fall around Svalbard. Our results further highlight the major role of SCM in supplying copepods with a long-lived and relatively stable source of food in the Canadian Archipelago. The presence of these subsurface layers is likely to mitigate or delay the negative effects of a potential mismatch between primary producers and consumers that could occur with climate change.

Environmental, taxonomic, and lipid data generated for this study are available online at https://dataverse.scholarsportal.info/dataset.xhtml?persistentId=doi:10.5683/SP3/BH7ZBR.

The supplemental files for this article can be found as follows:

Figures S1–S4. Tables S1–S3. Docx

The authors are grateful to the officers and crew of NGCC Amundsen and to Louis Fortier, to Jonathan Gagnon, Gabrièle Deslongchamps, Cyril Aubry, and Jeanette Wells for technical support, to Pascal Guillot for CTD data processing, and also to the contributors of free or open-source software (CRAN-R, Ocean Data View, LibreOffice, Inkscape, and Linux community).

This work was supported by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to J-ÉT and CCP via the strategic network CHONe (Canada Healthy Ocean Network), a grant from the Sentinel North program of the Canada First Research Excellence Fund to J-ÉT (project BriGHT), and a grant from the network center of excellence ArcticNet to J-ÉT and Louis Fortier. This work contributes to the scientific programs of the FRQNT strategic cluster Québec-Océan and the Institut Nordique du Québec.

The authors declare that the work submitted here was not carried out in the presence of any personal, professional, or financial relationships and thus there is no conflict of interest. Jean-Éric Tremblay is an associate editor at Elementa. He was not involved in the review process of this article.

Conducted the field sampling, performed the laboratory and statistical analysis, and wrote the manuscript: VM.

Created the project, designed the sampling, and revised the manuscript: J-ÉT, CCP.

Performed some lipid analyses and revised the manuscript: JFM.

How to cite this article: Marmillot, V, Parrish, CC, Tremblay, J-É, MacKinnon, JF. 2024. Lipid transfers within the lower food web of western Arctic seas. Elementa: Science of the Anthropocene 12(1). DOI: https://doi.org/10.1525/elementa.2022.00084

Domain Editor-in-Chief: Jody W. Deming, University of Washington, Seattle, WA, USA

Associate Editor: Julie E. Keister, University of Washington, Seattle, WA, USA

Knowledge Domain: Ocean Science