A multi-trophic marker approach reveals high feeding plasticity in Barents Sea under-ice fauna

Microalgae growing within and attached to the bottom of Arctic sea ice (sympagic algae) can serve as a nutritious food resource for animals inhabiting the sea-ice water interface (under-ice fauna), particularly during the bottom ice-algal bloom in spring. As a consequence, under-ice fauna is likely impacted by sea-ice decline and changes in ice-algal primary production. To investigate this, samples of pelagic ( = PPOM) and ice-associated particulate organic matter ( = IPOM) and the ice-associated amphipods Apherusa glacialis and Eusirus holmii , and polar cod ( Boreogadus saida ), collected below ridged sea ice at two locations with pronounced differences in productivity in the northern Barents Sea during May 2021, were assessed for their trophic marker content. Specifically, we investigated the composition of diatom-and dinoflagellate-produced fatty acids (FAs), pelagic and sympagic highly branched isoprenoid (HBI) lipids as well as sterols to determine the animals ’ dietary preferences and trophic association to the sea-ice habitat during spring. Relative proportions of FAs differed strongly between PPOM and IPOM, indicating differences in species composition and degradation state between pelagic and sympagic habitats, respectively. FA signatures and sterol content of the consumers largely resembled known diet compositions with a strong reliance on diatom-derived carbon in A. glacialis , a higher degree of carnivory in E. holmii and evidence of Calanus -feeding in polar cod. Sympagic HBIs were detected at either low concentrations or not at all, in both producers and consumers, likely as a result of the very low abundance of their source diatoms. Pronounced trophic marker variability in A. glacialis collected at the highly productive shelf slope station versus the less productive central Arctic Basin station suggests a surprisingly high flexibility in carbon-source composition with a stronger reliance on pelagic food when available versus a higher importance of ice algal carbon when pelagic production is low. Nevertheless and despite the general lack (below detection limit) of sympagic HBIs in our dataset, high ice-algal biomass and elevated proportions of polyunsaturated FAs in IPOM compared to other seasons indicate that ice algae constitute a valuable nutritional carbon source as alternative to pelagic carbon during spring.

The Barents Sea seasonal ice-cover has been subject to change during all seasons over the past decades (Årthun et al., 2012;Onarheim and Årthun, 2017;Barton et al., 2018).Decline in sea-ice presence, extent and concentration (Kumar et al., 2021) alters timing, duration and magnitude of pelagic and sympagic primary production events (Dupont, 2012;Dalpadado et al., 2020;Song et al., 2021) with direct and indirect cascading impacts on food-web structures and trophic function in Arctic ecosystems (Kovacs et al., 2011;Barber et al., 2015;Stige et al., 2019;Pagano and Williams, 2021).Some marine animals have shown a positive response towards these alterations, as e.g.increasing biomass of capelin (Mallotus villosus) was linked to reduced sea-ice coverage in the Abbreviations: CAO, Central Arctic Ocean; chl a, Chlorophyll a; FA, Fatty acids; HBI, Highly branched isoprenoid; IPOM, Ice-associated particulate organic matter; PPOM, Pelagic particulate organic matter; PUFA, Polyunsaturated fatty acid.
In the Arctic, the gammaridean amphipod Apherusa glacialis is one of the dominant sea ice-associated species (Hop et al., 2000(Hop et al., , 2021;;Poltermann, 2001).Such herbivorous sympagic amphipods link the base of the Arctic marine food-web to intermediate and higher trophic levels (Lønne and Gulliksen, 1989;Lønne and Gabrielsen, 1992;Jakubas et al., 2011;Ogloff et al., 2020) and couple sympagic with pelagic and benthic food-web processes (Scott et al., 1999;Kohlbach et al., 2016).Apherusa glacialis is typically regarded to spend all or most of its lifecycle associated with sea ice (autochthonous) (Lønne and Gulliksen, 1991;Beuchel and Lønne, 2002).Recent investigations, however, challenge this assumption and indicate that A. glacialis can also occur throughout the water column, irrespective of the season (Kunisch et al., 2020), which could ultimately make this species less vulnerable and more adaptable to continued sea-ice retreat than previously assumed (Berge et al., 2012;Drivdal et al., 2021).In the case of the predatory gammaridean amphipod Eusirus holmii, the strength and nature of the ice association is less clear (Macnaughton et al., 2007) and it seems to be less reliant on sympagic carbon than A. glacialis (Brown et al., 2017).However, E. holmii is frequently found in close proximity to the ice (Gradinger et al., 2010;David et al., 2015;Hop et al., 2021) and has been shown to utilize ice algae as an energy source (Kohlbach et al., 2016).
Apherusa glacialis can serve as an important food item in the diet of polar cod (Boreogadus saida) (Lønne and Gulliksen, 1989;Kohlbach et al., 2017), a key species for the energy transfer within the Arctic foodweb, which in turn is preyed on by many species of seals, whales and seabirds (Bradstreet and Cross, 1982;Hop and Gjøsaeter, 2013).Although polar cod usually spends only part of its life cycle closely associated with the ice (Lønne and Gulliksen, 1989;Gradinger and Bluhm, 2004;David et al., 2016), it might be particularly sensitive to sea-ice loss (Huserbråten et al., 2019) due to its reliance on the ice for spawning and the development of the larvae in the under-ice environment (Eriksen et al., 2015;Dahlke et al., 2018;Gjøsaeter et al., 2020).Other findings, however, suggest that reduced sea ice might have a positive effect on the growth of this species since low sea-ice concentration during winter potentially extends the period of suitable growth conditions (Dupont et al., 2020).
Based on the analysis of compound-specific stable isotopes, Kohlbach et al. (2016) concluded that up to 92 % of the dietary carbon of A. glacialis from the central Arctic Ocean (CAO) was sourced from ice algae during summer, and that the sea-ice signal was transferred to young polar cod inhabiting the under-ice environment, preying extensively on these amphipods (Kohlbach et al., 2017).Using the presence of sympagic highly branched isoprenoid (HBI) lipids (Brown and Belt, 2012;Brown et al., 2014;Belt, 2018), an approach also applied in this study (details in Schmidt et al., 2018;Kohlbach et al., 2021a), the importance of ice algae as a carbon source for A. glacialis was similarly high in individuals collected in Svalbard waters during summer (Brown et al., 2017).A somewhat lower contribution (up to 79 %) from sea-ice algae was found for the carbon budget of Eusirus holmii in the CAO during summer (based on the stable isotope method; Kohlbach et al., 2016), while Brown et al. (2017) identified a relatively low ice algae carbon content for this species sampled in the Nansen Basin, implying that it obtained the majority of its organic carbon from other sources.
Besides these previous results, information on carbon sources utilized by under-ice fauna using trophic marker approaches is relatively scarce (e.g., Budge et al., 2008), emphasizing the need for more data collection from different regions of the Arctic Ocean as well as from different seasons, especially as dietary requirements of marine animals can underlie seasonal trends (Kohlbach et al., 2021b).Changes at these lower trophic levels can have important implications for the entire carbon flow as well as carbon cycling (Ehrlich et al., 2021).As a region of extreme environmental change, the Barents Sea offers the opportunity to closely study the effects of climate change on marine life (e.g.Dalpadado et al., 2020;Koenigstein, 2020;Ingvaldsen et al., 2021).Sea-ice ridges have been suggested as ecological hotspots that host a particularly rich biodiversity and suitable habitat for sea ice-associated flora and fauna (Hop et al., 2000;Gradinger et al., 2010;Lange et al., 2017;Fernández-Méndez et al., 2018), but are likely to be altered in structure and distribution under increasing temperatures and wind (Wadhams and Toberg, 2012).To study carbon and food source compositions of marine organisms, a suite of biochemical methods can be used, including the analysis of fatty acids (FAs), HBIs and sterols, preferably applied as a multi-trophic marker approach (discussed in Kohlbach et al., 2021a and references there).Here, we investigated the FA composition of typical Arctic underice fauna species (A.glacialis, E. holmii, polar cod) collected in the (northern) Barents Sea to assess their dietary requirements (see Dalsgaard et al., 2003) during May 2021, when ice algae were expected to represent a nutritious food source.Furthermore, we traced the sea-ice signal from producers to consumers using the presence/absence of sympagic HBIs (i.e., IP 25 and IPSO 25 ; e.g.Kohlbach et al., 2019;Yunda-Guarin et al., 2020).Sterols reflected the importance of heterotrophic food items in the diet of the investigated species (Ruess and Müller-Navarra, 2019).The results contribute to our understanding of the importance of sea ice-derived food sources for the ice-associated food web under current sea-ice conditions.

Sampling
As part of the Norwegian Nansen Legacy project (arvenetternansen. com;Wassmann, 2022), samples of pelagic particulate organic matter (PPOM), ice-associated POM (IPOM) and typical representatives of the Arctic under-ice fauna were collected during the Nansen Legacy seasonal cruise Q2 (27 April to 20 May 2021) with RV Kronprins Haakon in the northern Barents Sea.Samples for this study were taken between 9 and 13 May 2021 at the ice-covered sampling stations P6 and P7 (Fig. 1A, Table 1).The sea ice encountered during our sampling campaign was ridged at both sampling stations (Fig. 1B, C).
PPOM was collected at 20 m depth with Niskin bottles attached to a CTD rosette.IPOM was sampled in sea-ice ridges by divers with a slurp gun at both stations (Fig. 2A, B).Both seawater (between 2.5 and 3 L) and IPOM samples (50 mL) were filtered through pre-combusted (3 h, 550 • C) 47 mm Whatman® GF/F filters.Filters were frozen at − 80 • C until further processing.
The under-ice sampling of ice fauna was facilitated by divers with varying equipment (Fig. 2C, D).The under-ice amphipods Apherusa glacialis and Eusirus holmii were collected with a suction pump at both stations (Fig. 2C, Table 2).Small individuals of A. glacialis (<10 mm) were pooled to obtain sufficient material for subsequent analyses (between 5 and 9 individuals per sample) and up to three individuals of E. holmii were pooled.Polar cod were caught with a spear (Fig. 2D) and immediately dissected to retrieve muscle and liver tissue for analyses.All samples were frozen at − 80 • C.
Chlorophyll (chl) a was sampled at 20 m with a CTD rosette, for the bottom 10 cm of level sea ice with a 9 cm diameter Kovacs ice corer and at ridge surfaces by divers with a slurp gun (Fig. 2B).All samples were filtered through 25 mm Whatman® GF/F filters under low vacuum pressure (~30 kPa).Filters were stored in polypropylene tubes with mL of methanol added for extraction (overnight at 0-4 • C).Chlorophyll a concentrations were measured on a Trilogy Laboratory fluorometer (Turner Designs, CA, USA) before and after acidification with two drops of 5 % HCl according to Holm-Hansen and Riemann (1978).

Fatty acids (FAs)
FA compositions of POM and animals were analysed at Akvaplanniva, Tromsø, Norway.Lipids were extracted after Folch et al. (1957) with methanol/dichloromethane (1:2).FAs were transesterified to fatty acid methyl esters (FAMEs) with 1 % sulfuric acid in methanol at 50 • C for 16 h and extracted with hexane/ether (1:1, v/v).FAMEs were analysed on an Agilent 7890 gas chromatograph with a capillary glass column (Agilent/Varian CP7419 Select FAME; 50 m × 0.25 mm ID × 0.25 µm film thickness), equipped with a flame ionization detector (FID) using a temperature program (60 Samples were injected split-less and hydrogen was used as a carrier gas.FAMEs were identified via reference standard mixtures (GLC96 and GLC68D from NuCheck-Prep) containing known amounts of the relevant compounds.FAs were quantified with an internal standard (21:0) that was added prior to lipid extraction, using individual FA response factors determined from the standard solutions.The proportions of individual FAs were expressed as mass percentage of the total FA content as well as concentrations (g 100 g lipid -1 ; Table A.1). Sample size for each sample type/species can be found in Table 4.

Highly branched isoprenoids (HBIs) and sterols
HBIs and sterols were analysed at the University of Plymouth, UK (polar cod: only muscle tissue was analysed).Detailed information on analytical procedure and equipment is given in Kohlbach et al. (2021a).
HBIs were eluted with hexane (five column volumes) before being dried under gentle N 2 stream (25 • C) and analysed by gas chromatography-mass spectrometry (GC-MS).The analysis was carried out on an Agilent 7890A gas chromatograph (GC), coupled to an Agilent 5975 Mass Selective Detector (MSD), fitted with an Agilent HP-5 ms column (30 m length × 0.25 mm internal diameter × 0.25 µm film thickness) with auto-splitless injection (300 • C) and helium carrier gas (1 mL min − 1 constant flow).Analysis was carried out by total ion current (TIC; m/z 50-500) and selective ion monitoring (SIM) techniques (70 eV) using a ramped temperature programme of 10 • C min − 1 from 40 to 300 • C followed by a 10 min isothermal at 300 • C. Identification of HBIs was achieved following analysis in selective ion monitoring (SIM; m/z 350.3 [IP 25 ], m/z 348.3 [IPSO 25 ] and m/z 346.3 [HBIs III and IV]; limit of detection = 10 ng mL − 1 ) by comparing relative retention times with those of purified standards.Quantification of HBIs was achieved by integrating individual ion responses in SIM mode and normalising these to the corresponding peak area of the internal standard (9-octyl-8-heptadecene (9-OHD): m/z 350.3) and an instrumental response factor obtained from purified standards (Belt et al., 2012).The two triunsaturated HBIs (HBIs III and IV) were used to assess the consumption of pelagic algal material, including that from the marginal ice zone (MIZ) (Brown et al., 2014;Belt, 2018).The mono-and di-unsaturated HBIs IP 25 and IPSO 25 , assumed to be produced exclusively by certain sea-ice diatoms, were assessed to provide information about the reliance of the consumers on sympagic carbon (Brown et al., 2018;Schmidt et al., 2018).
Sterols were eluted from the silica column using hexane:methylacetate (4:1,v/v), dried (N 2 ) and derivatised using N,O-bis (trimethylsilyl)trifluoroacetamide (BSTFA; 70 • C, 1 h) prior to analysis by GC-MS.Individual sterols were identified by comparison of the mass spectra of their trimethylsilylethers with published data (Belt et al., 2018).Quantification of individual sterols was carried out in the SIM mode utilising 5α-androstan-3β-ol (m/z 333) as an internal standard.
The GC-MS-derived masses of HBIs and sterols were converted to water column/sea-ice concentrations using the volume of filtered sea water (2.5-3 L) and sea ice (50 mL), respectively, and to concentrations in zooplankton body using the mass of the sample extracted (8-100 mg).
Sample size was n = 3 for PPOM, n = 10 for IPOM, n = 14 for A. glacialis, n = 3 for E. holmii and n = 4 for polar cod.Muscle, liver a wet weight per individual for FA analysis.b dry weight per individual for HBI analysis.c wet weight entire fish.

Data analysis
Variability in FA datasets of POM and the under-ice fauna species was visualized with correspondence analysis (CA), suitable for analysing compositional data (Greenacre and Primicerio, 2014;Greenacre, 2017).Differences of FA proportions between the two sampling stations in PPOM, IPOM and A. glacialis were assessed using Wilcoxon rank sum tests; 1-way ANOVAs followed by Tukey's HSD post-hoc tests were applied when assessing seasonal differences of FA proportions in PPOM and IPOM.A statistical threshold of α = 0.05 was chosen, and results with p ≤ 0.05 were considered significant.Prior to statistical analysis, the data were verified for normality of distribution.All measures of statistical variation are reported as means ± 1 SD.All statistical analyses and data visualization were done in R v.4.1.0(R Core Team, 2021) using the packages 'ggplot2 ′ (Wickham, 2016) and 'vegan' (Oksanen et al., 2020).

Results & discussion
The under-ice biota sampled by the divers was strongly dominated by Apherusa glacialis, which is in line with its pan-Arctic occurrence and importance as part of the sympagic fauna in Arctic ecosystems (David et al., 2015;Ehrlich et al., 2020;Hop et al., 2021).Eusirus holmii has been described as a hyperbenthic species, restricted to water masses with temperatures lower than 1 • C (Weisshappel, 2000), but with a wide depth range from 400 to 1600 m (Lörz et al., 2018).Eusirus holmii is, however, also well known to be associated with the sea-ice habitat (Tencati and Geiger, 1968) and has been identified as part of the underice fauna in the CAO (David et al., 2015) and the waters north of Svalbard (Ehrlich et al., 2020).Eusirus holmii is typically less abundant in the ice-fauna community than A. glacialis (Brown et al., 2017;Ehrlich et al., 2020), which was also the case in this study.Juvenile polar cod Boreogadus saida has frequently been described to inhabit the under-ice environment (Gradinger and Bluhm, 2004;Melnikov and Chernova, 2013;David et al., 2016;Huserbråten et al., 2019).Although generally abundant in the northern Barents Sea (Hop and Gjøsaeter, 2013;McBride et al., 2014), in this study, the divers collected only four individuals at station P6.If present, polar cod are generally hiding within ridges, in open spaces as present at P6 (but not at P7).With many demands for sampling by divers, polar cod were only searched for and caught in a few dives during the expedition.

Pelagic and ice-associated particulate organic matter
Despite their close proximity, the two sampling stations reflected different water mass regimes (Fig. 1A): P6 was more representative of the highly productive shelf-regime, while P7 was more representative of the low-productive central Arctic Basin.An under-ice phytoplankton bloom occurred at station P6, which was dominated by pelagic diatoms of the genus Thalassiosira (P.Assmy, pers.obs., and congruent with the literature: Moran et al., 2012), while the chl a concentrations in the water column at P7 were lower by one order of magnitude (Table 3).Chlorophyll a concentrations > 10 mg m − 3 in the bottom three cm of level sea ice at both stations (Table 3) indicated that the vernal ice algal bloom was in full progress during our sampling.Chlorophyll a concentrations in the bottom sections of level ice at the same sampling locations were distinctly lower during (late) summer (≤2.3 mg m − 3 ) and (early) winter (≤0.3 mg m − 3 ) (Vader et al., 2021), emphasizing seasonal variations in ice-algal productivity (Leu et al., 2011;Ji et al., 2013).The ridge at P7 was more consolidated (H.Hop, pers. obs.) and had fewer open spaces and less surface area compared to the ridge at P6, which  might explain the somewhat lower chl a biomass in the different ridge surfaces sampled at P7. Fatty acid (FA) profiles of PPOM and IPOM differed substantially (Fig. 3, Table 4) and also between the two sampling locations (Fig. 4), pointing, on the one hand, to differences in algal community composition and on the other hand, to differences in physiological state of the algal communities and bloom succession (Falk-Petersen et al., 1998;Wang et al., 2014).Consistent with some previous studies (McMahon et al., 2006;Kohlbach et al., 2021a), the diatom signal was overall stronger in IPOM than in PPOM; however, dinoflagellates contributed to both pelagic and sympagic algal communities in low quantities (Table 4).
In PPOM, the diatom-associated FAs 16:1(n-7) and 16:4(n-1) had higher average contributions to the FA profile at station P6 compared to P7, while the dinoflagellate signal was similar at both stations, which is in agreement with the dominance of pelagic diatoms at P6 (Fig. 4A).Concentrations of the diatom-associated FA 20:5(n-3) in PPOM were similar between the stations.IPOM collected at station P6 had, on average, twice as much 16:1(n-7) compared to P7, whereas the other two diatom-associated FAs 16:4(n-1) and 20:5(n-3) had significantly higher contributions at station P7 compared to P6 (Fig. 4B).As seen for PPOM, contributions of both dinoflagellate-associated FAs (18:4(n-3) and 22:6(n-3)) were more similar between the two sampling locations than diatom-associated FAs (Fig. 4B), indicating that diatoms were the main drivers of variability.
Polar lipids used for the stabilization of bio-membranes have a high contribution of polyunsaturated FAs (PUFAs), such as 20:5(n-3) and 22:6(n-3) (Lee et al., 2006), which are essential for successful growth and reproduction of Arctic animals (Sargent, 1995;Brett and Müller-Navarra, 1997).Concentrations are highest in the beginning of the bloom (Falk-Petersen et al., 1998), translating into a high food quality for the dependent food web (Gulati and Demott, 1997;Søreide et al., 2010), with a gradual shift to a dominance of neutral lipids for energy storage towards the end of the bloom (Smith et al., 1993).In accordance with the literature and the seasonal progression of blooming events, in the current study, IPOM had the highest PUFA production in spring and lowest nutritional value during advanced summer ice melt (Kohlbach et al., 2021a), while PUFA production by pelagic algae peaked during summer (Fig. 5).It has to be considered, however, that the FA content is not only a function of time, but also strongly dependent on the dominating algae species (Henderson et al., 1998).
At P6, IPOM collected by the divers contained all HBIs analysed for, of both pelagic (HBIs III and IV) and sympagic (IP 25 and IPSO 25 ) origin (Fig. 6A).In all P6 IPOM samples, the concentrations of pelagic HBIs were approx.ten times higher (sum up to 3272 pg mL − 1 ) than sympagic HBIs (sum up to 315 pg mL − 1 ).HBI III (1546 to 2413 pg mL − 1 ) was approx.three times more concentrated than HBI IV (571 to 859 pg mL − 1 ), while IPSO 25 concentrations (208 to 279 pg mL − 1 ) were roughly one magnitude higher compared to IP 25 concentrations (26 to 36 pg mL − 1 ).The presence of pelagic HBIs in IPOM is likely a consequence of the sampling technique and the under-ice bloom at P6 during the time of sampling.The divers collected bottom-ice algal material with slurp guns, which also sucked in some of the surrounding seawater at the icewater interface, a suggestion supported by microscopic identification of pelagic diatoms such as Thalassiosira spp. in some slurp gun samples (P.Assmy, unpubl.data).In contrast, no HBIs were detected in IPOM at P7.The sympagic HBI-producing algal taxa (Haslea crucigeroides, H. spicula, H. kjellmanii and Pleurosigma stuxbergii var rhomboides) are generally not among the dominant taxa in Arctic ice-algal communities (von Quillfeldt, 2000;Brown et al., 2014).In our study, Haslea spp.were not detected in taxonomic investigations of slurp gun samples, while Pleurosigma sp.occurred at very low abundances (P.Assmy, unpubl.data).Thus, the absence of HBIs in IPOM from P7 was likely the result of the lower algal biomass and thus lower contribution of HBI-producing algae.Moreover, the ridge at P6 was younger with a more complex surface area and thus potentially with more algal patches than the older more consolidated ridge at P7 (H.Hop, pers. obs.).No HBIs were detected in the bottom 3 cm of ice cores collected with a 9 cm ice corer (Kovacs) at both stations (n = 4), with the exception of IP 25 at low concentration in one ice core from station P7 (4.6 pg mL − 1 ).These differences in HBI content between material collected by divers and those collected with an ice corer might reflect differences in mode of collection.The distribution of ice algae is known to be heterogeneous (Rysgaard et al., 2001;Lange et al., 2016), leading to potentially high variability in sample material collected 'blindly' with ice corers.In contrast, divers targeted under-ice areas with visibly high levels of biomass, increasing the probability of collecting HBI-producing algae in higher concentrations.Another explanation could be the loss of bottom-ice algal material during the ice core retrieval.
Concentrations of phytosterols and zoosterols were positively correlated in the POM samples (Fig. 6B; Pearson's correlation: corr.
e Relative proportions of polyunsaturated FAs (PUFAs) can be used as an indicator of food quality of an algal community (Søreide et al., 2010).
D. Kohlbach et al. coeff.= 0.97, p < 0.001).The ratios of zoosterols/phytosterols were somewhat lower in PPOM (1.9 ± 0.6) than in IPOM (2.7 ± 1.0) and were rather variable between IPOM from P6 (3.4 ± 0.9) and P7 (2.0 ± 0.5).The higher concentrations of phytosterols in IPOM from P6 were in accordance with the higher algal biomass and the detection of sympagic HBIs at this station, likely resulting in a proportionally higher contribution of HBI-producing algal taxa (see above).Due to the sampling technique (no prefiltering of samples), POM samples are likely to contain heterotrophic material to some extent (i.e., components of under-ice fauna in PPOM and/or IPOM; in-ice meiofauna, microbial components in IPOM), which can explain the presence of zoosterols in the POM samples.Moreover, some zoosterols can be synthesized in both algae and animal cells (Belt et al., 2018).

Apherusa glacialis
All three FA ratios indicated the dominance of diatom-derived FAs over dinoflagellate-derived FAs (Table 4), irrespective of the sampling location, which is in agreement with previous studies identifying diatoms as an important food item for this species (Scott et al., 1999;Kohlbach et al., 2016).As found for the POM samples, there was large variability in individual FA proportions in A. glacialis between the two stations (Fig. 3), pointing to their ability to utilize carbon and food of different origin, and possibly mirroring their varying (trophic) sea-ice association.Contrary to both POM types, the average relative contribution of the diatom-associated FA 16:1(n-7) was twice as high in A. glacialis from station P7 compared to P6 (Fig. 4C), which was similar to proportions detected in A. glacialis from the CAO during late summer, where 16:1(n-7) accounted, on average, for ca.half of the FA content Fig. 4. Differences in the relative proportions (mean ± 1 SD %) of the trophic marker fatty acids 16:1(n-7), 16:4(n-1) and 20:5(n-3) (diatom-associated), 18:4(n-3) and 22:6(n-3) (dinoflagellate-associated) in (A) pelagic particulate organic matter (PPOM), (B) ice-associated POM (IPOM) and (C) the amphipod Apherusa glacialis at stations P6 and P7.Associated bars marked with asterisk "*" represent significant differences between the stations (Wilcoxon test p < 0.05).Fig. 5. Seasonal differences in the relative proportions (mean ± 1 SD %) of the trophic marker fatty acids 16:1(n-7), 16:4(n-1) and 20:5(n-3) (diatom-associated), 18:4(n-3) and 22:6(n-3) (dinoflagellate-associated) in (A) pelagic particulate organic matter (PPOM) and (B) ice-associated POM (IPOM).Dates of sampling: springthis study, summer-5 to 27 August 2019, winter-28 November to 17 December 2019.I-POM samples from summer and winter were collected via ice coring (details on sampling and stations: Kohlbach et al., 2021a, b).Considered are only samples collected at stations P6 and P7 in all seasons (exception: winter IPOM sampling at station P5 instead of P6; see Kohlbach et al., 2021b for map).Note: different scales.Asterisk "*" represents significant differences between the seasons listed below (ANOVA test p < 0.05).(Kohlbach et al., 2016).The opposite trend was found for the contribution of the diatom-associated FA 20:5(n-3), which was twice as high at station P6 compared to P7 (Fig. 4C).The third typical diatom-associated FA 16:4(n-1) was also more prominent in the P6 than in the P7 samples.Overall, the diatom signal was stronger in A. glacialis from P7 (sum of diatom-associated FAs: 53.6 %) compared to P6 (sum of diatomassociated FAs: 46.3 %).Both dinoflagellate-associated FAs 18:4(n-3) and 22:6(n-3) were significantly higher concentrated in A. glacialis at P6 than P7 (Fig. 4C).This station-specific difference could not be observed in either of the POM community samples, which could mean that the A. glacialis FA profile was not representative of the algal community at the time of sampling.The exact turnover rate of FAs in the amphipod is to date unclear but could be several weeks (Boissonnot et al., 2016(Boissonnot et al., , 2019)).The observed differences between baseline (i.e., PPOM and IPOM) and consumer can also reflect the (vertical and horizontal) mobility of A. glacialis, encountering varying carbon sources, i.e., algal communities, throughout the water column as well as at the underside of the ice.In A. glacialis from the CAO, there was a spatial variability in FA proportions, possibly linked to differences in seasonal progression of the ecosystem and different environmental conditions between the Nansen and Amundsen Basins (Kohlbach et al., 2016).In the CAO study, A. glacialis from the more northern Amundsen Basin had higher proportions of 16:1(n-7) and lower proportions of all other trophic marker FAs compared to individuals from the Nansen Basin, similar to station P7 versus P6, which was likely the result of lower nutrient and consequently chl a concentrations at P7.
In accordance with its previously reported strong association to the sea-ice primary production in both seasonal and perennial sea-ice systems (Søreide et al., 2006;Budge et al., 2008;Kohlbach et al., 2016), A. glacialis was the only under-ice fauna species investigated in this study containing detectable amounts of sympagic HBIs (Fig. 6A).The lack of sympagic HBIs IP 25 in all and IPSO 25 in about half of the samples (eight out of 14 samples) is likely a detection limit issue rather than the absence of these lipids in A. glacialis, and further mirrors the distribution of sympagic HBIs in IPOM (see above).Concentrations of the pelagic HBIs were close to the detection limit (0 to 57 ng g − 1 ), and since the sympagic HBIs occurred in distinctly lower concentrations in IPOM (Fig. 6A), the failure to detect sympagic HBIs in the consumers was not surprising, unless there was preferential uptake.The presence of IPSO 25 in A. glacialis at P7 (0 to 75 ng g − 1 ) but not the pelagic HBIs III and IV in most of the samples likely reflects a change in diatom species
2.6 ± 0.6 7.1 ± 3.9 18.4 ± 2.8 5.9 ± 0.9 D. Kohlbach et al. composition between P6 and P7 (see 3.1).The detection of sympagic HBIs in A. glacialis at station P7 but not P6 (Fig. 6A), points to a stronger reliance on pelagic carbon sources at P6, supported by the FA profiles, which was likely related to grazing on the under-ice bloom.In contrast, low pelagic productivity left few alternatives to ice algal carbon at P7.This dietary opportunism contradicts the common notion that A. glacialis represents a typical ice amphipod with an exclusively sympagic diet (Werner 1997).
The higher phytosterol content in combination with lower zoosterol concentrations compared to the other under-ice fauna species (Fig. 6B) further mirrors A. glacialis' predominantly herbivorous (detritivorous) lifestyle (Poltermann, 2001;Arndt et al., 2005), assuming that sterol distributions can be used to distinguish algae-versus animal-produced carbon to some extent (Drazen et al., 2008;Ruess and Müller-Navarra, 2019).In contrast to the highly variable FA profiles and HBI content, ratios of zoosterols/phytosterols in A. glacialis showed relatively low variability between P6 (40.3 ± 4.7) and P7 (34.5 ± 8.2), suggesting similarities in feeding behaviour, i.e., the reliance on algal-versus animal-derived material, irrespective of the sampling station, prevailing environmental conditions and carbon source/prey availability.

Eusirus holmii
Individuals of E. holmii covered a large size range (<10 to 40 mm) and thus likely multiple ontogenetic stages (juveniles and adults, respectively).The largest individual analysed for FAs (40 mm) had a substantially lower diatom signal compared to the other three individuals (particularly low content of 16:1(n-7): 2 %, others: 13 to 23 %), while containing more PUFAs 20:5(n-3) (20 %, others: 10-16 %) and 22:6(n-3) (30 %, others: 8-12 %), indicating a different diet and/or body condition than the smaller individuals.The two samples with individuals below 10 mm in size also had lower PUFA proportions than the second large individual (30 mm), indicating that the proportion of membrane lipids increased with size/age while lipid storage capacities decreased.In support of this assumption, the two adults had a lower percentage of extracted lipids per wet weight (0.6 and 1.3 %) compared to the juveniles (3.1 and 6.0 %).As expected, FA carnivory ratios of 18:1 (n-9)/18:1(n-7) (Falk-Petersen et al., 1990;Graeve et al., 1997) increased with size from 3.6 to 10.3, indicating the higher importance of heterotrophic food sources for adult E. holmii compared to the younger individuals.In accordance with that result, the ratio of zoo-to phytosterols was also substantially higher in a large individual (351.3)compared to two small animals (85.8 ± 3.1), further suggesting a more pronounced carnivorous lifestyle of older individuals.There was evidence of Calanus feeding, concurrent with E. holmii's reported predatory lifestyle (Macnaughton et al., 2007), but overall copepod-associated FA proportions were relatively low (Table 4).Interestingly, proportions of the copepod-associated FAs 20:1 and 22:1 were higher in smaller (younger) (up to 17 and 10 %, respectively) than larger (older) E. holmii (up to 6 and 3 %, respectively).Based on this, adult E. holmii likely targeted other (perhaps larger) prey than Calanus copepods, such as amphipods (Macnaughton et al., 2007).Spatial variability in FAs could not be assessed in this study since all individuals were collected at station P6.
The FA profile of the specimen in our study varied greatly from E. holmii collected during August in the CAO (Kohlbach et al., 2016).For example, the average relative proportion of 16:1(n-7) in individuals from the CAO was 36 % (here: 15 %).Moreover, in E. holmii from the CAO, 82 % of the lipids were storage lipids.In this study, we did not determine the levels of storage versus membrane lipids, but from the high proportions of membrane-bound PUFAs (Table 4), it can be assumed that polar membrane lipids had a greater contribution to the overall lipid content of the animals.These differences likely reflect seasonal differences in diet and body condition as well as geographically-driven dietary variation and/or availability of prey.The animals collected for this study made use of their lipid storage during the long polar winter, while the animals collected in the CAO were likely continuously feeding throughout the summer.
In our study, we did not detect sympagic or pelagic HBIs in E. holmii.Brown et al. (2017) suggested that E. holmii likely relied mainly on other food sources than sea-ice algae north of Svalbard, which could have also been the case here and has therefore a relatively low relevance for the transfer of ice-derived carbon through the food web.Contrastingly, results derived from stable isotopes indicated a strong reliance on ice-algal carbon for E. holmii from the CAO (Kohlbach et al., 2016).Animals inhabiting the CAO have possibly a stronger sea-ice association due to the presence of sea ice year-round and the generally low pelagic productivity that needs to be counterbalanced; during August/September, the time of sampling, ice algal primary production can represent 50 % of the total primary production (Fernández-Méndez et al., 2015).As mentioned above, the seemingly absence of these ice proxies can further point to limitations of the analytical system, i.e., concentrations below the detection limit.

Polar cod Boreogadus saida
As reported previously (Kohlbach et al., 2017), the FA profiles of polar cod tissue (muscle and liver) differed greatly from each other (Fig. 3, Table 4).The liver is known as the main lipid storage organ in polar cod (Hop et al., 1995), accordingly the proportions of storageassociated FAs such as 16:1(n-7) and the copepod-associated FAs were significantly higher in liver versus muscle tissue.In the liver tissue, nearly half of the FAs were copepod-associated FAs, reflecting their strong reliance on Calanus spp. as food (Bouchard and Fortier, 2020).This is in line with the diet composition of polar cod from the western Barents Sea during spring (Lønne and Gulliksen, 1989), where C. glacialis was found to be the main prey.In our study, there is evidence of a preference for C. hyperboreus over other Calanus species due to the similarly high proportions of 20:1 and 22:1, a pattern typically found in the largest of the Calanus species inhabiting the Barents Sea.Supporting this assumption, C. hyperboreus is known to be associated with deep waters (Hirche, 1997; bottom depth > 3000 m at P7) and contributed largely to the Calanus copepod biomass at stations P6 and P7.
The diet composition of polar cod can be quite diverse, with differences on a spatial and temporal scale as well as related to ontogeny, reflecting availability and accessibility of their preferred prey as well as their opportunistic feeding behavior (Ajiad and Gjøsaeter, 1990;Hop and Gjøsaeter, 2013;Matley et al., 2013;Gray et al., 2016;Nakano et al., 2016).In comparison to polar cod collected during summer in the CAO, relative proportions of the diatom-associated FA 16:1(n-7) were at least three times lower in this study in both tissues (CAO muscle: 16 %, liver: 28 %; this study muscle: 3 %, liver: 10 %).The stomach content of the CAO fish was dominated by A. glacialis (Kohlbach et al., 2017), which can explain the strong 16:1(n-7) signal, a FA that contributed to almost half of the FA content in the amphipod (Kohlbach et al., 2016).These differences could be explained by varying availability of their prey; A. glacialis occurred in high numbers in the CAO (David et al., 2015), and might have simply been less abundant or less easily accessible underneath the thinner, ridged ice in the Barents Sea.Furthermore, the fish in the present study was in a later ontogenetic stage, based on their larger size (CAO: average 78 mm, this study: average 118 mm) and wet weight (CAO: average 3 g, this study: average 11 g).Larger polar cod have a greater prey diversity, including amphipods such as Themisto libellula (Renaud et al., 2012;Dalpadado et al., 2016).
No HBIs could be detected in the polar cod muscle tissue, which might point to concentrations below the detection limit as mentioned above.The lack of sympagic HBIs could further be related to the seemingly lower reliance on A. glacialis, which likely transferred the seaice signal in the individuals from the CAO, where the ice algal contribution was quantified with up to 65 % in the muscle tissue based on FAspecific stable isotope analysis (Kohlbach et al., 2017).There might also be a moderate to low importance of sympagic food for polar cod in the sampling region, as polar cod sampled during August and November/ December in the Barents Sea did also not contain any detectable amounts of HBIs, with the exception of pelagic HBIs in low concentrations in few individuals (unpubl.data).Despite their presence underneath the sea ice, the polar cod in our study might have had a stronger reliance on pelagic food sources, as also suggested for juvenile polar cod from the Barents and Beaufort Seas (Hop and Gjøsaeter, 2013;Graham et al., 2014), presumably feeding on copepods utilizing the under-ice bloom at P6.As discussed above, these results might reflect the spatial variability in food composition with a larger probability and requirement to incorporate sympagic food items in a sea ice-dominated ecosystem such as the CAO.Sterol composition and high zoosterol/ phytosterol ratios (181.7 ± 25.5) showed relatively little variability between the four polar cod individuals (Fig. 6B).The general similarity in trophic marker composition among muscle and liver tissue, respectively, points to a resemblance in food sources utilized by the different individuals collected at station P6.

Conclusions
In this study, trophic marker evidence of the reliance on sympagic carbon was found in the under-ice amphipod Apherusa glacialis, confirming its previously reported trophic sea-ice association.The low concentrations of sympagic HBIs in A. glacialis along with their absence in the amphipod Eusirus holmii and polar cod possibly reflect a low contribution of HBI-producing taxa to the sea-ice diatom community.High spatial variability in FA profiles and in the presence of sympagic HBIs suggest that the dietary composition of A. glacialis might be rather flexible and opportunistic, and perhaps less ice-algal dependent compared to regions that are ice-covered year-round.In comparison to a perennially ice-covered system, a lower importance of sympagic carbon to the diet of the investigated under-ice fauna species and, in general, for trophic processes in the seasonally ice-covered Barents Sea might be expected since pelagic carbon sources are abundant during the productive season.In accordance with this assumption, an under-ice bloom occurring at station P6 likely offered sufficient carbon and food sources that might have been preferably utilized by the under-ice fauna, suggesting that such species are well adapted to exploiting pelagic food items.In contrast, the less productive pelagic system at P7 likely resulted in a stronger reliance on sympagic carbon, as shown for A. glacialis.These results imply that ice-associated species do not necessarily rely exclusively on sympagic carbon: When following the Transpolar drift, A. glacialis is likely to traverse different productivity regimes and adopt the ability to switch between sympagic and pelagic food items.This trophic plasticity might provide them with an advantage regarding continued sea-ice retreat and accompanied changes in sea-ice primary production but could potentially also impact their nutritional value for higher trophic levels depending on the food quality of their carbon sources.
Compared to level sea ice, sea-ice ridges as encountered in our study particularly facilitate the attachment of algae by providing greater surface area sheltered from harsh environmental surroundings (Fernández-Méndez et al., 2018), resulting in sympagic algae communities with high biomass and diversity, which in turn attract herbivorous ice fauna and consequently their predators (Hop et al., 2000;Gradinger et al., 2010).Sea-ice ridges are generally understudied despite their importance as structural features of the Arctic sea-ice habitat (Hutter et al., 2019;Lange et al. under review).Hence, it is unclear whether environmental changes of the Arctic icescape will eventually benefit or hamper the development of sea-ice ridges (Wadhams and Toberg, 2012;Fernández-Méndez et al., 2018;Lange et al. under review).Regardless of changes in sea-ice ridges in future decades, environmental change will have an impact on ice-associated fauna, their trophic interactions and the energy flow through the entire Arctic food web.Our study contributes to our current knowledge on the usage of sea-ice ridges as a feeding ground for ice-associated fauna and encourages further investigations of the sympagic habitat to gain information about potential differences in the importance of ice-associated algae for food-web functioning during different seasons, regions and in comparison to level sea ice.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Table 2
Dimensions of species analysed in this study (mean ± 1 SD, range is given in brackets).

Table 3
Chlorophyll a concentrations (A.Vader et al., dataset submittedfor publication) at 20 m depth and in sea ice at the two sampling locations.
a based on three pooled ice-core bottoms. .1.
a diatom-associated FA. b dinoflagellate-associated FA. c copepod-associated FA: isomers 20:1 and 22:1 are synthesized de novo by Calanus spp.; they typically indicate the importance of Calanus spp.and other copepods as a food source for a consumer