Under-ice availability of phytoplankton lipids is key to freshwater zooplankton winter survival

Shortening winter ice-cover duration in lakes highlights an urgent need for research focused on under-ice ecosystem dynamics and their contributions to whole-ecosystem processes. Low temperature, reduced light and consequent changes in autotrophic and heterotrophic resources alter the diet for long-lived consumers, with consequences on their metabolism in winter. We show in a survival experiment that the copepod Leptodiaptomus minutus in a boreal lake does not survive five months under the ice without food. We then report seasonal changes in phytoplankton, terrestrial and bacterial fatty acid (FA) biomarkers in seston and in four zooplankton species for an entire year. Phytoplankton FA were highly available in seston (2.6 µg L−1) throughout the first month under the ice. Copepods accumulated them in high quantities (44.8 µg mg dry weight−1), building lipid reserves that comprised up to 76% of body mass. Terrestrial and bacterial FA were accumulated only in low quantities (<2.5 µg mg dry weight−1). The results highlight the importance of algal FA reserve accumulation for winter survival as a key ecological process in the annual life cycle of the freshwater plankton community with likely consequences to the overall annual production of aquatic FA for higher trophic levels and ultimately for human consumption.

high content of polyunsaturated fatty acids (PUFA). PUFA are a major component of lipids that are considered high quality food 14 . However, in a majority of allochthony studies based on stable isotopes, these lipid reserves have not been taken into account 15, 16 and sometimes have been chemically removed 17 , as they are not considered indicators of the recent diet but have potentially been accumulated over a long time period 18 . They are thus not considered as part of baseline metabolism, but rather as latent energy reserves to be available for future metabolic requirements, enabling consumers to survive and reproduce during periods of food scarcity 4 . In many marine copepods, lipids are exhausted over the course of winter, suggesting that they have an important metabolic role under the ice 13,19 . Interestingly, the highest copepod lipid content also in lakes has been measured during winter 18 precisely when phytoplankton resources are the scarcest in the environment. During this time, dissolved organic carbon (DOC) of terrestrial origin is abundant in boreal lakes 20 . Although it cannot be directly consumed by zooplankton 21 , DOC may be assimilated in the microbial loop and trophically upgraded by heterotrophic protists 22 , thus constituting a potential resource for zooplankton to build lipid reserves or fuel day-to-day metabolism in an environment with low primary production and phytoplankton abundance.
On-going climate warming leads to the mobilization of terrestrial carbon pools, thereby increasing allochthonous carbon inputs from catchments into surface waters worldwide 23 . The fate of this terrestrial pool, including its use in sustaining secondary production of plankton and higher trophic levels in lakes is still largely debated 15,24 . Many studies demonstrated a significant contribution of terrestrial organic matter to the biomass 8,16 . However, whether allochthonous carbon sources contribute to the lipid reserve accumulation of zooplankton in winter, and to subsequent metabolic investment to growth, is not known. This information would contribute to our understanding of the roles, and relative importance of, autochthonous and allochthonous carbon sources in sustaining aquatic food webs.
The composition of lipid reserves differs depending on whether their origin is phytoplanktonic, bacterial or terrestrial. While same FA can be synthesized by numerous different species, aquatic food webs are characterized by a high abundance of long-chain n-3 PUFA 25 with a strong prevalence of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) while terrestrial ecosystems have a dominance of n-6 PUFA and higher linoleic acid (LNA) and alpha-linolenic acid (ALA) contents 26 . Moreover, several FA are typically synthesized by terrestrial plants, whereas others are only found in aquatic primary producers 27 and in many cases may be attributed to specific taxa 28,29 . Long-chain saturated fatty acids (LC-SAFA, i.e. C20:0, C22:0, C24:0) are associated with terrestrial plants 30 and are abundant in terrestrial leaves e.g. senescent beech leaves contain >40% of SAFA 31 . LC-SAFA are characteristic to various terrestrial plant species from temperate and boreal biomes, including birch, alder, cottonwood, maple and willow but they are common also in reed and peat 26,32 . Other FA, such as branched and odd numbered FA (iso-and anteiso-C15:0 and C17:0) are biomarkers of bacteria 33 . Among PUFA, eicosapentaenoic acid (EPA, 20:5n-3), docosahexaenoic acid (DHA, 22:6n-3) and arachidonic acid (ARA, 20:4n-6) are mainly synthesized by algae 34 and are considered essential molecules needed by zooplankton for growth, reproduction and regulation of membrane fluidity. Aquatic organisms have limited capacity to synthesize them de novo and thus need to acquire these specific PUFA mainly from their diet 35 . Decreasing algal production is therefore believed to influence the PUFA availability in the environment. This poses a challenge for winter-reproducing species to produce eggs 36 and winter-active species to maintain membrane fluidity 37 , which requires PUFA, although PUFA are at lowest availability.
To better understand how phytoplanktonic, terrestrial and bacterial compounds contribute to zooplankton lipid reserves and how they are used by zooplankton in the course of the year and particularly during winter, we designed a study combining a laboratory survival experiment with a 12-month field survey. First, we experimentally estimated how long accumulated lipid reserves can sustain copepod metabolism. These results were then completed by a one-year lake sampling, where we measured in detail the phytoplanktonic, terrestrial and bacterial fatty acid availability in seston and their accumulation in four species of copepods and cladocerans. Since many zooplankton are active in winter under the ice and use lipid reserves mainly for reproduction 4 , we hypothesized that terrestrial organic matter and/or bacteria would be used as energy sources to maintain zooplankton in the absence of primary production. To test this hypothesis, we followed the survival of a calanoid copepod Leptodiaptomus minutus during winter without food and with winter seston diet, and expected survival to be higher when the copepods had access to winter seston presumed to be dominated by heterotrophic resources. We also hypothesised that terrestrial and bacterial fatty acids contribute to the lipid reserves according to their abundance in the seston, and that their contribution in the copepod lipid reserves would be highest in winter.

Results
Seasonal abundances in zooplankton community. L. minutus, Cyclops scutifer and Daphnia spp. were active during the whole year ( Fig. 1). They were also the most abundant zooplankton in the lake with average annual abundances of 12.9 ± 4.9, 3.3 ± 1.7 and 4.9 ± 2.8 ind L −1 (mean ± SD), respectively. They represented 58%, 15% and 22% of the total community abundance. L. minutus spent the winter as an adult, but the C. scutifer population was entirely made of copepodite (C-IV) individuals from November to April. The Daphnia population was comprised of adults under the ice, with relatively high abundances until 12 December and then rather low until mid-May. No other copepods or cladocerans were present in the water column during winter except highly abundant Bosmina spp. in early winter and punctual presences of Eucyclops speratus and Tropocyclops prasinus. The cyclopoid Mesocyclops edax was absent from the water column from November to May although being the fourth most abundant crustacean zooplankton in the lake (1.0 ± 1.1 ind L −1 and 5% of the average annual community abundance).
Starvation experiment. The survival of L. minutus at first followed the same pattern in both treatments but after ten weeks started to differ strongly between starved and fed conditions (Fig. 2). Access to food significantly increased zooplankton survival (F (1,245) = 1682.6, p < 0.0001) which was affected by time (F (24,245) = 230.3, p < 0.0001) and both factors interacted (F (24,245) = 48.7, p < 0.0001). During the first 66 days of the experiment, the survival was not significantly different among the treatments (p > 0.13; see Supplementary Table S2). Only from day 66 (29 January) to day 73 (5 February) the number of individuals alive in the starved condition dropped from  Seasonality in water chemistry and putative food sources. The most pronounced change in water chemistry, followed by shifts in putative food sources, took place in autumn. The most important increase in nutrient concentrations was registered from September to October (Table 1) when TDN almost doubled (×1.8) and TDP more than doubled (×2.2). DOC increased a month before, reaching its maximum already in September (6.6 mg C L −1 ). SUVA 254 was lowest in August  Table 1). The six phytoplankton FA biomarkers (FAB) dominated the seston FAB composition with an annual average of 1.7 ± 1.3 µg L −1 (mean ± SD), followed by terrestrial (0.3 ± 0.1 µg L −1 ) and bacterial FAB (0.2 ± 0.1 µg L −1 ) (Fig. 3). When expressed as a percentage of total FA in seston, the distribution was the following: 8 Table 1. Environmental and biological variables from the epilimnion of Lake Simoncouche represented by temperature (Temp, °C), total dissolved nitrogen (TDN, mg N L −1 ), total dissolved phosphorus (TDP, µg P L −1 ),dissolved organic carbon (DOC, mg C L −1 ), chlorophyll-a (Chl-a, µg L −1 ), bacteria biomass (Bact Biom, µg C L −1 ) and SUVA 254nm (L mg −1 m −1 ).  Figure 3. Seasonal pattern of seston fatty acids divided by FA biomarkers that represent phytoplankton, terrestrial and bacterial sources. Note the different scales on the Y-axis. Grey shade represents the period when the lake was ice covered. The sampling dates are indicated on the X-axis. FA with negligible contribution are not displayed.
phytoplankton FAB, 1.5% (±0.9) terrestrial FAB and 1.2% (±1.3) bacterial FAB. All FAB of the three potential energy sources were more abundant in summer than in winter. Phytoplankton FAB had an average concentration of 0.8 µg L −1 under the ice and 2.1 µg L −1 in the ice-free period (t = −2.25, p = 0.04). The same pattern was observed for terrestrial biomarkers with 0.2 µg L −1 in winter and 0.3 µg L −1 in summer (t = −2.77, p = 0.02) as well as for bacterial biomarkers with 0.1 µg L −1 in winter and 0.2 µg L −1 in summer, although this difference was not significant (t = −1.99, p = 0.06).

Total lipids and fatty acid composition in zooplankton.
A broad range of lipid content was observed in copepods, especially in L. minutus, whose lipid content ranged from 18% (of dry weight) in mid-September to 76% in late January (see Supplementary Fig. S1). C. scutifer had its lowest lipid content (21%) in May and highest (50%) in January. The late January maximum of lipids for L. minutus and C. scutifer revealed that these copepods had accumulated lipids under the ice, while the decrease in their lipid content was measured starting only at the end of February. The lipid content of M. edax was stable around 28 ± 9% (mean ± SD) with a maximum of 47% in May. The lipid content of Daphnia was highest in winter (33% in February) although it remained relatively stable throughout the year (22 ± 5%).
Principal component analysis captured 74% of the total variation in zooplankton FA composition, with 57% to axis 1 and 18% to axis 2 (Fig. 5). PUFA, phytoplankton FAB, n-3 FA and SAFA contributed 22%, 21%, 21% and  17%, respectively, to axis 1. Terrestrial FAB and bacterial FAB contributed 49% and 47%, respectively, to axis 2. L. minutus and C. scutifer were associated with axis 1 corresponding to high PUFA and phytoplankton FAB. The seston samples were characterised mainly by the presence of SAFA as were also the majority of Daphnia samples. SAFA composition in seston and Daphnia were dominated by C16:0 and C18:0 (see Supplementary Table S3). Some of the seston samples were further associated with bacterial and terrestrial FA. The composition of FA among taxonomic groups (L. minutus, C. scutifer, M. edax, Daphnia spp. and seston) was significantly different (PERMANOVA, F (4,122) = 93.2, p = 0.001).

Discussion
Our results demonstrate that the usually overlooked winter season involves key ecosystem processes. The high availability of phytoplankton PUFA in seston and their efficient accumulation by copepods establish a critical link between primary producers and higher trophic levels. By use of fatty acid biomarkers (FAB), we showed that during times of low primary production in winter, contrary to our hypotheses, the prevalence of terrestrial organic matter and/or bacteria did not increase in copepod or cladoceran lipid reserves. Instead, early winter accumulation of phytoplankton-derived PUFA and their progressive decline in copepod lipid reserves in midand late-winter suggests that phytoplankton FA are critical for several species of zooplankton to survive and remain in an active stage until spring. Energy to sustain winter metabolism and survival can be obtained from lipid reserve accumulation that has been considered to take place in autumn before the lake freezes 12 . Our experiment on L. minutus demonstrated that also early winter is a critical accumulation period; the individuals had not accumulated enough FA at the moment of ice formation to be able to survive the entire winter without external food input. At the ice-on, when the experiment started and the L. minutus were collected from the lake, the copepods had accumulated lipid reserves to survive for two months in laboratory, which carried them without additional food to February. However, the seston FA composition in the lake showed that the phytoplanktonic FA were available until January, i.e. for about one month after the ice had formed (Fig. 3). If we extrapolate the survival time without food, obtained from the experiment (66 days), we can calculate that in order to survive until the ice melting (18 April), L. minutus could have stopped eating on 12 February. This corresponds well with the moment the lake population of copepods ceased to accumulate FA reserves (19 January, Fig. 4), indicating FA in seston were exhausted. Interestingly, they started to use the FA reserves only a month later (29 February), probably corresponding to a minimum metabolic activity in mid-winter. Lake Simoncouche phytoplankton production begins before the ice disappears 4 providing fresh food for the overwintering populations in late winter. We argue that had the L. minutus remained in the lake longer before the initiation of the experiment they would have been able to accumulate the reserves needed to carry them over the months when there was no food in the water column. Few studies have estimated the survival of similar species under starved conditions. Elendt and Storch 38 discussed that L. minutus can survive 15.4 days, M. edax 24.3 days and Daphnia magna only 7.6 days. However, these authors do not report season, lipid content or temperature at which these survival times were estimated, thus limiting a direct comparison with our experiment. Nevertheless, the shorter survival time of Daphnia under starved conditions is in accordance with our abundance data that showed that Daphnia abundance declined close to zero once the seston PUFA pool was exhausted (Figs 1, 3).
The concentration of phytoplankton FA in seston peaked just before the ice-in and just after the ice-out, and followed the seasonal pattern of phytoplankton FA production in Lake Simoncouche 36 . The autumn peak decreased gradually but remained at a relatively high level for about a month under the ice (Fig. 3). Historically, phytoplankton primary production, like every biological activity, is assumed negligible in winter, but several recent studies reported under-ice primary production in oceans 39 and in many lakes 40 . A seasonal pattern similar to our seston FA composition has also been shown for lakes in the northern USA, with a very high spring bloom associated with phytoplankton FA abundance and a relatively high abundance in autumn 41 . Usually eicosapentaenoic acid (EPA, 20:5n-3) and docosahexaenoic acid (DHA, 22:6n-3) content in phytoplankton are limited by nitrogen and phosphorous concentrations in water 42,43 . Summer TDN and TDP shortage in Lake Simoncouche (Table 1) was followed by a nutrient increase during lake mixing (October). It appears plausible that the autumn nutrient increase in the water column allows the phytoplankton community to synthesize new PUFA that the nutrient depletion in summer does not permit. Also, n-3 FA are synthesized by different phytoplankton species 29 , and vertical mixing has been shown to cause major changes in phytoplankton community composition 44 . These phenomena may explain the n-3 FA maximum in autumn.
All FAB increased in quantity when the ice formed, and then decreased to their annual minimum by mid-winter (Fig. 3). The winter terrestrial FA minimum coincided with the minimal values of SUVA, indicating decreased input of terrestrial matter from the frozen catchment soils 45 . The minimum concentration of bacterial FA coincided with the minimum in bacterial biomass in winter. The same observation was valid for the autumnal FA peak confirming that seasonal changes in terrestrial and bacterial FAB concentrations followed the presence of terrestrial or bacterial material in seston.
The high variability in FA content and composition among species indicate different adaptations to winter and different life strategies 12 with species that stay active in winter having higher lipid content and FA composition dominated by PUFA in autumn (Fig. 5). Earlier studies have assumed that reserve accumulation would take place in autumn before ice formation 12 and have suggested that zooplankton strongly reduce or even cease their food intake under the ice 17 . Here, the copepods L. minutus and C. scutifer showed a strong ability to accumulate lipids until January, up to 76% of the biomass (see Supplementary Fig. S1), emphasizing the importance of early winter in plankton ecology, a season that has been largely ignored in limnological studies. Yet, this lipid accumulation and over-wintering strategies have already been observed in herbivorous calanoids from marine polar environment 19 . We propose that the autumnal and early winter lipid accumulation allows certain zooplankton to spend the entire winter in an active form either as adults (L. minutus) or as C-IV stages (C. scutifer), which are the two strategies observed for these species 4,46 . As lack of food and cold temperatures are considered stress factors, Daphnia are usually believed to cope with this stress by producing resting eggs (ephippia; Hiruta and Tochinai 47 ). Our observations indicate that the majority of Daphnia in Lake Simoncouche followed this strategy (Fig. 1), however the individuals that overwintered actively did seemingly well even in late January when they were observed to carry parthenogenetic young (Grosbois, unpublished data, see Supplementary Fig. S3). The cyclopoid M. edax also disappeared from the water column to the sediments from November to May as C-V stage copepodites, possibly as a strategy to avoid predators or because they do not have the physiological ability to accumulate lipids.
When consumed, the branched FA common to bacteria are transferred to higher trophic levels without modification and can be measured in the copepod FA composition 48 . The branched and saturated FA have been shown to be associated to growth and membrane fluidity regulation in bacteria 49 but their physiological role in zooplankton is less known. Because of the lack of n-3 and n-6 PUFA, bacteria are commonly considered as a poor food source with negative effects on zooplankton growth and reproduction 50 . Similarly, the FA quality of terrestrial material is considered to be poorer than that of aquatic material 30 . However, terrestrial organic matter is known to be selectively assimilated in bacteria biomass 51 and then probably transferred to higher trophic levels via heterotrophic eukaryotic protists and rotifers, therefore making it available to crustacean zooplankton. During this transfer the nutritional quality of terrestrial organic material and bacteria can be trophically upgraded by heterotrophic flagellates 52 and ciliates 53 . Our results, however, show that despite the potential trophic upgrade of terrestrial and bacterial material, terrestrial and bacterial FAB were not substantially accumulated by zooplankton at any time of the year. The maximal contribution of terrestrial (2.1%) and bacterial FAB (2.9%) in zooplankton was measured for C. scutifer in December and February showing that these biomarkers played a small role in the total FA accumulation in the four studied zooplankton species. Our results therefore do not give support to earlier literature that has suggested that zooplankton switch to terrestrial organic carbon sources in lack of phytoplankton production 54,55 . It is important to note that our study does not exclude the possibility that zooplankton use terrestrial organic matter for respiration and/or to support cell and tissue renewal. However, the results confirm that terrestrial and bacterial molecules are not preferentially stored in zooplankton lipid reserves, even when terrestrial and bacterial FAB are present and available to zooplankton in seston. This observation is in accordance with the results of Mariash, et al. 56 , who showed an absence of SAFA trophic accumulation indicating a lack of terrestrial and bacterial FA accumulation in cladoceran lipid reserves.
Among all the FA accumulated by copepods in winter, PUFA from phytoplankton were the most accumulated demonstrating the predominant role of these FA in the reserve accumulation of organisms that stay active under the ice (Figs 4, 5). These FA are important regulators of membrane fluidity, which is reduced at low temperatures 57 , and are used by the organisms to cope with winter conditions. They have also been suggested to contribute to organism's metabolic maintenance as well as to investment of n-3 FA into reproduction in late winter and early spring 4 . Both L. minutus and C. scutifer accumulated these FA, especially DHA (see Supplementary Table S3), which is an essential FA for copepod reproduction 58 . As C. scutifer needed to mature from C-IV stage to adult in winter and then reproduce (Fig. 1), it is very likely that the accumulated PUFA were used primarily for growth and then reproduction. FA composition differed among taxa and characterized winter abundant species (L. minutus and C. scutifer copepods) with high dominance of PUFA from phytoplankton in their accumulated lipids and winter low abundant or absent species (Daphnia spp. and M. edax) with SAFA dominance (Fig. 5). None of these species accumulated bacterial or terrestrial FA substantially. SAFA dominance in Daphnia and M. edax were characterized by the generalist C16:0 and C18:0 FA rather than terrestrial LC-SAFA or bacterial Br-SAFA (Supplementary Table S3). Ephippia-producing Daphnia that disappear from the water column in winter do not accumulate FA to biomass but rather invest them to resting eggs (ephippia), in particular EPA 59 that was largely available in seston before winter. The concentration of phytoplankton FA in Daphnia started to increase in April indicating that phytoplankton FA were produced in the seston and available for consumers before the ice melted. M. edax predation on Daphnia (see Supplementary Fig S2) was confirmed by their similar FA composition (Fig. 5). Collectively, these results demonstrate that the use and accumulation of FA is closely related to the species life strategy.
To conclude, this study demonstrates that the availability of phytoplankton FA remained high in seston in early winter permitting copepods to accumulate and subsequently metabolise n-3 FA throughout winter, thereby providing a mechanism for under ice growth and reproduction 4 . Although the mechanism is known for some marine organisms especially in polar ecosystems, no information has earlier been available for freshwaters. To our knowledge, our study also reports for the first time seston and zooplankton bacterial and terrestrial FAB for a complete year. These terrestrial and bacterial FAB were not accumulated by zooplankton in winter, suggesting that terrestrial organic material and bacteria are not used as alternative resources during times of low primary production under the ice. The duration of ice cover is predicted to be shorter in the future with climate change 60 , phytoplankton with low PUFA content such as cyanobacteria are predicted to increase in the next decades 61 , and the worldwide FA source is predicted not to meet future human needs 62 . It is therefore essential to better understand how PUFA accumulation by zooplankton is seasonally regulated, how it will be affected by changing ice conditions, and how these changes will drive modifications in FA composition of long-lived consumers such as fish.

Methods
Study site and zooplankton community. We sampled Lake Simoncouche (48°23′N, 71°25′W), a medium size mesotrophic shallow lake (mean depth 2. The experiment was run with the calanoid copepod Leptodiaptomus minutus, a common species in eastern North-America that dominates many boreal lakes 64 . In Lake Simoncouche it makes up to 93% of the total zooplankton biomass (Grosbois unpublished). The copepods were sampled from the pelagic zone of the lake via vertical net tows and transferred to 1 L plastic containers with lake water and brought to the laboratory. Both treatment modalities were composed of six replicates (N tot = 12) of fifty copepods each, which were selected under a stereomicroscope (Discovery V12, Zeiss, Jena, Germany, ×8-x100) and transferred to 250 mL beakers using a pipette. Half of the beakers (6) contained GF/F-filtered lake water and were considered the starved treatment, the other 6 beakers had lake water (<50 µm) providing the copepods with a natural food supply present in the lake (control). The beakers were installed at 4 °C in the dark to simulate lake conditions in winter. Water was entirely renewed weekly with fresh Lake Simoncouche water that was either sieved (50 µm) or GF/F-filtered depending on the treatment, and the copepods were counted, survival noted, and dead individuals removed.
Water and zooplankton sampling. Lake water was collected monthly from 12 January 2011 to 16 May 2012 for total dissolved nitrogen (TDN), total dissolved phosphorous (TDP), dissolved organic carbon (DOC), specific UV-absorbance (SUVA 254 ), chlorophyll a (Chl-a), bacterial biomass (Bact Biom) and from 6 July 2011 to 7 May 2013 for fatty acid (FA) analysis. Samples were taken from the epilimnion at the deepest point of the lake using a 2 L Limnos sampler (Limnos Ltd, Turku, Finland). Water was collected from every meter, pooled to one integrated sample in a bucket and stored in a 4 L Nalgene container. Water for FA analysis was collected from the metalimnion as well (hypolimnion does not form in this shallow lake). In the laboratory, subsamples of water for TDN, TDP, DOC and SUVA were filtered through a cellulose acetate filter (0.2 µm) that had been pre-rinsed with Milli-Q water. TDN, TDP, and DOC were analysed using a Lachat Autoanalyser, a ThermoSpectronic spectrophotometer and a Shimadzu TOC-V, respectively, at the Institut national de la recherche scientifique (INRS), Quebec City, Canada. SUVA 254 , an index of DOC aromaticity associated with allochthonous (terrestrial) carbon sources 45 , was measured as absorbance at 254 nm using a Cary 100 UV-Vis spectrophotometer (Agilent, Santa Clara, U.S.A.) and normalized to DOC concentration. Samples for Chl-a (three replicates on each date) were filtered onto GF/F filters that were then folded, wrapped in aluminium foil and stored at −80 °C until spectrofluorometric analysis as in Mush 65 using a Cary Eclipse fluorescence spectrophotometer (Agilent, Santa Clara, USA).
Zooplankton was sampled weekly to monthly from 19-May-2011 to 23-May-2012 for identification, abundance, and FA. The samples for FA were collected by vertical net tows with a 50 µm mesh size net (diameter 25 cm) from the entire water column (0-6 m). To estimate zooplankton abundance, water samples (6 to 20 L) were collected weekly with a Limnos sampler from several depths and concentrated to one integrated sample using a 50 µm sieve. Formaldehyde (final concentration: 4%) was added to the abundance samples until identification 66,67 with an inverted microscope (Axio Observer A1, Zeiss, Jena, Germany, ×100) and Utermöhl chambers. Copepod identification was carried out for copepodites and adults further divided into males and females. Cladocerans were counted without considering developmental stages. When zooplankton density was too high, half or a quarter of the sample was counted after division in Folsom's sample divider. A minimum of 100 individuals were identified per sample except for 7 low density samples (N tot = 71) where about 60 individuals were counted. For zooplankton FA analyses, 100-200 individuals were carefully collected with forceps under stereomicroscope, placed in Eppendorf tubes (1.5 mL) and stored at −80 °C until freeze-drying. Fatty acids were analysed from the 4 most abundant taxa of the community: Leptodiaptomus minutus, Cyclops scutifer, Mesocyclops edax and Daphnia spp.
Fatty acid analyses. Lipids were extracted from freeze dried seston and zooplankton in chloroform-methanol mixture following Heissenberger, et al. 68 . Lipid concentration in zooplankton was calculated from lipid mass measured by gravimetry and from zooplankton mass obtained using a micro-balance (XP26 DeltaRange, Mettler Toledo, Greifensee, Switzerland). As such extraction also includes molecules such as lipid head groups, sterols and carotenoids we further quantified only the FA in the samples. Toluene was added to lipid extracts and each sample was trans-esterified at 50 °C during 15 hours with 1% methanolic sulfuric acid. The resulting fatty acid methyl esters (FAME) were separated from the rest of the material by adding KHCO 3 − water (2% v/v) and hexane. FAME were then identified and quantified by gas chromatography-mass spectrometry (GC-MS) using an Agilent 7890 A chromatograph (Agilent Technologies, Santa Clara, CA) equipped with an Agilent 5975 C mass spectrometer with triple-axis detector and an Agilent J&W DB-23 column (60 m length, 0.25 mm inner diameter, 0.15 µm film thickness). Samples were injected in splitless mode, temperature ramp was 70 °C for 1.5 min with different steps of 20 °C per minute until 110 °C, 12.5 °C per minute until 160 °C and 2.5 °C per minute until 230 °C with a hold time of 6.5 minutes. The gas flow was 1.0032 ml per minute. The instrument was run in Single Ion Monitoring mode to ensure lowest possible limits of detection and most precise quantification. The resulting retention time and ion composition were used for FAME identification using different FAME and bacterial acid methyl esters mix standards (Sigma-Aldrich), and the compound identity was confirmed with the limited mass spectrums of the standards. The peak area of the most abundant FA specific ion (m/z 74, 79, 81 and 87) versus an internal standard (nonadecanoic acid, Sigma-Aldrich, N5252) was used for FAME quantification using calibration curves based on known standard concentrations.
Statistical analyses. Treatments from the survival experiment were tested with repeated measures ANOVA and refined with contrasts for each date using a non-parametric Wilcoxon test, as the data did not meet the assumptions of homoscedasticity and normality. Kolmogorov-Smirnov and Bartlett's tests were used to test data normality and homoscedasticity respectively. Alpha level was set at 0.05. Principal component analyses (PCA) were applied to normalized FA composition data that was grouped according to FA biomarkers (FAB), polyunsaturated fatty acids (PUFA), n-3 FA, n-6 FA, monounsaturated fatty acids (MUFA) and saturated fatty acids (SAFA). Differences of FA composition among taxa were tested with PERMANOVA in PRIMER v.6.1.11 & PERMANOVA + v.1.0.1 69 . FAB concentrations were compared using student's t-test. All analyses except PERMANOVA were carried out in the R environment v.3.3.1 70 .

Data Availability
Most of the data generated or analysed during this study are included in this published article (and its Supplementary  Information files). Other datasets are available from the corresponding author on reasonable request.