The Arctic is shifting towards a prevalence of warm and more saline Atlantic-like waters. These changes in the marine environment pose significant challenges for the eco-physiology of marine invertebrates. Here we measured the metabolic enzyme activity of citrate synthase and lactate dehydrogenase, as well as the energy content and level of oxidative damage in 71 individuals (~ 10/station) of Thysanoessa inermis collected in six fjords in Svalbard that were characterized by different levels of influence of Atlantic water and, thus, temperature and salinity variability in the water column. T. inermis inhabiting fjords with strong influence of Atlantic water masses had lower lipid and protein content, and higher anaerobic metabolism compared to those from more Arctic fjord types. Moreover, T. inermis collected in fjords with high variability in both temperature and salinity had 0.8-fold lower lipid content than station with more stable temperature and salinity. Our results suggest that T. inermis in fjords influenced by Atlantic waters is possibly under stress leading to increased metabolism, consequently enhancing energy consumption. If the energy consumption is not compensated for, by an uptake, it could result in a decrease of total biomass of T. inermis with possible consequences for the entire Arctic food web.

Sampling

At each station, measurements of temperature, salinity and fluorescence were obtained by a ship-board conductivity, temperature, and depth (CTD) profiler (SBE911plus, SeaBird Electronics) with a SeaPoint fluorometer attached. The latitude and longitude of sampling locations are given in Table 1 as well as the average temperature and salinity of the sampling location from the bottom to 5m depth. We have also calculated the variation (i.e., delta = maximum value minus minimum value recorded) of both salinity and temperature from the bottom to 5 m depth to estimate how much variation in these variables euphausiids will encounter during their diel vertical migrations. These delta values were used in the statistical analyses.

To estimate chlorophyll a concentration at surface (5 m), seawater samples of 100 mL were filtered through 25 mm GF/F filters (Whatman) in triplicate, extracted in 100% methanol for 24h at 4°C, and measured fluorometrically with an AU10 Turner Fluorometer (Turner Design, Inc.) according to the method by Parsons et al. (1984).

Euphausiids were sampled using a macrozooplankton net (1 mm mesh size, 3.14 m2 opening area) designed for catching large and fast-swimming krill and amphipods. The net was deployed oblique from the surface to 20 m over the sea floor while the vessel moved at 2knts. Upon retrieval, the cod-end was immediately emptied into a container filled with ambient sea water, and 10-14 live and actively swimming krill were collected with forceps and placed in petri dishes filled with sea water (see sample size per station; Table 1). The species was identified under a Leica M50 stereo microscope, and an image of each individual was taken when possible. Length of each individual was measured from the tip of the rostrum to the end of the telson at an accuracy of 1 mm. The individuals were dried with tissue paper and stored in 1.5-mL plastic Eppendorf tubes, and frozen in -80°C while live. Samples were stored at -80°C until molecular analyses.

Biomarker methods

The frozen krill were weighed to get the wet weight of each individual and then crushed to powder in a tissue lyser (Qiagen) (twice/1 min at 30 shakes per second). Thereafter the frozen powder was divided into two sub-samples, one for the energy content measurements (lipids, proteins, glycogen and glucose) and one for the enzymatic (CS and LDH) and oxidative stress (LPX and Carb) measurements. For energy content measurements, samples were homogenized in 1:5 mg µl-1 0.1M citrate buffer (pH 5.0) using a Bullet Blender (Next Advance, speed 8, 5 minutes). The samples were aliquoted for protein, lipid, glucose and glycogen measurements. The aliquots for glucose and glycogen were boiled for 2.5 minutes.
For the enzymatic activity and oxidative stress measurements, samples were homogenized in 1:5 mg µl-1 of 100 mM K-phosphate buffer containing 150 mM KCI (pH 7.4) using a Bullet Blender (speed 8, 5 minutes). Aliquots were taken first for lactate dehydrogenase (LDH) activity measurements that were further diluted with 50 mM Tris solution (pH 7.4). Second aliquots were taken for citrate synthase (CS) activity measurements. The remaining samples were further diluted to reach 1:10 mg of tissue µl-1 of buffer and a third aliquot was taken for lipid peroxidation (LPX) analyses. Thereafter the leftovers of the samples were centrifuged at +4°C, 10,000g for 15 min and aliquoted for protein carbonylation (Carb) and protein measurements. All the samples were flash-frozen in liquid nitrogen and stored at -80°C until further analyses.

CS and LDH are enzymes involved in the aerobic and anaerobic metabolism of organisms, respectively. Thanks to the measurement of the activity of CS we could evaluate the level of aerobic metabolism increase or decrease in response to thermal and salinity variation. However, if there is a stressful condition due to exacerbation of those environmental conditions the organism could suppress aerobic metabolism in favour of the anaerobic metabolism, like glycolysis, whose LDH activity is taking part of. The CS enzyme activity measurements were done according to Anttila et al. (2013), having following substrate concentrations: DTNB 0.17 mM, oxaloacetate 1.9 mM and Acetyl CoA 0.14 mM. In the protocol the background signal was determined by measuring the absorbance without the addition of oxaloacetate. The LDH enzyme activity was also measured according to Anttila et al. (2013) having the following substrate concentrations: NADH 0.25 mM and pyruvate-Na 25 mM. In the protocol the background signal was determined by measuring the absorbance without the addition of pyruvate-Na. Both enzyme activities were done using the colour formation and activity was measured for 3 minutes at room temperature at a wavelength of 412 nm for CS and 340 nm for LDH. After the measurements, the protein concentrations were determined for each sample using a PierceTM BCA Protein Assay kit (ThermoFisher), which were used to normalise the enzyme activity The activities were calculated per mg of protein in samples.

Oxidative stress, reported here using lipid peroxidation and protein carbonylation, is a response of the cells to the negative effect of reactive oxygen species (ROS) produced by mitochondria respiration and that are not counteracted by antioxidant enzymes. If the production of ROS exceeds the antioxidant counteractive capacity, the biomolecules of the cell are depleted, resulting in membrane destruction and protein denaturation. Therefore, studying those damage would give as a measure of stress of the cell in response to environmental stress that is not well counteracted. Lipid peroxidation was determined based on the protocol from Vuori & Kanerva (2018), using as reagent 2.5 mM ammonium iron (II) sulfate in 0.25 M sulfuric acid and 0.111 mM xylenol orange in methanol. After two hours of incubation, the absorbance of the sample was measured at a wavelength of 570 nm and the lipid peroxidation was calculated per mg of protein. Protein carbonylation (Carb) was conducted following the instructions of the Protein Carbonyl Content Assay Kit from Sigma-Aldrich (MAK094-1KT). The kit is based on the principle that carbonyl content is determined by the derivatization of protein carbonyl groups with 2,4-dinitrophenylhydrazine (DNPH) leading to the formation of stable dinitrophenyl (DNP) hydrazone adducts. It was measured by spectrophotometer at 375 nm and the amount of carbonylation (in nmol) was expressed per amount of protein (in mg) in the samples (measured with BCA kit). Measurement of the energy content could provide information about the energy reserves that the organism is consuming in response to environmental variation. While glycogen and glucose measurement could provide information relative to the fast response to an environmental stressor, the lipid and protein content would give more insight on the quantity of energy storage depletion in response to a prolonged stressful condition. The total lipid content was measured using the phospho-vanillin method (Frings et al., 1972) with vanillin reagent and phosphoric acid, and the absorbance was measured at 540 nm. Total protein content was measured using a protein dye binding method (Protein Assay kit, Sigma-Aldrich) and the absorbance was measured at 562 nm. Total glycogen and glucose contents were measured using the amyloglucosidase method (Carr & Neff, 1984) based on a glucose standard curve and as a reagent an O-toluidine mix with acetic acid, measured at 650 nm. The energy contents were calculated per g of tissue. For the enzyme activity assay (CS and LDH) a plate control sample was used. For oxidative stress damage a standard curve was used for LPX assay, while a milliQ-water sample + blank + plate control sample were used for Carb assay. For the energy content a milliQ-water sample + blank + standard curve + a plate control sample were used for glycogen and glucose, while a milliQ-water sample + standard curve were used for lipids. For BCA protein assay a standard curve only was used.

Statistical analyses

All the statistical analyses were conducted in RStudio using the R version 4.2.2 (‘R: The R Project for Statistical Computing’). To see if mass, length, energy content, energy metabolism and oxidative damage of krill differed among stations, we ran a one-way ANOVA test. Thereafter, we investigated in more depth whether the type of water mass (Arctic, Mixed or Atlantic) shaped the biological responses. The fjords were divided into different types as indicated in Table 1. The definitions were based on the water temperatures and salinities during the sampling time; see Results section. Thereafter, a one-way ANOVA test was run using the water type as categorical variable (Atlantic, Arctic and Mixed). If there were significant differences among stations and water types, the pairwise contrasts were analysed using the Tukey’s test.

A general linear model was used to analyse how T. inermis was influenced by variation (delta, i.e., difference between highest and lowest observed value within station) in temperature and salinity that they encounter during diel vertical migration. The models were run for each biological response (lipid, protein, glycogen and glucose content, as well as CS and LDH activity, and LPX and Carb amount) using the environmental variables (delta salinity and delta temperature; see Table 1) as fixed effect, against each biological response as dependent variable. The models included the depth of the station, chlorophyll a concentration and the mass of individuals as co-factors. The depth was included into the model due to the high variability in the depth between the stations (123 to 326 m). Furthermore, since the mass of krill was found to be significantly different among stations and water masses, and because mass also could influence the energetics of krill, it was included in the model as covariate. Similarly, chlorophyll a concentration varied among the stations; thus, it was included into the model since food availability could influence the measured biological variables. We performed a model selection followed by a step-by-step approach based on the Akaike Information Criterion (AICc), which was obtained by running the function compare_performance in package performance (Lüdecke et al. 2021). First, we tested which combination of the co-factors in the model (having delta temperature and salinity as fixed effects in interaction) had the lowest AICc values. The best model was kept for the next step where it was evaluated whether the interaction effect or the additive model (i.e., without interaction effect) of delta salinity and temperature resulted in lower AICc value. The final model for each biological variable was determined by the lowest AICc value. Each model residual distribution was visually assessed using the function simulateResiduals in DHARMa R package (F. 2018). The model selection for each biological variable is shown in Supplementary Table S1. The final models are shown in Table 2. In all statistical tests significance was considered at p < 0.05. The model outputs were plotted using the ggplot2 package (Wickham, 2016).