2.4 Stable isotope analysis
Isotope data are expressed in delta (δ ) notation:\(\delta^{i}E_{\text{sample}}=\frac{\left(\frac{i_{E}}{j_{E}}\right)_{\text{sample}}-\left(\frac{i_{E}}{j_{E}}\right)_{\text{ref}}}{\left(\frac{i_{E}}{j_{E}}\right)_{\text{Ref}}}\)For the element E, the ratio of heavy (i) to light (j) isotope are measured in both sample and references (Coplen & Shrestha 2016). To express the isotopic data as per mil (‰), they are multiplied by 1000. The isotope ratios are expressed relative to international standards; Vienna Pee Dee Belemnite (VPDB) for carbon and atmospheric air for nitrogen.
All tissue samples for compound specific isotope analysis were freeze dried and then hydrolyzed in 6 N HCl at 110°C for 20 h before derivatizing the AAs to N -acetyl methyl esters (NACME, (Corr, Berstan & Evershed 2007) following the protocols by (Larsen et al. 2013; Larsen et al. 2016a). The samples were analysed at the Leibniz Laboratory at Kiel University. The average standard deviation for the samples, across all AAs was 0.3‰ for δ13C. Elemental content and bulk isotope values were determined at the Stable Isotope Facility of the Experimental Ecology Group, GEOMAR, Kiel. The overall standard deviation for the measurement range 5.0-15.0 µg N and 10.0-140 µg C was ±0.2 ‰ and ±0.15‰, respectively. We did not perform lipid extraction prior to stable isotope analyses of tissue samples because this can affect δ 15N values (Svenssonet al. 2016). Instead, we applied lipid correction toδ 13C values with C/N values larger than 3.3 (indicating elevated lipid content) following Post et al. (2007). For detailed CSIA and bulk SIA methods, see the Supplementary Information. AA See Supplementary Table S2 for δ13C values and Supplementary Table S3 for bulk δ13C and δ15N values.
2.5 Statistical analyses
All statistical analyses were performed in R version 3.5.1 (R-Development-Core-Team 2018). To assess whether EAA in consumers originate from bacteria, fungi or marine phytoplankton, we applied linear discriminant function analysis (LDA) (R: MASS ) using δ13CEAA training data from Larsenet al. (2013). To assess the power of differentiating among functional groups and among species with δ13CEAA data, we applied Principal Component Analysis (PCA, R: vegan ) using mean-centred δ13CEAA values to factor out baseline isotope variability. The mean-centred values were calculated by subtracting each individual δ13CEAAvalue from the mean δ13C values of all EAAs for each sample. Prior to the PCA, we applied LDA to find the most effective set of independent variables for predicting category membership. With this set of independent variables, we performed covariance matrix PCA that preserves variance as the range and scale of variables are in the same units of measure. Using the first and second principal component scores, we then applied Multivariate Analysis of Variance (MANOVA, R:manova ) in conjunction with Pillai’s trace to test the null hypothesis that groups have a common centroid in a dependent variable vector space. A rejection of this hypothesis entails that the groups have significantly different δ13CEAApatterns or fingerprints. The MANOVA tests were performed on groups with ≥5 specimens. To remove the effect of a covariate factor, we applied Multivariate Analysis of Covariance (MANOVA, R: jmv ). All data for multivariate comparisons were first assessed for homogeneity of variance by using Fligner-Killeen tests and visually checked for departures from normality on Q-Q plots. To test for species-specific δ13C differences for each EAA for consumers from Kiel Bight and the Arkona Basin, respectively, we used a One Way ANOVA with Tukey’s HSD test (R: aov; TukeyHSD ). Using scatterplots, we also investigated the power of differentiating niches with isotope values of the glycolytic AAs and bulk carbon and nitrogen, respectively. We used linear modelling to test the strength of linear associations (R: lm ).
3. RESULTS
3.1 Biosynthetic origins of the essential amino acids
According to our LDA using training data of broad phylogenetic groups, phytoplankton were the primary EAA source for all consumers in Kiel Bay and Arkona Basin; contributions from bacteria and fungi were small or possibly absent (Fig. 2). The discrete clustering of most functional groups indicates that they were supported by different phytoplankton sources, here listed in terms of association along the along the first linear discriminant: suspension feeders, benthic flatfish, scavengers, pelagic piscivores, planktivores and benthic predators.