Understanding the response of predators to ecological change at multiple temporal scales can elucidate critical predator-prey dynamics that would otherwise go unrecognized. We performed compound-specific nitrogen stable isotope analysis (CSIA) of amino acids on 153 harbor seal museum skull specimens to determine how trophic position of this marine predator has responded to ecosystem change over the past century. The relationships between harbor seal trophic position, ocean condition, and prey abundance, were analyzed using hierarchical modelling of a multi-amino acid framework and applying 1-, 2-, and 3- year temporal lags. We identified delayed responses of harbor seal trophic position to both physical ocean conditions (upwelling, sea surface temperature, freshwater discharge) and prey availability (Pacific hake, Pacific herring and Chinook salmon). However, the magnitude and direction of the trophic position response to ecological changes depended on the temporal delay. For example, harbor seal trophic position was negatively associated with summer upwelling, but had a 1- year delayed response to summer sea surface temperature, indicating some predator responses to ecosystem change are not immediately observable. These results highlight the importance of considering dynamic responses of predators to their environment as multiple ecological factors are often changing simultaneously and can take years to propagate up the food web.

Collagen samples were analyzed for using 5 mg of purified collagen from approximately 50 mg of bone. Preliminary analyses were conducted to determine the highest rate of collagen return from bone sampled from different parts of the skull to minimize destruction (mandible, internal occipital shelf, temporal process). All produced similar stable isotope measurements. Samples were primarily taken from the internal occipital shelf at the back of the skull to maintain external integrity. Bone was decalcified using 0.2 M HCl for 24-72 hours depending on bone thickness, followed by centrifugation and nanopure water rinse. Removal of humic acids was conducted using 0.125 M NaOH for 20 hours. Samples were washed to a neutral pH, then solubilized in 0.01N HCl. Once solubilized, samples were dried under a stream of N2 and freeze dried. Freeze dried collagen was analyzed for bulk isotopic composition of nitrogen by the UW IsoLab (isolab.ess.washington.edu) using a coupled elemental analyzer-isotope ratio mass spectrometer following the standard protocols of the laboratory. C:N ratios were available for most (n = 107) samples used as a measure of the quality for nitrogen analyses of bone collagen for stable isotope analysis (van Klinken 1999). No samples within this subset were outside the acceptable C:N range of 2.7-3.6 (by mass), indicating there was no substantial loss of glycine or addition of nitrogen due to microbial processing from mortality, decay, curation, and analysis. We therefore assumed samples without C:N data (n = 20) also were within the acceptable range as they were subject to the same storage and processing procedures at the same museum institution (University of Alaska Fairbanks, Museum of the North).

δ15N of eleven amino acids (alanine, glycine, proline, aspartic acid, leucine, isoleucine, valine, threonine, serine, glutamic acid, phenylalanine, tyrosine) were measured in the UW Facility for Compound-Specific Isotope Analysis of Environmental Samples. The composition of amino acids in bone collagen is highly variable across amino acids (Gauza-Włodarczyk et al. 2017). Tyrosine, isoleucine, valine, and threonine are not abundant in bone collagen, whereas glycine is 20 times more abundant in bone collagen than most amino acids on a gram amino acid per 100 g of protein basis. Samples were prepared following the procedures developed by Chikaraishi et al. (2007) and protocols by Rachel Jeffrey’s lab at University of Liverpool UK which are modifications of that published by Metges et al. (1996) and Popp et al. (2007). Briefly, proteins were hydrolyzed in 6N HCl and purified using a cation exchange column. Norleucine was added as an internal standard. Amino acids were esterified using isopropanol acetyl chloride, and derivatized via acylation with 4:1 toluene: pivaloyl chloride. Samples were brought up in ethyl acetate and analyzed using a coupled gas chromatography-combustion-isotope ratio mass spectrometer system (GC-C-irMA; Thermo Scientific Trace GC + GC IsoLink coupled to a Delta V irMS) in continuous flow mode monitoring masses (m/z) 28 and 29. A 30 m x 0.32 mm x 0.50 mm Agilent Technologies DB-35 capillary column with 35% Phenyl and 65% polysiloxane stationary phase and moderate polarity was used (Chikaraishi et al. 2010) with an inlet temperature of 260 C, column flow of 2 ml/min and oven ramp of 9 ˚C min-1.  Tyrosine and isoleucine for most samples were not discernable and thus were omitted from this analysis. Leucine and isoleucine also co-eluted for many samples and thus leucine stable isotope measurements were deemed unreliable and also omitted from this dataset.  For each run, a 12 amino acid external standard with known isotopic composition was injected four times to condition the column followed by sample injections. Samples were injected in triplicate, with the 12 amino acid standard mixture injected every two samples (or six injections). A two-hour column oxidation was performed after 6 samples (25 injections) followed by a 30-minute backflush. δ15N was measured as:

S1.   δ15N ‰ vs.  air = [(15N/14N)Sample / (15N/14N)air - 1] *1000

            For each machine run, a linear model was fit for each individual amino acid using the following equation:

S2.   Stdaa=maat+baa

Where m represents the slope of the precision drift, t represents the injection number since last column oxidation, and Std represents the δ15N of an individual amino acid aa for a standard observation. The data was then corrected using the following equations:

S3.   Daa, t=Stdaa,t- Trueaa

Where Daa,t is the difference between an observed standard δ15N (Stdaa,t) for a given amino acid (aa) at a given injection number (t) and the true δ15N for that standard. Then:

S4.   Samplecorrected,  aa, t=Sampleobs,aa,t- Daa,t

Wherethe drift value, Daa,t, is subtracted from the sample value for a given amino acid and a given injection to correct the observed sample values for precision drift since last column oxidation. Mean sample corrected values for the triplicate injections were used for all analyses and trophic position calculations and are reported in this dataset. Norleucine had lower precision in standards compared to phenyalanine, therefore no correction using the internal standard was applied. Mean precision for a given amino acid standard was calculated using the standard deviation of the external standard injections for a given run after drift correction and taking a mean of each run's standard deviation. Conditioning injections were omitted from this calculation.

For some sampled specimens, metadata was not available the museum institution. For these specimens, length, sex or latitude/longitude is reported as 'NA'. For some specimens there was not enough bone collagen to run both compound specific stable isotope analysis (CSIA) of amino acids and bulk stable isotope analysis. CSIA was prioritized and there is no bulk stable isotope data for carbon or nitrogen reported which is represented in the data as 'NA'. For some specimens, CSIA of an individual amino acid was deterimined to be poor quality due to peak area or co-elution with other peaks. Data that was not considered sufficient quality was omitted from both this dataset and analyses.