We assessed physiological health of glaucous-winged gulls wintering in urban and natural habitats in the Salish Sea, British Columbia. Although ecological context differed among sites, based on stable isotope analysis, physiological health did not vary with location or habitat. Our study provides physiological reference values for future biomonitoring.

Methods

Study area

The Salish Sea (49° 20' 10.4", -123° 50' 21.6") comprises two areas of sheltered marine waters harbouring large human population centres: the Strait of Georgia in Southern British Columbia and Puget Sound in Washington State, which are connected to the open Pacific Ocean by the exposed Strait of Juan de Fuca. In January and February of 2020 and 2021, we sampled adult GWGUs throughout the Canadian portion of the Salish Sea (Figure 1). We attempted to sample evenly between the following regions: ‘Lower Mainland’, ‘Greater Victoria’, ‘Southern Vancouver Island’, and the ‘Northern Salish Sea.’ Regions were categorized primarily by geographic proximity, but generally had similar beach substrate types and levels of anthropogenic influence throughout a given region (Supplemental Figures 1 and 2).

Within each region, we sampled as evenly as possible among various habitat types including landfills, ‘urban,’ and ‘natural’ areas (Table 1). However, levels of human population density (Supplemental Figure 1), urbanization, and other types of land use varied among regions (Supplemental Figure 2), therefore urban habitat types are over-represented in some regions (Table 1). For example, urban areas were comprised of beaches near high human population densities, as well as city parks, whereas ‘natural’ habitats included beaches in areas with considerably lower human population densities, and less industrial activity. Gulls using landfills were also sampled in the Lower Mainland, Greater Victoria, Southern Vancouver Island, and the Northern Salish Sea. Satellite imagery from the North American Land Change Monitoring System database was used to guide categorization of capture locations into ‘urban’ versus ‘natural’ habitat types (250 x 250 m resolution, North American Land Change Monitoring System, 2021) while human population density (people/km²) was obtained using census data (Statistics Canada, 2017). All three habitat types and all four regions of the Salish Sea were sampled in both years of study. Additionally, in 2021, ten gulls were sampled on the west coast of Vancouver Island near the small towns of Ucluelet and Tofino as an ‘outlier’ group for comparison with Salish Sea birds.

Data Collection

Adult Gulls were live-captured primarily using baited noose-mats (Liu et al., 2017) and occasionally with pneumatic CO₂ net guns, when bait was not an effective attractant (Edwards and Gilchrist, 2011). As soon as possible after capture, we collected no more than 6 mL of blood (<1% of body weight) from the brachial vein of one or both wings using a 27.5-gauge heparinized needle and syringe. We also collected approximately 20 µL of whole blood from tarsal veins using a non-heparinized lancet and capillary tube. Of this, half was stored in 95% ethanol for molecular sexing, and the rest was used to measure glucose levels (mmol/L) in the field using a handheld glucose meter (Accu-check Aviva; Roche, Basel, Switzerland). In the field, samples for haemoglobin analysis were prepared by adding 5 mL of fresh, whole blood to 1.25 mL of Drabkin’s reagent (D5941 Sigma-Aldrich Canada, Oakville, Ontario, Canada). Hematocrit was measured on two capillary tubes that were immediately filled with fresh, whole blood and stored at 4°C for up to 6 hours before centrifuging for three minutes at 13,000 g (Microspin 24; Vulcon Technologies, Grandview, Missouri, USA). All remaining blood was stored in a heparinized vacutainer at 4°C for up to 6 hours before being centrifuged for 10 minutes at 5000 rpm. Plasma (without lipid extraction) and red blood cells were separated and stored at -20°C for up to 1 month and then at -80°C until assayed. Handling time was calculated for each individual and defined as the number of minutes from capture time until blood collection was finished. Before release, all gulls were photographed, aged, banded, and morphometric measurements were collected including mass (± 20 g) and tarsus length (± 1 mm).

All research was conducted under Environment and Climate Change Canada (ECCC) Banding Permit #10667F, and ECCC Migratory Bird Sanctuary Permit #MM-BC-2020-0002. Animal use protocols were approved by ECCC’s Western and Northern Animal Care Committee (21MH03), as well as the Simon Fraser University Animal Care Committee (protocol no. 1318B-20). All personnel completed mandatory Animal Care training.

Plasma analysis

Plasma triglyceride levels (mmol L^-1) were analyzed with a colorimetric assay according to the manufacturer’s instructions (Sigma-Aldrich Co.; also see Fowler and Williams, 2017). Hematocrit was measured using digital calipers (± 0.1 mm) and determined as a percentage of packed red cell volume to total column height (plasma plus packed red cell volume). Haemoglobin was measured using the cyanomethemoglobin method (Drabkin and Austin, 1932) modified for use with a microplate spectrophotometer and absorbance read at 540 nm (Wagner et al., 2008). Total antioxidant titres (OXY; µmol HClO mL^-1) and reactive oxygen metabolites (dROMs; mg H₂O₂ dL^-1) in the plasma were measured using OXY and dROMs kits, respectively, from Diacron International (Grosseto, Italy). OXY absorbances were read at 490 nm, and dROMs at 546 nm, using protocols modified after Guindre‐Parker et al. (2013) and Casagrande et al. (2012), respectively. All assays were run using 96-well plates and a microplate spectrophotometer (BioTek Powerwave 340; BioTek Instruments, Winooski, Vermont). For quality control, samples with an intra-assay coefficient of variation (CV) > 10% when assayed in triplicate (haemoglobin and OXY), or CV > 12% if run in duplicate (triglycerides and dROMs), were re-assayed if no obvious outlier could be removed. Inter-assay variation was 4.11% (triglycerides), 2.20% (haemoglobin), 9.41% (OXY), 10.57% (dROMs), and intra-assay variation was 6.74% (triglycerides), 1.60% (haemoglobin), 4.27% (OXY), and 6.69% (dROMs).

Sexing

To determine gull sex, DNA was extracted from blood stored in 95% ethanol using a modified Chelex protocol (Burg and Croxall, 2001; Walsh et al., 2013). Individuals were sexed using the Z43BF/Z43BR Primer Pair (Dawson et al., 2016); the forward primer was modified with M13 to allow incorporation of fluorescent marker to run on Licor gel. All PCR reactions were conducted in 10 µL reactions with 1 µL of genomic DNA. PCR cocktails contained 2.0 µL ClearFlexi Buffer 5x (Promega), 2.5 mM MgCl₂, 200 µM dNTP, 1 µM each primer, 0.05 µM M13 primer, 0.5 units GoTaq (Promega). We used the following Thermocycler Conditions: 1 cycle of 30 seconds at 94°C; 35 cycles of 30 seconds at 94°C, and 45 sec at 55°C, and 45 seconds at 72°C, with a final extension for 5 minutes at 72°C, and 5 sec at 4°C. All PCR products were run on a 6% acrylamide gel. All gels included known positives (one male and one female) to maintain consistency across gels.

Stable Isotopes

For carbon and nitrogen stable isotope analyses, we weighed 1 mg of freeze-dried whole blood into pre-combusted tin capsules. Encapsulated blood was combusted at 1030°C in a Carlo Erba NA1500 or Eurovector 3000 elemental analyser. The resulting N₂ and CO₂ were separated chromatographcally and introduced to an Elementar Isoprime (Elementar; Langenselbold, Germany) or a Nu Instruments Horizon (Nu Instruments Ltd.; Wrexham, United Kingdom) isotope ratio mass spectrometer. We used two reference materials to normalize the results to VPDB and AIR: BWBIII keratin (δ¹³C = –20.18 ‰, δ¹⁵N = +14.31 ‰, respectively) and PRCgel (δ¹³C = –13.64 ‰, δ¹⁵N= +5.07 ‰, respectively). Within run (n = 5) precisions as determined from both reference and sample duplicate analyses were ± 0.1 ‰ for both δ¹³C and δ¹⁵N.

Data analysis

For all physiological biomarkers measured, sample distributions were examined for normality and whether values were biologically plausible based on reference values for other gulls (e.g. Laranjeiro et al., 2020; Minias, 2015; Newman et al., 1997). Based on these reference values, biologically implausible outliers were removed for haemoglobin (n = 6; > 24 g/dL) and OXY (n = 1; < 110 µmol HClO/mL). Log transformations were used for triglycerides, glucose, and dROMs (Fowler and Williams, 2017). Statistical analyses were performed, with significance determined using an alpha level of 0.05, in R version 4.3.3 (R Core Team, 2024). Pearson’s correlation coefficients were used to examine pairwise relationships between the six physiological biomarkers measured and to test the potential effect of handling time on each trait. To address potential bias due to sex differences in our data, we first determined whether the ratio of females to males sampled was significantly different, a) among capture years and b) between years, using the Chi-squared test. We also tested whether mass significantly varied with gull sex using ANOVA.

Using linear mixed-effects models, we determined whether variation in any of the six biomarkers measured was explained by a) sex, b) mass, c) sex + mass, or d) sex*mass.  Free fatty acids in plasma can impact dROMs assay results (Pérez-Rodríguez et al., 2015), so we additionally tested whether triglycerides, or any combination of triglycerides, sex, and mass explained significant variation in dROMs measurements. All models were run with year as a random effect, except for haemoglobin which was only measured in 2021. We used Akaike Information Criterion for small sample sizes (AICc) to determine the model of best fit using the MuMIn package (version 1.48.4; Bartoń, 2023). For a given trait, if the model with the lowest AICc score included mass as a significant effect, it was treated as a covariate, while a significant effect of sex was instead included as an interaction term with region or habitat in future models. This was to account for a likely skewed sex-ratio among regions and habitats, which could not be formally assessed due to uneven sample sizes. If neither mass nor sex were significant, no covariate or interaction term was included.

Next, we assessed whether each trait varied significantly by a) region or b) habitat type at capture using ANOVA (lme4 version 1.1-35.5; Bates et al., 2015) and post-hoc Tukey tests for pairwise comparisons. Least-squares means were calculated using the emmeans package (version 1.10.4; Lenth, 2023). To control for potential environmental variation between sampling years, year was included as a random effect for all models, excluding haemoglobin which was only measured in 2021. Covariates (mass, and/or triglycerides) and interaction terms (sex*region or sex*habitat, and sex*mass) determined previously by model selection (described above) were included as needed.

Principal components analysis (PCA) was used to examine the pattern of correlations and distributions amongst GWGU physiological biomarkers, and to provide indices of overall ‘health’ for gulls wintering in the Salish Sea. Using the same approach as described above for individual biomarkers, we determined if any significant covariates (i.e. mass and/or sex) should be controlled for with PC variables. Additionally, PCA scores for individuals were used to compare physiological health among regions and habitat types of capture locations (R Core Team, 2024). Specifically, we tested whether individual scores from the first principal component (PC1) or the second principal component (PC2) varied significantly among region or habitat type using ANOVA and post-hoc Tukey tests for pairwise comparisons.

For stable isotope data, we also determined whether to control for covariation due to sex and/or mass, and then tested if, a) δ¹³C and δ¹⁵N varied among sampled regions or habitats using ANOVAs with post-hoc Tukey tests, and then, b) if individual variation in δ¹³C and δ¹⁵N were correlated with each of the six physiological biomarkers we measured, as well as the two PCA-derived variables of overall ‘health’.