Clams are efficient vectors of potent algal neurotoxins, a suite of saxitoxin (STX) congeners collectively called paralytic shellfish toxins (PSTs), to higher trophic levels. The Alaskan Arctic is a region facing an expanding threat from PSTs due to ocean warming, yet little is known about PSTs in clams from this region. Quantifying total toxicity in bivalves requires analytical techniques, such as high-performance liquid chromatography (HPLC). Enzyme-linked immunosorbent assays (ELISAs) are an efficient but only semi-quantitative method for measuring clam toxicity. PSTs (STX eq.) were measured in split clam samples (n = 16) from the Alaskan Arctic using ELISA and HPLC methods to develop a preliminary linear model for conservatively estimating total toxicity in clams from ELISA toxin values (R²_adj = 0.95, p < 0.001). Profiles of PST congeners and total toxicity using HPLC were also assessed in additional clams (n = 36 additional, n = 52 total). Clams contained mostly potent PST congeners, and over half of the clams had PST concentrations above the seafood regulatory limit. These data will help assess the exposure risks of PSTs in Arctic marine food webs, as harmful algal bloom activity is predicted to increase as the Arctic continues to warm.

Sample Collection

Clams were collected during research cruises of opportunity (n = 5) during 2019, 2020, and 2022 along routinely sampled survey transects (Figure 2A). Van Veen Grabs were deployed at various stations along survey transects to collect clams from the sea floor. Grab contents were sieved with seawater to isolate clams, which were then sorted to the genus or species level, collected in sterile whirl-paks or zip-lock plastic bags, and stored at −20 °C until toxin analysis.

Clams were analyzed for PSTs using ELISA and HPLC methods based on sample availability at research cruise stations throughout the Alaskan Arctic (Figure 2A). Initially, one clam at each station was analyzed for PSTs (STX eq.) using an Abraxis^® ELISA kit. (Gold Standard Diagnostics, Horsham, PA, USA) Based on ELISA results, stations throughout the Alaskan Arctic with clams containing low (BDL) to high toxins (≥80 µg STX eq. 100 g⁻¹) were selected, and an additional clam sample (n = 1) was submitted for HPLC analysis. Additionally, one or more clams were pooled from either the same station or among similar stations along the same transect, homogenized using metal scissors and a spatula, and then the homogenate (n = 1) was split equally to perform ELISA (target sample mass; 1.0 g, minimum; 0.4 g) and HPLC (target mass; 1.0 g, minimum; 0.5 g) analyses (“split” sample, Figure 1A). These “split” samples were limited (n = 16) due to sample availability and achieving soft tissue mass required to perform both ELISA and HPLC analyses. A total of 52 samples were submitted for toxin analysis (n = 36, HPLC-only; n = 16, split samples for ELISA and HPLC analyses).

Paralytic Shellfish Toxin Measurements

Abraxis ELISA

Saxitoxin (STX eq.) was measured in clam samples using an Abraxis^® ELISA kit (product no: 52255B) following Lefebvre et al. [22,23]. Solvent (50% methanol/50% water) was added to homogenized samples using a 4:1 ratio (solvent (mL)/sample mass (g)), which were then homogenized for 1 min at 2100 rpm using a hand-held probe (GLH 850, 10 mm; Omni-International, Kennesaw, GA, USA). Samples were then centrifuged at 3063 × g for 20 min at 4 °C (Jouan CR3i centrifuge, Thermo Electron Corp., Waltham, MA, USA), extract transferred to 4 mL amber vials and stored at 4 °C prior to toxin analysis. Aliquots (200 µL) of extracts were filtered through a spin filter (Millipore Sigma Ultra-Free Centrifugal filters, 0.22 μm; Merck KGaA, Darmstadt, Germany) and further diluted to 1:50 prior to measuring STX concentrations (STX eq.) using the Abraxis^â ELISA kit. All manufacturer instructions were followed to obtain STX eq. concentrations (µg STX eq. 100 g⁻¹) from extracts. The cross-reactivity of the ELISA antibodies with PST congeners according to the Abraxis user manual are listed in Table 2. ELISA values that were below detectable limits (BDL) (n = 2 clams) were included in linear model development by assigning half the lowest quantifiable limit (LQL) (LQL = 0.70 µg STX eq. 100 g⁻¹, BDL= 0.35 µg STX eq. 100 g⁻¹) [44,45].

HPLC Methods

HPLC analysis of samples for paralytic shellfish toxins followed the post-column oxidation (PCOX) AOAC Official Method 2011.02 [46]. Extractions involved using equal volume of extraction acid to sample weight (1:1 ratio). Briefly, extraction acid (0.1N HCl) was added to homogenized sample in a 50 mL Falcon centrifuge tube (Falcon – BD, Franklin Lakes, NJ) and placed in boiling water for 5 min. Afterwards, the cooled sample was spun down via centrifugation, and proteins were removed by adding 30% trichloroacetic acid (TCA) followed by pH adjustment with NaOH. After filtration of the processed sample, it was ready for analysis.

The HPLC instrumentation was a Waters Acquity Arc system equipped with a refrigerated autosampler (4 °C) and a Waters 2475 FLR fluorescence detector (Waters, Milford, MA, USA). All extracts were injected at 10 µL and a flow rate of 0.8 mL/min. Attached to the HPLC was a Pickering Laboratories ONYX PCX post-column oxidation instrument (Pickering Laboratories Inc., Mountain View, CA, USA). Flow rates for both oxidant and acid were 0.4 mL/min, and column temperature was 30 °C. Standards used were Certified Reference Material (CRM) obtained from the National Research Council Canada (NRCC, Ottawa, ON, Canada). Toxin analogs analyzed were the following: GTX1, GTX4, GTX2, GTX3, dcGTX2, dcGTX3, GTX5, NEO, dcSTX, STX, C1, and C2. The saxitoxin standard used was saxitoxin dihydrochloride, and all results are expressed as STX·diHCl equivalents. Detection limit was established as 2 µg STX·diHCl 100 g⁻¹. Quality control of instrument and method performance was evaluated in several ways. Upon instrument startup, upper-level and lower-level calibration mix standards were run to ensure instrument response was sufficient. Then, after analysis of 4–6 samples, a mid-level standard mix was run to check that instrument performance was being maintained and not degraded. Post analysis, again, included analysis of upper-level and lower-level quality control standard mixes. Also, once a zero toxin sample was obtained, it was spiked with a standard mix and carried out through extraction and analysis to ensure no matrix effects were occurring by comparing those results with previous quality control standard mixes. Chromatograms for standard mixes and detectable congeners in a clam sample can be found in the Supplementary Material. Toxicity equivalency factors (TEFs) used in the equation below are listed in Table 2.

Calculations were carried out using the following equation to yield µg STX·diHCl eq. 100 g⁻¹:

μg STX∙diHCl eq= μM toxin ×TEF ×(372.2 g)/(1000 mL)×Fvol/(Ext.vol)×((Wt+Vol)/Wt)×100

where:

µM = concentration of toxin in extract;

Fvol = final volume of deproteinated extract (560 µL);

Ext. vol = volume of crude extract used (500 µL);

Wt = weight of sample used (g);

Vol = volume of acid extractant used (mL);

TEF = toxicity equivalence factor (Table 2).

The µg STX·diHCl eq. 100 g⁻¹ concentrations from HPLC analysis are referred to as µg STX eq. 100 g⁻¹ throughout the manuscript (including figures and tables) for ease in comparison of total toxicity concentrations across studies (e.g., EFSA [4]).

Statistical Analysis

All statistical analyses were conducted using the software programs R version 4.4.2 [47] and R Studio version 2024.09.01+394 [48]. All maps were generated using the software QGIS version 3.34 [49]. All samples (n = 52) analyzed for PSTs via HPLC were included in total toxicity analysis; however, one clam had zero measured toxins and was excluded from PST congener profile analyses (n = 51). An alpha of 0.05 was used as a significance threshold for all relevant analyses.

Correction Model—ELISA and HPLC PST Measurements

An initial linear regression was performed using split clam samples (n = 16) to predict total PST concentrations (µg STX eq. 100 g⁻¹, HPLC PST values) from ELISA STX eq. values. One clam was BDL for STX eq. using ELISA, but had a paired HPLC value of over 80 µg STX eq. 100 g⁻¹. This clam was a clear outlier and was removed from linear regression analysis, resulting in a total of n = 15 clams for model development (Table 1). Linear models were compared with square root transformed and non-transformed data using AIC criteria to determine best model fit. The selected model was verified for underlying assumptions by plotting model residuals with fitted values, covariates, and assessing the distribution of residuals.

Paralytic Shellfish Toxin Profiles and Total Toxicity Analyses

Paralytic shellfish toxin profiles for clams were constructed by using relative abundances (µM, molarity) of each congener measured during HPLC analysis. Profiles were assessed for similarities among regions using distance-based multi-variate analysis. Specifically, toxin profiles of clams (n = 51) were compared among different regions (Bering Strait [n = 16], Chukchi Sea [n = 22], and Beaufort Sea [n = 13]) using non-metric multidimensional scaling (NMDS), analysis of similarities (ANOSIM), and similarity percentage (SIMPER) analyses from the R package vegan . A Bray–Curtis transformation was applied to congener proportions in clams to calculate a distance matrix for NMDS, ANOSIM, and SIMPER analysis, which allowed for the assessment of which proportional differences of specific congeners contributed to overall dissimilarities in PST profiles among regions.

Differences in total toxicity (µg STX eq. 100 g⁻¹, HPLC PST values) in clams collected from different regions were tested by comparing linear models with and without (i.e., null model) region as a main effect and selecting the model that had the lowest AICc value and highest AIC weight. Models were generated using the R package lme4 ]. A square root transformation was applied to toxicity data to achieve normality (Shapiro–Wilks test, p = 0.08) for linear model analysis. The full and selected models were verified as described above for linear regression analysis. If the main effect (region) was selected, a Type II ANOVA (F-tests for linear models) with subsequent Tukey HSD tests compared estimated marginal mean (EMM) toxicity concentrations using the R package emmeans. Estimated marginal means were compared because they are dependent on model results (compared to ordinary means based on empirical data) and, therefore, take into account unbalanced sample sizes across regions .