Contrasting sea ice conditions shape microbial food webs in Hudson Bay (Canadian Arctic)

The Arctic, including Hudson Bay (HB), is undergoing rapid reductions in sea ice extent and duration, with the ice-free season having lengthened by over three weeks since the 1980s. Future scenarios predict continued warming and freshening, further extending the open-water season. Primary production in HB is tightly linked to ice phenology, with spring blooms typically occurring near the marginal ice zone during breakup, followed by persistent subsurface chlorophyll maxima (SCM) in open waters. These shifts alter light and nutrient regimes, influencing phytoplankton succession and cascading microbial interactions, yet the consequences for microbial food webs from surface to depth remain poorly resolved. The study aimed to determine how contrasting ice conditions structure microbial eukaryote and prokaryote communities and their associations across depths, and to test the hypothesis that surface community distributions driven by ice conditions propagate to deeper waters, affecting carbon and energy export to the shallow benthos of HB.

Prior work documents earlier Arctic blooms and increased SCM persistence with changing ice conditions, and shows that phytoplankton succession is governed by light and nutrients, with microzooplankton tracking prey dynamics. Phytoplankton blooms release dissolved organic matter (DOM) that sustains diverse, specialized heterotrophic bacteria and bacterivores; DOM quality and quantity can select for distinct bacterial lineages. In HB, under-ice blooms and subsequent SCM formation have been observed, with strong spatial variability driven by winds and freshwater inputs. Globally, co-occurrence networks have been used to infer microbial interactions, highlighting, for example, Syndiniales as key parasites influencing phytoplankton, and the role of particle-associated microbes in deep export. However, cascading effects of sea-ice retreat on integrated microbial food webs in HB from surface to depth had not been systematically examined.

Field sampling was conducted aboard CCGS Amundsen in June 2018 across 11 stations spanning northwestern to central Hudson Bay along an ice-cover gradient. Sea ice concentration (SIC) was obtained from the Canadian Ice Service Digital Archive. Hydrographic profiles (CTD) measured temperature, salinity, oxygen, chlorophyll fluorescence, CDOM, and transmission. Discrete water samples were collected at surface mixed layer, SCM, 70 m, and 10 m above bottom (3 depths at shallow stations), yielding 42 samples. Nutrients (NO3−, NO2−, PO4 3−, Si(OH)4) were analyzed onboard following GEOTRACES protocols. Flow cytometry quantified total phytoplankton (pico- <2 µm; nano- >2 µm; cyanobacteria) and bacterial cells, using chlorophyll and phycoerythrin fluorescence and SYBR Green staining. For nucleic acids, 6 L of prefiltered water (50 µm mesh) were sequentially filtered (3 µm and 0.22 µm). DNA and RNA were co-extracted (Qiagen AllPrep), RNA reverse-transcribed to cDNA, and absence of DNA in RNA was verified by PCR. Amplicon libraries targeted the V4 region: eukaryotes (18S rDNA and rRNA) with E572F/E1009R; prokaryotes (16S rDNA and rRNA) with 515F–806R. Libraries were purified, indexed, pooled, and sequenced on two Illumina MiSeq runs. Reads are deposited under PRJNA627250 (eukaryotes) and PRJNA721720 (prokaryotes). Bioinformatics used DADA2 within QIIME2 for denoising, merging, and chimera removal; taxonomy was assigned with PR2 v4.12 (eukaryotes) and SILVA 132 (prokaryotes). Large (3–50 µm) and small (0.22–3 µm) fractions were summed per sample; metazoan and chloroplast reads and unclassified phylum-level sequences were removed (phyloseq). Relative abundance transformation corrected sequencing depth; very low-abundance ASVs were filtered, including per-sample ASVs ≤0.003% and globally rare ASVs, yielding 1371 (euk rDNA), 1384 (euk rRNA), 3891 (prok rDNA), 4152 (prok rRNA) ASVs. Phylogenies were built with MAFFT and RAxML (GTR+GAMMA). Community analyses used Hellinger-transformed Bray–Curtis and GUniFrac distances, NMDS, Procrustes and Mantel tests (999 permutations). Distance-based RDA (db-RDA) used standardized environmental variables with forward selection (9999 permutations); phosphate and silicate were excluded due to covariance with nitrate. Z-scores were computed for the 50 most abundant ASVs (rDNA). Co-occurrence networks were constructed (CoNet in Cytoscape) using the 500 most abundant eukaryote and prokaryote rDNA ASVs, separately for euphotic (surface+SCM) and deeper (70 m+bottom) samples; deepest shallow inshore samples were excluded. Positive associations were inferred by Pearson, Spearman, mutual information, and Bray–Curtis; sparsity was reduced by removing ASVs with many zeros. Significance used permutations and bootstraps with Brown’s method and Benjamini–Hochberg correction; edges supported by ≥2 methods with p<0.01 and stable bootstrap scores were retained. Network metrics (degree, density, heterogeneity) were computed, and subnetworks were interpreted by region (central HB, northwestern HB, and their deep counterparts).

- Ice and hydrography: Central HB stations (st21, st16) were ~97% ice-covered; st24 and st15 had 20–50% (mobile pack); northwestern stations (st19, st17, st22, st23, st44, st28) were ice-free. Under-ice surface waters were colder (−1.3 to −1.4 °C) and moderately saline (31.1–31.5), while open waters were warmer (0.1–2.4 °C) with variable salinity (as low as 30.1 at st44). Nitrate and silicate were low above 50 m in open NW HB but higher under ice (e.g., SCM at st6: NO3− 3.75 µmol L⁻1, Si(OH)4 8.77 µmol L⁻1). Deep layers (70 m and bottom) were enriched (means: NO3− 6.21, PO4 1.08, Si(OH)4 14.33 µmol L⁻1). Chlorophyll a was mostly <1 µg L⁻1 in central HB (weak SCM: 1.71 µg L⁻1 at st24, 46 m); distinct SCMs formed at open-water stations with a maximum 4.81 µg L⁻1 at st28. - Cell abundances: Nano/micro-phytoplankton ranged from 8 mL⁻1 (bottom st16) to 4.34×10³ mL⁻1 (SCM st44). Pico-phytoplankton correlated with small phototroph ASVs; maximum 1.54×10⁵ cells mL⁻1 at SCM st18. Bacterial cells were higher in the euphotic zone (maximum 1.44×10⁶ mL⁻1 at surface st28) than depth (down to 5.12×10⁵ mL⁻1 at SCM st23). - Community structure: Eukaryote and prokaryote communities (rDNA and rRNA) showed congruent biogeographic patterns and strong similarity (Procrustes and Mantel tests). Bray–Curtis clustering delineated four clusters separating NW vs central HB and surface/SCM vs deeper samples; deep NW samples clustered together and apart from deep central HB. GUniFrac and Bray–Curtis gave highly similar clustering for prokaryotes. - Environmental drivers: db-RDA separated euphotic from deep samples along RDA1 (explained variance: eukaryotes 27.08%, prokaryotes 18.22%), reflecting higher deep nutrients. At equal depths, central HB samples aligned with higher nutrients; pico-phytoplankton abundance in ice-covered waters structured variation along the secondary axis. - Surface/SCM composition: Ice-covered central HB and the Narrows had higher relative abundances of small phototrophs (Phaeocystis pouchetii, Micromonas polaris, Bathycoccus prasinos). Northern HB (st17, st18) showed more diatom reads (Thalassiosira). Fragilariopsis sp. and Actinocyclus curvulatus were abundant in open NW HB but scarce under central ice. Choanoflagellates (Diaphanoeca undulata, Calliacantha natans, C. longicaudata) and Gyrodinium were enriched in NW HB (18.9% of ASVs vs 2.6% in central), reaching >18% at st22 and st28. Prokaryotes in open NW HB were enriched in Balneatrix, SAR11 clade Ia, Colwelliaceae, and Polaribacter; central HB favored Pseudohongiella, SAR86, and Flavobacteriaceae NS9; SAR92 ASVs partitioned between open and ice-edge conditions. - Deep composition: In NW deep waters, pelagic diatoms (Thalassiosira, Chaetoceros) were relatively more abundant (together up to 15.7% of ASVs); central/northern deep stations had increased Radiolaria and Syndiniales. Deep NW bacteria favored Polaribacter (7.2%), Nitrincolaceae (3.5%), and Colwellia (0.9%). Potential ammonia oxidizers (Candidatus Nitrosopumilus; Thermoplasmata groups II/III) were highly connected in deep central HB and together exceeded 15% of the community (rDNA and rRNA). - Network analysis: Surface–SCM networks showed higher heterogeneity (hub nodes), while 70 m–bottom networks had higher average degree and density (more connectivity). Regional subnetworks reflected clustering: NW HB hubs included Colwellia, Bacillariophyta, and choanoflagellates; central HB hubs included Syndiniales and Mamiellophyceae (Micromonas, Bathycoccus). Deep NW hubs included Polaribacter, Colwellia, and Bacillariophyta; deep central HB hubs included Syndiniales, Thermoplasmata II/III, and Nitrosopumilales. The diatoms Thalassiosira (ASV 5450) and Melosira arctica (ASV 2805) occurred at all depths; M. arctica was exclusive to ice-covered central HB, while Thalassiosira characterized open NW HB. - Synthesis: Co-occurrence networks support a <3 µm pico-phytoplankton-based food web under ice and a >3 µm nano–microphytoplankton-based food web in open waters. Patterns suggest enhanced SCM diatom-driven export in longer open-water seasons, linking surface bloom state to deep community structure.

Findings support the hypothesis that sea-ice conditions modulate surface phytoplankton communities and cascade to structure heterotrophic eukaryote and prokaryote assemblages at depth. Under ice and at the ice edge, post–under-ice bloom conditions and low surface nitrate favored pico-phytoplankton (Micromonas, Bathycoccus, Phaeocystis). In open NW waters, nutrient-replete SCMs supported larger nano–microphytoplankton, including pelagic diatoms (Thalassiosira, Chaetoceros). Co-occurrence networks revealed biotic interactions shaping communities: at the ice edge, Syndiniales showed high connectivity with pico- and some diatom taxa, consistent with parasitic top-down effects and potential roles in bloom collapse. In open waters, choanoflagellates co-occurred with Gammaproteobacteria, consistent with bacterivory linking bacterial recycling to higher trophic transfer. Bacterial communities reflected organic matter quality and light regime: ice-edge assemblages (Pseudohongiella, SAR86, SAR92, Flavobacteriaceae) associate with phytoplankton-derived DOM and phototrophy via rhodopsins; open-water communities (Colwelliaceae, Polaribacter, Balneatrix, SAR11) align with particle-associated degradation of larger phytoplankton products at the SCM. Despite compositional shifts, bacterial cell abundances remained similar in the euphotic zone, consistent with strong grazing pressure. At depth, more complex, densely connected networks and distinct regional niches indicate both local selection and imprinting by sinking particles from overlying waters. Deep NW networks linked diatoms with Colwellia and Polaribacter, suggesting export of pelagic diatom biomass. Deep central HB networks paired Syndiniales with Radiolaria (Chaunacanthida), potentially reflecting parasitism and contributions to export. Detection of sympagic diatoms (Melosira arctica, Nitzschia) at depth in central HB implies under-ice production and rapid export of ice algae. The prominence of archaeal ammonia oxidizers (Nitrosopumilales) and Thermoplasmata suggests nitrification contributes to the deep nitrate pool, consistent with nutrient accumulation from both riverine input and remineralization of exported OM. Collectively, results indicate that longer open-water periods will favor large-cell SCM communities and enhanced diatom-mediated export, altering carbon flow and microbial food web organization from surface to deep waters.

Sea-ice state indirectly structures microbial food webs in Hudson Bay from the surface to depth. Ice-edge conditions promote pico-phytoplankton-based networks with strong Syndiniales interactions, while open waters favor nano–microphytoplankton (notably diatoms) and associated heterotrophs, particularly choanoflagellates and particle-associated bacteria. Deep communities bear the imprint of overlying production via sinking particles, including export of pelagic diatoms in open waters and sympagic diatoms under ice, alongside active archaeal nitrifiers contributing to deep nitrate. With projections of earlier breakup and prolonged open-water seasons, HB is likely to experience longer periods of regenerated production and increased diatom-driven export from SCMs, influencing deep microbial communities and potentially shifting HB toward a CO2 source depending on SCM depth and efficiency. Future research should include time-series across the full seasonal cycle, finer-resolution size-fractionation to capture large aggregates and marine snow communities, targeted assays for Archaea, and integrative measurements (e.g., sediment traps, rate processes) to quantify export pathways and trophic impacts.

- Temporal scope: A late-spring snapshot; no full seasonal or interannual coverage to resolve succession dynamics. - Size biases: Prefiltration through 50 µm mesh may have excluded large chain-forming diatoms and aggregates, potentially underestimating their contribution; flow cytometry settings (22 µm core optimized for nanoplankton) limited detection/quantification of large diatoms. - Primer bias: Universal primers not optimized for Archaea likely underrepresented archaeal diversity and abundance. - Network inference: Co-occurrence reflects statistical associations, not direct interactions or causality; sparse data and compositional effects may influence inferred edges despite stringent filtering and multiple metrics. - Spatial coverage: Limited number of stations and removal of deepest samples from two shallow inshore stations for network analyses may affect representativeness. - Process rates: No direct measurements of production, grazing, parasitism, nitrification, or sinking/export fluxes; interpretations rely on community patterns and literature context.