Atlantic water influx and sea-ice cover drive taxonomic and functional shifts in Arctic marine bacterial communities

Abstract

The Arctic Ocean is experiencing unprecedented changes because of climate warming, necessitating detailed analyses on the ecology and dynamics of biological communities to understand current and future ecosystem shifts. Here, we generated a four-year, high-resolution amplicon dataset along with one annual cycle of PacBio HiFi read metagenomes from the East Greenland Current (EGC), and combined this with datasets spanning different spatiotemporal scales (Tara Arctic and MOSAiC) to assess the impact of Atlantic water influx and sea-ice cover on bacterial communities in the Arctic Ocean. Densely ice-covered polar waters harboured a temporally stable, resident microbiome. Atlantic water influx and reduced sea-ice cover resulted in the dominance of seasonally fluctuating populations, resembling a process of "replacement" through advection, mixing and environmental sorting. We identified bacterial signature populations of distinct environmental regimes, including polar night and high-ice cover, and assessed their ecological roles. Dynamics of signature populations were consistent across the wider Arctic; e.g. those associated with dense ice cover and winter in the EGC were abundant in the central Arctic Ocean in winter. Population- and community-level analyses revealed metabolic distinctions between bacteria affiliated with Arctic and Atlantic conditions; the former with increased potential to use bacterial- and terrestrial-derived substrates or inorganic compounds. Our evidence on bacterial dynamics over spatiotemporal scales provides novel insights into Arctic ecology and indicates a progressing Biological Atlantification of the warming Arctic Ocean, with consequences for food webs and biogeochemical cycles.

Fig. 1. Location of seafloor moorings and environmental conditions in the MIZ (2016–2018) and core-EGC (2018–2020).

a Example representation of monthly average (January 2020) current velocities at the approximate depth of sampling (78 m). White and dark red arrows indicate strongest and weakest velocities, respectively. b Example representation (December 2019) of sea-ice cover. Increasing opacity of white colour reflects increasing sea-ice cover (pure white = 100%). Current and sea-ice data were obtained from copernicus.eu under ‘ARCTIC_ANALYSIS_FORECAST_PHY_002_001_a’. c Variation in AW proportion, ice cover and water temperature at the two moorings. The bathymetric map was made using data from GEBCO.

Fig. 2. Community structure across water mass, sea-ice and daylight conditions.

Distance-based redundancy analysis based on Bray-Curtis dissimilarities of community composition along with AW proportion (blue), past ice cover (green) and daylight (orange) as constraining factors. The factors were selected using a stepwise significance test and combined into a single model (R² = 0.1, p = 0.01) that constrains 14% of the total variation. For ease of interpretation, the environmental conditions are visualised individually on the same ordination.

Fig. 3. Distribution dynamics and co-occurrence of ASVs.

a Occurrence of ASVs across samples in relation to their maximum relative abundances, along with categorisation into resident, intermittent and transient. b Average number of connections within the co-occurrence networks for resident, intermittent and transient ASVs. c Relative abundance dynamics of resident, intermittent and transient ASVs over time.

Fig. 4. Sparse partial least square regression (sPLS) linking community structure and environmental parameters.

a Heatmap showing eight major sPLS clusters, encompassing 430 ASVs with significant correlations to environmental conditions. b Representation of the most prominent genera per cluster. ASVs with <1% relative abundance were excluded, whilst the remaining were grouped by genus and the maximum abundance of each genus shown. Due to high collinearity with AW proportion, temperature and salinity were excluded. Thresholds: coefficients > 0.4, p < 0.05.

Fig. 5. Comparison and dynamics of MAGs across the Fram Strait and Arctic Ocean.

We compared metagenome-assembled genomes (MAGs) generated in this study (FRAM_EGC), and from samples previously collected in the Fram Strait (FRAM18) [23], in the Arctic Ocean during summer (TARA) [34], and in the Arctic Ocean during winter (MOSAIC) [35]. a Number of shared and unique species across the four MAG datasets, determined by comparisons at 95% average nucleotide identity threshold. b Number of metagenomes in which FRAM_EGC MAGs were detected with at least 1× coverage. The horizontal purple line represents the total number of samples (n = 67). c Average relative abundance of sPLS clusters (Fig. 4) across different ice cover, daylight and depth values determined by read recruitment from Arctic Ocean and Fram Strait metagenomes to the respective FRAM_EGC MAGs. The ~4000 m sample from MOSAiC was not included.

Fig. 6. Temporal dynamics of signature populations.

Signature populations were identified as ASV representatives from sPLS clusters that a corresponding MAG was recovered for (based on 100% identity threshold competitive read recruitment). The temporal dynamics visualised are derived from ASV data. The missing chlorophyll data in 2016−2018 is due to the lack of a sensor on the MIZ mooring.

Fig. 7. Selected genes involved in the metabolism of organic and inorganic compounds enriched under high- and low-ice conditions.

Enrichment is displayed as centred-log ratio transformed normalised gene counts. Where several genes of a single pathway or mechanism were identified as enriched, they were grouped into one and the term ‘utilisation’ used (e.g. “taurine utilisation” indicates the uptake and degradation of taurine). When single genes were identified, the corresponding gene names are included. AA amino acids, BCAA branched-chain amino acids, GH glycoside hydrolase.

Fig. 8. Bacterial communities under contrasting AW influx and ice cover conditions.

Illustration showing the ten taxonomic groups with highest average relative abundances under Atlantic vs. Arctic conditions, derived from the relative abundances of Int-ASVs (sPLS cluster C1) and Res-ASVs (sPLS cluster C8), respectively. Figure was generated using Biorender.com.