Marine nitrogen-fixers in the Canadian Arctic Gateway are dominated by biogeographically distinct noncyanobacterial communities

Abstract

We describe diazotrophs present during a 2015 GEOTRACES expedition through the Canadian Arctic Gateway (CAG) using nifH metabarcoding. In the less studied Labrador Sea, Bradyrhizobium sp. and Vitreoscilla sp. nifH variants were dominant, while in Baffin Bay, a Stutzerimonas stutzeri variant was dominant. In comparison, the Canadian Arctic Archipelago (CAA) was characterized by a broader set of dominant variants belonging to Desulfobulbaceae, Desulfuromonadales, Arcobacter sp., Vibrio spp., and Sulfuriferula sp. Although dominant diazotrophs fell within known nifH clusters I and III, only a few of these variants were frequently recovered in a 5-year weekly nifH times series in the coastal NW Atlantic presented herein, notably S. stutzeri and variants belonging to Desulfobacterales and Desulfuromonadales. In addition, the majority of dominant Arctic nifH variants shared low similarity (< 92% nucleotide identities) to sequences in a global noncyanobacterial diazotroph catalog recently compiled by others. We further detected UCYN-A throughout the CAG at low-levels using quantitative-PCR assays. Temperature, depth, salinity, oxygen, and nitrate were most strongly correlated to the Arctic diazotroph diversity observed, and we found a stark division between diazotroph communities of the Labrador Sea versus Baffin Bay and the CAA, hence establishing that a previously unknown biogeographic community division can occur for diazotrophs in the CAG.

Although marine diazotrophs occur globally, significantly less is known about their ecology and biodiversity within the Arctic Ocean; this study in the Canadian Arctic Gateway and Archipelago revealed geographically unique diazotroph community compositions in this polar region.

Introduction

The Arctic Ocean is undergoing major environmental change due to climate-related warming (Ardyna et al. 2014), yet this ocean and its nearby seas are among the most understudied oceanographic areas because of their remote location and relatively harsher climate (Boeuf et al. 2014). In recent years, considerable progress has been made in the elucidation of polar microbial communities (Deming 2002, Dinasquet et al. 2018, Edwards et al. 2020, Zhang et al. 2020); for instance, Arctic marine bacteria experience strong seasonal changes (Kellogg et al. 2019) due to differing light and sea ice regimes between winter and summer seasons (Dinasquet et al. 2018), and bacteria in the Arctic Ocean have structured biogeography (for e.g. between surface and deep waters and from Eurasian to Canadian Arctic basins; Galand et al. 2009). Early cloning studies (e.g. Pommier et al. 2007) have also shown that marine heterotrophic microbes in the Arctic Ocean include both cosmopolitan and polar-associated community members (summarized by Boeuf et al. 2014). The degree of cosmopolitanism and endemism between taxonomic groups also applies to intraspecific microbial diversity, as seen for example with SAR11 phylotypes that are restricted to the Arctic Ocean (Kraemer et al. 2019).

In the spring, melting sea ice and increased solar irradiance initiate an Arctic phytoplankton bloom that contributes to polar primary productivity (Comeau et al. 2011, Campbell et al. 2017, Schuback et al. 2017). Nitrogen is generally the limiting nutrient in the Arctic Ocean, particularly for photosynthesis/primary production (Comeau et al. 2011, von Friesen and Riemann 2020), with Fernandez-Mendez et al. (2016) recently showing that surface N:P ratios point to nitrogen limitation in the Central Arctic Ocean. Diazotrophs that can convert atmospheric dinitrogen (N₂) to ammonia via nitrogen fixation are an important microbial group that can overcome such limitation (Sohm et al. 2011, Tang et al. 2019); however, it is only recently that their presence within the Arctic Ocean has been fully recognized due to previously held assumptions regarding how inorganic nitrogen, temperature, and oxygen shape diazotroph biogeography in the ocean (Gallon 1992, Breitbarth et al. 2007, Fernández-Méndez et al. 2016, Sipler et al. 2017, Stal 2017, Shiozaki et al. 2018, Zehr and Capone 2020, 2021a, b).

Two main studies initially identified the presence of diazotrophs in the Arctic and showed that the dinitrogenase reductase gene (nifH) could be amplified in the Central Arctic Ocean (Damm et al. 2010) and that nifH community patterns could be recovered in the Arctic via next-generation sequencing (Farnelid et al. 2011). Interestingly, the latter work was a global survey of nifH and showed that Baffin Bay [within the Canadian Arctic Gateway (CAG)] was an extreme outlier relative to other regions in the global ocean with respect to its diazotrophic community composition (Farnelid et al. 2011). No cyanobacterial diazotrophs were recovered from the Baffin Bay microbial DNA samples, and no noncyanobacterial phylotypes were shared with other sites sampled across the globe, suggesting likely arctic endemism for these diazotrophs (Farnelid et al. 2011). Heterotrophic diazotrophs [or noncyanobacterial diazotrophs (NCDs), more precisely; Turk-Kubo et al. 2022] are now considered widespread (Riemann et al. 2010, Farnelid et al. 2011, Langlois et al. 2015, Bombar et al. 2016), with more recent surveys (e.g. Tara metatranscriptomes) confirming their dominance and nifH expression in the world’s oceans (Farnelid et al. 2011, Salazar et al. 2019).

While recent studies support the consensus that most diazotrophs from the Arctic Ocean are NCDs of clusters I and III (Blais et al. 2012, Díez et al. 2012, Fernández-Méndez et al. 2016, Shiozaki et al. 2018), notable exceptions include Trichodesmium detection in sea ice brine from Fram Straight near Greenland (Díez et al. 2012) and the observation of cluster IV diazotrophs in sea-ice and melt ponds near the Central Arctic Ocean (Fernández-Méndez et al. 2016). Although they do not represent a large proportion of the Arctic diazotrophic community, the symbiotic unicellular cyanobacterial diazotroph known as Candidatus Atelocyanobacterium thalassa (or UCYN-A) has been recently recovered in western Arctic waters of the Chukchi Sea (Harding et al. 2018, Shiozaki et al. 2018). UCYN-A exchanges fixed-N for fixed-C from its algal host (Martínez-Pérez et al. 2016), and consequently has lost the ability to fix carbon via photosynthesis (Tripp et al. 2010). Single cell nitrogen-fixation rates are comparable between polar and nonpolar UCYN-A (Harding et al. 2018, Zehr and Capone 2021a). In addition to UCYN-A, Richelia and Epithemia diatom–diazotroph associations (Caputo et al. 2019, Schvarcz et al. 2022), and Arctic-associated ultrasmall (< 0.22 micron) Arcobacter (Karlusich et al. 2021) have also recently been found at higher latitudes, further proving that diazotrophs are part of the Arctic marine microbiome. Measurable nitrogen fixation rates range from 0.02 nmol N l⁻¹ day⁻¹ in Baffin Bay (Blais et al. 2012) to 17.2 nmol N l⁻¹ day⁻¹ in the coastal Chukchi Sea [Sipler et al. 2017; also see Shiozaki et al. (2018)]. Sipler et al. (2017) estimated that during ice-free periods, Arctic shelves alone could account for as much as 2.7% of global nitrogen fixation. Although effects may also be regionally specific (von Friesen and Riemann 2020), the significant nitrogen fixation rates, therefore indicate that although diazotrophs represent a low percentage of bacterioplankton in the Arctic (Salazar et al. 2019, Karlusich et al. 2021) the input of new nitrogen from these microbes may still be important.

Although diazotrophs are now considered broadly distributed at polar latitudes, their geographic distributions and diversity across various regions of the Arctic Ocean are underexplored (Shiozaki et al. 2017, von Friesen and Riemann 2020, Karlusich et al. 2021, Meiler et al. 2022). We aimed to help bridge this critical knowledge gap by further investigating diazotroph diversity based on nifH amplicon sequencing from the Labrador Sea into Baffin Bay and onwards towards the Canadian Arctic Archipelago (CAA). Specifically, we identified dominant diazotrophs present within this Arctic region, along with their correlations to ocean conditions such as depth, temperature, size fraction, oxygen levels, and selected macro and micronutrients. Given recent reports of UCYN-A in the Arctic Ocean (Harding et al. 2018, Shiozaki et al. 2018), we further conducted quantitative PCR assays of UCYN-A ecotypes (A1 and A2) to assess their presence within eastern Canadian Arctic waters. Since dominant Arctic phylotypes have previously been found to be generally less similar to those of other oceans (Turk-Kubo et al. 2022), we also sought to further establish any degree of Arctic endemism for diazotrophs identified as important in the region. This was accomplished by comparison of CAG-dominant nifH amplicon sequence variants (ASVs) to: (i) a 5-year weekly nifH time series that we have established in a temperate NW Atlantic fjord (Bedford Basin, NS, Canada), (ii) more global nifH sequence sets that are publicly available [from Delmont et al. (2021) and Turk-Kubo et al. (2022)], and (iii) to known nifH sequences from the western Arctic Ocean (Shiozaki et al. 2018). Overall, it is critical to elucidate the diversity and ecology of diazotrophs within the Arctic sector, given that this fraction of the ocean microbiome can contribute to Arctic Ocean primary productivity via the generation of new fixed-nitrogen (von Friesen and Riemann 2020).

Materials and methods

Cruise samples and their environmental data

Water samples (n = 125) were collected via Niskin bottles aboard the CCGS Amundsen between 10 July 2015 and 20 August 2015 during the ArcticNet1502 (GN02) 2015 expedition. This expedition was part of an ArcticNet and Canadian Arctic GEOTRACES joint effort (Anderson et al. 2014) to study the Labrador Sea and Northwest Passage. DNA was collected at eleven stations spread throughout the Labrador Sea (K1 and LS2), Baffin Bay (BB1, BB2, and BB3), and CAA [CAA1–2 and CAA4–7; see Table S1 (Supporting Information) for coordinates]. At each station ∼4 l of seawater (average = 3.93 l) was subjected to peristaltic filtration to collect biomass onto 3 µm and 0.2 µm Isopore polycarbonate membrane filters (Millipore, Ireland) in series such that the large fraction represents ≥ 3 µm and the small fraction 0.2–3 µm. A map of station coordinates is provided in Fig. 1. Multiple water depths were collected at each station (Table S1, Supporting Information) and ranged from surface to ∼2300 m. DNA filters were immediately frozen and kept at –80°C until further processing.

Figure 1.

Sites sampled along the CAG during the ArcticNet1502 (GN02) GEOTRACES expedition. Samples are from July to August 2015. (A) Arctic Ocean region covered by nifH sampling during our study and the location of the weekly Bedford Basin Monitoring Program (BBMP; asterisk), dark box indicates boarders of map shown in panel B. (B) Detailed map of study sites with bathymetry (background color) from ETOPO1 dataset (Amante and Eakins 2009, NOAA National Geophysical Data Center 2009, Simons and John 2022). Light gray arrows depict the main circulation through and near the Davis Strait (DS) with Arctic outflows on the western side and inflow from the West Greenland and Irminger Currents on the eastern side, as well as the southern flowing Labrador Sea Current further south along the western side of Labrador Sea [after Fragoso et al. (2016), Colombo et al. (2020), and Lehmann et al. (2022)]. CAA stands for Canadian Arctic Archipelago.

Oceanographic data collected during the GN02 cruise, summarized by section plots provided in Fig. 2, were both CTD sensor-derived (for temperature, salinity, oxygen, and fluorescence) and bottle-derived (for nitrate, nitrite, phosphate, silicate, and dissolved and particulate trace metals). These data are publicly available as part of the GEOTRACES Intermediate Data Product 2021 Version 1 dataset (IDP2021; GEOTRACES Intermediate Data Product Group 2021), along with dissolved iron and manganese data accessible from the supplemental data reported in Colombo et al. (2020) and total particulate iron, manganese, vanadium, and phosphorus further accessible from the supplemental data of Colombo et al. (2021 and 2022). Due to the tight water budget on the GEOTRACES cruises, DNA samples and other bottle-derived oceanographic data were occasionally collected on different CTD casts at a given station. Therefore, nutrient and trace metal data were matched to the closest depth and time at which DNA samples were collected at each station.

Figure 2.

Oceanographic data for CAG during 2015 GN02 expedition. (A) Sensor-derived temperature (Temp), salinity, fluorescence (Fluor), and dissolved oxygen (Oxygen). (B) Bottle-derived nutrients and trace metals from IDP2021 dataset (GEOTRACES Intermediate Data Product Group 2021) and supplemental data in Colombo et al. (2020, 2021, 2022). Shown are NO₂⁻ (nitrite), NO₃⁻ (nitrate), PO₄³⁻ (phosphate), Si (Silicate), D-Fe (dissolved iron), and TP-Fe (total particulate iron). Dark regions outline the bottom bathymetry (ETOPO1 dataset; Amante and Eakins 2009) along the transect based on distance (km) starting from the outer Labrador Sea (at K1). No TP-Fe data available for station CAA2. Plots were generated with Ocean Data View (Schlitzer 2002, 2021). Dots are positions with oceanographic measurements, open circles are positions where eDNA was collected.

Coastal NW Atlantic time series samples and their environmental data

Intermittently during 2014 (once in each of March, June, September, and December; 10 samples total) and weekly for 5 years between January 2015 and December 2019, we collected DNA samples from the coastal NW Atlantic [in Bedford Basin at site BBMP (Fig. 1A)—this location (44.6936 LAT, −63.6403 LON) is within a fjord that is also the site of Halifax Harbour (Nova Scotia, Canada; Kerrigan et al. 2017)]. Seawater samples from 1, 5, and 10 m (surface) and 60 m (bottom) were collected using Niskin bottles. Biomass from 500 ml was filtered onto Isopore polycarbonate filters (Millipore) using a peristaltic pump such that size ranges were: 0.2–160 µm (i.e. ≥ 0.2 µm) for 2014–2015, 0.2–330 µm (i.e. ≥ 0.2 µm) for 2016–2017, and 0.2–3 µm (small fraction) and ≥ 3 µm (large fraction) for 2018–2019. Filters were immediately frozen at –80°C after each collection.

Measurements for temperature, salinity, and oxygen (all CTD sensor-derived) and nitrate (Niskin-derived) from the Bedford Basin are accessible from the Bedford Institute of Oceanography [BIO; see reference (BIO 2022)]. These oceanographic data from the Bedford Basin are collected as part of the BIO weekly Bedford Basin Monitoring Program (Li and Dickie 2001, Li et al. 2006).

DNA extractions, nifH amplicon sequencing, and UCYN-A quantitative-PCRs

All DNA filters from cruise and time series were processed using the same procedure. DNA was extracted (50 µl final volume) using a DNeasy Plant Mini kit (Qiagen; Zorz et al. 2019). A nested PCR method was used to amplify the nifH diazotroph marker gene from each sample using the nifH3/nifH4 [ATRTTRTTNGCNGCRTA/TTYTAYGGNAARGGNGG] and nifH1/nifH2 [TGYGAYCCNAARGCNGA/ADNGCCATCATYTCNCC] primer pairs from Zehr and McReynolds (1989) and Zani et al. (2000). PCRs (25 µl for PCR1, 10 µl for PCR2) were carried out with: 1x buffer (Qiagen), 4 mM (PCR1) or 3 mM (PCR2) MgCl₂ (Qiagen), 240 µg/ml BSA (NEB), 9.725 µl (PCR1) or 4.29 µl (PCR2) molecular biology grade H₂O (Invitrogen), 0.025 U/µl HotStar Taq (Qiagen), 2.5 µl template extracted DNA (PCR1) or 1 µl of nifH3/4 PCR product (PCR2), 10 nM dNTPs (Invitrogen), and 800 nM nifH primers (IDT). Thermocycler settings were: (PCR1) 95°C—15 min, 35 cycles of 95°C—1 min, 45°C—1 min, and 72°C—1 min, then 75°C—10 min; PCR2 increased the annealing temperature to 54°C and decreased the number of cycles to 28. Amplicons of 359 bp detected in second-round PCRs via agarose gel electrophoresis were further processed for Illumina next-generation sequencing of the nifH gene. Briefly, this involved repeating the nifH3/4 PCR1 (PCR1d) with 1/10 diluted DNA, and then combining equal amounts of the products from PCR1 and PCR1d as template for a modified PCR2 (PCR2f) with nifH fusion primers (Ratten 2017) that combined Illumina adaptors and barcodes with nifH1/nifH2. Thermocycler settings were as above with a 52°C annealing temperature and 35 cycles of amplification. Barcoded products were normalized and purified using a Just-A-Plate 96 kit (Charm Biotech). The final nifH library pool was sequenced on an Illumina MiSeq instrument (Integrated Microbiome Resource, Dalhousie University, Halifax, NS, Canada). NifH sequencing data are deposited under NCBI Bioprojects PRJNA930772 for the Bedford Basin time series and PRJNA931255 for the Arctic GEOTRACES 2015 expedition (NCBI Resource Coordinators 2018).

To enumerate UCYN-A, we used the quantitative-PCR (qPCR) assay of Langlois et al. (2008) for the UCYN-A1 ecotype: 1x TaqMan Universal PCR master mix (Applied Biosystems), 100 nM TaqMan MGB 6-FAM probe (Applied Biosystems), 900 nM of each forward and reverse primer (IDT), 400 µg/ml BSA (NEB), 2.232 µl of molecular biology grade H₂O (Invitrogen), and 5 µl of 0.56x DNA, with thermocycler settings of: 95°C—10 min then 45 cycles of 95°C—15 s and 60°C—1 min. For the UCYN-A2 ecotype we developed a new assay modified from Thompson et al. (2014): 1x TaqMan, 200 nM probe 5′-FAM-TCTGGTGGTCCTGAGCCCGGA-NFQ-3′, 2.5 µM each of forward primer 5′-GGTTACAACAACGTTTTATGTGTTGAA-3′ and reverse primer 5′-ACCACGACCAGCACATCCA-3′, 400 µg/ml BSA, 1.56 µl molecular biology grade H₂O, and 5 µl 0.56x DNA, with thermocycler settings of 95°C—10 min then 45 cycles of 95°C—15 s and 64°C—1 min. Both assays were run on a ViiA7 real-time PCR system with QuantStudio software (Applied Biosystems) used to assess the final number of gene copies per reaction. Assuming 4 l of filtered seawater per DNA sample, our limit of quantification (LOQ) for both assays was 5 copies/l, and average qPCR efficiencies were 106% ± 24% SD for the A1 assay and 91% ± 10% SD for the A2 assay (calculated using LinRegPCR (Ramakers et al. 2003, Ruijter et al. 2009)). A Qubit 4 fluorometer and 1x HS dsDNA kit (Invitrogen) were used to quantify gBlock gene fragments of the nifH gene (IDT) for generating A1 and A2 standard curves (standards were run in triplicate, with water (Invitrogen) used as a nontemplate control).

Data analyses

ASV identification

ASVs were generated from nifH sequencing data for both the Arctic research cruise and the Bedford Basin time series using QIIME 2 version 2019.7 (Bolyen et al. 2019) following a workflow modified from the Microbiome Helper pipeline of Comeau et al. (2017). For primer removal the nifH1/2 primers were used with cutadapt (Martin 2011). Reads were denoised into ASVs using deblur (Amir et al. 2017) with a nifH reference set (Gaby and Buckley 2014) and a consistent trim-length of 325 bp. ASVs with a frequency of < 0.1% of the mean sample depth were removed due to potential bleed-through between sequencing runs. Final ASV tables were converted to % relative abundances [calculated as: (# reads per ASV in sample ÷ total reads in sample) * 100] and to centered-log ratio (CLR) values using phyloseq and microbiome R packages (Gloor et al. 2017, Lahti and Shetty 2019, McMurdie and Holmes 2013, R Core Team 2021). The final mean sampling depth per sample was 2504 ± 907 SD reads.

Taxonomic identification

Diazotroph taxonomies were assigned by placing their reference sequences into the phylogeny described by Kapili and Dekas (2021). The workflow included: hmmer version ≥ 3.1 (see hmmer.org; Eddy 2011) for the alignment of sequences, EPA-ng (Barbera et al. 2019) for the placement of sequences into the Kapili and Dekas (2021) phylogeny, and gappa (Czech et al. 2020) to assign the final taxonomy based on the tree placements. The implementation of this pipeline required the use of EMBOSS software package (Barbera et al. 2019). To ensure the Kapili and Dekas (2021) sequences (n = 8877) would encompass currently known diazotroph genomes, we also subjected reference nifH genes (n = 3490) from the genome taxonomy database (GTDB; Chaumeil et al. 2022) to the above pipeline. Out of this GTDB set, 613 sequences were incorrectly placed as compared to their known taxonomy, while 977 sequences with known taxonomy were not placed in the tree at all, and the remaining 1900 sequences were correctly assigned. To account for the possibility that our Arctic ASVs might belong to the prior two groups, we used BLAST (Johnson et al. 2008) to align our Arctic nifH sequences to the above misidentified/unclassified reference sequences; seven ASVs were renamed as a result, of which six were rare, and the seventh (Pseudomonadales) was incorrect due to a recent name change within the genus Pseudomonas to Stutzerimonas (Lalucat et al. 2022).

As a final check of the taxonomy, we locally aligned all dominant ASV reference sequences against the NCBI nucleotide collection nr/nt (Acland et al. 2014) via BLAST (Johnson et al. 2008); although this was useful for inferring the current best-matches to our ASVs, we have defaulted to the tree placement taxonomy as it was typically more conservative by placing sequences mainly at higher classification levels. Scores for alignments to the NCBI collection (Acland et al. 2014) and to Kapili and Dekas (2021) are provided in Table S2 (Supporting Information). Note that we defined dominant ASVs as those that either: (a) had the most reads across samples (using the sum of reads across samples) and contributed to ∼80% of the total reads in the nifH dataset or (b) were within the top four ASVs per sample based on total reads, and then from this (c) were found in ≥ 3 samples and reached ≥ 1% relative abundance in at least one sample. Hence, dominant ASVs are those that are nonrare both in relative abundance and presence across samples.

Graphical analyses and statistics

R version 4.1.2, RStudio version 2021.09.1.372, and ggplot2 (Wickham 2016) were used for graphical data analyses (R Core Team 2021, R Studio Team 2021) along with additional R packages listed in Methods S1 (Supporting Information). Multivariate analysis was conducted using vegan (Oksanen et al. 2022) with its decostand function used to standardize oceanographic data and envfit function used to fit environmental vectors on the RDA (permutations = 999). Aitchison distances (dissimilarity index) between the nifH of the Bedford Basin time series and the Arctic cruise samples were calculated using the vegdist function in vegan (Oksanen et al. 2022); the Bedford Basin ASV table used for distance calculations was also prefiltered to only include ASVs with ≥ 25 total reads across all samples to reduce computation time. The canonical correlation analysis (CCA) presented and its associated heatmap visualization was completed using the mixOmics R package (Rohart et al. 2017) by assuming 12 components and the ridge method with Lambda1 = Lambda2 = 0 as CCA parameter settings. The multipatt function in the indicspecies R package was used to conduct a multilevel pattern analysis of species patterns versus site groupings (settings were function = r.g and number permutations = 9999) (De Cáceres and Legendre 2009), while the intersections of ASVs across samples was plotted using the UpSetR package (Conway et al. 2017). Maps were generated with rnaturalearth (South 2017), ggrepel (Slowikowski 2021), ggspatial (Dunnington 2022), and ggOceanMaps (Vihtakari 2022) packages with land and island data from Natural Earth Data (www.naturalearthdata.com) and also bathymetry data from the online ETOPO1 topography dataset (etopo180) accessed via ERDDAP and its griddap protocol/access form (Amante and Eakins 2009, NOAA National Geophysical Data Center 2009, Simons and John 2022). The institutions/creators listed for bathymetry data are the National Oceanic and Atmospheric Administration and the National Geophysical Data Center. Section plots of oceanographic data were generated using ETOPO1 data (Amante and Eakins 2009) and Ocean Data View (Schlitzer 2002, 2021).

Phylogenetic and sequence analyses were carried out in Geneious Prime version 2022.2.2 (www.geneious.com); therein, RAxML 8.2.11 was used to generate Maximum Likelihood (ML) trees for dominant NCD and cyanobacterial nifH sequences using rapid bootstrapping (1000 replicates), the GTR GAMMA model, and a search for the best-scoring ML tree (random seed = 12 345) (Stamatakis 2014). Trees were built from codon-aligned nucleotide sequences generated by first aligning translated nifH sequences using MUSCLE (Edgar 2004) and then converting back to nucleotides using PAL2NAL (Suyama et al. 2006). Phylogenetic visualizations were built using iTOL (Letunic and Bork 2021). The Bedford Basin time series nifH reference sequences (with singleton ASVs removed) were searched for Arctic nifH ASVs using their alphanumeric names given that both datasets were processed with identical parameters. We also used the standalone BLAST (Altschul et al. 1990) function within Geneious Prime to align our nifH ASV reference sequences to sequences reported in Kapili and Dekas (2021) (to help with the taxon identifications described above), as well as Shiozaki et al. (2018) (to compare to the western Arctic Ocean) and Delmont et al. (2021) and Turk-Kubo et al. (2022) (to assess general endemism). Note that we compare our ASVs to those in the study of Shiozaki et al. (2018) given that their sampling occurred from early September to early October 2015, and hence close to the same time period that our samples were collected.

Results

Diazotroph biogeography within the CAG

Summary plots for physical, chemical, nutrient, and trace metal data are provided (Fig. 2; Figure S1, Supporting Information). Iron and vanadium metals were selected as they are particularly relevant for the nitrogenase enzyme (Zehr et al. 2003); no data were available for molybdenum. Since detailed descriptions for the above measurements have been given elsewhere by others (Schuback et al. 2017, Colombo et al. 2019, 2020, 2021, 2022, Lehmann et al. 2019), trends will only be summarized here. Samples were collected towards the end of seasonally expected ice coverage (end of July–August) (Lehmann et al. 2019, Randelhoff et al. 2020). At sampling, the mixed-layer depth was distinct and extended to a maximum of ∼40 m at LS2 (Fig. 2; Schuback et al. 2017). Within this upper layer and extending to the subsurface chlorophyll maximum (SCM) (located submixed layer), an increased fluorescence signature reflected phytoplankton growth (highest SCM at 28 m for BB3) (Fig. 2; Schuback et al. 2017). Lower temperatures occurred below the mixed layer and nitrate and phosphate concentrations were lower in the mixed layer (< 10 µM for nitrate and < 1.35 µM for phosphate above 50 m) (Fig.   2; Schuback et al. 2017). Schuback et al. (2017) reported N:P ratios below 16:1 suggesting nitrate limitation in the mixed-layer throughout the Labrador Sea, Baffin Bay, and CAA. Colder saltier seawater occurred in the deep water of the Baffin Bay, and warmer saltier seawater in the Labrador Sea, whereas less saline seawater characterized the surface of the Baffin Bay and CAA stations where surface warming and freshwater inputs from the Arctic region were apparent (e.g. from glacial and sea ice meltwater) (Fig.   2; Colombo et al. 2019, Lehmann et al. 2019). In the deep and older Baffin Bay waters, reduced oxygen corresponded to higher nutrient concentrations (Fig. 2; Lehmann et al. 2019). Dissolved iron and manganese were higher in the CAA and lower in the Labrador Sea (Fig. 2B; Figure S1, Supporting Information; Colombo et al. 2020). In the CAA, dissolved trace metals come from the benthic layer and can be advected to Baffin Bay (Colombo et al. 2020, 2021). Total particulate iron and vanadium were higher near the bottom of the water column and at boundary currents, while total particulate phosphorus was reflective of surface primary productivity (Fig. 2B; Figure S1, Supporting Information; Colombo et al. 2022).

We identified 2490 nifH ASVs within the CAG (Fig. 3A). NifH genes were detected in the upper (> 100 m) and lower (< 100 m) water column at all stations (Fig. 3A) except for BB2 and CAA4, where they were found only at lower depths. Likewise, nifH genes were consistently in both large (> 3 µm) and small (0.2–3 µm) fractions (Fig. 3A). Due to the compositional nature of the nifH dataset, we converted reads to CLR values (Gloor et al. 2017). In our case, 0.1%, 1%, 4%, and 10% relative abundances correspond to approximate thresholds of 2.5, 4.5, 6, and 7 CLR values, respectively (Figure S2, Supporting Information). The entire set of ASVs fell within 13 broader taxonomic groups that could be separated by several proteobacterial classes, the phyla Bacteroidetes, Cyanobacteria, Nitrospirae, Planctomycetes, and Verrucomicrobia, as well as unknown proteobacteria and bacteria (Fig. 3A). Based on the total distribution of ASVs across stations, the readily observable trend characterizing nifH diversity at the time of sampling was the division of diazotroph communities into three distinct biogeographical groups: (i) the K1 and LS2 stations in the Labrador Sea, (ii) the BB stations in Baffin Bay and the CAA2 surface, and (iii) the remaining CAA stations from the CAA (Fig. 3A). The Labrador Sea group was predominantly composed of Betaproteobacteria, while the CAA group was predominantly composed of Gamma-, Delta-, and Epsilonproteobacteria, as well as ASVs belonging to Bacteroidetes (mainly at CAA1 and CAA5) and other ASVs grouped into Unknown Bacteria (Fig. 3A and B). Since ASVs with a higher relative abundance can obfuscate rarer diazotrophs, we also plotted individual ASV patterns across stations using the same taxonomic groupings (Fig. 3B). Values point to distinct spatial patterns among rarer diazotrophs, including: (i) Alphaproteobacteria predominant within the Labrador Sea group, (ii) diazotrophs belonging to Nitrospirae toward the surface at station CAA1, and (iii) Planctomycetes and Verrucomicrobia ASVs detected mainly within the CAA group (Fig. 3B). Cyanobacterial diazotrophs were infrequent and detected in only five samples; hence, NCDs were greatly overrepresented relative to cyanobacterial diazotrophs within the CAG (Fig. 3). A comparison of the number of shared ASVs between samples further confirmed that while individual sites had a large set of unique ASVs (with BB stations having the fewest total number of ASVs overall), site pairings within the CAA group and within the Labrador Sea group shared more ASVs with each other than for site pairings across these two groups (Figure S3, Supporting Information). Therefore, with respect to shared ASVs, stations in the Labrador Sea and the CAA are more similar to nearby stations than to those further away. At a lower taxonomic level specific Orders typified each broader taxonomic grouping (Figure S4, Supporting Information); for instance, Orders detected throughout the water column at some stations included: (i) Burkholderiales at station K1, (ii) Pseudomonadales at station BB1, and (iii) Vibrionales, Desulfobacterales, Desulfuromonadales, and Campylobacterales at multiple CAA stations (Figure S4, Supporting Information).

Figure 3.

Major nifH taxonomic groups present in CAG during GN02 Expedition. (A) Sum of CLR values for all ASVs within each major taxonomic group for each station and depth. (B) CLR values for individual ASVs at each station and depth by group. DNA filters were size-fractionated where L = ≥ 3 µm, and S = 0.2–3 µm. Analyses were limited to CLR values > 0 to account for data points where ASVs are absent. Rarefaction curves for datasets used herein shown in Figure S16 (Supporting Information). A total of 17 ASVs had stop codons and were, therefore, suggestive of pseudogenes, these were all very rare occurring in only one sample each with overall averages of three reads, 0.17% relative abundances, and 2.3 CLR scores across all samples. Based on taxonomic placement, these pseudogenes all fell within the numerous proteobacteria and unknown bacteria groups found in the Labrador Sea and CAA stations; these pseudogenes have been excluded from the analyses presented herein.

Total nifH diversity for all ASVs was further analyzed via redundancy analysis (RDA) and regression fitting of oceanographic measurements (Figure S5 and Table S3, Supporting Information). Oceanographic conditions significantly correlated to diazotrophic diversity during our study were temperature, depth, salinity, oxygen, and nitrate (all at P-value = .001), as well as fluorescence (P-value = .002) and phosphate (P-value = .006; based on α = 0.05; Table S3, Supporting Information). Nitrite and silicate were not significantly correlated to diazotrophic diversity within our samples (P-values > .05; Table S3, Supporting Information). Trace metals were not included in the RDA analysis given that their concentrations were not available for half of the DNA samples, but they are considered below with respect to dominant ASVs. The first seven axes of the RDA explained only ∼8.56% of the variance in the dataset, highlighting the ongoing need to sample more regions and environmental parameters within the CAG to continue resolving underlying structure/groupings within Arctic diazotroph communities. Specific RDA trends are summarized in Results S1 (Supporting Information).

Dominant CAG diazotrophs and their environmental links

We identified 106 nifH ASVs that were dominant throughout our dataset of the CAG [recall dominant ASVs had the most reads (contributed to ∼80% of the total nifH dataset) or were in the top four ASVs per sample and were in ≥ 3 samples and reached ≥ 1% relative abundance in at least one sample]. Although these dominant ASVs represented only 4.3% of the total observed ASVs, they accounted for 52% of the nifH reads. The remaining nondominant ASVs were less widespread across samples and belonged to the broader groups already mentioned (Fig. 3; Figure S4, Supporting Information) and will not be considered further. A phylogenetic tree for the 53 ASVs that occurred in more than seven samples shows the diversity encapsulated by the most frequently observed ASVs (53/106 = 50% of the dominant ASVs; Fig. 4). Spatial profiles for all dominant ASVs at each station and depth are provided (see Figures S6 and S7, Supporting Information), as are alignment scores for their final taxonomy (Table S2, Supporting Information). The dominant diazotrophs in the CAG during our study all belonged to NCD phylotypes within diazotrophic clusters I and III; there were no dominant ASVs detected for clusters IB, II, and IV (Fig. 4).

Figure 4.

ML tree of major nifH ASVs present in the CAG during the GN02 expedition. ASVs shown were those present in ≥ 7 samples. To increase legibility, branch lengths were not used [see Figure S17 (Supporting Information) for branch lengths]. Noncolored sequences are the closest reference sequences currently matching the ASVs as retrieved via BLAST from NCBI (Johnson et al. 2008, Acland et al. 2014) and from Kapili and Dekas (2021). The tree also includes other diazotrophs for which taxonomy and genomes are known. Support values are bootstrap values based on 1000 replicates. The tree is built using a codon-alignment. For colored labels that represent Arctic ASVs collected herein: large roman numerals (I–IV) shown in boxes are major diazotroph clades, outer labels around circle are taxonomic classes (or broader taxonomic groups based on colors in other figures), inner labels around branches are taxonomic orders (or more narrow taxonomic groups). Greek characters refer to groups of proteobacteria.

Consistent with earlier analyses of the total Arctic nifH dataset, dominant ASVs likewise showed spatially distinct distribution patterns and were correlated to a particular region and set of environmental variables (for e.g. depth) within the CAG based on a CCA (Fig. 5). In the Labrador Sea dominant ASVs belonged to the genera Bradyrhizobium and Vitreoscilla, as well as to the order Burkholderiales (all within cluster I; Figs 4 and  5). These ASVs were correlated with higher temperatures, higher oxygen levels, and generally lower trace metal concentrations (Fig. 5). The one ASV significantly associated mainly to Baffin Bay belonged to Stutzerimonas stutzeri (formerly Pseudomonas stutzeri) from cluster I; this ASV was present primarily throughout deeper (> 100 m) samples and was correlated with higher salinities, lower oxygen levels, and higher concentrations of nitrate and phosphate (Figs 4 and 5). In the CAA, various dominant ASVs were related to Desulfopila spp. (the Desulfobulbaceae), Desulfuromonas sp. or Geopsychrobacter sp. (the Desulfuromonadales), Sulfuricurvum spp. (Campylobacterales), Arcobacter sp. (Campylobacterales), Vibrio spp. (the Vibrionales), and Sulfuriferula sp. (Nitrosomonadales). The CAA region also contained several ASVs that were less taxonomically resolved including an Unknown Desulfovibrionaceae, two Unknown Deltaproteobacteria, an Unknown Nitrospiraceae, an Unknown Epsilonproteobacteria, and an Unknown Sphingomonadaceae (though this is most likely a Novosphingobium sp. (Figs 4 and 5). Although environmental conditions associated with dominant CAA ASVs appeared more variable than those observed for the Labrador Sea and Baffin Bay, these ASVs largely correlated with lower temperatures (Fig. 5) and could somewhat be divided by depth (Table S4, Supporting Information) and by salinity (Fig. 5).

Figure 5.

CCA between dominant ASVs in the CAG GN02 expedition and oceanographic environmental conditions. Analysis is limited to ASVs in ≥ 7 samples. Positive and negative scores indicate a positive and negative correlation, respectively, between a given ASV and an individual environmental condition. The heatmap has been organized by hierarchical clustering placing similar correlation profiles together amongst ASVs and amongst environmental conditions (clustering uses Euclidean distance and complete agglomeration method for hierarchical trees). Columns for region and size fraction show ASVs that were significantly associated with a CAG region or size fraction based on a multilevel pattern analysis (at α = 0.05; Tables S4 and S5, Supporting Information). Region categories in bold were significant in the multilevel pattern analysis, while those in italics could only be associated to a certain region from trends in spatial profiles (Figure S6, Supporting Information). For e.g. if an ASV was present at only one station per region it may not be significantly correlated to said region, in such cases, observing the raw data can help determine the region where the ASV was observed. Abbreviations: TP-Fe (total particulate Fe), TP-V (total particulate V), D-Fe (dissolved Fe), KS and LS (Labrador Sea), BB (Baffin Bay), CAA (Canadian Arctic Archipelago), Sm (0.2–3 µm), and Lr (≥ 3 µm). Colored dots correspond to taxonomy coloring in Fig. 4.

Interestingly, the CAA was home to three dominant ASV groups that could only be identified to the Bacteria domain (labelled A, B, and C) and one identified to the Proteobacteria phylum (Fig. 4). Unknown Bacteria A is most similar (86% pairwise identity; PI) to Kiritimatiellales, Unknown Bacteria B is most similar (∼79% PI) to Mangrovibacterium sp., and Unknown Bacteria C is similar to Nitrospiraceae (81% PI) and an “Uncultured Bradyrhizobium sp.” (87% PI); the latter Bradyrhizobium may be a horizontally transferred nifH gene due to its phylogenetic placement away from other Bradyrhizobium spp. (Fig. 4). The Unknown Proteobacteria is similar (80% PI) to Vibrio sp. and the group known as Gamma_1 (84% nucleotide PI; Turk-Kubo et al. 2022).

We also completed a supplemental multilevel pattern analysis, which generally agreed with the prior environmental correlations (those in Fig. 5), however, this set of analyses required the subjective categorization of environmental data into broad groupings (e.g. “high” versus “low” temperature, and so on). Hence, multilevel pattern results are more subjective, and must therefore be taken within the context of how groupings were defined (Results S2 and Table S4, Supporting Information).

To further capture any associations between dominant ASVs and environmental conditions, raw CLR distributions for each dominant ASV relative to raw environmental data were plotted (Figure S10, Supporting Information); these results confirm the broader Labrador Sea versus CAA divisions discussed above, but indicate: (i) the highest fluorescence values were mainly observed when Unknown Desulfobulbaceae nifH ASVs were present, (ii) higher nitrite concentrations were observed when S. stutzeri, Vitreoscilla sp., Unknown Burkholderiales, and a Bradyrhizobium sp. were present, and (iii) several instances of high silicate were observed when Novosphingobium/Sphingomonadaceae was present (Figure S10, Supporting Information). Dominant ASVs with the highest relative abundances and most frequent detections (each in ≥ 27 samples) included Bradyrhizobium (ASV93fd9) in the Labrador Sea, S. stutzeri (ASVb0b18) in Baffin Bay, and at least three ASVs from the Unknown Desulfobulbaceae in the CAA.

Collectively, the prior results demonstrate that the Labrador Sea and Baffin Bay were characterized by only a few dominant nifH ASVs during our study, whereas the CAA was characterized by a relatively greater number of dominant nifH ASVs (Fig. 5). Furthermore, dominant diazotrophs mirrored the biogeographic divisions seen for the entire diazotroph community (those trends observed in Fig. 3 and Figure S5, Supporting Information). Interestingly, the CAA also contained multiple dominant ASVs with poor taxonomic characterization suggesting that the CAA may be home to several yet-undescribed NCDs that were particularly important within this region of the Arctic Ocean during the time of sampling.

UCYN-A within the CAG

Although rare, cyanobacterial diazotrophs were identified in five samples and when present represented 1.8%–61% of the nifH reads. Cyanobacterial diazotrophs belonged to Trichodesmium, Pseudanabaena, and two Unknown Cyanobacteria that were phylogenetically closest to the Chroococcidiopsis and Euhalothece genera (Fig. 6A). UCYN-A was detected at one station in the Baffin Bay (Fig. 6A), prompting further quantification of the nifH gene for this species within the CAG. NifH gene copy numbers for two UCYN-A ecotypes, A1 and A2, are shown (Fig. 6B). Although UCYN-A1 and A2 were detected in the Labrador Sea, Baffin Bay, and the CAA, values were typically very close to the LOQ (Fig. 6B). Neither of the UCYN-A ecotypes were detected at station LS2 within the Labrador Sea (Fig. 6), while highest values (but still very low at < ∼30 nifH copies/l) were observed for both ecotypes at CAA7 at the top of the water column. No diazotrophs were detected in these surface CAA7 samples using nifH amplicon sequencing (Fig. 3A), suggesting that nifH amplification via widely used nested PCRs and UV imaging was not sensitive enough to detect UCYN-A at this station. Furthermore, at BB1 where the UCYN-A nifH gene was detected through amplicon sequencing, there were no detectable UCYN-A nifH copies via qPCR (Fig. 6); this disparity may arise from the greater number of PCR cycles associated with nifH sequencing (Zani et al. 2000). The UCYN-A1 ASV sequenced herein would be compatible with the UCYN-A1 qPCR assay. These data indicate that while UCYN-A was primarily found at CAA7 in the CAA, it was also detectable throughout the CAG during our study despite being found in the rare fraction of the diazotroph microbiome.

Figure 6.

Cyanobacterial diazotrophs detected in CAG during GN02 cruise. (A) ML phylogeny of all cyanobacterial diazotroph ASVs detected. Additional sequences in tree that are similar to the ASVs detected (nonbold labels) were collected from NCBI (Johnson et al. 2008, Acland et al. 2014) and Kapili and Dekas (2021). Support values (numbers mid branch) are rapid bootstrap values based on 1000 replicates. The tree is built using a codon-alignment. Each ASV was only detected in one sample and the % relative abundance within such samples is shown (gray boxes). UCYN-A was also detected based on its nifH (black box). (B) qPCR data for UCYN-A1 and -A2 ecotypes. Size fractions (Fraction) correspond to L = ≥ 3 µm, and S = 0.2–3 µm. Values with asterisks (*) are those where UCYN-A was detectable but not quantifiable (LOQ = 5 nifH copies/l).

Searching for dominant CAG diazotrophs outside of the CAG and polar realm

To assess whether the dominant ASVs found herein also occurred outside of the CAG, we searched for these nifH ASVs sequences within: (i) a nifH time series from a nearby fjord in the coastal NWA, (ii) previously published data for the western Arctic Ocean (Shiozaki et al. 2018), (iii) Tara samples reported for lower latitudes (Delmont et al. 2021), and (iv) a recently curated and comprehensive catalog of NCDs based on samples from across the world’s oceans (Turk-Kubo et al. 2022).

Seasonally through 2014, and weekly through 2015–2019, we sampled for diazotrophs in the Bedford Basin (BBMP in Fig. 1A); this sequencing effort amounted to 81 661 nifH ASVs that were observed more than once. The Bedford Basin displays the expected seasonal cycling of a temperate fjord in the North Atlantic, with lower surface temperatures in the winter and higher surface temperatures in the summer (Fig. 7A; Results S3, Supporting Information; Li and Dickie 2001). It is important to note that from 2014–2017 DNA samples were not size-fractionated, whereas from 2018 to 2019 samples were size-fractionated in the same way as the Arctic samples shown herein. Consequently, weeks falling between 2018 and 2019 of the time-series are more directly comparable to the Arctic DNA sampling presented in previous figures (when considering CLR values). However, the difference in the sampling regimes does not affect the comparison of the ASV DNA signatures for the diazotrophs. The entirety of the time-series is useful for comparing ASV DNA signatures across these environments.

Figure 7.

Dominant ASVs recovered from the CAG within a 5-year time series in the NWA. (A) surface and deep (60 m) ocean conditions of the Bedford Basin time series (2016–2019). Known shelf-water intrusion events are shown as dashed vertical lines (I14–I18b; Haas et al. 2021). (B) Number of dominant ASVs recovered [dark grey] and not recovered [light grey] in the time series by taxonomic group. (C) CLR values for top two taxonomic groups recovered, as well as a Pseudomonadales ASV that had a distinct temporal pattern. Note only 10 samples were collected in 2014 and size fractionation started in 2018, thus doubling observations for the last 2 years. Since Arctic samples were size fractionated only years 2018–2019 are directly comparable to Arctic samples. Earlier samples are useful from the perspective of determining presence/absence of ASVs over > 2 years. (D) Profiles for ASVs that comprise nearly all observations shown in the upper two panels of (C). (E) Comparison of S. stutzeri (Pseudomonadales) versus dissolved oxygen in the Bedford Basin at 5 m. (F) Distributions for Aitchison Distances between Bedford Basin 5 m and Arctic 5 m samples, gray highlighting shows months when Arctic GN02 cruise occurred (other depths shown in Figure S15, Supporting Information).

Based on identical DNA sequence identity (PI), ∼10% (or 247/2490) of the total ASVs observed in the CAG were recovered in the fjord (Figure S11, Supporting Information); this fraction was much higher for dominant Arctic ASVs at ∼42% (or 44/106 ASVs), with the most recovered dominant groups belonging to the Desulfobacterales (eight ASVs) and Desulfuromonadales (12 ASVs) [Fig. 7B and C; also see Figure S12 (Supporting Information) for the full set of ASVs]. Of these, five ASVs comprised most of the detections (Fig. 7D; Figure S13, Supporting Information). Based on phylogenetic placement (Fig. 4) one of the Desulfobacterales is related to nifH from Desulfopila (ASV81bee), while two of the Desulfuromonadales are most related to the nifH from Geopsychrobacter sp. (ASV009ac and ASV30da6; 88%–89% PI) (Table S2, Supporting Information). These ASVs were generally present throughout each year, mainly at 60 m, whereas one of the Unknown Desulfuromonadales (ASVef783) was detected in both surface and deep samples (Fig. 7D).

Further comparison to the dominant diazotrophs from the western Arctic Ocean given by Shiozaki et al. (2018) indicated four overlapping reference sequences (Table S6, Supporting Information). These all belonged to Desulfobacterales and Desulfuromonadales, however, “ASV30da6 Unknown Desulfuromonadales” was the sole ASV recovered at high frequency within the fjord (Table S6 and Figure S13, Supporting Information).

The Pseudomonadales group is also of special interest, given its very distinct presence within the fjord at one depth and time period (late 2018 at 5 m; Fig. 7C). This group was characterized by a single nifH ASV identical to S. stutzeri (ASVb0b18) found in Baffin Bay (Fig. 5). Interestingly, in the Bedford Basin this nifH sequence was present in near opposite conditions to that of the Baffin Bay; it was restricted to the surface with comparatively lower salinity relative to Baffin Bay and had highest relative abundance/CLR values when nitrate was low, and temperatures were above 10°C [Fig. 7 and Figure S10, Supporting Information; also see Figure S14a (Supporting Information) for easier interpretation at 5 m]. Other oceanographic data suggests that the presence of this ASV in the Bedford Basin may have been tied to oxygen levels. Although temperature ultimately controls the amount of oxygen that will dissolve at the surface (Figure S14b, Supporting Information), the ASV matching to S. stutzeri was detected mainly when surface dissolved oxygen dropped to concentrations more comparable to Baffin Bay (Fig. 7E versus oxygen in Fig. 2A).

Although temperatures in the fjord were higher than in the Arctic during the time of sampling, Aitchison Distances calculated across all sample pairs between fjord versus CAG indicated the highest diazotroph community similarities occurred between the two regions during the month in which the Arctic cruise samples were collected (Fig. 7F). This trend was consistent across all depths (Figure S15, Supporting Information). Hence, the Bedford Basin nifH time series appears to share some overall seasonal similarity to the CAG with respect to its diazotroph community, at least during the mid to late summer; further Arctic sampling is required to determine whether this trend holds true across other times of the year.

The diazotrophs that were dominant in the CAG did not match any of the nifH sequences belonging to the diazotroph MAGs reported by Delmont et al. (2021) with high sequence identity [at 99.69% coverage (cov.) and 100% PI to the ASVs herein]. Even at a lower threshold there was only one MAG (AON_82_MAG_70) that somewhat matched to one of our dominant Arctic ASVs (ASV41648 Unknown Sphingomonadaceae; most similar to Novosphingobium) (> 95% cov. and > 92% PI; Tables S2 and S6, Supporting Information). Such low recoveries suggest that the nifH ASVs we identified from the CAG were not well-represented in Tara Oceans diazotroph MAGs collected at mainly lower latitudes (Delmont et al. 2021).

Recently, a comprehensive catalogue of major NCD groups in the ocean has become available (Turk-Kubo et al. 2022). Matches to ∼13% (14/106 ASVs) of the dominant Arctic diazotrophs of the CAG at the time sampled were found within this catalog based on nucleotide sequences (at > 95% cov. and > 92% PI; Table S6, Supporting Information). When considering amino acid similarities for the best nucleotide matches to the catalog instead, this proportion increased to 44% (Table S6, Supporting Information). Dominant CAG ASVs well-recovered by the NCD catalog included members of the Burkholderiales, Desulfobacterales, Desulfuromonadales, Pseudomonadales, Rhizobiales, and Sphingomonadales (Table S6, Supporting Information). Dominant CAG ASVs poorly recovered by the catalog (< 90% amino acid similarity) were the Campylobacterales, the Nitrosomonadales (ASv9cb21 Sulfuriferula sp.), the Unknown Desulfovibrionaceae ASVef8b2, the Unknown Nitrospiraceae ASV5c36e, and two Unknown Deltaproteobacteria (ASVa6c8b and ASVa14c3) (Table S6, Supporting Information). With marginally better representation, the Unknown Bacterial Groups A, B, and C had at least one member within each group with 92%–95% amino acid similarity to an NCD within the catalog (Table S6, Supporting Information).

Overall, the above results suggest that while some dominant diazotrophs in the CAG were characterized by ASVs elsewhere in the western Arctic and coastal NW Atlantic (chiefly, Desulfobacterales and Desulfuromonadales), the majority (> 50%) were poorly characterized at the ASV-level by reference sequences earmarked as the most important NCDs known at lower-latitudes or elsewhere in the ocean (Delmont et al. 2021, Turk-Kubo et al. 2022). This finding is further underscored by the fact that several of the groups identified herein can currently only be identified at higher levels of classification based on comparisons to known nifH sequences (Table S2, Supporting Information) and current nifH taxonomic methods [for e.g. using the tree of Kapili and Dekas (2021) as was completed herein].

Discussion

Taxonomy and predicted ecology of NCDs within the CAG

Although the potential for lateral gene transfer and gene duplication requires that taxonomic assignments derived solely from nifH be interpreted with some caution (Zehr et al. 2003, Riemann et al. 2010), it is nevertheless still useful to consider the known ecologies for the taxa observed. In the Labrador Sea, Burkholderiales, Vitreoscilla, and Bradyrhizobium ASVs were important. The group Burkholderiales has previously been associated to the Mackenzie River and other rivers near the Beaufort Sea (Ortega-Retuerta et al. 2013, Kellogg et al. 2019), as well as sediments of Baffin Bay (Algora et al. 2015), and to the dark ocean more generally (Orcutt et al. 2011). The dominant Vitreoscilla recovered were most like Vitreoscilla filiformis, an aerobic/microaerophilic species known to grow on a variety of carbon sources (Strohl et al. 1986). Meanwhile, Bradyrhizobium is a genus that has been found at lower latitude (e.g. east Indian Ocean; Wu et al. 2021), as well as in the coastal NWA (Tang et al. 2019). Although Bradyrhizobium are mainly described from their symbiotic association with legumes, free-living members with genes for oxygen tolerance are becoming more appreciated in the marine environment (Tao et al. 2021).

In Baffin Bay, S. stutzeri (previously P. stutzeri) was important. Stutzerimonas stutzeri is a model organism for denitrification and marine cultures of this organism have been isolated from suboxic waters in the Baltic Sea (Lalucat et al. 2006, Bentzon-Tilia et al. 2014); the species S. stutzeri is also known to fix nitrogen and can be found terrestrially (Desnoues et al. 2003, Zhang et al. 2019). Others have shown that benthic denitrification occurs in bottom waters of the Baffin Bay (Lehmann et al. 2019). While it is tempting to connect the presence of S. stutzeri to denitrification processes in the Baffin Bay, this microbe occurred mid-depth and just above where benthic denitrification is expected (Lehmann et al. 2019). For S. stutzeri, more research will be required to fully resolve any links to other nitrogen-related pathways or oxygen preferences in the CAG and Bedford Basin, and to also confirm whether the nifH ASV detected in the CAG and Bedford Basin are indeed from the same organism and not a pattern resultant from lateral gene transfer between different taxa (Zehr et al. 2003). Similar work will also be needed to fully resolve why members of the Burkholderiales were detected in conjunction with high nitrite levels during our study.

In the CAA, several groups were important including Vibrio, Desulfuromonadales, Desulfobacterales, Sulfuricurvum, and Arcobacter. The genus Vibrio is geographically widespread and has been previously shown to increase proportionally to rising temperature in the North Sea (Vezzulli et al. 2012). Members of this genus (e.g. Vibrio natrigens) are known to be fast growing and capable of fixing nitrogen under anaerobic conditions (Hoff et al. 2020). Desulfuromonadales and Desulfobulbaceae are iron and sulfur reducers known to occur in Arctic surface sediments (Jabir et al. 2021) and were the only two groups with identical ASVs recovered in the dominant diazotrophs reported by Shiozaki et al. (2018), where they were likewise associated with sediments. The occurrence of these groups toward the bottom of the Bedford Basin is consistent with a sediment-related origin, thus it may be that these diazotrophs were possibly present in all three study regions (western Arctic, eastern Arctic, and coastal NW Atlantic) due to their shared association with sediments. Considering that the bottom waters of the Bedford Basin typically range from oxic to hypoxic throughout the seasonal cycle, with hypoxia to near-anoxia occurring in the late summer to fall prior to winter mixing or any oxygenating events caused by shelf water intrusions (Subhadeep et al. 2023), it is also possible that the Desulfobulbaceae and Desulfuromonadales, which are generally considered obligate anaerobes may require specialized growth conditions within the more aerobic water column (Kuever et al. 2005a & 2005b). For example, it is known that particles in the water column can provide areas where oxygen may be drawn down further, hence creating a less aerobic environment for diazotrophs (Pedersen et al. 2018, Hallstrøm et al. 2022). It is feasible that resuspended sediments may provide one source for such particles within the water column. A final group, Sulfuricurvum and Arcobacter (the Campylobacterales) are typical members of the dark ocean (Orcutt et al. 2011) and have been previously reported in Arctic sediments (Jabir et al. 2021) and under Arctic sea ice (Fernández-Méndez et al. 2016), suggesting their prevalence within the Arctic marine environment. Note that the recently discovered ultrasmall Arcobacter (Karlusich et al. 2021) was not similar to the dominant Arcobacter reported herein (∼82% nucleotide similarity; ∼91% amino acid similarity; Turk-Kubo et al. 2022).

Considering the above lifestyle preferences, future studies should: (i) consider more closely how dissolved oxygen may play a role in diazotroph diversity within the CAG (even if this role is indirect via temperature’s influence on dissolved oxygen), and (ii) aim to collect more vertical profiles that encompass multiple layers of the water column and sediments to further elucidate how resuspended sediments shape diazotroph diversity in the Arctic Ocean. Furthermore, the topic of resuspended sediments and vertical ocean profiles is also closely related to how diazotrophs are structured by ocean depth either through export from the photic zone (Bonnet et al. 2023) or through their direct occurrence within deeper aphotic zones (Hewson et al. 2007, Bonnet et al. 2013, Benavides et al. 2018, Karlusich et al. 2021).

Previous work has shown little overlap between diazotrophic communities in the Arctic and elsewhere (Farnelid et al. 2011, Blais et al. 2012), a conclusion supported by our findings as the majority of the dominant ASVs identified herein were poorly represented at the ASV-level by other known NCD nifH reference sequences (Turk-Kubo et al. 2022). While on a broader scale the finding that most diazotrophs in the Arctic Ocean belong to proteobacterial members (Karlusich et al. 2021) of the phylogenetic clusters I and III agrees with our results (Farnelid et al. 2011, Blais et al. 2012, Díez et al. 2012, Fernández-Méndez et al. 2016, Shiozaki et al. 2018), on a finer scale (at the genus-level) our results further indicate that there are consistent genera being recovered via nifH amplicon sequencing from within the marine Arctic sector. To summarize, overlaps between our study and other marine-derived nifH signatures previously found in the Arctic included Bradyrhizobium, Vibrio, Desulfuromonas, Sulfuricurvum, Arcobacter, and Desulfobulbaceae (Blais et al. 2012, Díez et al. 2012, Fernández-Méndez et al. 2016, Salazar et al. 2019, Jabir et al. 2021).

Overall, our results present a snapshot of diazotrophic diversity in the CAG during the time of year when one would expect a putative ecological niche for diazotrophs (late summer after the phytoplankton bloom and when N:P ratios reflect nitrate limitation in the mixed-layer; Schuback et al. 2017). While lower levels of DIN in the mixed layer could directly create a niche for diazotrophs, it is also possible that the primary productivity generated by the spring bloom may have supported NCD groups deeper in the water column (for e.g. via DOC/DOM; Bombar et al. 2016, Fonseca-Batista et al. 2019). Indeed, elevated florescence and TP-P values at surface depths found during the time of sampling would indicate a source of energy from primary production (Schuback et al. 2017, Colombo et al. 2022). Although additional temporal sampling is needed to confirm if the dominant ASVs that we found are annually reoccurring in the CAG, our results nevertheless contribute to our understanding of diazotrophic diversity in the CAG given the sparsity of previous sampling in this region [reviewed by von Friesen and Riemann (2020)]. Our results especially advance the spatial coverage for nifH community sequencing within the CAG, as previous work has mainly employed different methods, including nifH clone libraries in Baffin Bay and the CAA (Blais et al. 2012), pyrosequencing of one sample from Baffin Bay (Farnelid et al. 2011), and a few metagenomics samples (Salazar et al. 2019). To our knowledge, the Labrador Sea has never been investigated via nifH amplicon community sequencing.

UCYN-A and other cyanobacterial diazotrophs within the CAG

Our few detections of cyanobacterial diazotrophs in the CAG extend previous observations of their presence in the Arctic. Although others have reported on Nodularia-like nifH within sea ice (Fernández-Méndez et al. 2016) and Trichodesmium in Arctic sea ice brine (Díez et al. 2012), with so few observations it still remains difficult to decipher the environmental parameters and mechanisms shaping the occurrence of cyanobacterial diazotrophs within the Arctic sector. Considering that we detected ASVs likely belonging to Trichodesmium, Chroococcidiopsis, and Pseudanabaena below 200 m, it is possible these nifH signatures may have been transported to such depths. This is consistent with the hypothesis that transport/advection from lower latitudes may be a main source for marine cyanobacterial diazotrophs within the Arctic Ocean (von Friesen and Riemann 2020, Zehr and Capone 2021a). In the Labrador Sea deep convection occurs during the winter and this could provide one way for surface phytoplankton cells to be transported to deeper depths (Koelling et al. 2022). Also of relevance is that members of Chroococcidiopsis are known to exist in colder environments (Caiola et al. 1996).

Like the other cyanobacteria detected, UCYN-A were infrequent at the time of our study. Although it is possible that UCYN-A may be more abundant at other stations at different times, plausible reasons for higher UCYN-A abundances at CAA7 include this station’s coastal status and its proximity to the western Arctic. Selden et al. (2022) recently showed that UCYN-A growth could be stimulated by coastal upwelling in the nearby Beaufort Sea. If coastal upwelling is a driving factor for UCYN-A in the Arctic, then this would be more evident at the coastal CAA sites versus the more pelagic sites in Baffin Bay and the Labrador Sea. The status of CAA7 as one of the two most westward sites sampled within the CAA (CAA6 being the other) also suggests that advection from the western Arctic may have been responsible for the higher UCYN-A abundances at this station. Although CAA6 and CAA7 are both closer to the Beaufort Sea, published work tracking dissolved lead from the Pacific/Canadian Basin versus the Atlantic (using datasets collected during the same GN02 expedition presented herein) point to CAA7 as being more influenced by Canadian Basin waters than by Baffin Bay and Atlantic waters (Colombo et al. 2019). Isotopic nitrate tracer studies from the GN02 cruise also showed that transport occurred from west to east on the southern side of Parry Channel within the CAA (Lehmann et al. 2019, 2022, Sherwood et al. 2021). Hence, if advection was driving the presence of UCYN-A at CAA7, it would likely be because it was originating from the western Arctic where it is known to occur (Harding et al. 2018, Shiozaki et al. 2018). Although our findings do not suggest a high abundance for UCYN-A within the CAG at the time sampled, they nevertheless provide important spatial data that can be used to further refine models of global diazotroph distributions where polar data are critically needed (Tang and Cassar 2019).

Biogeographic division of diazotrophs between the Labrador Sea and Baffin Bay/CAA

We detected a major division between the diazotrophic communities found in the Labrador Sea versus those found at more northern sites within Baffin Bay and the CAA at the time of sampling, with contemporaneous oceanographic data indicating that samples from the Labrador Sea were mainly distinguished by their higher temperatures. Although we cannot discount that the Labrador Sea may have been selective to ASVs preferring slightly higher temperatures, it is important to note that samples across the CAG during the time of sampling were all relatively cold at 0–6°C. Consequently, temperature differences may also be reflective of other ocean processes more strongly influencing the separation of the two communities. In support of the latter scenario is the fact that while both areas are considered polar biomes, the Baffin Bay/CAA and the Labrador Sea represent separate biogeochemical provinces corresponding to the Boreal Polar province (BPLR) versus the Atlantic Arctic province (ARCT), respectively (Longhurst 1995, Reygondeau et al. 2013). BPLR includes the Arctic Ocean and nearby areas (e.g. CAA) that are influenced by its surface waters, while ARCT includes the Labrador Sea, Irminger Sea, and western Greenland Sea (Longhurst 1995). In the CAG, this separation can be especially distinguished by the Davis Strait between the Labrador Sea and Baffin Bay where the shallowing bathymetry influences major currents (Belkin et al. 2009); in this area (Fig. 1B) the West Greenland and Irminger Currents move west/southwest off of the western coast of Greenland, encircling the Labrador Sea and contributing to differences between the two regions (more northern Baffin Bay versus Labrador Sea; Wu et al. 2012, Lacour et al. 2015, Lehmann et al. 2019). Furthermore, Baffin Bay had dissimilarity to the CAA. Unlike the CAA, the water column of Baffin Bay is structured such that its surface, mid-layer (for e.g. at 700 m), and bottom waters, are more so influenced by Arctic waters, Atlantic waters, and isolated waters, respectively (Lehmann et al. 2019, Colombo et al. 2020). At present, the influence of currents moving through the Labrador Sea (and consequently the effects this has on the water masses present) has been studied with a greater emphasis on phytoplankton [for e.g. see Fragoso et al. (2016) and Lacour et al. (2015)]. Our results would argue that in future similar attention should be given to how these physical features may also shape the diazotrophic communities within the region. In this regard, more strategic sampling over time, as well as from the Labrador and West Greenland Currents (versus the central Labrador Sea as was targeted herein), would help place further spatiotemporal limits on the NCD division that we have identified for the CAG.

Conclusions

Given the urgency of climate-driven changes occurring within the Arctic (for e.g. decreasing sea ice; Fernández-Gómez et al. 2018, Zehr and Capone 2020) and the potential for marine diazotrophs to contribute to Arctic nitrogen fixation (Sipler et al. 2017), it is now pressing that we establish detailed baseline data for marine diazotrophs in the Arctic. Our findings help address this need and have shown that in this area of the Arctic there is a previously unrecognized biogeographic separation that can occur between diazotrophs within the Labrador Sea and those in Baffin Bay/the CAA—a separation that appears reflective of the physico-chemical features that shape the water masses within this region of ocean (Longhurst 1995, Reygondeau et al. 2013). Despite low-level detection, our results also provide initial data for UCYN-A within the CAG, thus confirming its broader presence within the Arctic on the eastern side of the CAA. Although our results cannot directly demonstrate that nitrogen-fixation was occurring, they nevertheless provide a baseline for future studies that may seek to reconcile nitrogen fixation rates with historical diazotroph occurrences and relative abundances within the region. We also provided a detailed overview of the dominant diazotrophs that were detected during our study of the CAG. With some exception, the majority of the dominant nifH ASVs we observed in the CAG were poorly represented at the ASV-level in the known reference sequences for major NCDs groups globally and at lower latitudes (Delmont et al. 2021, Turk-Kubo et al. 2022). Taken together, our findings highlight that the CAG may be a prime location for future studies on NCDs and lesser-known taxonomic groups that fall within this fraction of the ocean's microbiome.

Authors’ contributions

Brent M. Robicheau (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft), Jennifer Tolman (Data curation, Investigation, Methodology, Project administration, Validation, Writing – review & editing), Sonja Rose (Data curation, Investigation, Validation, Writing – review & editing), Dhwani Desai (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Writing – review & editing), and Julie LaRoche (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Writing – review & editing).

Conflict of interest

The authors declare that there is no conflict of interest.

Funding

This work was supported by the Ocean Frontier Institute (through funding awarded to J.L.R.); a Discovery grant (awarded to J.L.R.) and a CGS-doctoral award (granted to B.M.R.) from the Natural Sciences and Engineering Research Council of Canada; a Canada Foundation for Innovation grant (to J.L.R.); and Killam Predoctoral and NS Graduate scholarships awarded to B.M.R.