Need for focus on microbial species following ice melt and changing freshwater regimes in a Janus Arctic Gateway

Oceanic gateways are sensitive to climate driven processes. By connecting oceans, they have a global influence on marine biological production and biogeochemical cycles. The furthest north of these gateways is Nares Strait at the top of the North Water between Greenland and Ellesmere Island (Canada). This gateway is globally beneficial, first by supporting high local mammal and bird populations and second with the outflow of phosphate-rich Arctic waters fueling the North Atlantic spring bloom. Both sides of the North Water are hydrologically distinct with counter currents that make this Arctic portal a Janus gateway, after Janus, the Roman god of duality. We examined oceanographic properties and differences in phytoplankton and other protist communities from the eastern and western sides of the North Water (latitude 76.5°N) and found that species differed markedly due to salinity stratification regimes and local hydrography. Typical Arctic communities were associated with south flowing currents along the Canadian side, while potentially noxious Pseudo-nitzschia spp. were dominant on the Greenland side and associated with greater surface freshening from ice melt. This susceptibility of the Greenland side to Pseudo-nitzschia spp. blooms suggest that monitoring species responses to climate mediated changes is needed.


Results
Oceanographic characteristics. At the time of sampling in mid-august, the temperature-salinity (TS) profiles showed fresher surface waters and more saline deep waters on the Greenland side compared to the Canadian side (Fig. 1B). On the Canadian side, there was a slight increase in salinity between the surface and the weak subsurface chlorophyll maximum (SCM) (31.45 to 32.35). This was in contrast to fresher surface waters (salinity of 30.51) and a sharp halocline on the Greenland side with salinity of 33.32 at the SCM (Table 1, Fig. 1B). The surface chlorophyll (Chl a) concentrations from satellite data were similar on the two sides and both sides had been ice free since the end of June ( Supplementary Fig. S1). Total Chl a from samples collected at the surface were also similar on the two sides (Supplementary Fig. S2; ANOVA, p-value = 0.304). At the bottom of the PML starting at 20 m depth, the two sides differed in nearly all respects. On the Canadian side, in situ chlorophyll fluorescence (Fig. 2, upper pannel) and total Chl a concentrations ( Supplementary Fig. S2A) changed little down the water  (Table 1 and Supplementary Table S2) compared to the Canadian side. For nutrients, phosphate concentrations were higher down the water column on the Canadian side. Nitrate concentrations were depleted to at least 20 m on both sides and increased with depth below the PML with higher concentrations on the Greenland side. There was a nitrite maximum detected on the Greenland side at 40 m (Fig. 2).
Our repeated sampling indicated that nitrate and silicate concentrations at 20 m were similar on both sides of the North Water (Supplementary Table S2). However, phosphate concentrations and fluorescent colored dissolved organic matter (fCDOM) showed significant differences between the two sides at 20 m. On the Greenland side, phosphate (avg. 0.29 µM) and fCDOM (avg. 2.42 fluorescence units) were significantly lower compared to the Canadian side (phosphate avg. 0.49 µM and fCDOM avg. 3.08 units) ( Table 1, Supplementary Table S2).
Microbial communities. For T 1 and T 4 , both DNA and RNA templates were sequenced. The relative proportions of taxa inferred from the DNA (rDNA, Supplementary Fig. S3) was similar to proportions of the same taxa from the samples inferred from the RNA (rRNA, see bars for T 1 and T 4 in Fig. 3). A general exception was a higher proportions of dinoflagellates in the rDNA ( Supplementary Fig. S3, Supplementary Table S3). Both templates showed the marked difference in the microbial eukaryote communities on the two sides of the North Water (Fig. 3, Supplementary Fig. S3). Unweighted clustering of the rRNA dominants Operational Taxonmic Units (OTUs with >100 read occurrences) also clearly separated communities on the two sides ( Supplementary  Fig. S4). Environmental variables (Supplementary Table S2) that were significantly correlated with Principle Coordinates Analysis (PCoA) community values were used for the Redundancy Analysis (RDA). The Greenland and Canadian sides clearly separated along the RDA1 axis explaining 20.9% of the total variance (Fig. 4). Higher phosphate, fCDOM and temperature were associated with the Canadian communities, with nitrite, DO and salinity associated with the Greenland communities. Analysis of variance (ANOVA) repeated measures and Kruskal-Wallis tests supported the RDA results (Supplementary Table S2).
Over the ca. 24 hours of 20 m repeated sampling, the Canadian side communities varied substantially, with the T 1 community predominantly the small diatom, Chaetoceros aff. gelidus (up to 72% of reads at T 1 ). Chaetoceros had largely disappeared 4 hours later; replaced by Phaeocystis, which accounted for 43% of the small fraction reads at T 2 . Subsequently we detected a mix of small photosynthetic flagellates, which co-occurred with another Hylochaete Chaetoceros species; C. neogracile (Fig. 3, Supplementary Table S4A). The communities on the Greenland side at 20 m were less variable (Supplementary Table S5), with all samples containing the potentially toxic pennate diatom Pseudo-nitzschia (up to 52% of rRNA reads at T 1 ). The two dominant diatom genera Chaetoceros and Pseudo-nitzschia were genetically diverse with multiple OTUs found on both the Canadian and Greenland sides ( Supplementary Fig. S5). The presence of both genera was confirmed microscopically (Supplementary Fig. S6 plates a,b for Chaetoceros, and Fig. S6 plates g,h,i,j for Pseudo-nitzschia).
Alveolates represented the second most abundant major group from rRNA on both sides of the North Water. Ciliates made up 28% of the total reads on the Canadian side and 30% of the total reads on the Greenland side (Fig. 3 unidentified Litostomatea detected on both sides. Urotricha was common in the large fraction on the Canadian side but less on the Greenland side. In contrast, the genus Monodinium was more abundant on the Greenland compared to the Canadian side (Supplementary Table S4). Core dinoflagellates accounted for ca. 30% of total reads on the Canadian side and 21% of total reads on the Greenland side. For both sides, unclassified dinoflagellates accounted for 12% to 18% of reads (Supplementary Table S4). The most common dinoflagellate OTUs on the two sides were Gymnodiniales. At the species level Gyrodinium spirale was common on the Canadian side and Torodinium robustum, which was also identified microscopically (Supplementary Fig. S6 plates c,d) was common on both sides. The majority of dinoflagellate taxa were morphologically diverse gymnodinoids and gyrodinoids between 6-50 µm and not identified further using microscopy. From HTS, Syndiniales were predominantly in the Duboscquellidae in Group I marine alveolates (MALV I) 27 and the still poorly known environmental MALV III, associated with the environmental clone OLI11005 28 (GenBank AJ402349). The relative abundances of smaller flagellated taxa differed between the two sides. For example, several haptophyte OTUs were more common on the Canadian side with Phaeocystis accounting for 8.8% of reads in the small size fraction, and Chrysochromulina with 1.2% in the large fraction (Fig. 3, Supplementary Table S4). The Arctic Micromonas CCMP 2099 was relatively more abundant on the Canadian side with 6.1% of the small fraction reads, compared to the Greenland side with 3.8% of the reads (Fig. 3 Fig. S6 plates i,j), accounted for nearly 3% of small fraction reads on the Greenland but only 0.1% on the Canadian side. Two heterotrophic flagellate taxa; Picozoa and group 2 uncultivated marine stramenopiles (MAST 2), were common on the Greenland side with each accounting for 1.9% of reads in the small fraction. There were marked differences in the dominant chrysophytes on the two sides ( Fig. 3 and Supplementary Table S4) with Dinobryon balticum identified from the Canadian side ( Supplementary  Fig. S6, plates e,f) and D. belgica on the Greenland side ( Supplementary Fig. S6 plates k,l).

Discussion
Within the PML the microbial eukaryotic communities from the two sides of the North Water differed taxonomically, with Pseudo-nitzschia dominated communities persistently found on the Greenland side, and more mixed communities on the Canadian side. The Canadian communities changed several times over the nearly 24 hours of sampling (Fig. 3). The high similarity among samples collected on the Greenland side (Supplementary Table S5) suggested that a single sample could have been used to represent the community in mid-August. In contrast, no single typical community was detected on the Canadian side. The active communities detected using rRNA,  were consistent with the physical oceanography of the respective sides. Salinity determines Arctic Ocean physical oceanographic structure 29 and on the Greenland side, Atlantic waters contribute to the higher salinity and sensible heat at depth 11 , which leads to shoaling of the upper mixed layer 30 . During our study the fresh surface waters on the Greenland side originated from first year and multi-year sea ice melt and also likely from iceberg and glacier melt from around Greenland 16,31 . A high external input of surface fresh water originating from the Greenland Ice Sheet (GrIS) was previously suggested as an explanation for anomalous values for alkalinity and delta O 18 data from the same station on the same cruise 16 . The Canadian side reflected the more general freshening of the Arctic Ocean that is due to multi-year sea ice melt 32 . This overall freshening on the Canadian side could be seen in the weak halocline and lower average water column salinity (Fig. 1B, Fig. 2). The higher phosphate and fCDOM concentrations ( Table 1, Table S2), on the Canadian side were also characteristic of Arctic Water 33 , in keeping with the Nares Strait Gateway directly contributing excess Arctic Ocean phosphate to the Western Labrador Sea and North Atlantic 7,8 .
At the surface, the two sides of Northern Baffin Bay were similar in terms of ice cover and surface bloom phenology ( Supplementary Fig. S1), with surface chlorophyll concentrations on both sides within the long-term norm for mid-August 18 . However, below the surface, along with the marked differences in physical structure, differences in biological indicators, such as the prominent SCM and overall higher Chl a concentrations at depth on the Greenland side, were striking. The formation of the Arctic SCM below the PML depends on the availability of both light and nutrients 34 and the high nitrate concentrations below 30 m on the Canadian side, suggest that phytoplankton may have been light limited at that depth. If so, light limitation could be explained by water column instability associated with mixing below the halocline, moving phytoplankton out of the euphotic zone to depths below irradiance levels sufficient for net growth. Alternatively, or in addition, the lack of formation of a prominent SCM on the Canadian side may have been due to the hydrologic complexity of the region, where multiple water-masses with similar densities and different recent histories could converge and interleave 35 . Such interleaving would be consistent with temperature-salinity (TS) profiles (Fig. 1B) and rapid community changes (Fig. 3) at the single depth sampled. In contrast, there was a prominent SCM on the Greenland side, which was associated with a well stratified water column and presumably continuous nutrient supply to the bottom of the euphotic zone beneath the sharp halocline. Saline Atlantic water below the fresher PML is a characteristic of the Greenland side 30 , and such ideal conditions for SCM formation 36 (Fig. 2, Supplementary Fig. S2) would be expected, suggesting annual formation of an SCM on the Greenland side.
The RDA (Fig. 4) supported the notion that the distinct communities on the two sides were associated with the physical oceanography of the region 17 . Higher phosphate and fCDOM concentrations, characteristic of Arctic surface waters 23 were associated with samples collected from the Canadian side. In contrast, Greenland side communities were associated with greater DO concentrations consistent with local upward diffusion of DO from the photosynthetic activity at the much more robust SCM.
The active community from the RNA template (Fig. 3, Supplementary Table S4), documented by microscopy ( Supplementary Fig. S6), revealed the presence of Hyalochaete Chaetoceros predominantly at T 1 on the Canadian side, when the colder temperatures and the high fCDOM indicated the highest contribution of unmodified Arctic waters 33 . These small diatoms were earlier reported (identified as C. socialis) as ubiquitous in the summer phytoplankton in the North Water 24 . The abundant Chaetoceros populations suggested that the 20 m sample at T 1 may have originated nearer the surface under high irradiances 37 . At T 2 on the Canadian side, the community changed, with a high proportion of reads from Phaeocystis, which is another common polar species. For example, Phaeocystis is reported from the Ross Sea, Antarctica 38 . Several studies suggest that Phaeocystis rapidly responds to increased light and nutrient inputs 39,40 consistent with changing conditions on the Canadian side of the Nares Strait gateway. Another small flagellate Micromonas polaris, which was recently described as an endemic arctic species 41 , and almost universally reported from Arctic 18 S rRNA surveys as Micromonas CCMP 2099 42,43 , was retrieved on both sides but with highest relative abundance after T 5 on the Canadian side (ca 14% in the small fraction; Fig. 3). Other typical Arctic flora on the Canadian side included chrysophytes such as Dinobryon balticum, which contributes to carbon flux in cold waters 44 . In contrast, the morphospecies Dinobryon belgica, which has smaller colonies ( Supplementary Fig. S6 plate e versus l) was noted on the Greenland side. The 30-fold difference in the abundance of Pyramimonas on the Greenland versus Canadian side may have been facilitated by the low salinity surface waters. For example, Daugbjerg 45 indicated that some Pyramimonas species are able to tolerate a wide range of salinity and Pyramimonas was also reported as a dominant nanoflagellate in surface waters with high freshwater input from the Mackenzie River and surface ice melt in the Beaufort Sea 43 . In addition, Pyramimonas is associated with meltwater pools on sea ice 46 and found during the winter-spring transition in the fjord and ice influenced Disko Bay in Greenland south of our study region 47 .
The molecular data (Fig. 3, Supplementary Fig. S3, Supplementary Table S3, Supplementary Table S4) verified using microscopy ( Supplementary Fig. S6) showed that a major difference between the two sides was the overwhelming dominance of Pseudo-nitzschia spp. on the Greenland side, accounting up to 52% of reads, but <0.1% of reads on the Canadian side. The presence of Pseudo-nitzschia is of concern, since the genus can produce the neurological toxin, domoic acid (DA) 48 . In contrast, more typical summer Arctic communities 49 were found on Canadian side, with variability consistent with changes in water masses with different recent histories and species 35 . Pseudo-nitzschia has long been present in this region, but high concentrations have not been previously reported. During the 1998 North Water Polynya (NOW) study, Lovejoy et al. 22 observed a diverse diatom community on the Greenland side in July, including Pseudo-nitzschia cf. seriata. A Pseudo-nitzschia strain (CCMP 2093) now classified as P. arctica 50 was isolated from the North Water (78°35.87′N, 74°29.53′W) in 1998 and other Pseudo-nitzschia species have been isolated from Greenland fjords south of our study region 51 . Pseudo-nitzschia have wide salinity tolerances, are able to use a variety of nitrogen sources, and are favored under conditions of high dissolved Fe concentrations 52 . The nitrite peak at 40 m on the Greenland side suggests high biological activity and an additional nitrogen source for Pseudo-nitzschia 53 . Overall, we suggest that GrIS inflows could favor blooms of Pseudo-nitzschia spp. and potentially DA production 48 . DA producing Pseudo-nitzschia blooms are detrimental to marine ecosystems since DA causes memory impairment and is harmful to marine mammals 54 . To date, DA has not been reported from the North Water, but there have been few or no DA surveys in the region, to our knowledge. Locally isolated Pseudo-nitzschia strains from Disko Bay (Greenland) produce DA when fed to Arctic copepodites 51 and Pseudo-nitzschia OTUs here clustered with species such as P. brasiliana ( Supplementary  Fig. S5A), which is known to produce DA 52 , suggesting the potential of DA production. Recently, large-scale phytoplankton blooms in the Labrador Sea were reported to coincide with freshwater discharge from glaciers and meltwater from the GrIS, however information on species was lacking 55 , and given that the bloom was attributed to increased Fe and organic nutrients, it would not be surprising if Pseudo-nitzschia was present, and such phenomena warrant investigation.
In sum, although classified as a single eco-region for modeling studies 56 , The Greenland side of the gateway was more saline below PML, but fresher on the surface (Fig. 1B), consistent with recent increases in melt water from the Greenland Ice Sheet (GrIS) 57 . Freshwater from the GrIS, glacial melt and multi-year sea-ice melt enter the North Water via Nares Strait as surface water along the Greenland side 17 . Additional GrIS melt from Northeast Greenland 58 could also contribute to surface freshwater on the Greenland side of the North Water since it flows southward along the east coast of Greenland (via the East Greenland Current) and at the southern tip of Greenland, becomes entrained into the northward coastal flow (West Greenland Current) 57 . Any increase in GrIS meltwater and associated high iron (Fe) and other micronutrient concentrations 59 has implications for the biogeochemistry of Baffin Bay 55 and would create conditions that could favor blooms of Pseudo-nitzschia. This in turn, suggests a threat to the historic role of the North Water in supporting marine bird and mammal populations including endangered Greenland populations of Bowhead whales, which migrate to this latitude in summer and graze on copepods 60 . Finally, given the increasing cyclonic circulation in Northern Baffin Bay, low-phosphate waters and potential toxic species might flow south to the North Atlantic along the Baffin Bay meridional ridge 17 , making this a true Janus gateway. This work highlights the need to consider drivers that select for species and the biological diversity of gateways, if we are to fully understand impacts of climate change in the Arctic and elsewhere.

Materials and Methods
Satellite-derived products. Satellite-derived level-3 data sets of Chl a concentration (mg m −3 ) and particulate back-scattering coefficients at 443 nm (b bp , m −1 ) (Supplementary Fig. S1) were obtained from the European Space Agency's GlobColour project (http://www.globcolour.info). Eight-day composite Chl a concentration and b bp coefficients were determined, respectively using standard Case 1 water algorithms 61 and the semi-analytical Garver Siegel Maritorena (GSM) merging algorithm 62 . The sea-ice concentration was derived from the Advanced Microwave Scanning Radiometer -Earth Observing System (AMSR-E) sensor, and made available by National Snow and Ice Data Center (NSIDC; https://nsidc.org). Actual sea ice conditions northward and just prior to sampling were further verified by inspecting a range of additional satellite products (https://worldview.earthdata. nasa.gov/). Field sampling. Oceanographic data was collected aboard the CCGS Amundsen using a rosette system equipped with a conductivity, temperature, depth (CTD) profiler (Sea-Bird SBE-911 CTD), relative nitrate (In-Situ Ultraviolet Spectrometer, ISUS, Satlantic), oxygen (Seabird SBE-43), chlorophyll fluorescence (Seapoint), fluorescent colored dissolved organic matter (fCDOM; Wetlabs ECO) and photosynthetically available radiation (PAR, 400-700 nm; Biospherical Instruments QDP2300) sensors. The oxygen sensor was calibrated onboard against Winkler titrations 36 .
The ship followed a buoy with a drogue suspended to 20 m, with the goal of following the bottom of the PML and sampled approximately every 4 hours beginning at solar time 05:15 for Stn 101 (Canadian side) and 06:00 for Stn 115 (Greenland side), ending after 23 hours at and 22.5 hours, respectively (Supplementary Table S1). At Stn 101, the drift started (T 1 ) at lat 76°23.242′N, long 077°23.412′ W, and ended (T 7 ) at lat 76°17.372′N, long 077°46.264′W. The Stn 115 drift started (T 1 ) at lat 76°20.297′N, long 071°11.514′W and ended at (T 7 ) lat 76°29.858′N, long 071°26.615′W (Fig. 1A and Supplementary Table S1). Water samples were collected on the upcast from 12-L Niskin-type bottles mounted on the rosette. Nutrient samples were collected every 10 m from the surface to 60 m. Nutrients were analyzed on board within 4 h after collection as using a Bran-Luebbe 3 autoanalyzer 63 . Optical depths were derived from vertical PAR profiles taken from the shadow-free side of the ship using a PNF-300 radiometer (Biospherical Instruments). At time T 1 for Stn 101 and time T 3 for Stn 115, water samples for duplicate Chl a were collected at seven optical depths (100, 50, 30, 15, 5, 1, and 0.2% of surface irradiance), the SCM determined from the downward cast of the CTD, and at 80 m and 100 m. Chl a samples were also collected and analyzed from 20 m along with nucleic acids (see below) during the drift. Samples for total Chl a were filtered onto 25 mm Whatman GF/F filters (TChl a; >0.7 µm) and onto 5 µm pore size polycarbonate (PC) membrane filters (Nuclepore ™ ) to estimate the large fraction (L-Chl a; >5 µm) and analyzed on board as in Parsons et al. 64 . Concentrations in the smaller fraction (S-Chl a; 0.7-5 μm) were from subtraction. For nucleic acids, five independent samples from 20 m were collected from each the Canadian and Greenland sides (Supplementary Table S1). Six L of sample water prefiltered through a 50 μm nylon mesh, and then sequentially filteredthrough a 47-mm diameter 3 μm pore size PC membrane filter, and a 0.2 μm Sterivex ™ Unit (Millipore). Material on the filters was preserved in RNAlater ™ (ThermoFisher) then frozen at −80 °C. The different size fractionation protocols for Chl a and nucleic acids was due to logistic constraints aboard the ship, and small and large designations should be considered as indicative only.
Whole water samples for fluorescence microscopy were collected at 20 m at T 1 for Stn 101 and T 2 for Stn 115. Single aliquots of 48 mL were preserved and filtered onto black 0.8 µm PC filters, stained with 4,6-diamidino-2-phenylindole (DAPI) following 65 . The slides were inspected at 1000X magnification using an Olympus IX71 microscope equipped with UV and blue excitation filter blocks. Images were captured with the integrated QImaging Retiga 2000R CCD Camera and QCapture software version 2.9.11. Laboratory procedures. DNA and RNA were extracted from the same filters using the All-Prep DNA/ RNA Minikit (Qiagen). DNA was used to verify the overall community including dead and dormant cells at two time points 12 hours apart on either side (T 1 and T 4 ). RNA was converted to cDNA using the High Capacity Reverse Transcription Kit (Applied Biosystems). The V4 region of 18 S rRNA gene and rRNA (from cDNA) was amplified using the eukaryote specific primers E572 (forward) and E1009 (reverse), see Comeau et al. 66 , coupled with a MiSeq© specific linking primer. To decrease potential PCR bias, 1, 5 and 10-fold diluted template was used for PCRs for each sample. The PCR products of the three dilutions were pooled together and purified using the Axygen ® PCR cleanup kit (Axygen) and then quantified spectrophotometrically with the Nanodrop 1000 ™ (ThermoFisher Scientific). Unique pairs of barcodes (tags) were added to the sample amplicons using the TruSeq ® and Nextera ® (both Illumina) barcode sets in a nested PCR as described in Comeau et al. 67 . Equimolar concentrations of the sample barcoded amplicons were sequenced on the Illumina MiSeq at the Plate-forme d' Analyses Génomiques (IBIS, Université Laval, Québec, QC, Canada). All reads are deposited in NCBI GenBank Sequence Read Archive (SRA) under the BioProject number PRJNA383398 (GenBank: SRX2745611 to SRX2745642).
Data and statistical analysis. Paired end reads were processed with UPARSE 68 , with microbial eukaryote operational taxonomic units OTUs retained following Wu et al. 69 . Based on a 99% similarity the reads clustered into 2069 OTUs. We then used an arctic specific reference database for taxonomic assignment of OTUs 70 . The resulting OTU classifications are available at https://zenodo.org/record/1205261#.WrOAMJPwb-Y and their associated reads at https://zenodo.org/record/1205255#.WrN_RJPwb-Y.
Data sets were further processed in QIIME 71 . Rarefied datasets (11,500 sequences per sample) were used as input for the UniFrac unweighted distances 72 , with beta diversity and principal coordinates (PCoA) as output. Environmental variables that were significantly correlated with PCoA community values were used for the redundancy analysis (RDA) 73 computed using the rda function in the Rstudio package 'vegan' (v2.4-1) 74 . Parameters that best explained variability in the PCoA were selected using the ordistep function in the 'vegan' package to build an optimal model (highest adjusted coefficient determination). One-way ANOVA (parametric) and Kruskal-Wallis (non-parametric) tests were carried out in PAST v3.0 75 to test for differences between the two sides of the North Water. ANOVA repeated measures tests were carried out using PAST v3.0 75 . Statistical tests of physical and nutrient data and profile plots were carried out in Sigma Plot (v11).
With the aim to focus on more abundant organisms, we selected OTUs that had a minimum of 100 occurrences in at least one rRNA sample from both sides of the North Water Clusters and heatmaps were constructed using the heatmap.2 function in gplots package in R. Rarefied rRNA read counts were first log transformed and then clustered by site (Y axis) and by OTUs (X axis) using Euclidean distances. Dendograms were generated using hierarchical clustering UPGMA.
To explore the diversity of the two dominant diatom genera in our samples, reference sequences related to Pseudo-nitzschia were retrieved from Percopo et al. 50 and those related to Chaetoceros were retrieved from Chamnansinp et al. 76 . The V4 18 S rRNA sequences were aligned separately with MAFFT (v7) 77 .