Submesoscale physicochemical dynamics directly shape bacterioplankton community structure in space and time

Submesoscale eddies and fronts are important components of oceanic mixing and energy fluxes. These phenomena occur in the surface ocean for a period of several days, on scales between a few hundred meters and few tens of kilometers. Remote sensing and modeling suggest that eddies and fronts may influence marine ecosystem dynamics, but their limited temporal and spatial scales make them challenging for observation and in situ sampling. Here, the study of a submesoscale filament in summerly Arctic waters (depth 0–400 m) revealed enhanced mixing of Polar and Atlantic water masses, resulting in a ca. 4 km wide and ca. 50 km long filament with distinct physical and biogeochemical characteristics. Compared to the surrounding waters, the filament was characterized by a distinct phytoplankton bloom, associated with depleted inorganic nutrients, elevated chlorophyll a concentrations, as well as twofold higher phyto‐ and bacterioplankton cell abundances. High‐throughput 16S rRNA gene sequencing of bacterioplankton communities revealed enrichment of typical phytoplankton bloom‐associated taxonomic groups (e.g., Flavobacteriales) inside the filament. Furthermore, linked to the strong water subduction, the vertical export of organic matter to 400 m depth inside the filament was twofold higher compared to the surrounding waters. Altogether, our results show that physical submesoscale mixing can shape distinct biogeochemical conditions and microbial communities within a few kilometers of the ocean. Hence, the role of submesoscale features in polar waters for surface ocean biodiversity and biogeochemical processes need further investigation, especially with regard to the fate of sea ice in the warming Arctic Ocean.

Microbial communities perform key functions in marine biogeochemical cycles and energy fluxes (Azam and Malfatti 2007;Buchan et al. 2014). Linking these microbial activities with their distributional patterns is required for a better understanding of organic matter transformations and associated ecological dynamics. To date, the regional biogeography of pelagic microbial communities is mostly attributed to larger-scale (> 50 km) processes of physical mixing (Fuhrman et al. 2006). However, there is increasing evidence that mesoscale eddies and frontal systems have important ecological functions, featuring elevated phytoplankton growth (Lévy et al. 2001;Mouriño 2004;Thomsen et al. 2016), unique bacterioplankton communities (Nelson et al. 2014;Baltar et al. 2016;Djurhuus et al. 2017) and high microbial activity (Baltar et al. 2007;Baltar et al. 2010;Baltar and Arístegui 2017). Thus, physical oceanic dynamics on small spatial scales presumably influence the composition and activity of bacterioplankton communities and therefore directly impact biogeochemical processes.
In recent years, observations by satellite remote sensing, autonomous profiling floats, and towed instruments revealed the ubiquity of submesoscale features in the global ocean, confirming previous hypotheses based on numerical models (Ferrari 2011;Lévy et al. 2012; Thompson et al. 2016). These oceanic motions in surface waters are due to horizontal density gradients, taking the shape of eddies and filaments with a lateral spatial range between a few hundred meters and some tens of kilometers (Sasaki et al. 2014). Their motions persist for up to several days, characterized by strong lateral currents along the filament and an increased vertical circulation (McWilliams 2016). The relatively small spatial scale and short-lived nature of such phenomena make them challenging for in situ biological observations (e.g., molecular sampling). To date, biological observations from oceanic submesoscale filaments have been focused on phyto-and zooplankton (reviewed in Lévy et al. 2018), whereas almost nothing is known about bacterioplankton communities in these filaments despite their ecological relevance (Baltar et al. 2009(Baltar et al. , 2016Baltar and Arístegui 2017).
The Fram Strait is the main gateway between the North Atlantic and Arctic Oceans. The Strait harbors of two major surface current systems: (1) the East Greenland Current flowing southwards along the Greenland shelf, transporting cold Polar water and sea ice to the North Atlantic (de Steur et al. 2009); and (2) the West Spitsbergen Current flowing northwards along the Svalbard Archipelago, transporting relatively warm and saline Atlantic water into the Arctic Ocean (Beszczynska-Möller et al. 2012;von Appen et al. 2016). The interactions between these distinct water masses result in a baroclinic instability in the central Fram Strait which facilitates subduction of Atlantic water below the Polar water Wulff et al. 2016;Wekerle et al. 2017). Investigating oceanographic dynamics in the Fram Strait is of importance considering the relevance of interacting water masses for the "Atlantification" of the Arctic Ocean (Randelhoff et al. 2018;Wang et al. 2020). In addition to physicochemical differences, the Polar and the Atlantic water masses harbor distinct bacterioplankton communities (Wilson et al. 2017;Fadeev et al. 2018;Müller et al. 2018). Thus, this region provides an excellent opportunity to investigate how bacterioplankton communities are shaped by submesoscale processes between distinct water masses and how these patterns correspond to the general microbial distribution in the Fram Strait.
Here, we explored biological responses to the physicochemical processes within a submesoscale filament in the central Fram Strait, formed as a result of a strong salinity gradient between Atlantic and Polar waters close to the ice edge (von Appen et al. 2018). Using a set of measured parameters, embedded into a high-resolution physical analysis of the filament (von Appen et al. 2018) which we briefly summarize below, we investigated how the submesoscale processes impacted nutrient distribution, phyto-and bacterioplankton community composition, and downward export of particulate organic matter (POC) from the filament. We studied freeliving (0.22-3 μm) and particle-associated (>3 μm) bacterial and archaeal communities (further addressed as bacterioplankton) inside and outside the filament, at six depths from surface (10 m) to mesopelagic (400 m) waters. This allowed comparisons of bacterioplankton communities inside and outside the filament, and evaluation of the extent the observed differences resulted from mixing between water masses or internal succession within the communities. Specifically, we tested the hypothesis that submesoscale variations in phytoplankton and bacterioplankton communities directly correspond with the observed environmental gradients within the filament and surrounding waters. Further insights into the biological dynamics associated with such ubiquitous submesoscale features are needed to upscale their impact on ecosystem productivity and functioning in polar seas.

Water sampling and metadata collection
Sampling was carried out in July 2017 during RV Polarstern expedition PS107 using 12 L Niskin bottles mounted on a CTD rosette (SBE 911 plus probe; Sea-Bird Electronics, Bellevue, WA) equipped with double temperature and conductivity sensors, a pressure sensor, altimeter, chlorophyll a fluorometer and transmissometer. Hydrographic data, including temperature and salinity are available at PANGAEA (doi: 10.1594/ PANGAEA.894189). At all stations, water samples were collected at five depths: 10, 20-30, 50, 100, 200, and 400 m (Supporting information Table S1).

Concentrations of inorganic nutrients
Inorganic nutrients were analyzed with a continuous flow autoanalyzer (Evolution III, Alliance Instruments, Salzburg, Austria) directly on board after sampling. Phosphate, silicate, nitrate, nitrite, and ammonium were measured in technical duplicates of unfiltered CTD samples simultaneously on five channels, following Grasshoff et al. (2009). Ammonium was quantified using fluorescence spectroscopy after Kérouel and Aminot (1997). All measurements were calibrated with a five nutrients standard cocktail (traceable to SRM from NIST; Merck, Darmstadt, Germany) diluted in artificial seawater (ASW), which was also used as wash-water between the samples. Each 20th run we checked our standards with Reference Material for Nutrients in Seawater (CRM 7602-a; National Meteorology Institute of Japan, Japan). Standards and methods have been proven by inter-calibration exercises (e.g., ICES and Quasimeme).  Concentrations of chlorophyll a, particulate organic carbon and biogenic silica Concentrations of chlorophyll a (chl-a) were determined from 0.5 to 2 L of seawater filtered onto glass fiber filters (GF/F; Whatman plc, Maidstone, UK) under low vacuum pressure (<200 mbar). Filters were stored at À20 C until analysis. Pigments were extracted, using plastic vials, with 10 mL of 90% acetone. The filters treated with an ultrasonic device in an ice bath for <1 min, and then further extracted in the refrigerator for 2 h. Subsequently filters were centrifuged for 10 min at 5000 rpm at 4 C prior to measurement. The concentration was determined using glass vials fluorometrically (Turner Designs, CA), together with total phaeophytin concentration after acidification (HCl, 0.1 N) based on Edler (1979) and Evans and O'Reilly (1982), respectively. The standard deviation of replicate test samples was <10%.
Particulate organic carbon (POC) was determined from 0.75 to 2 L of seawater filtered onto pre-combusted (4 h at 500 C) glass fiber filters (GF/F; Whatman plc, Maidstone, UK) at low vacuum (<200 mbar). After filtration, the filters were stored frozen at À20 C until analysis. Prior to analysis, filters were soaked in 0.1 N HCl for removal of inorganic carbon and dried at 60 C. POC concentrations were determined with a CHN Elemental Analyzer by Carlo Erba (ThermoFisher Scientific, Rockford, IL).
Particulate biogenic silica (PbSi) was determined following von Bodungen et al. (1991). Subsamples of 0.5-1 L were filtered on cellulose acetate filters (0.8 μm pore size), processed using the wet-alkaline method (with 0.2 M NaOH pretreated 12 h at 85 C in an oven), and extracted for 2 h at 85 C in a shaking water bath.

Microscopic counts of phytoplankton groups
Seawater samples from 10 and 20-30 m depths were preserved in hexamethylenetetramine-buffered formalin (final concentration 0.5%) and stored in brown glass bottles. Phytoplankton cells were counted by light microscopy. For microscopic analyses an aliquot of 50 mL was transferred to settling chambers where cells were allowed to settle for 48 h. Phytoplankton cells were identified to genus level and at least 500 cells (50-100 per major group; up to 1000 for Phaeocystis) were counted with an inverted microscope at three different magnifications using phase contrast.
Phytoplankton carbon content (PPC) was calculated by multiplying cell counts with the carbon value for individual cells following Edler (1979).

Bacterioplankton cell abundances
Cell abundances were determined by FACSCalibur flow cytometry (Becton Dickinson, Franklin Lakes, NJ). Samples were fixed with glutaraldehyde at 1% final concentration and stored at À20 C. The samples were stained with SYBRGreen I (Invitrogen, Carlsbad, CA) prior to analysis and enumerated in side scatter vs. green fluorescence cytograms. Polyscience fluorescent latex beads (Becton Dickinson, Franklin Lakes, NJ) were used to normalize the counted events to volume (Gasol and Del Giorgio 2000).

Export of particulate organic carbon
Export flux of particulate organic carbon (POC) was determined via free-drifting sediment traps at 100, 200, and 400 m at the centre (station T1) and the south-eastern edge of the filament (station T3). The samples were split into eight subsamples using a McLane Wet Sample Divider (McLane Research Laboratories, Falmouth, MA) with deviation between aliquots <5.0%. One subsample per depth and sediment trap deployment for POC analyses. Zooplankton swimmers were removed manually under a microscope before filtration on pre-combusted GF/F filters (Whatman plc, Maidstone, UK). Filters were fumed with concentrated hydrochloride acid (HCl, 37%) for 24 h and dried for 48 h at 40 C before packing into tin cartridges and analysis on a EuroEA Elemental Analyzer (precision of AE0.7 μg C or AE0.3%; HEKAtech, Wegberg, Germany). Blank filters were used to correct for contaminations.

DNA isolation and 16S rRNA gene amplicon sequencing
For assessing archaeal and bacterial community composition 2-4 L of seawater were filtered with a peristaltic pump (Masterflex; Cole Parmer, Vernon Hills, IL) through successive membrane filters of 3 μm (Nucleopore 47 mm polycarbonate; Whatman plc, Maidstone, UK) and 0.22 μm (Sterivex; Merck, Darmstadt, Germany) pore sizes to obtain particle-associated (PA) and free-living (FL) communities, respectively. All samples were stored at À20 C until DNA isolation. Genomic DNA was isolated by a combined chemical and mechanical procedure using the PowerWater DNA Isolation Kit (QIAGEN, Hilden, Germany). Sterivex cartridges were cracked open and filters transferred to kit-supplied bead beating tubes. The isolation was continued according to the manufacturer's instructions, and DNA was stored at À 20 C. Library preparation was performed according to the standard instructions of the 16S Metagenomic Sequencing Library Preparation protocol (Illumina, San Diego, CA). The hypervariable V4-V5 region of 16S rRNA genes was amplified using primers 515F-Y (5 0 -GTGYCAGCMGCCGCGGTAA-3 0 ) and 926R (5 0 -CCGYCAATTY MTTTRAGTTT-3 0 ; Parada et al. 2016). Sequences were obtained on the Illumina HiSeq platform in a 2 Â 250 bp paired-end run (CeBiTec, Bielefeld, Germany) following the standard instructions of the 16S Metagenomic Sequencing Library Preparation protocol (Illumina, San Diego, CA).
Sample data matrices were managed using R package "phyloseq" v1.28.0 (McMurdie and Holmes 2013) and plots were generated using R package "ggplot2" v3.3.0 (G omez-Rubio 2017). Rarefaction analysis was conducted using R package "iNEXT" v2.0.20 (Hsieh et al. 2016). Calculation and visualization of shared and unique ASVs was conducted using R package "venn" v1.9 (Dusa 2020). Statistical tests were conducted using R package "vegan" v2.5.6 (Oksanen et al. 2019). Prior to downstream analysis, a prevalence threshold (i.e., in how many samples did an ASV appear at least once) of 5% was applied on the ASV abundance table (McMurdie and Holmes 2014). The fold-change in abundance of each ASV was calculated using the R package "DESeq2" v1.16.1 (Love et al. 2014). The method applies a generalized exact binomial test on variance stabilized ASV abundance. The results were filtered by significance, after correction for multiple-testing according to Benjamini and Hochberg (1995) with an adjusted p value < 0.1.
In other locations sampled during PS107, bacterioplankton communities were collected without size fractionation and sequenced separately using the same protocol as described above. In order to compare the communities of the submesoscale filament to these other Fram Strait locations, we agglomerated the sequences of free-living and particleassociated communities. Due to these methodological discrepancies, comparison between the communities was conducted on genus level as the lowest mutual taxonomic rank.

Data accession numbers and analyses repository
Raw paired-end, primer-trimmed reads were deposited in the European Nucleotide Archive (ENA; Silvester et al. 2018) under project accession number: PRJEB34666. The data were archived using the brokerage service of the German Federation for Biological Data (GFBio; Diepenbroek et al. 2014). Scripts for processing data can be accessed at "https://github.com/ edfadeev/submesoscale_analysis."

General characteristics of the submesoscale filament
In late July 2017, sea ice was present in the Fram Strait northeast and west of 0 EW/80 N (>15% concentration) while the eastern part of the Fram Strait was ice-free (Fig. 1a). On 26th of July 2017, satellite radar reflectivity showed a nearly straight streak of sea ice in the marginal ice zone at 2.5 E/79 N, which extended over 50 km from northeast to southwest and had a width of only 500 m (Fig. 1b). Below this sea-ice streak, we found a ca. 8 km wide filament characterized by a thin surface meltwater layer, above a layer of denser water that extended to >250 m depth.
Based on defined temperature and salinity characteristics of the main water masses in the Fram Strait (Rudels et al. 2013), both Atlantic water (temperature maximum > 5 C) and Polar water (temperature minima of ca. À1 C) were observed (Fig. 2). Outside of the filament, the surface ocean was strongly stratified by meltwater that occupied the top 15 m of the water column. Polar water was located directly below the meltwater layer at 20-40 m depth, while Atlantic water extended below the Polar water between 50 and 250 m depth. By contrast, inside the filament the meltwater layer covered the top 25 m, below which a relatively shallower and denser Atlantic water layer (25-250 m) was found (Fig. 2). Highresolution physical characterization revealed a frontal system of two strong currents flowing in opposite directions along the filament (Fig. 2a). These currents led to a horizontal inward-flow from both sides and mixing mostly in the upper 100 m inside the filament, apparent in the temperature and salinity characteristics of station T5 (Fig. 2b). Furthermore, von Appen et al. (2018) suggested that the denser waters (27.9 kg m À3 ) at ca. 100 m depth inside the filament were in the process of subduction, with an estimated vertical velocity of up to 100 m d À1 (Fig. 2a). Together, these mixing processes apparently enhanced phytoplankton growth in the filament waters, most likely by providing nutrients.

Biogeochemical conditions in the filament and the bounding currents
For biogeochemical and molecular analyses, water samples were collected inside the filament at stations T1, T2, and station T5 where strong mixing between Atlantic and Polar waters was observed (Fig. 2). Water samples from stations T3 and T4 represented the bounding currents on either side of the filament (Fig. 2). At each station, water samples were collected at six representative layers: surface (10 m), chlorophyll a maximum (chl-a max.; 20-30 m), below chl-a max. (50 m), epipelagic (100 m), and mesopelagic (200 and 400 m).
The ice melt-influenced surface waters inside the filament (measured at 10 m) were strongly depleted in nitrate (NO 3 ), with (NO 3 + NH 4 )/PO 4 ratios of 2.2-3.3, compared to 5.5-11.8 outside of the filament (Table 1). Partial pressure of carbon dioxide (pCO 2 ) in surface waters inside the filament was below 250 μatm, compared to ca. 300 μatm outside of the filament, recording the enhanced primary production. At the chl-a max. Depth (20-30 m), chl-a inside the filament ranged from 0.6 to 2.9 mg m À3 , peaking at station T5 (Table 1). In contrast, outside of the filament chl-a showed intermediate concentrations at the south-eastern station T3 (0.6-0.9 mg m À3 ), and low concentrations at the north- western station T4 (<0.3 mg m À3 ). Furthermore, in the upper 30 m, the POC concentrations and cell abundances of Phaeocystis-a major phytoplankton group-were also ca. twofold higher inside compared to outside of the filament (Table 1).
We deployed free-drifting sediment traps to quantify particulate organic carbon (POC) flux to 100, 200 and 400 m depth, both inside (near station T1), and outside (near the southeastern station T3) of the filament (Table 2). Both inside and outside the filament, we observed a similar POC flux of 50 and 55 mgPOC m À2 d À1 to 100 m, respectively (Table 2). However, the POC flux to 200 and 400 m depth inside the filament was almost twofold larger (48-56 mgPOC m À2 d À1 ) compared to outside of the filament (28-37 mgPOC m À2 d À1 ). This supports the hypothesis of von Appen et al. (2018) of a substantial subduction inside the filament, speeding up the transport of particles from the mixed upper 100 m to higher depths.

Abundance and diversity of bacterioplankton communities in the filament and the bounding currents
Bacterioplankton cell abundances correlated with chl-a concentrations (Spearman's rank correlation, r s = 0.8, p < 0.001). Inside the filament bacterioplankton cell abundances reached up to 1.1 Â 10 6 cells mL À1 (station T5), compared to twofold lower abundances outside of the filament, with well below 0.5 Â 10 6 cells mL À1 (Table 1). Beneath the upper 50 m, no differences in total cell abundances were observed (Supporting information Table S2). At all stations and sampling depths, we characterized the free-living (FL) and the particle-associated (PA) bacterioplankton communities using 16S rRNA gene amplicon sequencing. The final dataset consisted of 5,770,041 sequences in 59 samples that were assigned to 3980 amplicon sequence variants (ASVs; Supporting information Table S1). Rarefaction curves did not reach a plateau in any of the sampled communities, however, estimated asymptotic extrapolation to double amount of sequences showed only few additional ASVs (Supporting information Fig. 1). Thus, our sequencing depth represented most of the bacterioplankton diversity (Hsieh et al., 2016). Overall, bacterioplankton inside and outside of the filament was dominated by typical seawater taxonomic groups, i.e., the classes Alphaproteobacteria, Bacteroidia, and Gammaproteobacteria (Supporting information Fig. 2). Furthermore, surface water (10 m depth) communities outside of the filament had higher similarity to bacterioplankton communities in the western Fram Strait (i.e., Polar water; Supporting information Fig. 3). On the other hand, communities inside the filament had higher similarity to bacterioplankton communities in the eastern and northern parts of the Fram Strait (i.e., Atlantic water; Supporting information Fig. 3).
Alpha-diversity estimators of both free-living and particleassociated bacterioplankton revealed ca. twofold higher community richness in the epipelagic and mesopelagic depths (>100 m), compared to the upper 50 m (t-test, p < 0.05;  Supporting information Table 1). Furthermore, in the upper 50 m inside the filament, the number of observed ASVs and the Chao1 richness were lower (t-test, p < 0.05; Supporting information Fig. 4). Below the upper 50 m, there were no significant differences between bacterioplankton alpha diversity inside and outside the filament (t-test, p > 0.05; Supporting information Fig. 4). The composition of bacterioplankton communities showed a separation of the free-living and particle-associated communities at all depths (PERMANOVA, R 2 = 0.22, p < 0.001; Supporting information Fig. 5). In both fractions there was a clear separation between the communities in the upper 50 m and below (PERMANOVA, R 2 = 0.16, p < 0.001). In the upper 50 m (surface, chl-a max. and below chl-a max.), where most of the biogeochemical variability was observed, the composition of the bacterioplankton communities significantly differed inside and outside of the submesoscale filament (PERMANOVA, R 2 = 0.13, p < 0.001; Fig. 3). However, the surface (10 m) bacterioplankton community at station T3 (outside of the filament) showed compositional similarity to the respective communities inside the filament (Fig. 3). In contrast, in the Atlantic water masses below the upper 50 m (lower epipelagic and mesopelagic communities), there was no significant difference between the composition of the bacterioplankton communities inside and outside of the filament (PERMANOVA, p > 0.05; Supporting information Fig. 5).

Taxonomic composition of bacterioplankton communities
To investigate which taxonomic lineages accounted for the observed community differences, we performed an ASV presence/absence analysis. We revealed that 727 ASVs (18% of total ASVs) were shared among all free-living and particleassociated communities (Fig. 4). These shared ASVs constituted ca. 90% of all bacterioplankton sequences in the upper 50 m and below, and were mainly associated with the order Flavobacteriales (Bacteroidia; 112 ASVs) and the SAR11 clade (Alphaproteobacteria; 95 ASVs). In addition, 265 ASVs, also mainly associated with Flavobacteriales and the SAR11 clade, were unique to the upper 50 m bacterioplankton communities  inside the filament. On the other hand, 723 ASVs were unique to the upper 50 m bacterioplankton communities outside of the filament, and were mainly associated with the clades SAR11, SAR202, and SAR406. These unique ASVs comprised ca. 1% and 2% of the bacterioplankton sequences inside and outside of the filament, respectively. Below the upper 50 m, a similar amount of unique ASVs were observed inside and outside the filament (799 and 735 ASVs, respectively), mostly associated with Flavobacteriales and the clades SAR11, SAR406 and SAR202. We further investigated the differential sequence abundance of the shared ASVs in both free-living and particleassociated communities of the upper 50 m inside and outside of the filament (Fig. 5). The ASVs associated with Flavobacteriales (mainly families Flavobacteriaceae and Cryomorphaceae) and the SAR11 clade (mainly ecotype I and II) were significantly enriched inside the filament (Supporting information Table 3) and comprising 23%-56% of the sequences, in contrast to 6%-28% outside of the filament (Supporting information Fig. 6). Outside the filament, the communities were significantly enriched in various taxonomic groups, including the SAR202 clade and the archaeal order Nitrosopumilales (Supporting information Table 3), which comprised 3%-21% of the sequences in bacterioplankton communities outside of the filament compared to <3% inside the filament.

Discussion
Previous observations of low-latitude frontal systems revealed that they trigger phytoplankton growth (Taylor et al. 2012;Clayton et al. 2014;von Appen et al. 2020). In this study, we conducted the first biogeochemical and molecular characterization of an Arctic surface submesoscale frontal filament, observed in the Fram Strait (von Appen et al. 2018).
The distinct biogeochemical differences in the upper 30 m inside and outside of the filament were a result of locally enhanced primary production, enabled by horizontal mixing between nutrient-rich Atlantic and nutrient-depleted Polar waters. The (NO 3 + NH 4 )/PO 4 ratio indicated nitrate depletion by ongoing phytoplankton growth in the filament, and at its south-eastern edge (station T3). The chl-a and POC concentrations, as well as phytoplankton cell abundances, were more than twofold higher inside the filament. The observed chl-a  concentrations (up to 2.9 mg m À3 ) inside the filament were also relatively high in comparison to typically observed 1-2 mg m À3 during seasonal phytoplankton blooms in the region (Nöthig et al. 2015;Fadeev et al. 2018;Randelhoff et al. 2018). Combined with the enhanced drawdown of pCO 2 , nitrate depletion and the overall higher export flux, our evidence indicate a substantial stimulation of phytoplankton growth and export associated with the filament. These observations support the hypothesis that phytoplankton of high-latitudes may rapidly exploit submesoscale processes and produce substantial biomass, especially when nutrient-rich water masses are mixed by fronts such as the one observed here (Boyd et al. 2000;Taylor and Ferrari 2011;Lévy et al. 2012Lévy et al. , 2018. Despite the twofold higher concentrations of chl-a, POC, and phytoplankton inside the filament compared to its surrounding waters, we observed a similar flux of ca. 50 mgPOC m À2 d À1 to 100 m depth (i.e., the uppermost sediment trap) both inside and outside of the filament. This POC flux corresponded to previous measurements during seasonal phytoplankton blooms in the Fram Strait (Lalande et al. 2013;Fadeev et al. 2020), and suggested an export of fresh POC from the top meltwater layer inside and outside the filament at the time of sampling. A potential explanation for this pattern is the higher primary production in the deeper surface layer inside the filament, and the stronger mixing rate in the Atlantic water layer below. Previous observations in the region showed that particle sinking velocities are 20-60 m d À1 , and that there are faster sinking particles in ice-associated conditions (Fadeev et al. 2020). Therefore, it is likely that the POC formed at the surface inside the filament was then more rapidly exported to depths > 100 m by higher sinking rates and by physical water subduction. Furthermore, a time lag between production rates and export to deep waters cannot be excluded. In any case, the observed submesoscale process increased the export efficiency resulting in twofold larger POC export to depths below 100 m in the filament compared to the surrounding waters.
During the summer, surface waters bacterioplankton communities are mostly shaped by seasonal phytoplankton bloom conditions, with only small diversity variation between Atlantic and Polar water masses (Wilson et al. 2017;Fadeev et al. 2018). The response of bacterioplankton communities to blooming phytoplankton is often associated with elevated cell abundances and a change in taxonomic composition (Bunse and Pinhassi 2017). In the upper 50 m inside the filament, we observed higher bacterioplankton cell abundances and lower diversity of the communities, indicating the stimulation of specific taxonomic groups. The bacterioplankton communities inside the filament were enriched (in terms of sequence abundance) by ASVs of the order Flavobacteriales (class Bacteroidia), which specializes on degrading phytoplankton-derived organic compounds (Teeling et al. 2012) and is typically found during phytoplankton blooms (Buchan et al. 2014;Teeling et al. 2016). Furthermore, surface bacterioplankton communities inside the filament showed higher compositional similarity to communities sampled during the same cruise in Atlantic water of the West Spitsbergen Current, where elevated chl-a values indicated phytoplankton growth (Liu et al. 2019). In comparison, the upper 50 m bacterioplankton communities outside of the filament had lower cell abundances combined with higher taxonomic diversity. These communities were enriched in ASVs from SAR406 (phylum Marinimicrobia), SAR202 (class Dehalococcoidia), and the archaeal order Nitrosopumilales. Thus, surface bacterioplankton communities outside the filament resembled the communities in the Polar water of the East Greenland Current, where chl-a values did not show a pronounced phytoplankton growth (Liu et al. 2019). Taken together, our results suggest that overall composition of the bacterioplankton communities in the filament reflected both the origin of the water mass (mainly observed in the presence/ absence of taxa), and a response to the enhanced phytoplankton growth. Combined with the strongly depleted nitrate, the higher ammonia concentrations (Table 1) and the overall enhanced export flux, we further speculate that there was already an elevated heterotrophic microbial activity inside the filament as typical for the later bloom stage (Piontek et al. 2014;Fadeev et al. 2018).
The total "lifetime" of the observed submesoscale filament is unknown. Taking into account lateral advection velocities of ca. 10 cm s À1 (Wekerle et al. 2017) and the observed pronounced biological signal, the scenario of the filament being a result of a lateral transport from a remote phytoplankton bloom is unlikely. Based on the assumption that physicochemicalbiological dynamics originate from localized temporal evolution, rather than advected biomass, the age of the filament at the time of sampling can be estimated from changes in microbial cell abundances. With phytoplankton doubling times of 2-3 d at 4-6 C water temperatures (Kremer et al., 2017), reaching up to threefold higher cell abundances inside the filament would require up to 9 d. This temporal scale is consistent with the time required to reach the twofold higher cell abundances inside the filament (doubling time every 10 d; Kirchman et al. 2009). These growth-rate estimates do not take into account potential losses due to grazing by zooplankton or viral lysis, and thus represent rather the lower time limit. Based on this we hypothesize that, at the time of sampling, the age of the observed filament was already at least 10 d. Although temporal dynamics are not resolved by our sampling strategy, the approximated lifetime matches recent estimates of mesoscale eddies lifetime in Fram Strait based on high-resolution ocean-sea ice models . The boundaries of these mesoscale eddies may harbor submesoscale filaments such as the one studied here.
In conclusion, we showed that the mixing between Atlantic and Polar waters formed a spatially restricted water filament with distinct biological and biogeochemical characteristics. Its presumed persistence for at least 10 d is projected to have direct implications for organic matter export, to Fadeev et al. Submesoscale dynamics of bacterioplankton some extent enhanced via vertical subduction. We also showed that higher phytoplankton biomass inside the filament was associated with a pronounced change in the abundance and composition of bacterioplankton communities, towards dominance of heterotrophic "master recyclers." Taken together, our observations support the hypothesis that submesoscale features with sizes from kilometers to tens of kilometers promote phytoplankton blooms by nutrient supply (Lévy et al. 2001;Mouriño 2004;Mahadevan 2015), and that these features directly shape microbial diversity and community structure on small spatial and temporal scales (Lévy et al. 2012(Lévy et al. , 2018. Altogether, corresponding to estimates of the regularity of such phenomena in the ocean (Lévy et al. 2018), our observations support the notion that submesoscale processes play an important role in ocean productivity, biogeochemical fluxes and biodiversity patterns. Most submesoscale filaments and previously studied frontal systems result from temperature gradients between two semistationary water masses (e.g., Baltar et al. 2009Baltar et al. , 2016. The Arctic submesoscale filament observed here is the result of the strong salinity gradient between Atlantic and Polar waters. To produce similar density gradients, water masses would have to have temperature differences of several tens of C, which does not occur in the studied Arctic region. However, such strong salinity gradients can be found not only in the Fram Strait, but also in other ice-or freshwater-influenced regions worldwide, such as estuarine systems (Jaeger and Mahadevan 2018). Despite the fundamental nonphysical differences between such environments, similar physical submesoscale dynamics driven by the density gradients are to be expected. We therefore suggest that similar pronounced localized biological responses to physical processes as well as physical-biological interactions may also occur in other aquatic systems. To assess their impact on regional and global ocean productivity, high resolution methods would be needed, such as long range autonomous underwater vehicles or towed vehicles in combination with remote sensing by satellite. However, especially in polar regions, the ice cover limits quantification by remote sensing, challenging to understand and predict how the regional productivity is impacted by such submesoscale processes. This study provides a combination of methods that improves such predictions.