Spatial dynamics in Arctic bacterioplankton community densities are strongly linked to distinct physical and biological processes (Fram Strait, 79°N)

The Arctic is impacted by climate warming faster than any other oceanic region on Earth. Assessing the baseline of microbial communities in this rapidly changing ecosystem is vital for understanding the imminent implications of Arctic warming and sea ice retreat on ecosystem functioning. Using CARD-FISH and semi-automated counting, we quantified 14 ecologically relevant taxonomic groups of bacterioplankton (Bacteria and Archaea) from surface (0– 30 m) down to deep waters (2500 m) in summerly ice-covered and ice-free regions of the Fram Strait, the main gateway for Atlantic inflow into the Arctic Ocean. Cell abundances of the bacterioplankton communities in surface waters varied from 105 cells mL−1 in ice-covered region to 106 cells mL−1 in the ice-free region and were overall driven by variations in phytoplankton bloom conditions across the Strait. In surface waters the bacterial classes Bacteroidia and Gammaproteobacteria showed several-fold higher cell abundances under late phytoplankton bloom conditions of the ice-free regions. Other taxonomic groups, such as the Rhodobacteraceae, revealed a distinct association of cell abundances with the surface Atlantic waters. With depth (> 500 m) the total cell abundances of the bacterioplankton communities decreased by one to two orders of magnitude, while largely unknown taxonomic groups (e.g., SAR324 and SAR202 clades) maintained constant cell abundances throughout the entire water column (103 cells mL−1). This suggests that some enigmatic taxonomic groups may occupy a specific ecological niche in the entire water column. Our results provide the first quantitative spatial variations assessment of bacterioplankton in summerly ice-covered and ice-free Arctic water column, and suggest that further shift towards ice-free Arctic summers with longer phytoplankton blooms can lead to major changes in the associated standing stock of the bacterioplankton communities.


Introduction
Atmospheric and oceanic warming has a substantial impact on the Arctic Ocean already today (Dobricic et al., 2016;Sun et al., 2016;Dai et al., 2019). The strong decline in sea ice coverage (Peng and Meier, 2018;Dai et al., 2019) and the increasing heat transfer by the Atlantic water inflow (Beszczynska-Möller et al., 2012;Rudels et al., 2013;Walczowski et al., 2017) will affect stratification of the water column and can lead to an increased vertical mixing of the deep Atlantic core water, a process also termed "Atlantification" (Polyakov et al., 2017). The main inflow of Atlantic water into the Arctic Ocean occurs through the Fram Strait (Beszczynska-Möller et al., 2011), making it a sentinel region for observing the ongoing changes in the Arctic marine ecosystem (Soltwedel et al., 2005(Soltwedel et al., , 2016. The Fram Strait is also the main deep-water gateway between the Atlantic and the Arctic Oceans. It is a spatially dynamic region (Wekerle et al., 2017), with two distinct hydrographic regimes: the West Spitsbergen Current (WSC) that carries relatively warm and saline Atlantic water northwards along the Svalbard shelf (Beszczynska-Möller et al., 2012;von Appen et al., 2015), and the East Greenland Current (EGC) that transports cold polar water and sea ice southwards from the Arctic Ocean along the ice-covered Greenland shelf (de Steur et al., 2009).
Sea ice conditions have a strong impact on the seasonal ecological dynamics in the Arctic Ocean (Wassmann and Reigstad, 2011), and sea-ice algae can make up a significant proportion of the annual productivity (Leu et al., 2011;Boetius et al., 2013;Fernández-Méndez et al., 2014). However, the presence of sea ice and variations in its snow cover can repress the seasonal phytoplankton bloom in the water column through light limitation (Mundy et al., 2005;Leu et al., 2011), or change its timing, e.g. by increasing stratification of the surface waters once the ice melts . Previous summer observations in Fram Strait already suggested that total cell abundances and productivity of bacterioplankton communities in surface waters is driven by environmental parameters associated with phytoplankton bloom dynamics (Fadeev et al., 2018), such as the availability and composition of organic matter (Piontek et al., 2015;Engel et al., 2019), with differences between ice-covered and ice-free regions (Piontek et al., 2014;Fadeev et al., 2018).
Long-term summer observations in the region, conducted in the framework of the Long-Term Ecological Research (LTER) site HAUSGARTEN, revealed strong ecological variations associated with the Atlantic Meridional Overturning Circulation (AMOC) (Soltwedel et al., 2016). Warming events during the past decades influenced seasonal phytoplankton blooms by causing a slow but continuous increase in biomass, and a shift from diatom-to flagellate-dominated communities Engel et al., 2017;Basedow et al., 2018). It has been also observed that phytoplankton blooms of recent years show an increasing partitioning of the produced organic carbon into the dissolved phase (Engel et al., 2019), which may result in a more active microbial loop in the upper ocean and less export of particulate matter (Vernet et al., 2017). In times of a rapidly changing Arctic ecosystem, investigating the structure and the dynamics of bacterioplankton communities remains a key component to understand the ongoing changes on this environment. However, so far, an assessment of associated responses of the key bacterial taxa responsible for increased recycling is missing, especially with regard to shifts in standing stocks.
To date, the majority of Arctic bacterioplankton studies are performed using high-throughput sequencing of the 16S rRNA gene, which cannot directly be converted to absolute standing stock abundances of specific taxonomic groups due to PCR and other quantitative biases (Gloor et al., 2017;Kumar et al., 2018;Piwosz et al., 2020). Here we used semi-automatic CAtalyzed Reporter Deposition-Fluorescence In Situ Hybridization (CARD-FISH; (Pernthaler et al., 2002)), power of which lies in the ability to acquire absolute abundance of the targeted taxonomic groups free of compositional effect (Amann et al., 1990). While specific taxonomic groups can be targeted and quantified, also the positive hybridization signal provides additional information that the analyzed cells were alive and active before fixation (Amann et al., 1990;DeLong et al., 1999). Automatization of the microscopic examination and counting procedure allowed us to reach a high-throughput standard (Schattenhofer et al., 2009;Teeling et al., 2012;Bižić-Ionescu et al., 2015;Bennke et al., 2016). Using this technique, we quantified cell abundances of 14 taxonomic groups in 44 samples, collected at 4 different water layers from surface (15 -30 m) to the deep ocean (2500 m) in both icefree and ice-covered regions of the Fram Strait. The main objective of this study was to assess the standing stocks of key taxonomic groups in summerly Arctic bacterioplankton across the Strait. Using the high-throughput data of bacterioplankton cell abundances we tested previously established hypotheses that: 1) in surface waters the abundance of various taxonomic groups is associated with phytoplankton bloom conditions and is linked to the abundance of specific phytoplankton populations (Fadeev et al., 2018); 2) with an increasing depth the differences between bacterioplankton communities in ice-covered and ice-free regions decrease, and various taxonomic groups that are associated with winterly bacterioplankton dominate the deep sea communities.

Bacterioplankton communities strongly change in cell abundance and composition with depth
We investigated a total of 44 water samples from 11 stations that represented different hydrographic and biogeochemical conditions across Fram Strait ( Figure 1; Table S1). At each station, 4 different water layers were targeted: the surface mixed layer (0 -30 m), epipelagic waters (100 m), deep mesopelagic waters (500 -1000 m), and bathypelagic waters (1200 -2500 m; Table S2). Based on the known hydrography of the Fram Strait , and observed sea-ice conditions, the sampled stations were grouped into three distinct regions ( Figure 1): the eastern ice-free region associated with the WSC (Beszczynska-Möller et al., 2012), the western ice-covered region associated with the EGC (de Steur et al., 2009), and the partially ice-covered region in the north-east that is associated with the highly productive marginal ice zone (further addressed as North; i.e. "N") (Hebbeln and Wefer, 1991;Perrette et al., 2011). At the time of sampling in June-July 2016, chlorophyll a concentrations showed a developed phytoplankton bloom across the Strait (Table S1). Thus, fitting the interannual trend of summerly phytoplankton bloom conditions in the western icecovered EGC and the eastern ice-free WSC Fadeev et al., 2018).
We found that in all three regions, total cell abundances of the entire bacterioplankton community were highest at the surface with 10 5 -10 6 cells mL -1 , and significantly decreased with depth down to 10 4 cells mL -1 at meso-and bathypelagic depths (Figure 2a; Kruskal-Wallis test; Chi square = 554.39, df = 3, p-value < 0.01). Members of the domain Bacteria dominated the communities throughout the entire water column, with highest cell abundances in surface waters (10 5 -10 6 cells mL -1 ), and significantly lower abundances at depth (10 4 cells mL -1 ; Figure 2b; Kruskal-Wallis test; Chi square = 35.27, df = 3, p-value < 0.01). Archaeal cells had overall an order of magnitude lower abundance than bacterial cells throughout the entire water column, ranging from 10 4 cells mL -1 at the surface down to 10 3 cells mL -1 in the bathypelagic waters ( Figure 2c). However, unlike Bacteria, archaeal communities doubled their absolute cell abundances from the surface to 100 m depth (reaching up to 6 × 10 4 cells mL -1 ), followed by a significant decrease in cell abundance at meso-and bathypelagic depths (Kruskal-Wallis test; Chi square = 29.04, df = 3, p-value < 0.01). This pattern fits the known global trend of archaeal enrichment in epipelagic waters (Karner et al., 2001;Herndl et al., 2005;Kirchman et al., 2007;Varela et al., 2008;Schattenhofer et al., 2009) and was also observed in other regions of the Arctic Ocean (Amano-Sato et al., 2013). Altogether, the observed cell numbers in the surface waters were well within the range of previous observations in the Fram Strait waters (Piontek et al., 2014;Fadeev et al., 2018;Engel et al., 2019). Below the upper 100 m the observed bacterioplankton cell abundances were lower than the ones previously reported from the eastern Fram Strait by Quero et al. (2019). The discrepancy might be associated with methodological differences in the duration of cell staining incubation and the usage of automated vs. manual counting approach. Nevertheless, both studies showed a similar pattern of a strong decrease in bacterioplankton cell abundances with depth, which also matches observations in other oceanic regions (Karner et al., 2001;Church et al., 2003;Teira et al., 2004;Schattenhofer et al., 2009;Dobal-Amador et al., 2016).
In the surface waters, around 60% of the DAPI-stained total bacterioplankton community was covered by the Bacteria-specific probes in all stations (Table S2). At depth (>100 m) the total coverage of the Bacteria-specific probes strongly decreased to 16-40% of DAPI-stained cells (ANOVA; F3 = 15.39, p < 0.01; Table S2). Similar decrease in detectability of the domain-specific probes (EUB388I-III) was previously observed in other bacterioplankton microscopy studies (Karner et al., 2001;Herndl et al., 2005;Varela et al., 2008;Schattenhofer et al., 2009), and is potentially explained by a decrease in ribosomal nucleic acids within the bacterial cells (i.e. lower activity) towards the oligotrophic depths, or an increasing fraction of microbial cells that are not captured by the currently existing probes (Hewson et al., 2006;Sogin et al., 2006;Galand et al., 2009a;Agogué et al., 2011;Salazar et al., 2016).
Interestingly, the Archaea-specific probe showed a different trend. In surface waters, the coverage of the probe was higher in the EG stations (8% of DAPI-stained cells), compared to HG and N stations (1-2% of DAPI-stained cells; Table S2). At depth (>100 m), the coverage of the Archaea-specific probe increased significantly to 13-17% of DAPI-stained cells (ANOVA; F3 = 34.31, p < 0.01), which implies an increase in relative abundance with depth (Müller et al., 2018b). Taken together, our findings confirm previously observed higher abundance of Archaea in bacterioplankton communities of ice-covered waters (Wilson et al., 2017;Müller et al., 2018b), and correspond to the globally observed trend of an increasing archaeal importance at depth (Herndl et al., 2005;Teira et al., 2006;Galand et al., 2009b)

Surface water bacterioplankton communities are affected by distinct phytoplankton bloom conditions
Phytoplankton blooms in surface waters lead to an increasing cell abundance of heterotrophic bacteria that are specialized on degradation of organic matter from algal exudates and phytodetritus (Buchan et al., 2014;Teeling et al., 2016). Previous observations in the Fram Strait, acquired using high-throughput sequencing of the 16S rRNA gene, revealed a strong influence of the summerly phytoplankton bloom conditions on the bacterioplankton communities (Wilson et al., 2017;Müller et al., 2018a), differing between the ice-covered and ice-free regions of the Strait (Fadeev et al., 2018).During the sampling period, distinct phytoplankton bloom communities in surface waters across the Strait were observed, with a Phaeocystis-dominated bloom in the ice-free HG stations, a diatom-dominated bloom in the ice-covered EG stations, and a mixed diatoms and Phaeocystis populations bloom in the ice-margin stations N (data not included). Surface water total bacterioplankton communities showed significantly higher cell abundances in the HG and N stations (13-21 ×10 5 cells mL -1 ), compared to the EG stations (0.5-2 ×10 5 cells mL -1 ; Kruskal-Wallis test; Chi square = 81.85, df = 2, p-value < 0.01). The communities were dominated by bacterial cells that comprised 8-11 ×10 5 cells mL -1 in the HG and N stations, and 2 ×10 5 cells mL -1 in the EG stations.
Within the bacterial communities, a group of bacterial classes that are functionally associated with phytoplankton blooms in the region (Bacteroidia, Gammaproteobacteria, and the phylum Verrucomicrobia) (Fadeev et al., 2018) showed together several-fold higher cell abundances in the HG and N stations (2.3-10 × 10 5 cells mL -1 ), compared to the EG stations (0.5-1.5 × 10 5 cells mL -1 ). Jointly comprising up to 50% of the bacterioplankton communities (Table 1). However, other taxonomic groups, which were previously known to be associated with the polar water masses and winter communities in the Fram Strait (e.g., the class Deltaproteobacteria and the Thaumarchaeota) (Wilson et al., 2017;Fadeev et al., 2018;Müller et al., 2018b), showed higher cell abundances in the ice-covered EG stations, compared to the ice-free HG and ice-margin N stations (Table 1). Thus, suggesting that the cell abundances of different taxonomic groups may be associated with different environmental conditions (i.e., phytoplankton bloom vs. water masses) To test whether the observed differences in cell abundances are a result of distinct biological conditions (i.e., phytoplankton bloom) or distinct physical conditions (Atlantic vs. Arctic water masses) we conducted specific correlation tests between the cell abundances of different taxonomic groups and various environmental parameters. For that purpose, we defined physical characteristics of the water as a proxy for the distinct surface water masses . We identified that the cell abundances of the phylum Verrucomicrobia and the order Opitutales within it, as well as the SAR11 clade and the Rhodobacteraceae family (both within the class Alphaproteobacteria), showed strong positive correlation with statistical significance to water temperature (Pearson's correlation; r > 0.5, p-value < 0.05; Table S3), suggesting a potential association with the warmer Atlantic waters of the eastern Fram Strait. The phylum Verrucomicrobia has been previously shown to be a major polysaccharide-degrading bacterial taxonomic group in Svalbard fjords (Cardman et al., 2014), and therefore may also be associated with the outflow from the Svalbard fjords into the Atlantic waters of the WSC (Cottier et al., 2005). The SAR11 clade and the family Rhodobacteraceae, both have been previously shown to correlate with temperature at high latitudes (Giebel et al., 2011;Tada et al., 2013), and are known to have distinct phylotypes in water masses with different temperatures (Selje et al., 2004;Sperling et al., 2012;Giovannoni, 2017). However, the family Rhodobacteraceae is also known for their broad abilities in utilizing various organic compounds (Buchan et al., 2014;Luo and Moran, 2014), and thus one cannot rule out that their higher cell densities in warmer waters of the WSC and N regions are associated with the developed phytoplankton bloom in these regions. In addition, the SAR324 clade (class Deltaproteobacteria) showed strong positive correlation with statistical significance to salinity (Pearson's correlation; r > 0.5, p-value < 0.05; Table S3). During summer with increasing melt of sea ice, a low-salinity water layer is formed in surface waters. The strong stratification of this water layer enhances the development of the phytoplankton bloom (Fadeev et al., 2018). Thus, the correlation of these taxonomic groups with salinity rather suggests that their cell abundances are lower in surface waters with strong phytoplankton bloom (i.e., in WSC and N regions).
Because distinct water masses in the region differ not only in their physical characteristics but also in their inorganic nutrient budgets (Wilson and Wallace, 1990), we disentangled the effect of physical and biological parameters by calculating the seasonal net consumption of inorganic nutrients as the proxy for phytoplankton bloom state (Table S2) (Fadeev et al., 2018). Consumed nitrate (ΔNO3) and phosphate (ΔPO4) revealed a very strong positive correlation between them with statistical significance (Pearson's correlation; r = 0.86, p-value < 0.05). These two nutrients are required for phytoplankton growth in a constant ratio ('The Redfield ratio'), and thus the strong correlation between their consumption is expected. The consumed silica (ΔSiO3), which is used only by diatoms, did not show significant correlation to ΔPO4 and ΔNO3, which further supports that there are different phytoplankton populations across the Strait (i.e., diatoms vs. Phaeocystis). Phytoplankton bloom associated environmental parameters (chlorophyll a concentration and the consumed inorganic nutrients) revealed weaker relationships with cell abundances of the different taxonomic groups (Table S3). This might be explained by smaller differences in these parameters between the regions, due to an overall ongoing phytoplankton bloom across the entire Strait (Table S2). The class Gammaproteobacteria showed strong correlation with statistical significance to ΔPO4 (Pearson's correlation; r = 0.74, p-value < 0.05; Table S3). The correlation suggests a strong uptake of phosphate by Gammaproteobacteria that confirms their strong growth and high activity during phytoplankton blooms (Joint et al., 2002;Buchan et al., 2014).
The complexity of surface waters in Fram Strait, with different ice-coverage, the opposing water regimes (Atlantic and Polar water masses) along with complex mesoscale mixing events (Wekerle et al., 2017), and a dynamic ice-melt water layer, challenges the identification of specific associations between microbial cell abundances and environmental parameters. Nevertheless, our results showed that elevated cell abundances of some taxonomic groups (e.g., Gammaproteobacteria) are more likely to be associated with the phytoplankton bloom conditions observed at the site (e.g. through a link with algal exudates and nutrients as main source for growth) (Tada et al., 2011;Teeling et al., 2012), while other taxonomic groups (e.g., SAR11 clade) might be more influenced by the distinct Arctic water masses (Kraemer et al., 2020).

Enigmatic microbial lineages increase in cell abundance towards the deep ocean
The deep waters of the Fram Strait basin (> 500 m) have a rather homogeneous hydrography (von Appen et al., 2015), and are less affected by the seasonal dynamics governing the surface layers (Wilson et al., 2017). Previous molecular observations of the deep waters bacterioplankton communities showed high sequence abundance of largely unknown taxonomic groups, such as the SAR202 (class Dehalococcoidia), SAR324 (class Deltaproteobacteria), and SAR406 (phylum Marinimicrobia) (Wilson et al., 2017;Quero et al., 2019). There was also higher archaeal sequence abundance at depth, with the class Nitrososphaeria reaching up to 15% of the sequences in mesopelagic waters (> 200 m) (Wilson et al., 2017;Müller et al., 2018b). However, it has been also recently shown that in ice-covered regions of the Strait surface dominant taxonomic groups, such as Gammaproteobacteria and Nitrososphaeria, are exported via fast-sinking aggregates from surface to the deep ocean (> 1000 m), where they may realize an ecological niche. We observed that in all regions the total cell abundances of the bacterioplankton communities were in the range of 10 4 cells mL -1 in both meso-and bathypelagic waters (Figure 2), matching previous observations in other regions of the Arctic Ocean (Wells and Deming, 2003;Wells et al., 2006). Bacterial taxonomic groups that dominated the surface water communities (e.g., Bacteroidetes, Gammaproteobacteria and Verrucomicrobia), in both ice-free and ice-covered regions of the Strait, decreased by two orders of magnitude in their cell abundances at meso-and bathypelagic depths (Kruskal-Wallis test; p-value < 0.01; Table 2). This trend strongly correlated along the water column with the total bacterioplankton cell abundances (Pearson's correlation; r > 0.8, p-value < 0.05; Figure S1). In contrast, other bacterial groups, such as the SAR202 clade (class Dehalococcoidia) and SAR324 (class Deltaproteobacteria), proportionally decreased in cell abundances with depth, and maintained overall constant cell abundances of ca. 0.5 ×10 4 cells mL -1 throughout the entire water column (Table  2). Previous molecular studies of bacterioplankton communities in Fram Strait suggested a proportional increase of these largely understudied bacterial lineages in the deep ocean (Wilson et al., 2017). The cell abundances that are presented here indicate that their increasing proportional abundance at depth is due to stronger decrease in the cell abundances of other groups (Table 2; Table  S4). Very little is currently known about these two taxonomic groups, but previous genetic observations suggest that they possess distinct metabolic capabilities, and may be involved in the degradation of recalcitrant organic matter (SAR202; (Landry et al., 2017;Colatriano et al., 2018;Saw et al., 2020); or in sulfur oxidation (SAR324; (Swan et al., 2011;Sheik et al., 2014), which is however a highly unlikely energy source for the fully oxidized, oligotrophic polar waters. Their homogeneous distribution from the stratified surface to the homogenous deep ocean of the Fram Strait suggests that these enigmatic bacterial groups fulfil an ecological niche that exists in the entire water column, and thus may have unique roles in oceanic nutrient cycling.
Proportionally, the decrease with depth of archaeal cell abundances was less than of Bacteria (Table  S4). The Thaumarchaeota phylum strongly correlated with the pattern of the archaeal cell abundances (Pearson's correlation; r = 0.76, p-value < 0.05; Figure S1), with a two-fold increase in cell abundances from surface to epipelagic depth (100 m), followed by a substantial decrease towards meso-and bathypelagic waters (Table S4). This general trend matches deep ocean observations in previous quantitative studies of other oceanic regions (Karner et al., 2001;Church et al., 2003;Herndl et al., 2005). It has been shown in molecular studies that Thaumarchaeota comprise a large proportion of the bacterioplankton communities in the Fram Strait, especially in the epipelagic waters (Wilson et al., 2017;Müller et al., 2018b). In our study, the Thaumarchaeota exhibited the highest cell abundances at 100 m in the ice-free HG and ice-margin N stations (3 ×10 4 cells mL -1 ), where they comprised half of the total archaeal community (Table S2). The strong absolute decrease of Thaumarchaeota cell abundances towards the meso-and bathypelagic waters may represent a decrease in activity with depth (Herndl et al., 2005;Kirchman et al., 2007;Alonso-Saez et al., 2012) and thus lower detection of the cells. Alternatively, the relative decrease within the total archaeal cells may suggest that other pelagic archaeal groups, such as the Euryarchaeota, increase in abundance at depth (Galand et al., 2010).

Conclusions
Using state-of-the-art semi-automatic microscopy counting, we quantified the absolute cell abundance of 14 key taxonomic groups in summer bacterioplankton communities of the Fram Strait. Our observations covered both the ice-free and the ice-covered regions of the Strait, which at the time of sampling were characterized by different seasonal phytoplankton bloom conditions. Our results showed that in surface waters, abundance of some taxonomic groups (e.g., Rhodobacteraceae) was to certain extent related to the distinct water masses of the Strait. However, in general the abundance of different taxonomic groups was strongly positively (e.g., Gammaproteobacteria) and negatively (e.g., SAR324 clade) associated to the progressing states of the seasonal phytoplankton bloom across the Strait. Based on our observations, we postulate that summer surface bacterioplankton communities in Arctic waters are shaped mostly by the phytoplankton bloom rather than the distinct water masses they are associated with. This suggests that currently predicted longer seasonal phytoplankton blooms, as well as the increasing Atlantic influence on the Arctic Ocean (i.e., 'Atlantification'), may have a strong impact on the biogeographical distribution of certain bacterioplankton taxonomic groups in the surface Arctic waters. This study also provides the first extensive quantitative overview of bacterioplankton communities in the deep Arctic water column (> 500 m). We showed that at depth, largely unknown taxonomic groups, such as the SAR202 clade, maintained similar abundance throughout the entire water column (2500 m depth), where other taxa decline by severalfold. This observation suggests that despite their low abundance, some taxonomic groups may potentially realize a unique ecological niche throughout the entire water column. Altogether, our results on cell abundances of ecologically relevant taxonomic bacterioplankton groups expanded previous molecular observations in the Fram Strait and provided new insights into the quantitative dynamics of pelagic bacterioplankton communities from surface to the deep waters of the Arctic Ocean.

Sampling and environmental data collection
Sampling was carried out during the RV Polarstern expedition PS99.2 to the Long-Term Ecological Research (LTER) site HAUSGARTEN in Fram Strait (June 24th -July 16th, 2016). Sampling was carried out with 12 L Niskin bottles mounted on a CTD rosette (Sea-Bird Electronics Inc. SBE 911 plus probe) equipped with temperature and conductivity sensors, a pressure sensor, altimeter, and a chlorophyll fluorometer. On board, the samples were fixed with formalin in a final concentration of 2% for 10 -12 hours, then filtered onto 0.2 µm polycarbonate Nucleopore Track-Etched filters (Whatman, Buckinghamshire, UK), and stored at -20°C for further analysis.
Hydrographic data of the seawater including temperature and salinity were retrieved from PANGAEA (Tippenhauer et al., 2017), along with measured chlorophyll a concentrations   (Table S1).

Catalyzed reporter deposition-fluorescence in situ hybridization (CARD-FISH)
We quantified absolute cell abundances of 14 key bacterioplankton groups (Table S5), based on their relatively high sequence abundance and recurrences in previous molecular studies of Arctic waters (Bowman et al., 2012;Wilson et al., 2017;Müller et al., 2018b). CARD-FISH was applied based on the protocol established by (Pernthaler et al., 2002), using horseradish-peroxidase (HRP)-labelled oligonucleotide probes (Biomers.net, Ulm, Germany; Table S2). All probes were checked for specificity and coverage of their target groups against the SILVA database release 132 (Quast et al., 2012). All filters were embedded in 0.2 % low-gelling-point agarose, and treated with 10 mg mL -1 lysozyme solution (Sigma-Aldrich Chemie GmbH, Hamburg, Germany) for 1 h at 37°C. Filters for enumerating Archaea and Thaumarchaeota were treated for an additional 30 min in 36 U mL -1 achromopeptidase (Sigma-Aldrich Chemie GmbH, Hamburg, Germany) and 15 µg mL -1 proteinase K at 37°C. Subsequently, endogenous peroxidases were inactivated by submerging the filter pieces in 0.15 % H2O2 in methanol for 30 min, before rinsing in Milli-Q water and dehydration in 96 % ethanol. Then, the filters were covered in hybridization buffer and a probe concentration of 0.2 ng µL -1 . Hybridization was performed at 46°C for 2.5 h, followed by washing in pre-warmed washing buffer at 48°C for 10 min, and 15 min in 1x PBS. Signal amplification was carried out for 45 min at 46°C with amplification buffer containing either tyramide-bound Alexa 488 (1 μg/mL) or Alexa 594 (0.33 μg mL -1 ). Afterwards, the cells were counterstained in 1 µg/mL DAPI (4′,6-diamidino-2phenylindole; Thermo Fisher Scientific GmbH, Bremen, Germany) for 10 min at 46°C. After rinsing with Milli-Q water and 96 % ethanol, the filter pieces were embedded in a 4:1 mix of Citifluor (Citifluor Ltd, London, United Kingdom) and Vectashield (Vector Laboratories, Inc., Burlingame, United States), and stored overnight at -20°C for later microscopy evaluation.

Automated image acquisition and cell counting
The filters were evaluated microscopically under a Zeiss Axio Imager.Z2 stand (Carl Zeiss MicroImaging GmbH, Jena, Germany), equipped with a multipurpose fully automated microscope imaging system (MPISYS), a Colibri LED light source illumination system, and a multi-filter set 62HE (Carl Zeiss MicroImaging GmbH, Jena, Germany). Pictures were taken via a cooled chargedcoupled-device (CCD) camera (AxioCam MRm; Carl Zeiss AG, Oberkochen, Germany) with a 63x oil objective, a numerical aperture of 1.4, and a pixel size of 0.1016 μm/pixel, coupled to the AxioVision SE64 Rel.4.9.1 software (Carl Zeiss AG, Oberkochen, Germany) as described by . Exposure times were adjusted after manual inspection with the AxioVision Rel.4.8 software coupled to the SamLoc 1.7 software (Zeder et al., 2011), which was also used to define the coordinates of the filters on the slides. For image acquisition, channels were defined with the MPISYS software, and a minimum of 55 fields of view with a minimum distance of 0.25 mm were acquired of each filter piece by recoding a z-stack of 7 images in autofocus.
Cell enumeration was performed with the software Automated Cell Measuring and Enumeration Tool (ACMETool3, 2018-11-09; M. Zeder, Technobiology GmbH, Buchrain, Switzerland). Cells were counted as objects according to manually defined parameters separately for the DAPI and FISH channels.

Calculation of consumed inorganic nutrients
Following (Fadeev et al., 2018) the nutrient consumption (Δ) at each station was calculated by subtracting the mean value of all collected measurements above 50 m from the mean value of all collected measurements between 50 and 100 m (below the seasonal pycnocline).

Conflict of Interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Author contributions
MC, EF and VS-C designed and conducted the study. MC, EF and VS-C wrote the manuscript with guidance from AB. All authors critically revised the manuscript and gave their approval of the submitted version.