Compositions of dissolved organic matter in the ice-covered waters above the Aurora hydrothermal vent system, Gakkel Ridge, Arctic Ocean

. Hydrothermal vents modify and displace subsurface dissolved organic matter (DOM) into the ocean. Once in the 20 ocean, this DOM is transported together with elements, particles, dissolved gases


Introduction
Dissolved organic matter (DOM) in the vicinity of high temperature hydrothermal vents shows discrepancies in 35 concentration compared to ocean background concentrations, mainly due to thermal decomposition of DOM during hydrothermal circulation (Hawkes et al., 2015;Lang et al., 2006). During the formation of hot hydrothermal fluids in convective circulation systems, cold deep ocean water percolates through permeable pathways (faults/fissures) in the crust and is progressively heated and modified during high temperature alteration of igneous (e.g. mafic, ultramafic) rocks (German and Seyfried, 2014;Simoneit et al., 2004). Continuous exposure to extremes of temperature (up to 400°C) and low 40 pH either completely degrade DOM to volatile species (e.g., CO2) and dissolved hydrocarbon gases (predominantly methane and trace C2+ hydrocarbons) or alters originally stable, high molecular weight, refractory DOM to highly aromatic, unsaturated, oxygen poor, petroleum-like compositions in the hydrothermal vent fluids (Hawkes et al., 2015(Hawkes et al., , 2016Rossel et al., 2015Rossel et al., , 2017. In relatively low temperature hydrothermal systems (<150°C), DOM is not degraded by cracking (Lang et al., 2006), and production of DOM may be more prevalent (Hawkes et al., 2015). DOM may therefore be a significant source 45 of mixed forms of complex reduced carbon in the vent fluids, which can include dissolved free amino acids (Fuchida et al., 2014;Haberstroh and Karl, 1989;Horiuchi et al., 2004), lipids (McCollom et al., 1999a(McCollom et al., , 2015 and bio-labile organic carbon (Hansen et al., 2019;Longnecker et al., 2018;Rossel et al., 2015). Considering that the deep ocean DOM is refractory and primarily unavailable for organisms (Dittmar and Stubbins, 2014;Hansell, 2013), specialized microbial communities may therefore utilize hydrothermally modified DOM as a carbon source and support hydrothermal fauna in the vicinity of 50 hydrothermal vents (Bart et al., 2020;Folkers and Rombouts, 2020;Hestetun et al., 2016;Yahel et al., 2003).
The chemical composition and temperature of venting fluids emanating at the seafloor and contributing to hydrothermal plume formation are mainly controlled by the geological setting and physicochemical constraints including phase separation, water-rock interaction (mineral precipitation/dissolution) and biological processes (German and Seyfried, 2014;Nakamura and Takai, 2014). Hydrothermal fluids emitted from the vents are diluted with ambient water by factors of 10 4 -10 5 and form 55 incubation experiments showed that the thermal degradation of deep-sea recalcitrant DOM elevates the microbially accessible composition and stimulate the growth of prokaryotic communities (Hansen et al., 2019).
More than 300 high-temperature venting sites have been identified at mid-ocean ridges (Hannington et al., 2011;Nakamura and Takai, 2014) and around 800 are estimated to exist, roughly half of which are expected to be found in slow/ultraslow 70 spreading mid-ocean Ridges (Beaulieu et al., 2015). Because of the limited number of expeditions in the ice-covered Artic Ocean, surveys of hydrothermal vents at very high latitudes have been extremely rare. However, considering the amount of fluid venting at hydrothermal vents, and the far-reaching area of impact through plume dispersion, their influence on marine DOM may be substantial in the oligotrophic Arctic Ocean. There, hydrothermally derived DOM could potentially influence mesopelagic and epipelagic layers, and surface ocean sea ice interactions. 75 In this paper, we investigate the influence of hydrothermal activity on the DOM composition in the Arctic Ocean water column at the Aurora hydrothermal vent system. We suggest that the hydrothermal plume distributes thermally altered DOM to greater areas, where its composition will be altered through admixture with background sea water and through microbial processes. We characterise the water column DOM composition from the sea floor to the sea ice to assess the vertical extent of hydrothermal intrusion and its confluence with hydrophysical and chemical parameters. 80 2 Methods

Study area
The Aurora seamount (82.897 N, 6.255 W) is located at the southern tip of the ultraslow spreading (< 12.0 mm/year) Gakkel Ridge that extends for 1800 km across the Eurasian basin in the Arctic Ocean (DeMets et al., 2010). The seamount has a height of 300 m above the seafloor and is elongated in a southeast-northwest direction (Fig.1). The water column at the 85 mount's summit has a depth of 3800 m and it reaches 4500 m at the southern side of the ridge flank. The sea surface in this region is capped by a perennial ice cover. The Aurora seamount was explored by two research icebreakers in 2001 (Expedition AMORE) and evidence for high levels of hydrothermal activity were observed (fresh sulfide chimney structures, shimmering water, abundant biological activity) and anomalies in temperature, dissolved manganese, and light scattering were detected Michael et al., 2003). An active 'black smoker' hydrothermal vent at the Aurora site 90 (named after Aurora seamount) was later located ~100 m southwest of the summit in 2014, and high methane concentrations and temperature anomalies were documented along with a plume layer extending towards northwest of vent field (Boetius et al., 2014(Boetius et al., ,2015German and Boetius, 2017). As a continuation of the investigation in the area, and to further understand the role of the Gakkel Ridge in the scope of the global biogeography of chemosynthetic ecosystems (Vanreusel et al., 2009), the first HACON (Hot vents in an ice-covered ocean, HACON19) expedition attempted to reach the site again in 2019 (Bünz et 95 al., 2020). During the cruise, both active and inactive hydrothermal vents at the vent site were visually observed in much greater detail, using the Ocean Floor Observation and Bathymetry System (OFOBS; Purser et al., 2019). The active vent area https://doi.org/10.5194/bg-2021-350 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License. is evidently characterized by several vigorously venting black smokers, sulphide mounds, bacterial mats, thinly sedimented outcrops, and benthic communities typical of hydrothermal vent ecosystems (Bünz et al., 2020).

Sampling/ sub-sampling
We collected water samples from six stations near the Aurora hydrothermal vent field during the HACON19 research cruise on-board R/V Kronprins Haakon (September -October 2019) (Fig.1). Sea ice and the upper 1000 m of the water column were sampled at the Upper Layer (UL) station southwest of the vents. Plume (PL) and non-plume (NP1, NP2, NP3) stations were designated to observe the spreading of the hydrothermal plume and were measured from 2000 m depth down to 10 m 105 above the seafloor (<4500 m water depth). A Background station (BG) was sampled further away from the seamount ( Figure   1e). At this station, measurements were done from 2000 -4500 m water depth similarly to the PL and NP stations (Table 1).
Seawater samples were collected at selected sampling depths using Niskin bottles attached to a rosette (Error! Reference source not found.). Sea ice was sampled at the surface of UL, thawed in a high-density polyethylene (HDPE) container at room temperature and further treated similarly to the seawater samples collected with the rosette sampler. 110 Duplicate samples for methane concentration and stable C isotope ratios (δ 13 C-CH4) were collected immediately after the recovery of the rosette-sampler into 120 ml airtight syringes and pre-evacuated rubber-stoppered serum vials, respectively.
Care was taken to purge sampling lines of any bubbles, and methane samples were typically the first collected upon opening of Niskins. Methane concentrations from 2 independent samples (i.e. Niskin bottles closed at the same depth) typically agreed to within ±10-20% (2s; note that we could take duplicate samples only occasionally). For DOM and nutrient 115 characterization, seawater was first transferred into acid-washed (2% HCl) glass bottles (2 x 1135 ml) and filtered through GF/F filters (Whatman) within 2 hours after collection using a low-pressure vacuum. Nitrate, phosphate, silicate, ammonium, total nitrogen, total phosphorus, and dissolved organic carbon (DOC) samples were collected in 60 ml HDPE bottles from the filtrate and stored at -20 o C. Samples for dissolved inorganic carbon (DIC) and δ 13 C-DIC measurement were taken and poisoned with 10 µl HgCl2 and stored at 4 o C. Seawater aliquots for δ 18 O-H2O analysis were sampled in 20 ml clear 120 glass vials and stored at 4 o C. For solid phase extraction of DOM, 1 L of filtrate was acidified to pH 2 with HCl (37% v/v, Merck) and extracted with preconditioned (with 6 ml methanol + 12 ml pH 2 water) 500 mg PPL cartridges (BondElut, Agilent Technologies) as described previously (Dittmar et al., 2008). Loaded cartridges were dried under air vacuum for ~30 minutes and any organics retained in the cartridges were then eluted into pre-combusted, amber glass vials with 2 ml methanol and stored at -20 o C until analysis in the Fourier-transform ion cyclotron resonance mass spectrometer (FT-ICR 125 MS).

CTD and ADCP measurements
Sensor-based profiling of the water column was conducted with a Sea-Bird 911 plus CTD (Conductivity Temperature Depth) profiler (accuracies of 0.3 db, 0.001 o C, 0.002 for salinity (practical salinity scale, unitless)), SBE 43 coupled with dissolved oxygen sensor (calibrated by Winkler (1888) (Qian and Mopper, 1996). Deep Ocean DOC reference samples (Hansell Laboratory, University of Miami) were used to monitor precision and accuracy. DIC concentrations, δ 13 C-DIC and δ 18 O-H2O were measured on a 160 Thermo Scientific MAT253 isotope ratio mass spectrometry (IRMS) and presented in the delta notation as δ 13 C relative to Vienna Pee Dee Belemnite (VPDB) and δ 18 O relative to Vienna standard mean ocean water (VSMOW). The IRMS was calibrated with international standards VSMOW2, GISP; Greenland ice sheet precipitation, standard light Antarctic precipitation; SLAP2, VPDB and uncertainty of single measurements is ≤0.01 ‰. Dissolved methane concentrations were measured on-board following head-space extraction with nitrogen using a gas chromatograph (GC) equipped with a Flame 165 Ionization Detector following the procedure used in German et al. (2010). For δ 13 C-CH4 measurements, samples were  analyses, 100 µl aliquots of DOM samples were diluted 1:10 (v/v) with methanol (HPLC grade). ESI was selected as the ionization technique due to its ability to ionize polar, oxygen-containing compounds present in the DOM samples (Mopper et al., 2007). The samples were directly infused into the ion source by a syringe pump, operating at a flow rate 2 µL/min. Dry nitrogen was used as the drying and nebulizing gas. The ESI-generated ions were accumulated in the hexapole ion trap and 175 transferred into the ICR cell for trapping, excitation, and detection. For each spectrum, 100 scans were co-added over a mass/charge ratio (m/z) range of 150-2000. The Bruker Compass ftmsControl 2.1 software package was used for instrument control and data acquisition.

Fourier-Transform Ion Cyclotron Resonance Mass Spectrometry analyses and molecular formula assignments
Initial spectral post-processing was done with the Bruker DataAnalysis 5.0 SR1 software, including an internal mass recalibration with an in-house calibration list for the DOM samples. The data were then transferred to the PetroOrg IS-18.0.3 180 software (Omics LLC, Tallahassee, FL, USA) for molecular formula assignments. Only mass peaks with a signal-to-noise ratio (S/N) ≥ 5 were considered. In the assignments of the molecular formulae, monoisotopic compositions were limited to 12 C1-100 1 H1-200 14 N0-4 16 O0-30 32 S0-2 with a double bond equivalent (DBE) of 0-50 and a mass error of ≤ 1.0 ppm. Assigned molecular formulas were generated for each sample at positive and negative ion modes and combined to obtain a single molecular formula list. The combination of negative and positive mode ESI provides a better representation of DOM 185 composition for the carboxylic and acidic compounds (negative ESI) and hydrogen saturated aliphatic compounds (positive ESI) (Ohno et al., 2016;Sert et al., 2020).  Hodgkins et al. 195 (2016) as explained previously (Sert et al., 2020). For calculating percentages, the number of formulas that associated with a given category was divided by the total number of formulae in the sample and multiplied by 100. For example, the sum of the percentages of LPD, CAR, LGN and UHC is equal to 100, and similarly, the sum of the percentages of CHO, CHON and CHOS is equal to 100 for each sample.

Statistical analyses of DOM samples 200
Statistical analyses were performed in R (R Core Team, 2018). Diversity indices for DOM samples were calculated by the 'diversity' function analogous to biodiversity in ecology using the Shanon-Weawer formulation (Oksanen, 2019) as explained previously (Sert et al., 2020). A Bray-Curtis dissimilarity (distance) matrix (Bray and Curtis, 1957) was constructed by using relative abundances of formulas from 11 DOM samples that were collected >2000 m in stations BG, PL https://doi.org/10.5194/bg-2021-350 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License. and NP2. Hierarchical cluster analysis was applied for DOM compositions using a Ward clustering algorithm ('hclust' 205 function in Vegan) on a Bray-Curtis dissimilarity matrix. A heat map was constructed to visualize the sample clusters and the relative dissimilarities. Principal coordinate analysis (PCoA) was applied by using 'pcoa' function in the R package APE (Paradis and Schliep, 2019) on the Bray-Curtis dissimilarity matrix.

Water column physical and chemical properties 210
Water circulation in the Eurasian part of the Arctic Ocean is mainly characterized by Atlantic inflow and Arctic outflow through the Fram Strait, with a monthly mean volume of 9.5 and 11 Sv, respectively (Fahrbach et al., 2001). We categorize the water column in three water layers as epipelagic (0- m. From 450 to 1400 m, the temperature decreases from 0.9 to -0.5°C in a steep thermocline, forming AIW. AAW and AIW together form the East Greenland Current that carries Arctic waters southwest along the Greenland coasts and gradually 220 mixes with the convected surface waters (Rudels, 1995). Below the AIW layer, water temperature decreases to -0.73°C at around 2500 m where DW fills the Eurasian basin. Here, the density is largely controlled by salinity rather than temperature and the shape of the density profile is almost identical to the salinity profile ( Fig. 2b-c). Despite the dominant sea ice cover at our study area, preventing solar radiation from reaching the water column, a chlorophyll a (Chl a) fluorescence signal was detected down to 200 m depth (Fig. 2d). At all stations, the first 40 m below the sea surface showed a Chl a maximum with 225 an average concentration of 0.17 µg/L, followed by a decrease to 0.10 µg/L at 50 m. From this depth, concentrations gradually decreased further until the detection limit (0.01 µg/L) was reached at the bottom of the epipelagic layer (Fig. 2d).
Nutrient profiles, nutrient ratios, δ 13 C-DIC and δ 18 O all display a typical deep ocean surface-to-bottom gradient and were identical at all stations (Fig. 3). Nitrate, phosphate, and silicate concentrations were lowest at the surface (5 m) with 1, 0.44, and 6.1 µM, respectively, and increased gradually to average bathypelagic concentrations of 14.8, 0.96 and 12.0 µM at 2000 230 m (Fig. 3a-c). On the contrary, dissolved organic nutrients were highest in the epipelagic layer with the maximum concentrations of 92 µM for DOC at 100 m and 6.2 µM for DON at the surface ( Fig. 3e-f). Similarly, ammonium and DOP were detectable only in the epipelagic layer with average concentrations of 0.2 and 0.1 µM, respectively (see Fig. 3d for ammonium, DOP is not plotted). Nitrate, phosphate, and silicate concentrations in sea ice were below detection limit, however detectable concentrations of ammonium (0.6 µM), DOC (4.0 µM), and DON (0.6 µM) were measured in sea ice.     (Fig. 3h). This corresponds to water mass properties with contributions from sea ice melt waters and Arctic river 255 runoff with negative δ 18 O values in the upper 100 m within PSW, compared to highly uniform deep-water ratios. Our δ 18 O measurements are comparable to previous measurements in the area that reported ~0.3 ‰ in deep water and -2.6‰ to -1.6 ‰ in the upper 100 m (Bauch et al., 1995;Östlund and Hut, 1984).
A δ 13 C-DIC value of -13 ‰ was measured in sea ice. δ 13 C-DIC values decreased from 0.86 ‰ at the surface to 0.44 ‰ at 250 m and gradually increased to 0.62 at 1000 m. δ 13 C-DIC changed in a narrow range from 2000 m to bottom with average 260 composition of 1.03 ‰ (Fig. 3g). Sea ice coverage in the Arctic Ocean has been decreasing by ~9 % per decade, leading to a longer growing period for phytoplankton and associated increase in net primary production (Arrigo and van Dijken, 2015 and references therein). Taken together with the imprint of the anthropogenic CO2 (with δ 13 C value of ca. -8.5 ‰), δ 13 C-DIC values obtained at the sea ice and the upper layer of the water column are linked to contributions of several factors, including biological productivity, air sea gas exchange and sea ice or brine formation, which admixes DIC with a lighter δ 13 C signature 265 compared to that of deep waters. E.g., deep water δ 13 C-DIC values of 1.0 ±0.2 ‰ were found in the Canada Amundsen and Nansen basins (Bauch et al., 2015;de la Vega et al., 2019).
No substantial primary production occurred in the top 50 m layer of the water column, as shown by the nutrient minima and the Chl a maxima at ~40 m water depth. During the Arctic summer (from March to September), this layer gets depleted in inorganic nutrients and enriched in dissolved organic matter. In the deeper layers, nitrate/phosphate and nitrate/silicate ratios 270 are almost identical to the Redfield ratios of 16:1 and 1:1, respectively (Redfield, 1958) (Fig. 3i-j). Ratios of nitrate/phosphate (2.3:1 -5.2:1) and nitrate/silicate (1:5 -1:2.5) in the 0-100 m layer indicate that nitrate limits primary production. With the contribution of the epipelagic DON, however, surface/subsurface TDN/Phosphate (16.3:1) and TDN/Silicate (1.2:1) ratios became comparable with the mesopelagic/bathypelagic ratios (Fig. 3l). This indicates that DON may be used as a nitrogen source to support bacterial growth, considering that such low primary production cannot sustain 275 the bacterial demand in relation to limited light and low water temperature (Fouilland et al., 2018). Besides nitrogen, silicate was also limiting at the surface waters when compared to phosphate (silicate/phosphate ratios of ~13:1). Thus, perhaps due to diatom growth, silicate demand accompanies bacterial phosphate regeneration below the surface, lowering the silicate/phosphate ratios (7.6:1 -9.6:1) compared to the deep ocean (12.4:1) (Fig. 3k).

Dispersion of the hydrothermal plume 280
Turbidity anomalies, elevated methane concentrations and varying δ 13 C-CH4 values in the vertical profiles imply that the plume rises from the ~3900 m deep vent up to ~3000 m in the water column, and spreads laterally northward in the direction of the prevailing water current (Fig. 1d). Irrespective of the vent setting, tectonic structure, and background currents, previous studies and models imply that the fundamental physics behind the spreading of hydrothermal plumes are commonly https://doi.org/10.5194/bg-2021-350 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License. characterized by buoyant flow dynamics that have been described in depth for atmospheric systems (Tao et al., 2013). Upon 285 entering to the water column, a multi-phase plume rises and spreads laterally and carries dissolved gases and entrained particles that alter turbidity and methane concentrations in the non-buoyant layer (Fig. 4). Fundamental anomalies of venting fluids such as elevated temperature or low salinity can, on the other hand, become rapidly overprinted by ambient seawater background hydrography, making potential anomalies in these parameters undetectable with our instrumentation (Fig. 2).
Similarly, parameters that are not strongly influenced by the vent plume intrusion (i.e., nutrients, DOC and DON 290 concentrations) do not indicate any anomaly in relation to plume dispersion distant from the vent due to dilution with seawater (Fig. 3). This seems contradictory to molecular changes in DOM compositions; however, it must be stated that solid phase extractable DOM represents only a portion of bulk DOM (43 to 62 %; Dittmar et al., 2008), and compositional changes are not necessarily always coupled to large or substantial concentration changes. In essence, DOM molecular compositions may, more subtly, preserve some broad variations in endmember DOM traits than DOC and DON 295 concentrations (Osterholz et al., 2016).
Continuous drifting (often ~0.2 kts or more) of massive (km-scale, ~1 m thick) ice floes prevented the vessel from keeping stable positions on station while taking samples, and, together with time constraints, related to other operational needs of the expedition precluded us from directly sampling the buoyant stem of hydrothermal plume immediately above the active Aurora site. Due to the extreme depths (near 4 km), it furthermore typically took 3-4 hours to complete a single CTD cast, 300 so that the distance between start and end positions of a given cast were often 300-500 m apart. Nonetheless, the exact location of sampling was determined based on HiPAP data. The thickness of the non-buoyant plume also varied depending on the position of the sampling station in relation to the vent location, current direction, and bathymetric features. Based on turbidity anomalies, methane concentrations, and δ 13 C-CH4 contents, the plume layer was most evident at station PL, which was, counterintuitively, the station furthest away from the seamount (Fig. 4a-c), yet consistent with the prevailing current 305 direction (see Fig. 1d). Here, methane concentrations reached a maximum of 5.1 nM at 3363 m, which is >10-fold higher compared to ambient seawater concentrations (0.3 nM) as measured at similar depths at BG and coincide with the sharp maximum in turbidity. A small increase in turbidity was also measured at the other stations from 3000 to 3250 m implying some level of perturbation from the hydrothermal intrusion (Fig. 4c), but these were not as strong as in PL and not accompanied with an increase in δ 13 C-CH4 values or methane concentrations. We found one exception to this at NP1, which 310 was also at the northern side of the vents. Here, methane concentration (1.7 nM) and δ 13 C-CH4 (-43‰) were elevated at 3350 m, i.e., just below the turbidity anomaly.
Deep water methane stable carbon isotope compositions in the Arctic Ocean have previously been reported with values 320 ranging from -63 ‰ to -58‰ in the Beaufort and Chukchi Seas, along with higher concentrations of methane of up to 55 nM (Damm et al., 2008;Kudo et al., 2018;Lorenson et al., 2016). These are attributed to microbial methanogenesis overprinting the atmospheric methane equilibrium signature (typically ~3.5 nM and -47 ‰). We observed a similar range of δ 13 C-CH4 values with much lower concentrations of methane (e.g., -64.7 ‰ and 0.3 nM) at BG (3250 m) and argue that these values may also be caused by methanogenesis, possibly in sediments or even in the water column (Damm et al., 2010;Grossart et 325 al., 2011;Karl et al., 2008) but we suggest that methanotrophs also consume methane there (Reeburgh, 2007). In contrast, we found relatively heavy methane isotope values together with maximum methane concentrations in the plume layer at PL (e.g., -33 ‰ and 5.1 nM) (Fig. 4). Methane at hydrothermal vents is typically produced non-biologically, either by abiogenic (C reduction) or thermogenic (breakdown) processes and is generally characterised by much higher δ 13 C -values (Whiticar, To further probe the dynamics of water column methane dynamics at Aurora, we applied a 'Keeling' mixing model approach (Keeling, 1961;Keir et al., 2006). This end-member isotope mixing model is widely used in the fields of 340 atmospheric chemistry/physics to analyse admixtures of two isotopically distinct gas sources, and to determine the source isotope composition (e.g., Pataki et al., 2003). In a Keeling plot, the reciprocal of concentration is plotted against isotope composition (Fig. 5). In case of a linear mixing, the values scatter along a straight line and the intercept with the y-axis denotes the isotope value of the source signal. Our values of methane concentration and isotope composition show predominantly linear behaviour, indicating that the mixing and dilution dominates the observed concentration and isotope 345 systematics, as has been found in other Arctic settings (Damm and Budéus, 2003;Graves et al., 2015). More complex mixing scenarios involving mixing and methane oxidation were also suggested to explain methane isotope patterns in hydrothermal plumes (Keir et al., 2009). Though we cannot completely rule out some extent of microbial methane oxidation (particularly in the most distal samples), this process is associated with a significant kinetic isotope effect (Whiticar, 1999), and can even result in extremely high δ 13 C-CH4 values in microbially 'aged' hydrothermal plumes (Keir et al. 2006;Cowen 350 et al. 2002) . If such oxidation were pronounced, it would lead to substantial curvature in a Keeling plot by shifting the residual methane towards heavier isotopic values at lower concentrations (higher 1/[CH4]).  Considering a 'mixing without oxidation' scenario and under the assumption of linear mixing (Keir et al., 2009), we estimate 360 the δ 13 CH4 value of the hydrothermal fluid source to be about -36 ‰ at the Aurora hydrothermal field (Fig. 5). The estimate is interesting, since the Aurora hydrothermal system has, thus far, been postulated to be hosted in ultramafic rock Marcon et al., 2017). The δ 13 CH4 estimate presented here is substantially lower than both plume and endmember δ 13 CH4 values of many Mid Atlantic ridge hydrothermal vents known to be hosted in or influenced by ultramafic rock (ca. -5 to -20, e.g., Keir et al., 2006;Wang et al., 2018). Such ranges are assumed to be characteristic of putatively 365 'abiotic' methane (e.g., Wang et al., 2018), most likely derived from leaching of rock-hosted gas inclusions (e.g. Grozeva et al., 2020). A δ 13 CH4 value -13.1‰ within this range of has also been reported for a 'closest approach' in a CH4and H2-rich hydrothermal plume further east on Gakkel Ridge at 87°N (McDermott et al., 2017; confirmed by the same cavity ring-down spectrometry approach used here), also postulated to reflect ultramafic-influence. In contrast, δ 13 CH4 values in hydrothermal fluids either heavily influenced by (or hosted in) sediments undergoing substantial thermal alteration of associated 370 sedimentary organic matter tend to be generally more 13 C-depleted (ranging from -25 to -56‰; Baumberger et al. 2016 and references therein). Indeed, similarly 'intermediate' values of δ 13 CH4 (-27 to -29‰) compared to Mid-Atlantic Ridge hydrothermal systems were also obtained in the bare-rock hosted, but sediment-influenced, Loki's Castle hydrothermal system farther to the south on the ultraslow spreading Mohns-Knipovich ridge junction of the Arctic Mid Ocean Ridge system (Baumberger et al., 2016;Pedersen et al., 2010). There, exceptionally high methane (15.5 mM) and ammonium (6.1 375 mM) concentrations together with low C1/C2+ ratios are attributed to subsurface hydrothermal fluid-sediment interactions (linked to sediment accumulation at ridge flanks and ridge valley, Baumberger et al., 2016;Pedersen et al., 2010).
A greater extent of plume dispersion and somewhat higher methane concentrations (≤ 32 nM) were previously reported within ~200 m of the Aurora hydrothermal vent area (Boetius, 2015;German and Boetius, 2017) but the maximum concentrations reported here (2.8-5.1 nM, in station PL) do at least coincide with the highest δ 13 CH4 values (-31 to -33‰) 380 and maximum water column turbidity. Hence, we tentatively suggest that hydrothermal fluids at Aurora might, on the basis of these closest approach plume δ 13 CH4 values alone, potentially contain a thermogenic organic matter influence. However, as yet, the endmember methane concentration and carbon isotope composition of the hydrothermal fluids venting at Aurora are not known, precluding a proper assessment of the geochemical processes influencing methane there.

Composition of DOM in the sea ice, water column and in the hydrothermal plume layer 385
Analyses of DOM extracts revealed 8093 unique formulas in the whole dataset, with the molecular mass range between 150 and 1000 Da. From all unique formulas, 2614 and 2391 were uniquely obtained by negative and positive ESI, respectively and 3088 formulas were assigned in both (see Fig. S3). In agreement with previous findings (Ohno et al., 2016;Sert et al., 2020), unique positive ESI formulas were markedly more hydrogen saturated ( carbohydrates, lignin-like and aromatic compounds were more abundant in negative mode. Considering the features obtained in different modes, combining positive and negative ESI datasets provides a considerable advantage for differentiating samples.

Sea ice 395
The DOM composition of sea ice was determined from a single sample and indicates that it was predominantly more hydrogen-saturated relative to surface water composition (i.e., higher H/C ratios), with lower number of formulae, average abundances, average molecular weights, and considerably high LPD content (Fig. 6a-k, Fig. S2). Differences in molecular composition between seawater and sea ice are in agreement with the previous observations in sea ice DOM (Longnecker, 2015;Retelletti Brogi et al., 2018) and lake ice DOM (Liu et al., 2020;Santibáñez et al., 2019;Xu et al., 2020). implying a 400 selective transport of DOM with less complex, aliphatic compounds -rather than larger aromatic oligomers, during ice formation. The hydrodynamics of DOM transport are not known between phases, but previous observations show that DOM segregates between water and ice, favouring lower MW formulae and less aromatic fractions in the ice phase (Belzile et al., 2002;Santibáñez et al., 2019). In the Arctic Ocean, this partition process increases the bioavailability of terrestrial DOM from rivers during initial ice formation and contributes to its removal by retaining biologically active components (Jørgensen 405 et al., 2015). In addition to the abiotic fractionation and transport, biofilm mediated microbial degradation and photooxidation were previously shown to be responsible for fast degradation of petroleum biopolymers in sea ice (Vergeynst et al., 2019). High LPD abundance and low average MW in sea ice DOM may also be attributed to a fresh production by ice algae or heterotrophic degradation of DOM (Hill and Zimmerman, 2016;Li et al., 2019;Retelletti Brogi et al., 2018).

Upper 1000 m of the water column
From the sea surface down to 1000 m, changes in DOM indices such as the number of formulas, molecular diversity, and 420 average abundances are consistent with changes in nutrient and DOC showing their maxima at 5 m ( Fig. 6a-d, Fig. S2). This is in agreement with the expected pace and diversity of biological processes in this water layer, i.e., dissolved nutrient uptake by phytoplankton and the contribution to the synthesis of dissolved and particulate organic matter (Benner et al., 1992;Hedges, 1992). Below this layer, DOM abundances decreased and showed a less diverse composition below the photic zone with decreasing light availability and heterotrophic consumption (Fig. 6b-d, Fig. S2). 425 The uptake of bioavailable DOM alongside nutrients could explain the distribution of formulas of different chemical classes in the 0-1000 m layer. For example, lower abundances of LPD and LGN at the surface against the higher CAR and UNC contents ( Fig. 6e-h) indicate the selective use of the more hydrogenated aliphatic compounds compared to oxygenated or aromatic formulas. Similarly, CHON and CHOS heteroatom contents were maximal at the surface ( Fig. 6i-k), indicating the increase in the molecular variability and abundance of fresh DOM (Coch et al., 2019;Hertkorn et al., 2016;Jaffé et al., 430 2012). In contrast to bulk dissolve organic nitrogen (DON) concentrations and CHOS heteroatom contents, percentages of CHON formulas systematically increased from subsurface (~33%) to 1000 m to reach ~50 % of deep-water abundances (Fig.   6j). Taking this together with the CAR and LPD percentages at sub surface layers, surface derived carbohydrates are presumably used with increasing depth by heterotrophs to build up cell materials where N containing proteins dominate intracellular biochemicals (Aluwihare and Meador, 2008) . 435 For a detailed examination of the DOM structure in the upper layer (i.e., 0-1000 m) a more extensive data set would be required. However, changes in the percentages of the molecular formulas seem to indicate the gradual transformation of semi-labile DOM at the surface to semi-refractory DOM in the mesopelagic zone (200-1000m) (see Hansell, 2013). This trend was evident in most of the profiles displaying molecular percentages (Fig. 6e-k). For instance, when microbial growth is limited by the availability of inorganic nutrients (nitrate in our data) in oligotrophic (e.g. Arctic Ocean) and ultra-440 oligotrophic regions (e.g. Mediterranean), semi-labile DOM (e.g. polysaccharides and large proteins) resist rapid degradation and temporarily accumulate in the surface layer (Dittmar and Stubbins, 2014). Since, the bacterial community in the surface waters has limited capacity to degrade semi-labile DOM contrary to mesopelagic bacterial communities, which have the required metabolic mechanisms to metabolize semi-labile fractions (Repeta, 2015 and the references therein). While semilabile DOM gradually gets decomposed with depth, semi-refractory DOM may accumulate with same principle of changing 445 bacterial communities and molecular reactivities along with the microbial pump of refractory DOM to the deep ocean (Jiao et al., 2010).

Hydrothermal plume and the surroundings
The influence of the plume dispersion on DOM composition was investigated in 11 samples from three stations below 2000 m: BG (background), PL (plume) and NP2 (non-plume). The approximate locations of these samples relative to hydrothermal vents and the possible route of the plume dispersion are depicted in Fig. 7a. PCoA and hierarchical clustering indicate a similar composition in the samples obtained from the same station (Fig. 7b-c). Exceptions are samples PL-3000 and BG-2000 which display higher molecular diversity and number of formulas and are therefore more similar to NP2 samples (Fig. 7b). Nevertheless, samples from different depths but from the same station prominently group into three separate clusters (Fig. 7a-c). 455 The detailed DOM formulae compositions as depicted by van Krevelen diagrams reveal compositional differences between BG, PL and NP2 samples (Fig. 8) Considering that the plume dispersion mainly occurs along a lateral transect, we collected samples at the same depths along this lateral transect to be able to compare the differences in the DOM composition (e.g., BG-3250 vs NP2-3250 m and PL-3500 vs NP2-3500; Fig. 8b, 8h, and 8j-k, respectively). Formulas that were unique (i.e., formulae that were not present in 465 samples from the same layer in other stations) or distinct molecular indices (i.e., percentages in different molecular classes like LPD or CAR) were used to evaluate changes in DOM composition. Variations in these molecular properties might be related to distinct biogeochemical settings associated with the plume dispersion, as found with turbidity, methane concentrations, and δ 13 C-CH4 contents. For example, low molecular diversity, relative abundances and LPD content in PL-3500 and PL-3400 ( Fig. 5a-e) indicate the effect of hydrothermal intrusion or the molecular composition carried by the 470 plume, in line with previous experimental and environmental data suggesting preferential removal of higher molecular weight and oxygen rich compounds under hydrothermal conditions (Hawkes et al., 2016;Rossel et al., 2017).  Given that the DOM composition throughout the region is thermally altered, the nature of thermal alteration of organic matter at the hydrothermal systems has previously formulated as (i) reduction of organic compounds in low temperature regions (~60-150 o C) to generate products from weaker bonds (e.g. formation of petroleum, aliphatic hydrocarbons), (Simoneit, 1992) (ii) oxidation at higher temperature conditions (e.g. PAHs and alkanones) (Dittmar and Koch, 2006;495 Simoneit, 1995) and (iii) abiotic synthesis of organic matter by thermocatalytic (Fischer-Tropsch-type) reactions (McCollom et al., 1999b;McCollom and Seewald, 2007). Therefore, it is difficult to categorize thermal processes as either a source or a sink mechanism for marine DOM over a range of temperatures from ~60 o C to ~150 o C due to for instance, formation of petroleum products (Simoneit, 1992), water soluble vitamins and amino acids (Longnecker et al., 2018), elevated concentrations of formate, acetate concentrations (Lang et al., 2010;McDermott et al. 2015), fractions of labile 500 compositions (Hansen et al., 2019), and formation of iron and sulphur bearing organic ligands (Yücel et al., 2011). During high temperature hydrothermal circulation, however, ~94 % of the solid phase extractable DOC is typically removed (Hawkes et al., 2015). Findings at several Mid-Atlantic Ridge hydrothermal systems, and temperature-controlled hydrothermal experiments, showed that DOM is highly unstable and effectively removed during hydrothermal circulation, with a reduction in the abundance of recalcitrant DOM, and a decrease in number of formulas and average molecular mass 505 (Hawkes et al., 2015(Hawkes et al., , 2016Rossel et al., 2015Rossel et al., , 2017. This agrees with our observations of a decreasing numbers of molecular formulae, molecular diversity, average abundance and LPD composition in the samples where plume-related hydrochemistry was manifested by a higher level of turbidity, five times higher methane concentration, and markedly heavier methane carbon isotope compositions. We can therefore assume that the compositional differences between background seawater in BG samples and DOM in the plume layer (PL-3400) 510 may be due to hydrothermal degradation, or some other chemical combination of reactions that result in a molecular signal similar to those reported for hydrothermal fluids. Due to the difficult sampling situation in the high Arctic marine environment, we were unfortunately unable to take samples directly from the buoyant plume of the vent to properly illustrate the distribution and composition of hydrothermal endmember DOM. Our analysis, therefore, is solely based on compositional comparison between samples that were different in their relative position to the vent and its non-buoyant 515 plume. However, we do suggest that a portion of the observed composition in the PL is associated with the hydrothermal fluids, which are efficient at scavenging hydrothermal organic compounds from the vent area, and transporting them from the heated subsurface (Simoneit, 1992). Accordingly, the temperature profile at PL implies that the hydrothermal fluids must have been cooled down to the ambient temperature before reaching the sampling point. Therefore, any thermal modifications in DOM compositions must have occurred either before entering the lithosphere or in areas adjacent to the vents themselves, 520 being merely carried along with the plume-transported particles, dissolved gasses, or water masses and gradually diluted with the surrounding waters until it reached the sampling location.
Alternatively, unique formula compositions in NP2 with higher oxygen saturation (i.e., O:C≥ 1) or ~2% higher CAR content (see Fig. 5f) could be a product of microbial enrichment attributed to chemosynthetic activity at the southern part of the https://doi.org/10.5194/bg-2021-350 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License. seamount where the faunal communities, sponges, amphipods and traces of biological activity were more abundant as logged 525 by video and photographic transects from OFOBS dives (Bünz et al., 2020). Nevertheless, based on our data, a clear and direct link between activity of faunal communities and DOM molecular compositions could not be made. Because of this, we cannot yet suggest any mechanism that can transfer modifications of benthic communities to pelagic DOM compositions.
One other plausible explanation for the differences in molecular composition in PL and NP2 compared to BG could be a change in the current direction. The current direction could have been different in the past compared to what we observed on 530 our cruise. If this were the case, the observed composition at the NP2 station could be an imprint of a previous composition of the hydrothermal vent or some other form of environmental gradient that modified and diversified the DOM composition at NP2, while the PL samples reflect the current composition at the sampling location. The higher abundance of CHOS heteroatom composition in the NP2 samples could be evidence of that. A very recent study in Kairei and Pelagia hydrothermal vents in the Indian Ocean indicate 17% more CHOS molecular formulas in DOM from hot vent fluids along 535 with the elevated hydrogen sulphide concentrations compared to surrounding seawater (Noowong et al., 2021). Similarly, reproduction of sulphur containing compounds are found to occur under experimental hydrothermal conditions at several selected temperatures (Hawkes et al., 2016) and at less acidic conditions (Rossel et al., 2017).

Summary and Conclusion
Our investigations of the water column above the Aurora hydrothermal vent field (Gakkel Ridge, Arctic Ocean) show that 540 hydrothermal venting in this region displaces thermally degraded DOM compositions within the hydrothermal plume, along with 13 C enriched methane and other vent-derived properties. We show that DOM compositions in the plume layer samples had lower numbers of molecular formulae, molecular diversity, average abundance, and percentage of formulas assigned to the LPD molecular class. However, at a site with higher vent related biological imprinting and lower influence from plume dispersion, DOM compositions had higher numbers of formulae, molecular diversity, and CAR percentages. We 545 characterized the water column hydrophysical and biochemical properties, observing that nutrient distributions and DOM compositions mainly reflect vertical changes in primary production, atmospheric gas exchange, riverine influence and water mass distributions in the upper 1000 m layer, while the water column below 2000 m was quite uniform. DOM composition in sea ice was more labile, with markedly high LPD and low molecular weight compositions than the ocean surface waters, suggesting selective transport during ice formation and new production by ice algae. 550 DOM compositions in the Arctic Ocean are prone to changes in the upper ocean layers, in parallel to the changes in global climate and decline in sea ice extent. However, deep ocean DOM is mostly refractory, displaying remnants of ancient biological processes. Hydrothermal vents in the deep ocean could therefore be a dynamic source of organic matter and perhaps influence overlying pelagic ecosystems. There are several experimental studies on how DOM is modified during hydrothermal processes under different physical and chemical conditions, however, there is a gap of knowledge on to what 555 extent hydrothermally and chemosynthetically modified DOM spreads along plume layers, altering water column https://doi.org/10.5194/bg-2021-350 Preprint. Discussion started: 3 January 2022 c Author(s) 2022. CC BY 4.0 License.
biogeochemistry. To the best of our knowledge, our study is the first attempt to differentiate hydrothermal plume and nonplume DOM compositions in the deep Arctic Ocean; clearly more studies should address this in different hydrothermal systems and other plume-generating deep-sea environments.

Data and code availability 560
The data and the codes for generating figures and statistical analyses are available in UiT Open Research Data repository (Sert et al., 2021).

Author contributions
MFS designed the study and performed data acquisition with supervision from AS, FG and HN. EPR and KPH analyzed shipboard methane concentrations and methane carbon isotope compositions, respectively. TK and JJ analyzed dissolved 565 organic matter samples, generated formula lists and wrote mass spectrometry methodology. MFS wrote the manuscript with contributions from all coauthors.

Competing Interests
The authors have the following competing interests: One of the (co-)authors is a member of the editorial board of Biogeosciences. 570 ICR MS project; grant agreement 731077. The FT-ICR MS facility is also supported by Biocenter Finland, Biocenter Kuopio and the European Regional Development Fund (grant A70135). KPH contributed this research through support from the Exo-Ocean Worlds project within the Astrobiology Program at NASA, and through support from the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (80NM0018D0004). The publication charges for this article have been funded by a grant from the publication fund of UiT 580 The Arctic University of Norway.