Variations of microbial communities and substrate regimes in the eastern Fram Strait between summer and fall

.


Introduction
Marine phytoplankton release species-specific organic matter composed of high carbon and nitrogen, which is remineralized by heterotrophic microbes (Biersmith and Benner, 1998;Hedges et al., 2002).Organic matter is constantly produced and degraded, allowing it to be classified into either particulate organic matter (>0.7 μm) or dissolved organic matter (DOM; <0.7 μm) (Benner et al., 1992).Particularly DOM, as the largest carbon reservoir, can further be partitioned into the low-molecular-weight (<1 kDa) or high-molecular-weight fraction (>1 kDa) respectively (Hansell et al., 2009).The variable reactivity of low-molecular-weight DOM includes monomers such as free carbohydrates and amino acids, whereas high-molecular-weight DOM includes biopolymers such as dissolved combined carbohydrates (DCCHO) and dissolved hydrolyzable amino acids (DHAA).
DCCHO and DHAA serve as substrates for bacteria and archaea.For instance, Flavobacteriaceae are specialized in polysaccharide degradation suggesting a possible link to their prevalence during phytoplankton blooms in temperate and polar habitats (Kirchman et al., 2010;Wilson et al., 2017;Fadeev et al., 2018;Cardozo Mino et al., 2021).Gammaproteobacteria such as Porticoccaceae can be abundant in response to algal decay, as the release of high-molecular-weight DOM is particularly prominent during and towards the end of a phytoplankton bloom (Engel et al., 2011;Teeling et al., 2012).
In the Arctic Ocean, phytoplankton and microbes are controlled by pre-existing environmental conditions and are increasingly influenced by sea ice loss, glacial runoff, or permafrost melt (Boetius et al., 2015).For example, the seasonal cycles in light and nutrient availability are closely coupled to phytoplankton biomass in the Fram Strait (Randelhoff et al., 2018).As the productive season progresses, phytoplankton release biopolymers that promote the growth of microbial communities, which remineralize this organic matter (Piontek et al., 2014;von Jackowski et al., 2020).To continue understanding the substrate requirements of microbes, this study expands on demonstrated seasonal changes in biopolymer concentrations and microbial composition in the Fram Strait as the primary inflow to the Arctic Ocean (von Jackowski et al., 2020;Wietz et al., 2021).We hypothesized that the pronounced seasonal change in labile biopolymers would considerably shift the relative abundances of biopolymerdegrading microbes between summer and fall.Furthermore, given the overlap of biochemical and microbial diversity datasets, we investigated whether individual DCCHO and DHAA components are linked to specific microbial taxa.Assessing relationships between the microbial community and the biopolymer pool over the seasonal cycle is important for understanding carbon cycling in the Arctic Ocean.The approach establishes a baseline of substrate regimes and their re-mineralization in the Fram Strait between summer and fall.

Sampling
Samples for biochemical and microbial analyses were collected in the upper 100 m of the water column using a rosette sampler equipped with 24 Niskin bottles.The rosette sampler was coupled to a CTD (SBE 911plus, Sea-bird, USA) equipped with two temperature probes, two conductivity probes, one Digiquartz pressure sensor, one WET Labs ECO-AFL/FL fluorometer, one WET Labs C-Star transmissometer and one altimeter.The sampling depths were chosen based on the output of the WET Labs ECO-AFL/FL fluorometer that was used to estimate phytoplankton biomass and identify the deep chlorophyll maximum (DCM).Specifically, four depths were of particular interest: surface (5 or 10 m), the DCM, below the DCM (BDCM), and 100 m.In summer, surface water was consistently sampled at 10 m, DCM at 20-43 m, and BDCM at 30-52 m.In fall, the surface water was sampled at 5 m, DCM at 20-34 m, and BDCM at 40-50 m.The DCM was distinct at all stations during summer and less clear during fall, but to be consistent, all mid-water column peaks of fluorescence were handled as the DCM during both seasons.The CTD data are archived in the PANGAEA World Data Center (von Appen et al., 2019;von Jackowski and Engel, 2019).

Particulate and dissolved organic matter
Samples for particulate organic carbon (POC) were collected by filtering 1 to 4 L of seawater onto 0.7 μm poresized pre-combusted GF/F filters (500 C, 4 h) and stored at À20 C. Back in the laboratory, the thawed filters were soaked in 0.1 M HCl to remove inorganic carbon, dried at 60 C for 12 h, and measured using a EURO EA CHNS-O Elemental Analyser (HEKAtech GmbH, Germany) (Sharp, 1974).

Microbial production
Rates of primary production (PP) were measured in situ using the 14 C method modified after Engel et al. (2013).The seawater was incubated in duplicates with additional dark controls for a duration of 24 h.To also account for the changing diurnal cycle, the samples were incubated under constant light during summer, while the hours of light roughly matched the given day in fall.For example, incubation times decreased from 14 h (e.g. on 16.09.2018)at the beginning of the MSM77 expedition to 9.5 h (e.g.04.10.2018) at the end of the MSM77 expedition.
Each incubation was fractionated and terminated in three subsamples: total PP (PP-TOC), particulate PP (PP-POC) and dissolved PP (PP-DOC).The PP-TOC fraction was taken directly from the incubation flask, the PP-POC fraction was filtered onto a 25 mm 0.4 μm-pore-sized Nucleopore track-etched polycarbonate filter (Whatman, GE Healthcare Life Sciences, UK), and the PP-DOC fraction was subsampled from the filtrate.To convert the activities into a rate, we converted total alkalinity into dissolved inorganic carbon using the package 'seacarb ' (v.3.3.0).Data for the corresponding bacterial production (BP) based on the 3 H-microcentrifuge method (Smith and Azam, 1992) were incorporated from von Jackowski et al. (2020).

Cell abundance
Samples for cell abundance were fixed on board with glutardialdehyde at 2% final concentration and stored frozen (À80 C) until analysis by flow cytometry (FACSCalibur, Becton Dickinson, USA).The flow cytometer was calibrated and standardized with TruCount beads (Becton Dickinson).Due to a detection limit of 50 μm, samples were filtered through a mesh before counting using the Cell Quest 3.3 software with a DL of 2000 events s À1 .Orange autofluorescence was used to detect the phycoerythrin of cyanobacteria (Synechococcus) and cryptophytes, whereas red fluorescence was used to detect and distinguish picoeukaryotes (<2 μm) from nanoeukaryotes ($2-20 μm) (Read et al., 2014).Samples for heterotrophic cell analysis were filtered and stained with SybrGreenI (Invitrogen, USA), with data from the corresponding samples incorporated from von Jackowski et al. (2020).

Microbial community analysis
Seawater samples (1-4 L) were filtered through 0.22-μm Sterivex cartridges (Merck Millipore, USA) using a peristaltic pump within 1.5-2 h after retrieval of the CTD rosette and stored frozen (À80 C) until extraction.Filters were transferred from cartridges into kit-supplied tubes, and genomic DNA was isolated using a combined mechanical and chemical procedure using the PowerWater ® DNA Isolation Kit (QIAGEN, Germany).Amplicon libraries were prepared according to the 16S Metagenomic Sequencing Library protocol (Illumina, USA) using universal 16S rRNA gene primers 515F and 926R that covered the V4-V5 hypervariable region (Parada et al., 2016).Sequences were acquired using a 2 Â 300 bp paired-end run on a MiSeq platform (Illumina) at CeBiTec (Bielefeld, Germany).Sequence data have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB43926, using the data brokerage service of the German Federation for Biological Data (GFBio; Diepenbroek et al., 2014) in compliance with MIxS standards (Yilmaz et al., 2011).

Statistical analyses
Parameters that showed significant differences between the BDCM and 100 m were subsequently grouped into surface-to-BCDM ('surface', 'DCM' and 'BDCM') and 100 m.Detailed results of statistical analyses are documented in Table S2.Scripts are publically available at https://github.com/anabelvonjackowski.
Statistical analyses applied to the biogeochemical data included a Wilcoxon Rank Sum Test, analysis of variances (ANOVA) and a mixed model.If the interactive terms of the ANOVA were significant, they were fed into multiple contrast tests (Laird and Ware, 1982;Verbeke and Molenberghs, 2000) that included season ('summer', 'fall') and depth ('surface', 'DCM', 'BDCM' and '100 m') with the station as the random factor.

Study area
This study focused on eight stations within the Long-Term Ecological Research observatory HAUSGARTEN in the eastern Fram Strait (Soltwedel et al., 2016).Samples were collected during the expeditions PS114 with RV Polarstern from July 16th to July 23rd, 2018 (herein referred to as summer) and MSM77 with RV Maria S. Merian from September 16th to October 4th, 2018 (herein referred to as fall; Fig. 1A; Table S1).Four depths were of particular interest during this study: surface (5 or 10 m), the DCM, below the DCM (BDCM), and 100 m; in summer, the DCM at 20-43 m and BDCM at 30-52 m; in fall, the DCM at 20-34 m and BDCM at 40-50 m.
In the upper 100 m, temperature increased slightly from 4.53 AE 1.45 C in summer (n = 32) to 5.35 AE 0.94 C in fall (n = 30, Fig. S1) (von Appen et al., 2019;von Jackowski and Engel, 2019).The salinity was 34.83 AE 0.48 PSU in summer (n = 32) and 34.90 AE 0.26 PSU in fall (n = 30) (von Appen et al., 2019;von Jackowski and Engel, 2019).These warm and saline conditions are characteristic of the Atlantic water masses in the West Spitsbergen Current (WSC; Aagaard et al., 1985) and likely displaced the ice edge north of 80 N during summer and fall.The intensity of the WSC is seasonally variable and influences the advection of carbon into the eastern Fram Strait, the major gateway between the Atlantic and Arctic Ocean (von Appen et al., 2016;Vernet et al., 2019).

Seasonal variability in POC and autotrophic microbes
POC decreased threefold in the upper 100 m from 18.07 AE 10.05 μmol L À1 in summer (n = 32) to 5.25 AE 3.78 μmol L À1 in fall (n = 32, Wilcoxon Rank Sum Test p < 0.001; Fig. 1B).The spatial variability of POC within the upper 100 m showed a significant twofold decrease below the deep chlorophyll maximum (BDCM; 30-52 m) and 100 m in summer, while no difference was observed during fall (ANOVA Season: Depth F 3,56 = 5.39, p < 0.01; Multiple Contrast Test p < 0.05; Table S2).The seasonal change in POC concentrations was clearly related to phytoplankton dynamics, with a decline in chlorophyll-a (von Jackowski et al., 2020) and total biovolume (Lampe et al., 2021) from summer to fall in 2018.PP decreased more than twofold in the dissolved organic carbon fraction (PP-DOC), with a significant threefold decrease in the particulate fraction (PP-POC) from summer to fall (Wilcoxon Rank Sum Test p < 0.05).PP-DOC declined from 1.74 AE 0.86 μmol C L À1 d À1 in summer (n = 8) to 0.62 AE 0.69 μmol C L À1 d À1 in fall (n = 10).PP-POC declined from 1.42 AE 1.68 μmol C L À1 d À1 in summer (n = 8) to 0.29 AE 0.39 μmol C L À1 d À1 in fall (n = 10).Simultaneously, cell abundances of cryptophytes and picoeukaryotes significantly decreased between summer and fall (Wilcoxon Rank Sum Test p < 0.001, Table S2).Cryptophyte abundances declined from 0.7 AE 0.4 cells L À1 in summer (n = 32) to 0.2 AE 0.2 cells L À1 in fall (n = 32), overall being significantly less abundant in the BDCM and 100 m (ANOVA Season:Depth p < 0.05, Multiple Contrast Test p < 0.05 and p < 0.01 respectively, Table S2).Similarly, picoeukaryote abundances declined from 4.5 AE 3.8 Â 10 6 cells L À1 in summer (n = 32) to 2.3 AE 3.9 Â 10 6 cells L À1 in fall (n = 32).Converting picoeukaryote abundances into carbon, by assuming a carbon biomass conversion factor of 530 fg C L À1 (Worden et al., 2004), showed that picoeukaryotes contributed 2.4 AE 2.0 μg C L À1 or 1.1% POC to the carbon pool in summer (n = 32) and 1.2 AE 2.1 μg C L À1 or 1.3% POC in fall (n = 32, Fig. 1B).Nanoeukaryotic abundance showed the least seasonal change and decreased from 5.1 AE 4.1 Â 10 6 cells L À1 in summer (n = 32) to 4.9 AE 5.2 Â 10 6 cells L À1 during fall (n = 32).It is likely that the PP-DOC fraction decreased as a result of declining cryptophyte, picoeukaryote and nanoeukaryote cell numbers.In particular, picoeukaryotes might be less influenced by temperature and salinity but instead driven by the seasonal light intensity and duration within the WSC (Paulsen et al., 2016).All in all, our data support the observed decline in relative biovolume from summer to fall in the eastern Fram Strait (Lampe et al., 2021).
Absolute and relative abundances of the cyanobacterium Synechococcus significantly increased from summer to fall, evident in both flow cytometry and 16S rRNA data (Wilcoxon Rank Sum Test p < 0.001, Table S2).Synechococcus abundances increased 34-fold from 0.4 AE 0.5 Â 10 6 cells L À1 in summer (n = 32) to 14.6 AE 22.5 Â 10 6 cells L À1 in fall (n = 32, Fig. 1B and  C).Converting the Synechococcus abundances into carbon, by assuming a biomass conversion factor of 109.5 fg C L À1 (Kana and Glibert, 1987), showed that Synechococcus contributed 0.05 AE 0.05 μg C L À1 or 0.02% POC to the carbon pool in summer (n = 32) and subsequently increased 117-fold to 1.6 AE 2.5 μg C L À1 or 2.2% POC in fall (n = 32, Fig. 1B and C).According to our study and Paulsen et al. (2016), the seasonal pattern in Synechococcus abundances is as follows in the upper 100 m of the Fram Strait: 0.43 AE 0.45 Â 10 6 cells L À1 in July (this study), 3.02 AE 3.79 Â 10 6 cells L À1 in August  (Paulsen et al., 2016), 14.60 AE 22.54 Â 10 6 cells L À1 in September/October (this study), and 0.48 AE 0.08 Â 10 6 cells L À1 in November (Paulsen et al., 2016).This seasonal variability, as described here, could have farreaching implications for the Arctic epipelagic carbon pool, given the low carbon-to-nitrogen ratios of Cyanobacteria (Finkel et al., 2016) and the biopolymer pool if Synechococcus employ an osmotrophic strategy (Yelton et al., 2016).Overall, there is a need for more continuous measurements to verify the seasonal variability and possible temperature-induced northward expansion of phytoplankton communities entering the Arctic Ocean through the Fram Strait (Orkney et al., 2020).

Quality of the biopolymer pool
To assess the carbohydrate content of the biopolymer pool, we differentiated DCCHO into neutral sugars, acidic sugars and amino sugars.Concentrations of neutral and acidic sugars significantly differed between summer and fall (Wilcoxon Rank Sum Test p < 0.001), particularly between the BDCM (30-52 m) and 100 m during summer (ANOVA Season:Depth p < 0.01; Multiple Contrast Test p < 0.05; Table S2).Neutral sugars accounted for the largest biopolymer proportion to the dissolved organic carbon pool (DCCHO%DOC) and molecular composition of the DCCHO pool (mol.%DCCHO,Table 1).The dominance of hydrolyzed glucose and mannose/xylose in summer (29 AE 11 and 31 AE 9 mol.%respectively) and fall (31 AE 8 and 37 AE 5 respectively) confirms that the North Atlantic transports relatively degraded DOC into the Arctic Ocean (Rich et al., 1997;Amon and Benner, 2003;Piontek et al., 2020).In addition, this study is among few others to report on the hydrolyzable acidic and amino sugar compositions in the North Atlantic and the Fram Strait (Engel et al., 2012;Grosse et al., 2021).Similar to neutral sugars, acidic sugars displayed a seasonal decrease in their concentration, DCCHO%DOC, and mol.%DCCHO (Table 1), which suggests a strong decline in freshly excreted DOC throughout the upper 100 m from summer to fall (Borchard and Engel, 2015).Amino sugars showed a twofold increase of mol.%DCCHO throughout the upper 100 m (Table 1), which might be explained by the presence of galacturonic acid and glucuronic acid in bacteria-derived DOC (Benner and Kaiser, 2003).As bacterial numbers decline towards fall, cell death or viral lysis release cellular-derived amino sugars into the surrounding water (Fig. 1B).The low decay coefficients of amino sugars make them resistant to decomposition and increase their residence time (Kawasaki and Benner, 2006), which results in the twofold increase of mol.%DCCHO until fall.
To assess the protein content of the biopolymer pool, we differentiated DHAA into essential amino acids (EAA) and non-essential amino acids (NEAA).Phytoplankton can synthesize both EAAs and NEAA depending on their nutrient limitation, while higher trophic levels rely on the acquisition of EAA through their diet and acquisition or de novo synthesis of NEAA (Arts et al., 2009;Grosse et al., 2019;Larsen et al., 2022).Concentrations of EAA and NEAA decreased significantly between summer and fall (Table 2; Wilcoxon Rank Sum Test p < 0.001).The significant reduction in the amino acid reservoir is indicative of the utilization of nitrogen-rich compounds from summer to fall.Furthermore, EAA and NEAA concentrations during fall are representative of post-bloom conditions; i.e.EAA: 89.7 nmol L À1 , 0.5 DHAA%DOC, and 50.4 mol.%DHAA;NEAA: 90.6 nmol L À1 , 0.5 DHAA% DOC, and 49.6 mol.%DHAA in upper 100 m of the eastern Fram Strait in 2017 (Grosse et al., 2021) (Table 2).
Around 81% ASVs were shared between summer and fall, illustrating a considerable number of taxa that do not respond to the changing biopolymer pool and other seasonal changes (Wietz et al., 2021).Common taxa were defined as ASVs with >3 counts, >3% sequence abundance, and occurred in all the samples.Common taxa significantly differed between the surface-to-BCDM and 100 m (ANOSIM p < 0.01, Table S2).Within the surface-to-BCDM, some common taxa like Planktomarina and SAR11 clades significantly correlated with hydrolyzed acidic sugars (+, summer), serine (+, fall) and glycine (À) during both seasons (Figs 3 and 4).At 100 m, common taxa exhibited significant correlations with isoleucine (+) and glucose (À) during both seasons (Fig. 4C and D).
The predominance of labile biopolymers like fucose, rhamnose and threonine coincided with lower microbial diversity, illustrated by 3% of ASVs that were exclusively detected during summer (Fig. 2, Fig. S3).In particular, the SAR92 clade was significantly enriched in the surface-to- BDCM in summer (Fig. 3A, log 2 FC = 3.7 q < 0.01), which might be linked to the abundance of Phaeocystis colonies in the Fram Strait (Lampe et al., 2021;Wietz et al., 2021).Additionally, we observed high relative abundances of the Flavobacteriaceae genera Aurantivirga, Formosa, Polaribacter and Ulvibacter (Fig. 3A), with significant positive correlations to labile compounds like fucose and threonine (Fig. 4A).Fucose significantly constrained the ordination space of the redundancy analysis (RDA), suggesting the importance of fucose containing polymers for polysaccharide-degrading taxa (forward selection, F = 3.6, q < 0.05; Fig. 5) (Cottrell and Kirchman, 2000 et al., 2014;Reintjes et al., 2019).Furthermore, Formosa and Polaribacter show significant positive correlations to GABA (Fig. 4A).Hence these taxa might be actively growing in the surface-to-BDCM, since GABA can serve as an indicator for microbial activity and might portray amino acid turnover more accurately than 3 H-leucine-derived BP, which can substantially underestimate bacterial growth (Popendorf et al., 2020).At 100 m, we observed a significant increase of Candidatus Nitrosopumilus, LS-NOB and unclassified Marine Group II archaea (all log 2 FC, q < 0.01, Figs 3 and 4; Table S2).The shift towards taxa that are typically observed in low light waters suggests that the microbial community is already beginning to target lowmolecular-weight DOM.16% of ASVs were exclusively detected during fall (Fig. 2, Fig. S3).The significant enrichment of Synechococcus in the surface-to-BDCM corresponds to flow cytometry-derived cell counts (log 2 FC q < 0.05, Figs 1 and 2; Table S2).Additionally, the relative abundance of Candidatus Nitrosopumilus and Marinimicrobia (SAR406 clade) significantly increased from summer to fall (log 2 FC q < 0.05, Fig. 3; Table S2), which negatively correlated with more labile compounds (i.e.fucose, isoleucine, phenylalanine) but positively correlated with more refractory compounds (i.e.alanine, aspartic acid and glycine; Fig. 4).Candidatus Nitrosopelagicus and Candidatus Nitrosopumilus showed the highest abundances in 100 m during fall (log 2 FC q < 0.05, Fig. 3; Table S2).Generally, more refractory compounds were the biogeochemical divers for the microbial community in fall (Fig. 5).Glycine, for example, has been shown to have low microbial degradation rates (Veuger et al., 2012), which suggests that more refractory biopolymers either accumulate or are synthesized during fall.Particularly, Candidatus Nitrosopumilus have been linked to continuously releasing glycine but also alanine, valine, leucine, isoleucine, phenylalanine during ammonia oxidation (Bayer et al., 2019).Moreover, the presence of ammonia-oxidizing archaea and nitrite-oxidizing bacteria, like LS-NOB, indicate beginning nitrate replenishment at the start of the polar night (Fig. 4B).

Conclusion
Our assessment of microbial communities in the context of the organic matter pool identified autotrophic and heterotrophic seasonality in the Fram Strait, and their association with different biopolymers.Seasonal variability of autotrophic microbes was closely related to a decline in POC, production rates and cell abundances.The most notable seasonal shift was observed in Synechococcus that can use their osmotrophy strategy to compete for biopolymers with heterotrophic microbes (Yelton et al., 2016).Among the heterotrophic community, the lower alpha-diversity suggests that specialized groups target labile biopolymers in summer, while a higher alpha-diversity suggests that taxa including ammonia and nitrite oxidizers scavenge for more refractory substrates in fall.Our study highlights seasonally driven associations between biopolymers and microbial community, yet studying these associations under varying environmental conditions (e.g.sea ice versus ice free) and higher resolution approaches (e.g.transcriptomics), could truly explain the microbial substrate regimes in the Arctic Ocean.

Fig. 1 .
Fig. 1.Sampling sites, enrichment and relative abundance of phytoplankton in the upper 100 m during summer and fall.A. Samples were taken onboard the RV Polarstern from July 16th to July 23rd, 2018 (orange) and the RV Maria S. Merian from September 16th to October 4th, 2018 (black).Arrows illustrate the main currents in the Fram Strait after Soltwedel et al. (2016).B. Enrichments in the upper 100 m towards summer (left) and fall (right).C. Relative abundance of autotrophs measured using flow cytometry in the surface (5 or 10 m) to BDCM (<50 m) and 100 m.Abbreviations: EGC, East Greenland Current; RAC, Return Atlantic Current; SB, Svalbard Branch; WSC, West Spitsbergen Current; YB, Yermark Branch; DOC, dissolved organic carbon; SLDOC, semi-labile DOC; POC, particulate organic carbon; PP; primary production; BP, heterotrophic bacterial production; BDCM, below deep chlorophyll maximum.Biogeochemical data (DOC, DCCHO, DHAA) and ecological data (BP, BA) from corresponding samples of von Jackowski et al. (2020).

Fig. 3 .
Fig. 3. Relative abundance with enrichments in the upper 100 m during summer and fall.A.Relative abundance of top 30 genera in the surface -BDCM (5-50 m) were averaged due to the lack of significant differences between these depths (see text).B. Relative abundance of top 30 genera in 100 m.The enrichments (coloured bars) between the seasons were calculated using log 2 Fold-Change (orange = summer, black = fall, grey = no seasonal enrichment) with significant differences marked by an asterisk (p adj < 0.05) -Abbreviation: BDCM, below deep chlorophyll maximum.

Fig. 4 .
Fig. 4. Spearman rank correlation matrix of genera and biopolymers in the upper 100 m during summer and fall.A. Correlations of top 30 genera in surface -BDCM (10-50 m) due to lack of significant differences between these depths during summer.B. Correlations of top 30 genera in surface -BDCM (5-50 m) due to lack of significant differences between these depths during fall.C. Correlations of top 30 genera in 100 m during summer.D. Correlations of top 30 genera in 100 m during fall.The clusters were performed using 'complete' clusters analysis and the spearman correlation (blue to red) was performed and associated with adjusted p-values for multiple testing.Abbreviations: BDCM, below deep chlorophyll maximum.Corresponding biopolymer data were retrieved from von Jackowski et al. (2020).Significance codes are shown by asterisks: '***' < 0.001,'**' < 0.01,'*' < 0.05 and >0.05.

Fig. 5 .
Fig. 5. RDA based on Hellinger transformed samples.The continuous arrows correspond to the significant forward-selected variables, while the dashed arrows correspond to the remaining parameters included in this study.The boxes were manually added to the figure to add the seasonal separation.Abbreviations: GABA, gammaaminobutyric acid.Data for dissolved parameters and gel particle data from corresponding samples by von Jackowski et al. (2020).

Table 1 .
Concentration, the contribution of DOC (%DOC) and relative composition (mol.%) of DCCHO in the water column.

Table 2 .
Concentration, the contribution of DOC (%DOC) and relative composition (mol.%) of DHAA in the water column.DHAA were differentiated into EAA and NEAA.The DOC concentrations were incorporated from von Jackowski et al.