Phytoplankton dynamics shape bacterioplankton community structure and metabolism during the austral summer-autumn transition in the Western Antarctic Peninsula Authors: Mariia Pavlovska, Yevheniia Prekrasna-Kviatkovska, Andrii Zotov, Artem Dzhulai, Evgen Dykyi, Bruno Huettel, Bernhard M Fuchs, Rudolf I Amann, Hanno Teeling, Chandni Sidhu, Alexandre Anesio Published: 2026 ABSTRACT AbstractSeasonal changes in Antarctic coastal waters trigger pronounced shifts in microbial community composition and function, yet sparse spatial and temporal coverage currently limits our understanding of phytoplankton-bacterioplankton coupling. This study combines metagenomic and metatranscriptomic analyses of marine bacterioplankton with environmental data to address the functional dynamics of planktonic communities off the Western Antarctic Peninsula during the austral summer–autumn transition. Diatoms dominated the phytoplankton community, with generally low biomass and abundance, yet a species-specific succession was observed. The bacterioplankton community structure shifted from dominance of copiotrophic taxa (e.g. Polaribacter) towards oligotrophic lineages (e.g. SAR11) adapted to low-nutrient conditions, accompanied by a decrease in microbial carbohydrate-degradation activity. The capacity to degrade algal-derived polysaccharides varied between community members, with ß-glucan, α-glucan, chitin, and host glycan utilization present in all, and fucose, β-galactan and trehalose degradation restricted to specific taxa. DMSP metabolism also showed taxonomic specificity and was shaped by both physical (ice melt and fluctuations in solar irradiation) and biological factors (phytoplankton succession). Together, these findings reveal a complex, taxon-specific coupling between bacterioplankton and phytoplankton communities in the Western Antarctic Peninsula, linking community structure to likely functional gene expression and highlight how Antarctic bacterioplankton drives carbon and sulfur turnover in a polar marine ecosystem.Keywords: Western Antarctic Peninsula, diatom blooms, metagenomics, transcriptomics, carbon metabolism, DMS, DMSP, carbohydrates, CAZymes Introduction The marine biological carbon pump (BCP) encompasses all processes that transport a portion of the organic matter produced in the sun-lit (euphotic) epipelagic zone to deeper waters, where the more refractory fraction can be sequestered for years to centuries. In the euphotic layer, photosynthetic microalgae (phytoplankton) dominate primary production (Seymour et al. 2017) and supply the carbon footing for two tightly linked food webs, namely (i) the food web of higher trophic levels, ranging from small zooplankton (protists and metazoan larvae) to nekton (crustaceans, fish), and (ii) the heterotrophic bacterial loop, in which bacterioplankton recycle dissolved organic matter (DOM). A multitude of biotic and abiotic factors influence the BCP, which is why its efficiency varies considerably among oceanic regions. The Antarctic Southern Ocean exerts a disproportionate influence on the global climate, despite accounting for only ∼5% of the total ocean surface (Eakins and Sharman, 2010) . It is responsible for ∼40 ± 5% of oceanic CO₂ sequestration and 44 ± 2% of total oceanic heat uptake (Williams et al. 2023). At the same time, this polar ecosystem is of low-diversity and therefore highly vulnerable to environmental perturbations. Over recent decades, the Southern Ocean has experienced rapid physicochemical changes, including regional surface warming (González-Herrero et al. 2022, Mar Flexas et al. 2022), sea-ice fluctuations (Cook et al. 2005, Sun et al. 2023), and increasing vertical stratification (Pellichero et al. 2017, Rogers et al. 2020). These trends are not uniform across Antarctica. East Antarctica has shown a modest cooling trend during 1979-2018 (Turner et al. 2020), whereas the Western Antarctic Peninsula (WAP) and its adjacent Southern Ocean waters rank among the fastest-warming regions on Earth (Steig et al. 2009). Even within the WAP, temperature records differ: some studies report warming in the latter half of the twentieth century followed by cooling in the early twenty-first century (Turner et al. 2020, Zhang et al. 2023), while others document continued rapid warming (Gorodetskaya et al. 2023). Despite these regional nuances, an overall warming of the WAP and Southern Ocean is unequivocal (Cai et al. 2023), making this basin a suitable natural laboratory for examining how rising temperatures affect carbon turnover in a fragile yet globally important ecosystem. Carbon uptake in the ocean is driven centrally by food-web interactions, which orchestrate biogeochemical cycles from phytoplankton production to the remineralization of organic matter by heterotrophic bacteria (Dutkiewicz et al. 2024, Lin et al. 2021). Phytoplankton primary production fluctuates and peaks during bloom events that dramatically reshape nutrient fluxes via intense phyto-bacterioplankton coupling (Depeller and Davidson 2017). Most particulate organic matter (POM) aggregates and sinks, whereas the labile fraction of dissolved organic matter (DOM) is rapidly remineralized by surface heterotrophic bacteria (De La Rocha and Passow 2007). Consequently, phytoplankton supply ≈ 50% of the carbon required for bacterial growth (Seymour et al. 2017). Bacteria-phytoplankton interactions are diverse, complex, and vary in between algal bloom stages. Bacteria can stimulate algal growth by provision of growth-promoting compounds (e.g. Cooper and Smith 2015), but also increase algal mortality by means of algicidal compounds (Paul and Pohnert 2011, Shi et al. 2013) or even by direct predation (Wang et al. 2020). Conversely, phytoplankton provide bacteria with organic substrates (carbohydrates, amino acids, lipids) but also compete for limiting nutrients such as nitrogen and phosphorus (Lutzu and Dunford 2018). Certain bacterial clades have evolved specialized catabolic pathways for the degradation of specific algal exudates, thereby reducing intra-species competition through resource partitioning (Ferrer-González et al. 2021). In the WAP, bacterioplankton dynamics are shaped by both bottom-up controls (availability of phytoplankton-derived substrates and micronutrients, particularly iron) (Church et al. 2000, Debeljak et al. 2023) and top-down pressures (grazers and viruses). Specifically, availability of organic carbon from phytoplankton and predator and phage abundances play significant roles (Bowman et al. 2017). Likewise, biological interactions exert a major effect on microbial community composition in the Southern Ocean, where bacterioplankton is often dominated by Pelagibacterales and Rhodobacterales (Alphaproteobacteria, phylum Pseudomonadota), Oceanospirillales and Alteromonadales (Gammaproteobacteria, phylum Pseudomonadota), and Flavobacteriales (Bacteroidota) (Grzymski et al. 2012, Delmont et al. 2015, Cao et al. 2019, Zhang et al. 2023, Cho and Lee 2024, Pavlovska et al. 2025). Tight coupling between bacterioplankton and phytoplankton and their interplay in the ocean’s carbon cycle has been recognized in a number of studies in both polar (Cordone et al. 2022, Piontek et al. 2022, Thiele et al. 2022) and temperate (Teeling et al. 2012, Teeling et al. 2016, Sidhu et al. 2023, Wang et al. 2024) regions. Yet, this functional association is exceptionally pronounced in the Southern Ocean waters due to strong seasonality and nutrient limitation (Signori et al. 2014). Temporal coupling between free-living bacterioplankton and phytoplankton has been documented from mid-winter 2013 to late summer 2014 (Luria et al. 2016), where blooms were accompanied by rises in Polaribacter (Flavobacteriales) and Rhodobacteraceae, followed by a return to pre-bloom assemblages. In a study in summer 2013 and 2014 in Chile Bay, an increase in bacterioplankton diversity and community complexity was shown to be related to higher phototrophic productivity (Fuentes et al. 2019). In particular, higher diatom abundances were shown to be associated with the proliferation of specific heterotrophic bacteria, e.g. members of the genus Polaribacter. Similarly, a study conducted during the austral summer of 2017 in the Ross Sea showed that haptophyte abundances were coupled with an increase in oligotrophic and mixotrophic SAR11 (Cordone et al. 2022). Biological interactions with phytoplankton of in particular Bacillariophyceae, Coccolithophyceae, and Dinophyceae were identified as the driving factor of bacterioplankton succession in Wilhelm Archipelago waters (Pavlovska et al. 2025). Bacterial biomass production was shown to be directly related to Chl a concentration, and specific taxonomic signatures of bloom-responsive, neutral and low Chl a associated taxa were detected in Weddell Sea bacterioplankton communities (Piontek et al. 2022). On the other hand, phytoplankton biomass is rich in polysaccharides, which can represent ≥75% of their dry biomass depending on species and physiological state (Patel et al. 2023). Bacterial degradation of these polymers relies on carbohydrate-active enzymes (CAZymes), including glycoside hydrolases (GH), polysaccharide lyases (PL), carbohydrate esterases (CE), glycosyltransferases (GT), and auxiliary activity (AA) proteins (Drula et al. 2022). In particular in members of the Bacteroidota, CAZymes are oftentimes arranged in operon- or regulon-like loci known as polysaccharide utilization loci (sg. polysaccharide utilization locus, abbr.: PUL) (Sonnenburg et al. 2010). While most PULs target a single polysaccharide class, some substrates are degraded by multiple PULs (Sastre et al. 2024) and co-regulated genes may reside outside of the canonical loci (Unfried et al. 2018, Lapébie et al. 2019). For successive Phaeocystis antarctica blooms in the Amundsen Sea (austral summers 2010-2014), metagenomic analyses have revealed that members of Polaribacter (Bacteroidota) and the SAR92 (Gammaproteobacteria) possessed the capacity to consume high-molecular-weight dissolved organic carbon (HMW-DOC), whereas members of the gammaproteobacterial Ant4D3 and SUP05 clades specialized on low-molecular-weight DOC (LMW-DOC) (Lim et al. 2019). Genes encoding GH families targeting laminarin, chitin, and mannose were abundant. Ant4D3 activity in LMW-DOM uptake was also confirmed for summer/autumn 2010-2012 waters of the WAP (Nikrad et al. 2014). Besides polysaccharides, phytoplankton synthesizes larger quantities of organosulfur compounds, such as dimethylsulfoniopropionate (DMSP). In polar regions DMSP functions in cryoprotection (Bullock et al. 2017), osmoregulation (Vairavamurthy et al. 1985), and defense against grazers and oxidative stress (Sunda et al. 2002, Fredrickson et al. 2003). Enzymatic cleavage of DMSP yields dimethylsulfide (DMS), a climatically active gas that promotes cloud formation (Bora et al. 2026). The Southern Ocean is therefore a hotspot for DMS emissions (Stefels et al. 2018). Dinophyceae and Prymnesiophyceae are known to be the main DMSP producers, but DMSP is also synthesized in significant amounts by polar diatoms (Bullock et al. 2017, Sheehan and Petrou 2020). The exudation of DMSP differs between phytoplankton species and depends on salinity and temperature (Stefels et al. 2007). Sea ice algae, which are released upon meltdown events, are considered the main source of elevated DMSP in Southern Ocean waters (Stefels et al. (Stefels, Van Leeuwe and Jones 2018). Many bacterioplankters can use DMSP as a carbon and sulfur source (Kiene et al. 2000), and some even synthesize it de novo as an osmolyte (e.g. Zheng et al. 2020). Bacterial DMSP catabolism proceeds via either cleavage by DMSP lyases (encoded by dddD, dddL, dddK, dddP, dddQ, dddW, and dddY) or demethylation (encoded by dmdABCD), respectively (Bullock et al. 2017). Seasonal DMSP dynamics is well documented in Antarctica, yet most culture-based and in situ studies are limited to chemical and phytoplankton analysis. Phaeocystis antarctica is considered the main DMSP producer in the Southern Ocean (Vance et al. 2013, Asher, Dacey and Stukel 2017) and elevated concentrations of dissolved DMSP are observed during spring and summer, especially during the ice-melt (Liu et al. 2023). Meanwhile, the functional repertoire of DMSP utilization by Antarctic bacterioplankton has not yet been evaluated in situ. Combined metagenomic-metatranscriptomics investigations of Southern Ocean bacterioplankton are still scarce and limited in spatiotemporal coverage. Moreover, while the association between microalgae and bacterioplankton has been repeatedly documented (Cordone et al. 2022, Piontek et al. 2022, Thiele et al. 2022) taxon-specific coupling between phytoplankton dynamics and bacterioplankton taxonomic composition and functional activity remains poorly resolved in polar regions. To address this gap, we used a combined OMICs approach with samples collected during the transition from austral summer to autumn 2021 at the Wilhelm Archipelago. Our specific aims were to (i) characterize bacterioplankton community composition across this seasonal window, (ii) quantify expression of polysaccharide utilization loci (PULs) and assess how their activity tracks phytoplankton dynamics, and (iii) determine the prevalence and transcriptional activity of DMSP catabolic genes, thereby linking bacterial sulfur metabolism to seasonal phytoplankton production. By aligning community structure with functional gene expression, this study provides insights into how Antarctic bacterioplankton mediate carbon and sulfur fluxes in a polar marine ecosystem. Material and methods Sampling Sampling was carried out in coastal waters of the Southern Ocean near Akademik Vernadsky station [Wilhelm Archipelago, 65°14′24.0″ S, 64°12′45.7″ W; (Neverovsky et al. 2015, Brovchenko et al. 2020)] during the transition from summer to autumn 2021 (Fig. 1). Surface water (∼1 m depth) was collected with a 5 l Niskin bottle (Hydro-bios, Altenholz, Germany) on 10 occasions: 17.02.2021, 23.02.2021, 04.03.2021, 07.03.2021, 11.03.2021, 19.03.2021, 24.03.2021, 30.03.2021, 05.04.2021, and 10.04.2021 (dd.mm.yyyy). Each water sample was first passed through a 10 µm plankton net to remove the bulk of eukaryotes. For each time, 2 l of the resulting filtrate were passed through a 47 mm diameter 0.22 µm pore-sized polycarbonate filter (Millipore, Billerica, MA, USA) using a Combisart jet vacuum pump (Sartorius, Göttingen, Germany) to collect bacterioplankton. The filters were stored in RNAlater stabilizing solution (Sigma-Aldrich, St. Louis, MO, USA), transported at -20°C, and kept at -80°C until further processing in the laboratory. Sampling location in the waters of Wilhelm Archipelago (Western Antarctic Peninsula). Sampling location is denoted by Red color. For CARD-FISH (catalyzed reporter deposition-fluorescence in situ hybridization) bacterial cell counting, 50 ml of unfiltered seawater were fixed with 37% formaldehyde (v/v) up to a concentration of 1% (v/v) for 1 h at room temperature. Afterwards, the samples were filtered through 0.22 µm pore-sized polycarbonate membranes (Millipore). The filters were stored at -20°C in sterile Petri dishes until CARD-FISH analysis. Phytoplankton was collected by reverse-filtration of 5 l seawater samples through 1.2 µm pore-sized 142 mm nitrocellulose filters (Millipore) until a concentrate of 30-50 ml remained on the filters. Concentrates were fixed with a 1 : 1 (v/v) mixture of 40% formaldehyde and glacial acetic acid to a final formaldehyde concentration of 4%. Physicochemical analysis Vertical profiles of temperature, salinity, fluorescence (chlorophyll proxy) and dissolved oxygen were recorded in situ with an RBR Concerto CTD (RBR Ltd., Ottawa, Canada). Water samples for nutrient analyses (NH₄⁺, NO₂⁻, NO₃⁻, PO₄³⁻, Si(OH)₄, dissolved Fe) were measured with a Hach Lange DR3900 spectrophotometer (Hach, Loveland, CO, USA) following the manufacturer’s standard protocols. Full data are listed in Table S1. Phytoplankton taxonomy and enumeration Phytoplankton samples were examined with a Zeiss Primo Star light microscope (Zeiss, Oberkochen, Germany) at 400-1,000x magnification. For each sample, three replicate counts were performed in a 0.05 ml Nageotte chamber. Taxa larger than the Nageotte volume were counted in a 1 ml Sedgwick-Rafter chamber. The wet biomass was estimated via best-fitting geometric shapes, assuming cell densities of 1 mg/µl. Chlorophyll a analysis For each sample 1 l of seawater was vacuum-pumped through 47 mm diameter 0.45 µm pore-sized nitrocellulose filters (Millipore). Pigments were subsequently extracted using ethanol (90%, 78°C) and concentrations were measured using a Hach Lange DR3900 spectrophotometer (Hach). Bacterioplankton cell enumeration Total bacterial abundances were determined after staining with DAPI (4′,6-diamidino-2-phenylindole) on a ZEISS Axio Imager 2 microscope (ZEISS) as described previously (Thiele, Fuchs and Amann 2011). Cells of abundant bacterioplankton groups were likewise counted after staining with group-specific probes (SAR11-mix, RCA1000, POL740, POL1270, POL180, ULV995, SAR86-1245, SAR92-627) using the CARD-FISH technique. All probe sequences and corresponding references are provided in Table S2. Nucleic-acid extraction and sequencing DNA Polycarbonate filters containing bacterioplankton biomass were processed with the Higher Purity Soil DNA Isolation Kit (Canvax Biotech, Valladolid, Spain) according to the supplier’s instructions. Extracted DNA yielded concentrations ≥ 57 ng µl⁻¹ and A260/280 ratios ≈ 2.0. rRNA amplicon sequencing The V4 region was amplified with primers S-D-Bact-0341-b-S-17 and S-D-Bact-0785-a-A-21 (Klindworth et al. 2013) and sequenced on an Illumina NovaSeq 6000 (2 × 250 bp, paired-end). Ten samples (one per sampling date) were processed. Metagenomes Bacterioplankton samples from March and April were selected for detailed analysis because they encompass the transition from austral summer to austral autumn, a temporally narrow but ecologically significant period, underrepresented in Antarctic studies in comparison to spring and summer season. Two representative samples (11.03.2021 and 05.04. 2021) were subjected to long-read sequencing on a PacBio Sequel II system (Pacific Biosciences, Menlo Park, CA, USA) with one SMRT Cell per sample in HiFi mode. RNA Total RNA was extracted from five samples (11.03.2021, 19.03.2021, 24.03.2021, 30.03.201, and 05.04.2021) at the Max Planck Genome Centre (Cologne, Germany). Ribosomal RNA was depleted with custom oligonucleotide probes (Nobori et al. 2018). Libraries were sequenced on an Illumina HiSeq 3000 (2 × 150 bp). Sequencing summary metrics are provided in Table S3. Bioinformatic processing rRNA amplicons Paired-end reads were merged using FLASH v1.2.7 (Magoč and Salzberg 2011). Barcodes, homopolymers, short (<150 bp) and chimeric sequences were removed in QIIME 2 (Bolyen et al. 2019). OTUs were clustered de novo at 97% similarity and taxonomically assigned against SILVA v138.1 (Quast et al. 2013). Community composition plots were generated with the vegan and ggplot2 packages in R (R Core Team 2021) . Metagenomes The two HiFi datasets were co-assembled with Flye v2.8.3 (Kolmogorov et al. 2019) using the “-meta” and “-pacbio-hifi” flags. ORFs were predicted by FragGeneScan (Rho et al. 2010) using w1 and sanger_5 parameters and annotated via (i) BLAST+ (Camacho et al. 2009)) searches against KEGG r58.0 (Kanehisa and Goto 2000) and GenBank r239 (Clark et al. 2016), (ii) Diamond v0.9.10.125 (Buchfink, Xie and Huson 2014) searches against eggNOG v4.5 (Huerta-Cepas et al. 2016), and (iii) HMMER3 v3.1b2 (Eddy 2009) searches against Pfam 33.0 (Finn et al. 2016). CAZymes were predicted together with their target substrates using the dbCAN3 server (Zheng et al. 2023). Metatranscriptomes Ribosomal RNA reads were removed with SortMeRNA v3.0 (Kopylova et al. 2012). Remaining mRNA reads were quality trimmed (qtrim = 20) using BBMap’s bbduk and repaired with repair scripts (https://sourceforge.net/projects/bbmap/). Because the metagenomic assembly did not yield high-quality MAGs (≥70% completeness, ≤5% contamination), a metagenome co-assembly (Table S4) was used as a reference for mapping the retained mRNA sequences. Bowtie2 (Langmead and Salzberg 2012) was used for mapping from within SqueezeMeta v1.3.1 (Tamames and Puente-Sánchez 2019). Transcript abundances were calculated as transcripts per million (TPM), where TPM is defined as the number of times that a specific feature is found when randomly sampling one million features, given the abundances of the different features in the sample (Wagner et al. 2012, Puente-Sánchez et al. 2020): (reads per kilobase of each function / total reads per kilobase of all functions) × 106. Phylogenetic validation of DMSP-catabolism genes Protein sequences of candidate DMSP metabolism genes were aligned with reference sequences using MAFFT v7.526 (Katoh and Standley 2013), trimmed with TrimAl, (Capella-Gutiérrez et al. 2009), and subjected to maximum-likelihood tree reconstruction in IQ-TREE (automatic model selection, 1000 replicates, Minh, Schmidt and Chernomor 2020). Trees were visualized with iTOL (Letunic and Bork 2024) (Fig. S7A-E). All downstream statistics and figure generation were performed in Python v3.12 using the packages numpy v1.26.1 (Harris et al. 2020), seaborn v0.13.0 (Waskom et al. 2021), matplotlib v3.8.0 (Hunter 2007), and scipy.stats v1.11.3 (Virtanen et al. 2020). The sequence and accession numbers of ratified protein sequences are provided in Table S5. Results Physical and chemical conditions: a melt-driven event Temperature and salinity records at the sampling site (Fig. 1) indicated a melt-water pulse between March 11 and March 30. During this time, the salinity fell from 32.3 to 30.0 and the temperature dropped from +1.4°C to -0.7°C (Fig. S1A). The highest concentrations of dissolved iron (March 19) and dissolved inorganic nitrogen (DIN, March 24) were also measured within this period (sampling interval: ∼5-8 days). Silicate dropped from 4.5 mg/l (February 17) to 3.2 mg/l (March 24) with a subsequent gradual increase to 3.4 mg/l (April 5) (Table S1). Chlorophyll a (Chl a) concentrations remained low throughout the summer-autumn transition, peaking at 0.3 µg/l on March 11th and March 24th (Fig. S1B). Solar irradiance varied between 9 W/m2 (April 7) and 131 W/m2 (March 21) with a mean of 51 W/m2 (Fig. S1C). Few diatom species dominated the phytoplankton Phytoplankton abundance and biomass declined steadily during the sampled summer-autumn transition period (pale turquoise area in Fig. 2A and B), and the community was dominated by three classes: Bacillariophyceae (diatoms), Chlorophyceae, and Cryptophyceae. Diatoms contributed the largest share of both biomass and cell numbers (Fig. 2A and B), yet they were outnumbered by Cryptophyceae at the end of the sampling campaign. The maximum diatom biomass was observed on March 19 (224.4 mg/m3), and the abundance peak on March 24 (5,700 cells/l). The most prevalent species were Corethron criophilum (1,500-2,800 cells/l) and Fragillariopsis, which was mostly low in abundance (< 729 cells/l) but peaked on March 24 (4,700 cells/l). Minor diatoms (Pseudo-nitzschia, Thalassiosira) increased in early April to similar numbers of ≈735 cells/l (Fig. 2C). Cryptophyceae biomass peaked on March 4 (3.7 mg/m3; 7,600 cells/l) and April 10 (1.043 mg/m3; 10,440 cells/l) and Chlorophyceae biomass peaked on March 24 (0.52 mg/m3; 3,400 cells/l) (Table S6). Plankton dynamics at the sampling location of the Wilhelm archipelago during the summer–autumn transition: A) phytoplankton classes’ abundance, B) phytoplankton classes’ biomass, and C) dominant phytoplankton species abundances; D) cell counts of SAR11; E) bacterioplankton cell counts of other clades. The pale turquoise areas in panels A-C indicate the period of bacterioplankton metatranscriptome analysis. Gradual shift towards a higher proportion of oligotrophic bacterioplankton The study targeted bacterioplankton ranging from 0.22 to 10 µm, and thus included both free-living and particle-attached bacteria. Total bacterial cell counts ranged from 500,000 cells/ml on (March 30) to 720,000 cells/ml (April 5) (Fig. 2D, E; Table S7). Amplicon sequencing showed that the community was largely composed of Gammaproteobacteria, Alphaproteobacteria, and Bacteroidia (88.0%-98.8% of total reads); Actinobacteria, Bacilli, Clostridia, and Acidimicrobiia together contributed 1.2% and 12.0 % of total reads on February 23 and March 30, respectively. A shift in community composition was observed during the sampled period (Fig. 3A, pale turquoise area). While Gammaproteobacteria, Bacteroidia, and Acidimicrobiia relative abundance decreased, Alphaproteobacteria relative abundances increased from 20.4% on February 17 to 51.0% on April 5. Bacilli (mainly Fusobacteria) accounted for 1.9% and 2.7% of the total community on March 19 and 24, respectively, and members of the Clostridia class reached 5.3% relative abundance on March 30. Wilhelm archipelago bacterioplankton taxonomic structure assessed by amplicon sequencing of the 16S rRNA genes at the (A) class and (B) genus level. In graph B), the colors representing different bacterial clades are arranged in the same order as they appear in the legend At the genus level, the gradual rise in Alphaproteobacteria relative abundances was apparent, driven by an increase of members of the free-living alphaproteobacterial SAR11 clade Ia from 6.1% (end of February) to 22.2% (April) (Fig. 3B). Likewise, Planktomarina (alphaproteobacterial RCA clade) relative abundances increased from 2.3% to 10.9%, and members of the gammaproteobacterial SUP05 clade (Thioglobaceae) from 0.9% (February 17) to 5.1% (April 10). Conversely, relative abundances of other abundant bacterial genera showed a decreasing trend (Fig. 3B). In particular, Nitrincolaceae (Gammaproteobacteria) that dominated the studied community at the beginning of the sampling decreased from 33.2% (February 17) to 10.0% (April 5). Likewise, Polaribacter (Bacteroidia) dropped from 14.3% (February 17) to 2.2% (April 5), Candidatus Prosiliicoccus (Bacteroidia) [reclassified from Ulvibacter (Francis et al. 2019)] from 15.2% (March 11) to 1.4% (April 10) and Sulfitobacter (Alphaproteobacteria) from 8.0% (February 23) to 1.2% (April 5) (Fig. 3B). Lower abundances were observed for members of the gammaproteobacterial SAR92 and SAR86 clades. SAR92 relative abundance dropped from 3.1% (mid-March) to 0.9% (early April), while SAR86 increased from 0.4% (February 17) to 2.3% (April 5) (Fig. 3B). Further genera of lower abundances (Fig. S2) included members of the gammaproteobacterial OM43 and OM182 clades, the bacteroidotal NS4, NS7 and NS9 clades (Alonso et al. 2007), and the alphaproteobacterial Yoonia-Loktanella (Rhodobacteraceae). CARD-FISH cell counts of bacterioplankton groups of interest corroborated amplicon sequencing data. The SAR11 clade was confirmed as the most abundant group throughout the sampling period with up to ∼40,000 cells/ml on April 5 (Fig. 2D). Conversely, Polaribacter declined from an average of 28,000 to 11,000 cells/ml and Candidatus Prosiliicoccus from 23,000 to 11,000 cells/ml during the period of March 24 to April 5. During the same periods, cell numbers of SAR92 and SAR86 members ranged between 3,600 and 13,000 and 16,000 and 22,000 cells/ml, respectively (Fig. 2E). Correlations between bacterioplankton and phytoplankton during a summer–autumn transition Correlation analysis revealed positive correlations between Bacillariophyceae biomass, in particular Corethron criophilum, and Ca. Prosiliicoccus relative abundances (R = 0.9, P < 0.05), and between Cryptophyceae and Sulfitobacter relative abundances (R = 0.9, P < 0.05). Negative correlations were observed between Cryptophyceae biomass and SAR11 clade Ia, Planktomarina, SAR86 and SUP05 relative abundances (Fig. S3A). Further negative correlations included Thalassiosira sp. and Sulfitobacter (R = -0.9, P < 0.05), and Pseudo-nitzschia delicatissima and Nitrincolaceae, Polaribacter, Cetobacterium, and Sulfitobacter (R = -0.89, P < 0.05) (Fig. S3B), whereas Thalassiosira and Pseudo-nitzschia delicatissima biomasses were positively correlated with SUP05, SAR86, Planktomarina, SAR11 clade II and clade Ia relative abundances (Fig. S3B). Metatranscriptome analysis Metatranscriptome read numbers (Table S3) ranged from 29,253,190 (March 30) to 91,507,610 (April 5) with residual rRNA proportions of 11% (March 24) to 30% (April 5). Genes responsible for key functions in the bacterioplankton community were selected based on assessments of metagenomes from Antarctic summer and winter microbial communities by Grzymski et al. 2012. This included stress-response and organic matter degradation genes coding for (i) carbohydrate metabolism (genes in PULs), (ii) lipid metabolism, (iii) carotenoid biosynthesis, (iv) sulfatases, (v) iron uptake, transport & storage, (vi) nitrogen metabolism, (vii) bacteriorhodopsin, (viii) photoprotective pigments synthesis, (ix) oxidative stress response, and (x) DMSP metabolism (Fig. S4, Table S8). The highest expression of these genes was detected within the categories of carbohydrate, nitrogen and iron metabolism (Fig. S4), with a decreasing proportion of the total TPM from March to April. The highest total PUL expression (calculated as cumulative expression of GHs present in identified PULs) was detected on March 19 with 0.25% of the TPM and dropped to 0.15% on April 5 Nitrogen metabolism genes decreased from 0.30% on March 11 to 0.16% on April 5. Similarly, the highest expression of sulfatase and iron metabolism genes was observed on March 11 and 19 with 0.09% and 0.23%/0.26% of the TPM, respectively (Fig. S4). The expression of photoprotective pigment and carotenoid biosynthesis genes increased over the sampling period, from initial 0.1% and 0.08% to 0.14% and 0.13% of the TPM on April 5, respectively (Fig. S4). Bacteriorhodopsin genes were actively transcribed throughout the sampling period ranging from 0.02% (March 19) to 0.03% (March 24), whereas oxidative stress genes peaked at 0.04% (March 19 and 24) (Fig. S4). Expression of iron metabolism genes was positively related with seawater temperature (R = 0.95, P = 0.05), whereas the TPM of genes coding for photoprotective pigments had negative correlation (R = 0.95, P = 0.05). Carbohydrate metabolism genes’ expression correlated positively with Gammaproteobacteria abundance (R = 1.0, P = 0.02), and with Bacteroidia abundance and phytoplankton biomass [yet with a low p-value support (R = 0.87, P = 0.67, R = 0.9, P = 0.08, respectively)]. Taxon-specific dynamics of select expressed functional genes Carbohydrate degradation We analyzed bacterial genes for carbohydrate-active enzymes in terms of predicted target substrates. TPM values for PULs were generally low (Table S9), with the highest values for β-glucan (GH3, GH5, GH9, GH16, GH17, GH30, GH55, GH81, GH128, GH149) and chitin & peptidoglycan (CBM50, GH18, GH104) reaching 390 TPM on April 5 and 303 TPM on March 11, respectively. Alginate (PL6, PL17, PL7, PL5) and cellulose (GH7, CBM9) degradation transcripts were more abundant at the end of the sampling period. Within the sampling period, a higher activity of carbohydrate degradation transcripts was detected on March 11 and on April 5. Members of the Bacteroidia (formerly Flavobacteriia) exhibited the highest CAZyme TPM, targeting the majority of predicted carbohydrate substrates with the constant expression of β-glucan, exo-polysaccharide, peptidoglycan, α-glucan and xylan utilization genes (Fig. 4A). Degradation of a wide range of carbohydrate substrates was also shown for Gammaproteobacteria and Alphaproteobacteria with the highest TPMs for the transformation of β-glucan, chitin, peptidoglycan and xylan (Fig. 4A). Cytophagia represented by Flammeovirgaceae were found to be active on α-glucan (GH13), β-glucan (GH3), chitin (GH18), host glycans (GH20, GH109, GH31) (Fig. 4A). Dynamics of bacterioplankton carbohydrate metabolism activity during the survey period: A) relative abundance of carbohydrate degradation genes across bacterioplankton taxa (based on the substrate they target); (B, C). in relation to fluctuations in phytoplankton biomass and abundance. Generally, genes targeting β-glucans, such as laminarin (glycoside hydrolase families GH16, GH17, GH55, GH149), were more active at the beginning of April, whereas α-glucan utilization genes (glycoside hydrolase families GH65, GH31, GH13) were more active at the beginning of March, coinciding periods of lower phytoplankton biomass and lower Bacillariophyceae, Chlorophyceae and Cryptophyceae abundances (Fig. 4B, C). Similar to β-glucan, xylan degradation genes (GH43, GH10, CE1, CE4) were highly expressed on April 5. Apart from β-glucan and chitin /peptidoglycan, Alphaproteobacteria were less active in utilizing most polysaccharide substrates. Expression was detected for genes responsible for α-galactan (GH36) and fucose degradation (L-fucose mutarotase). Carbohydrate degradation activity was present in members of Rhodobacteraceae (xylan, β-glucan, α-galactan and α-glucan transformation), Roseobacteraceae (β-glucan transformation), and Sphingomonadaceae (trehalose transformation with α-trehalase) families (Fig. S5A). Rhodobacteraceae were characterized by the highest activity towards β-glucan and xylan degradation, which peaked on April 5. Similarly, α-galactan genes (α-galactosidase) were the most abundant at the end of the sampling period (Fig. S5A). Differential expression of genes for the degradation of particular carbohydrates was detected also within the gammaproteobacterial families Oceanospirillaceae, Porticoccaceae (SAR92), and Moraxellaceae. β-glucan utilization was detected in all three families, highest in Porticoccaceae at the end of the sampling campaign (Fig. S5B), whereas arabinogalactan (β-galactosidase) and fructan (GH32) utilization was detected only in Oceanospirillaceae. Similarly, only Porticoccaceae were found to utilize polyphenol, lignin/phlorotannins, β-galactan (β-galactosidase), and α-glucans. A large portion of Gammaproteobacteria with carbohydrate degradation activity could only be identified at class level, precluding a more detailed analysis. The majority of Bacteroidia with carbohydrate degradation activity were identified as Flavobacteriaceae, and a small proportion as Cryomorphaceae members. The latter possessed genes targeting chitin, β-galactans, β-glucans, xylan, and α-glucans (Fig. S5C). The overall Bacteroidia activity towards polysaccharide degradation was the highest on April 5 except for β-galactan and fucose (α-L-fucosidase), which were more actively utilized in March. Cytophagia were exclusively represented by Flammeovirgaceae members. These were mainly active in carbohydrate degradation at the beginning and at the end of the sampling campaign, with the highest TPMs recorded for α- and β-glucan utilization genes (Fig. S5D). DMSP degradation Bacterial DMSP demethylation prevailed over DMSP cleavage during the studied summer-autumn transition. Transcription of genes coding for dimethylsulfonio-propionate demethylase (dmdA), 3-(methylthio)propionyl-CoA ligase (dmdB), 3-(methylthio)propanoyl-CoA dehydrogenase (dmdC), and (methylthio)acryloyl-CoA hydratase (dmdD) was observed at all sampling dates. Generally, the transcription of demethylation genes increased towards the end of the sampling period after the initial drop on March 19 (Fig. 5A-C). DmdA was characterized with significantly higher TPM, which increased from 21,766 on March 19 to 117,150 on April 5 (Fig. 5A-C). DmdB was the second most actively transcribed of these genes with a maximum TPM of 11, 859 on March 30. Maximum of dmdC and dmdD transcription was observed at the beginning of the sampling campaign on March 11 - 18, 289 TPM and 6 ,489 TPM, respectively (Fig. 5A-C). In contrast to DMSP demethylation, the only detected expressed DMSP cleavage gene was dddL, coding for dimethylpropiothetin dethiomethylase with a TPM ranging from 130 (March 19) to 2 ,372 (March 11) (Fig. 5A–C) Dynamics of the most abundant DMSP utilization genes A). dmdABCD; (C, D) in relation to fluctuations in phytoplankton biomass and abundance. Unlike carbohydrate degradation, DMSP demethylation was only detected in Alpha- and Gammaproteobacteria. Further differences were apparent on the class level. Pelagibacteraceae expressed exclusively dmdA genes, whereas Rhodobacteraceae expressed dmdA, dmdB, and dmdD, yet with lower TPM (Fig. 5A). Three of the demethylation genes, dmdB, dmdC, and dmdD, were detected in Porticoccaceae with the highest TPM for dmdC - 8,660 on March 11. DmdA and dmdC were transcribed in Oceanospirillaceae, with dmdA transcriptional values significantly exceeding those of dmdC (Fig. 5A). Discussion The functional and taxonomic shifts of planktonic communities observed in the current study and multi-annual patterns detected in our previous research (Pavlovska et al. 2025) allowed to develop a conceptual scheme that captures the coupled dynamics of phytoplankton and bacterioplankton during the seasonal transition from austral summer to austral autumn (Fig. 6). This scheme illustrates a shift toward increasingly oligotrophic conditions, characterized by a decline in phytoplankton biomass, a corresponding reduction in copiotrophic bacterial groups, and a decrease in bacterial transcriptional investment in polysaccharide degradation, and is discussed in details below. A conceptual scheme illustrating taxonomic and functional remodelling of Wilhelm Archipelago planktonic communities during austral summer–autumn transition. Phytoplankton community composition Phytoplankton abundances during the sampled period were generally lower than those reported in our previous studies from the same locations (Zotov et al. 2024, Pavlovska et al. 2025). This was reflected in the measured seawater chlorophyll a concentrations, as well as the phytoplankton abundance and derived biomass data. Diatoms dominated, which is in line with previous data from the studied region (Ducklow et al. 2013, Mendes et al. 2013, Nardelli et al. 2023, Zotov et al. 2024, Pavlovska et al. 2025). Cryptophyceae were the second most abundant phytoplankton group, which is also characteristic for the WAP in recent decades, where a shift from large-celled (>20 μm) to smaller-celled phytoplankton (5-20 μm) has been observed in response to surface water warming (Nardelli et al. 2023). Species-specific dynamics were revealed within the Bacillariophyceae class throughout the summer-autumn transition, with Corethron criophilum dominating on March 19 and 30 and Fragillariopsis sp. peaking on March 24. These dynamics may result from different ecological strategies, as Corethron was previously shown to respond more favorably to intense light than low-light adapted Fragillariopsis in Antarctic Peninsula waters (Bowman et al. 2021). This corroborates the indications for a melting event on March 30, which likely led to increased light penetration as a consequence of reduced ice-cover. An alternative explanation would be that the Fragilariopsis peak reflects the initial release of ice-associated algae (Eickhoff et al. 2023) during an earlier phase of sea-ice melt, which was subsequently followed by the measured decrease in temperature and salinity. Given the high diversity, temporal variability and species-specificity of phytoplankton carbohydrate production (e.g. Gügi et al. 2015), the observed dynamics likely had an effect on the bacterioplankton functional response. Coupled dynamics between phytoplankton and bacterioplankton A pronounced trend towards an increase in abundance of oligotrophic lineages (e.g. SAR11 clade Ia) was observed throughout the summer-autumn transition period, whereas copiotrophic taxa—Nitrincolaceae (Thiele et al. 2023), Polaribacter (Avci et al. 2020), Ulvibacter (Piontek et al. 2022), Ca. Prosiliicoccus (Francis et al. 2019), Sulfitobacter (Choi et al. 2016, Costanzo et al. 2023), and Yoonia-Loktanella (Mayali et al. 2023) predominated during peaks in surface productivity. Higher abundance of Nitrincolaceae, Polaribacter, and Ca. Prosiliicoccus at the end of February and at the beginning of March were likely a response to higher diatom abundances on February 23 and Cryptophyceae on March 4 (Table S6). Dominance of SAR11 clade Ia members is typical for oligotrophic waters (Morris et al. 2002, Wilkins et al. 2013) and is usually observed before and after phytoplankton mass development (Liu et al. 2019). Bacteroidota (Polaribacter, Ca. Prosiliicoccus) have been previously found during Antarctic (Liu et al. 2019, Cordone et al. 2022, Piontek et al. 2022) and North Sea algal blooms (Teeling et al. 2012, Teeling et al. 2016, Sidhu et al. 2023, Wang et al. 2024) and have been associated with high Chl a concentration in the Northern Barents Sea (Thiele et al. 2022 and 2023). Polaribacter spp. have been shown to constitute a broad clade with members that have the potential to utilize different algal glycans (e.g. Avcı et al. 2020). Unique bacterioplankton dynamics, contrasting those observed during diatom-dominated blooms, were found during Phaeocystis antarctica bloom in Antarctic waters, where Polaribacter completely dominated algal high molecular weight (HMW) organic matter degradation (Lim et al. 2019). A decrease in phytoplankton abundance as in late bloom stages is frequently associated with higher ratios of opportunistic Gammaproteobacteria clades such as SAR92, SAR86 and SUP05 (Park et al. 2020, Thiele et al. 2023, Pavlovska et al. 2025). Indeed, both SAR86 and SUP05 were found to correlate with low-chlorophyll stations in the Weddell Sea (Piontek et al. 2022). The SUP05 clade was detected in Antarctic bacterioplankton communities throughout a year, since its metabolic versatility enables shifting from heterotrophy in summer, early autumn and spring to chemolithoautotrophy in winter (Grzymski et al. 2012, Williams et al. 2012, Morris and Spietz 2022, Piontek et al. 2022). In our data, abundances of SAR92 clade members were not fluctuating as much as those of Nitrincolaceae and Polaribacter. This could result from the known adaptation of SAR92 to both LMW and HMW DOM (Lim et al. 2019). Bacterioplankton-mediated degradation of algal polysaccharides Polysaccharide degradation represents a considerable genetic investment for marine bacteria, as it requires large suites of specialized enzymes, transporters, and regulatory elements, which together increase genome size and metabolic costs. Therefore, distinct bacterioplankton taxa have likely evolved metabolic specializations towards distinct algal exudates (Mühlenbruch et al. 2018). Hence, the succession of phytoplankton communities has likely created distinct substrate niches that drove the dynamics of microbial communities during the investigated summer-autumn transition. Potential activity towards the degradation of a range of polysaccharides was detected in different bacterioplankton groups, most prominently in Bacteroidia. This supports the known ecological role of many members of the Bacteroidia as specialists for algae-derived DOM that becomes available either by algal exudation, cell lysis or even direct interaction with algal cells (Andrew et al. 2022). Many Bacteroidia are outright specialists for algal polysaccharides and possess complex and diverse PUL uptake systems (e.g. Kappelman et al. 2019). In our study, Polaribacter was shown to be capable of ß-mannan, ß-glucan and host glycan utilization, whereas much lower PUL expression was detected in members of the Cryomorphaceae family, in line with the reports from the North Sea (e.g. Sidhu et al. 2023). Seasonal dynamics of CAZyme gene expression was found in Bacteroidia, Gammaproteobacteria, Alphaproteobacteria, and Cytophagia with the highest TPMs at the beginning and at the end of the sampling season. This pattern could arise from competition with phytoplankton during periods of higher algal biomass. The interim sampling period was characterized by higher iron and light availability (Fig. S1B, C), which could have offered competitive advantage for phytoplankton over bacteria (Ratnarajah et al. 2021). In contrast, the availability of algal necromass during phytoplankton senescence could increase the release of metabolites and consequently polysaccharide availability, fueling the expression of PULs (Olofsson et al. 2022). Indeed, both Bacteroidetes (Cryomorphaceae, Flavobacteraceae) and Gammaproteobacteria (Porticoccaceae, Vibrionaceae) have been shown to proliferate during the algae die-off phases utilizing α- and β-glucans (Pontiller et al. 2022, Wang et al. 2024, Beidler et al. 2024). Taxon-specific variation in the expression of glycoside hydrolase genes was detected suggesting differential capacities for polysaccharide degradation both within and between bacterioplankton classes. This was the most pronounced in Gammaproteobacteria, as the ability to utilize certain substrates was exclusive to some of its members. For instance, the expression of arabinogalactan and fructan (GH32, GH42) degradation genes was specific to Nitrincolaceae (Oceanospirillaceae), in line with the diversity of carbon utilization metabolism of this group (Thiele et al. 2023, Zühlke et al. 2026). A wide range of metabolic strategies has been previously found for Gammaproteobacteria members (e.g. Francis et al. 2021) and similar polysaccharide preferences have been detected in other marine bacterioplankton communities (e.g. Sidhu et al. 2023), driven by fine-tuned substrate prioritization mechanisms (Thomas et al. 2021). Alphaproteobacteria were less active in glycan degradation and occurred mostly in Rhodobacteraceae known to be positively correlated with diatom blooms, in particular with Thalassiosira (Tisserand et al. 2020, Mönnich et al. 2020, Pontiller et al. 2022). Alphaproteobacteria generally have less diverse polysaccharide utilization mechanisms compared to Gammaproteobacteria and Bacteroidia and can be outcompeted for ß-glucans (Francis et al. 2021, Wang et al. 2024, Sidhu et al. 2023). Polysaccharide lyase genes involved in alginate degradation were found in all studied taxonomic groups, except for Cytophagia, unlike pectin degradation, which was specific to Bacteroidia. This corroborates the data from other water bodies and is in line with preferential utilization of alginate over pectin (Bunse et al. 2021, Wang et al. 2024). Fucose degradation was most active in March. Fucose likely derives from phytoplankton EPS, fucosylated N-glycans from animal sources (Thomès and Bojar 2021) and from algal glycoproteins (Gügi et al. 2015), and can be putatively released by Fragillaropsis, inhabiting Antarctic sea ice (Dall'Osto et al. 2022). It has previously been shown that the genes responsible for fructose conversion to fucose are upregulated in Antarctic Fragillariopsis cylindricus during the periods of higher photosynthetic activity (Rossi and De Philippis 2016, Aslam et al. 2018). Alternatively, fucose may originate from the phytoplankton development happening before our sampling campaign and was utilized later due to its recalcitrant nature (Vidal-Melgoza et al. 2021). Utilization of ß-glucan, α-glucan, chitin, and xylan was universal and inherent to all major bacterial taxonomic groups, albeit with differences in transcript abundance. ß-glucan and α-glucan utilization dynamics followed a similar pattern, yet, the expression of α-glucan degradation genes was the highest at the beginning of the sampling campaign and was increasing more slowly after an initial drop in the middle of March (Figs. 4B and C & S6 A, B). Steeper increase in ß-glucan degradation compared to α-glucan might result from the fact that bacteria get ß-glucan directly from phytoplankton, whereas α-glucan can be re-used after the death and lysis of bacterial cells. Indeed, recent findings indicate that during a North Sea spring diatom bloom, Bacteroidia utilized excess glucose from algal laminarin degradation to produce internal storage α-glucans, subsequently recycled by other microbial community members from the dead bacterial cells (Sidhu et al. 2023, Beidler et al. 2024). Viral lysis and grazing by heterotrophic flagellates might orchestrate the release of α-glucan storage polysaccharides from bacteria (Deppeler et al. 2020, Biggs et al. 2021, Bartlau et al. 2022). More detailed data on the seasonal dynamics of local protistan and viral community might shed light on the reasons and trends in bacteria mortality. Thus, this factor should be accounted for and investigated in further studies. Microbial processing of DMSP Similar to carbohydrate degradation, DMSP metabolism was taxon-specific. The relevant genes were only transcribed in two classes: Alpha- and Gammaproteobacteria, which is in line with the previous data (Vila-Costa et al. 2007, 2010). Differential expression of demethylation genes was observed within and between bacterioplankton classes. The gene encoding for dimethylsulfoniopropionate demethylase was most actively transcribed in Pelagibacterales, which is consistent with the previous data indicating active expression of this gene in SAR11 (Varaljay et al. 2012, Liu et al. 2022). DmdA was also transcribed in Rhodobacteraceae and Oceanospirillaceae, indicating the capacity of initial DMSP demethylation (Liu et al. 2022). A dmdA transcription increase was observed from March to April, which can be associated with the meltdown event at the end of the sampling season accompanied by the spike of Fragillariopsis species. The dynamics of dmdB and dmdC was characterized by a different trend, which can be explained by the lower substrate specificity of these genes and their resulting broader functions (Bullock et al. 2017). It has been previously shown that DMSP is accumulated in Antarctic sea ice due to its role in cellular osmoregulation (Kameyama et al. 2020), and its content is higher at lower temperatures (Dawson et al. 2020). Therefore, DMSP becomes available to bacterioplankton when it is released from sea ice. Moreover, the increased solar irradiation associated with sea ice retreat has also been linked to more intense DMSP metabolism in the sea surface layer (Varaljay et al. 2012), especially in oligotrophic environments, where reduced sulfur assimilation is energetically advantageous (Vile Costa et al. 2010). A higher expression of demethylation genes was observed after the peaks in biomass of Cryptophyceae and Bacillariophyceae on March 11 and April 5, respectively. Even though DMSP production is usually associated with Phaeocystis antarctica blooms (Sheehan and Petrou 2020) positive correlation has been previously detected with cryptophyte and diatoms abundances in Antarctic waters (Gibson, Swadling and Burton 1996, Speeckaert et al. 2018), Sheehan and Petrou 2020) Both Fragillariopsis and Thalassiosira species have been shown to produce DMSP when cultured from Antarctic waters (Sheehan and Petrou 2020). Therefore, an increase in demethylation might also have been triggered by the peak in Fragillariosis abundance on March 24. Thus, it can be speculated that seasonal DMSP metabolism dynamics are driven by both physical (ice melt and solar irradiation increase) and biological (phytoplankton) factors in the studied bacterioplankton community. No specific DMSP transporters encoded by dddT (Todd et al. 2010, Li et al. 2021) and dmdT (Schroer et al. 2023) genes were detected. However, the genes coding for BCCTs (BetT/BetS, betaine-choline-carnitine transporters) and ABC transporters with broader substrate specificity (ProVWX) were found, which might be involved in DMSP transport (Fig. S8). Nitrogen cycling and photoheterotrophy in bacterioplankton Most expressed nitrogen metabolism genes coded for nitroreductases (nirK, nfnB, nfsB, nfrA1) (Table S8) involved in denitrification, a strictly anoxic process that can occur inside of particles, like those formed from decaying phytoplankton cells (Ciccarese et al. 2023). Another possible source would be an ornithogenic impact from the Gentoo penguin colony that expands on the islands of the Wilhelm Archipelago (Parnikoza et al. 2018, Prekrasna-Kviatkovska et al. 2024). Genes for the biosynthesis of carotenoid and photoprotective pigments exhibited an opposite trend, with the highest expression observed in April, coinciding with the decrease in surface seawater temperature and salinity (Fig. S4). We could not establish a clear connection between the expression of these genes and the overall solar irradiance over the daytime duration (Fig. S1C), which may indicate that the response was likely due to the ice melt and increase in light penetration during the sampling campaign. Bacteriorhodopsin genes were actively transcribed throughout the sampling period (Fig. S4), which corroborates previous reports from Antarctic waters, showing a prevalence of proteorhodopsin-bearing bacteria in surface waters during the late summer-early autumn period (Cifuentes-Anticevic et al. 2021). The taxonomic affiliation of the transcribed genes was similar to the one detected in the Ross Sea region during the austral summer, where photoheterotrophic members of Alphaproteobacteria (SAR11), Bacteroidia (Polaribacter), and Gammaproteobacteria (SAR92) were found (Grzymski et al. 2012). Study limitations Temporal resolution and biological replication was limited due to challenging Antarctic logistics. Thus the results of the current study should be treated as exploratory, focusing on broad functional patterns and their consistency with environmental parameters, rather than on gene-level differential expression. Furthermore, the top-down control by protist grazing pressure and viral lysis on both phyto- and bacterioplankton was not addressed. As such, our study provides background data for future higher resolved investigations of local marine microbial diversity. Additionally, the conclusions of the current study are based on the correlative transcriptomic patterns and thus no direct mechanistic links are established between the phyto- and bacterioplankton communities. Together with a deeper phytoplankton and zooplankton analysis, a higher temporal resolution and controlled mesocosm experiments will validate the mechanisms inferred here and provide deeper insights into bacterioplankton-phytoplankton dynamics which are of utmost importance to the fragile marine ecosystem of the Antarctic Southern Ocean. Conclusion Bacterioplankton community structure and gene expression was studied during the summer–autumn transition in the waters of Wilhelm Archipelago in the fragile and sensitive Southern Ocean ecosystem. Integrating metagenomic, metatranscriptomic, CARD-FISH, and contextual environmental data revealed dynamic taxonomic and functional links between phytoplankton and bacterioplankton communities. Despite the dominance of diatoms and low overall diversity and abundance, genus-specific dynamics within the phytoplankton community influenced the bacterioplankton composition. Corethron criophilum biomass was positively associated with Ca. Prosiliicoccus, whereas Pseudo-nitzschia delicatissima correlated with Nitrincolaceae, Polaribacter, Cetobacterium, and Sulfitobacter. Over the study period, the relative abundance of oligotrophic lineages such as SAR11 increased, accompanied by a decline in copiotrophic taxa. Functionally, the expression of carbohydrate-degradation genes decreased from March to April, possibly reflecting shifts in community metabolism. Expression of ß-glucan, α-glucan, chitin, lignin, and host glycan utilization genes was observed across all microbial taxonomic groups, while fucose, ß-galactan, arabian, and trehalose degradation was inherent to certain taxa only. Likewise, DMSP metabolism was taxon-specific and driven by both physical (ice melt and solar irradiation increase) and biological (phytoplankton dynamics) factors. Supplementary Material Contributor Information Mariia Pavlovska, State Institution National Antarctic Scientific Center, Tarasa Shevchenka Blvd., 16, 01601 Kyiv, Ukraine; Max Planck Institute for Marine Microbiology, Celsiusstraße 1, 28359 Bremen, Germany. Yevheniia Prekrasna-Kviatkovska, State Institution National Antarctic Scientific Center, Tarasa Shevchenka Blvd., 16, 01601 Kyiv, Ukraine. Andrii Zotov, Institute of Marine Biology, National Academy of Sciences of Ukraine, Pushkinska St. 37, 65125 Odesa, Ukraine. Artem Dzhulai, Graduate School of Oceanography, University of Rhode Island, Narragansett, RI 02882, United States. Evgen Dykyi, State Institution National Antarctic Scientific Center, Tarasa Shevchenka Blvd., 16, 01601 Kyiv, Ukraine. Bruno Huettel, Max Planck Genome Centre Cologne, Carl von Linné-Weg 10, 50829 Köln, Germany. Bernhard M Fuchs, Max Planck Institute for Marine Microbiology, Celsiusstraße 1, 28359 Bremen, Germany. Rudolf I Amann, Max Planck Institute for Marine Microbiology, Celsiusstraße 1, 28359 Bremen, Germany. Hanno Teeling, Max Planck Institute for Marine Microbiology, Celsiusstraße 1, 28359 Bremen, Germany. Chandni Sidhu, Max Planck Institute for Marine Microbiology, Celsiusstraße 1, 28359 Bremen, Germany. Author contributions Mariia. Pavlovska (Conceptualization, Formal Analysis, Investigation, Methodology, Project administration, Visualization, Writing – original draft, Writing – review & editing), Yevheniia. Prekrasna-Kviatkovska (Investigation, Validation, Writing – review & editing), Andrii. Zotov (Formal Analysis, Investigation, Validation), Artem. Dzhulai (Investigation, Writing – review & editing), Evgen. Dykyi (Resources, Supervision, Writing – review & editing), Bruno. Huettel (Data curation, Investigation, Methodology), Bernhard. M. Fuchs (Methodology, Validation, Writing – review & editing), Rudolf. I. Amann (Resources, Writing – review & editing), Hanno. Teeling (Conceptualization, Formal Analysis, Investigation, Resources, Supervision, Validation, Writing – review & editing), and Chandni. Sidhu (Formal Analysis, Investigation, Methodology, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing) Conflicts of interest None declared. Funding Mariia Pavlovska was supported by the Scientific Committee on Antarctic Research (SCAR) Fellowship Programme in 2019 and by the National Antarctic Scientific Center of Ukraine (State Special-Purpose Research Program in Antarctica for 2011–2025). Bernhard Fuchs, Hanno Teeling and Rudolf I. Amann were supported by the German Research Foundation (Deutsche Forschungsgemeinschaft-DFG) within the framework of the transregional CRC 420 CONCENTRATE (project 542264307). Chandni Sidhu was supported by the DFG (Project 569718716). Data availability All raw sequencing data have been deposited in the NCBI Sequence Read Archive (PRJNA1347791). Supporting environmental data is also available as Zenodo repository https://doi.org/10.5281/zenodo.17776272 References Associated Data Supplementary Materials Data Availability Statement All raw sequencing data have been deposited in the NCBI Sequence Read Archive (PRJNA1347791). Supporting environmental data is also available as Zenodo repository https://doi.org/10.5281/zenodo.17776272