Alpha-glucans from bacterial necromass indicate an intra-population loop within the marine carbon cycle

Sampling site

Subsurface seawater (1 m depth) was collected at 52 time points between 2nd of March and 26th of May 2020 at the station Helgoland Roads near Helgoland in the southern North Sea. Since 1962 bucket water samples have been taken as part of a long-term monitoring program Helgoland Roads (54°11’N 7°54’E; DEIMS.iD: https://deims.org/1e96ef9b-0915-4661-849f-b3a72f5aa9b1)⁴².

Sequence analysis

Sequence analysis was carried out on the basis of the existing annotation²⁰ with additional reannotation as describes previously²¹. Sequences were aligned using ClustalO (v2.1)⁴³ in Unipro UGENE (v47)⁴⁴. Trees were visualized using iTOL (v6.8.1)⁴⁵.

Chlorophyll a measurements and cells counts of total bacteria and dominant bacterial clades

Sample filtration was carried out under dim light to avoid the pigment loss during the filtration procedure. For 2016 and 2018 samples, pigment extraction and analysis was carried out using a combined protocol from Zapata et al.⁴⁶ and Garrido et al.^4,47. For 2020 samples, we followed the extraction and analysis method as described previously⁴⁸. Subsequently, pigments were separated via high-performance liquid chromatography (HPLC) (Waters 2695 Separation Module), and detected with a Waters 996 Photodiode Array Detector (Waters, Milford, MA, USA). Total bacterial cell counts (TCC) and cell numbers of the dominant clades Aurantivirga (CARD-FISH probe AUR452) and Polaribacter (POL740) of the 2020 spring phytoplankton bloom were described and published previously¹⁷.

18S rRNA gene sequencing

Sampled water was sequentially filtered onto polycarbonate membrane filters with different pore sizes (10 μm, 3 μm and 0.2 μm). For 18S rRNA gene amplicon sequencing, the two larger size fractions, >10 μm and 3–10 μm were analyzed. DNA was extracted from the filters using the DNeasy PowerSoil kit for DNA (Qiagen GmbH, Hilden, Germany). Mechanical lysis was achieved by bead beating in a FastPrep 24 5G (MP Biomedicals LLC, Irvine, CA, USA). The V7 region of the 18S rRNA gene was amplified using the primers [F-1183mod: 5’-AATTTGACTCAACRCGGG-3’, R-1443mod: 5’-GRGCATCACAGACCTG-3’]⁴⁹ coupled to custom adapter-barcode constructs. PCR amplification and Illumina MiSeq library preparation (Illumina Inc., San Diego, CA, USA) and sequencing (V3 chemistry) was carried out by LGC Genomics in Berlin. Sequences have been submitted to the European Nucleotide Archive under the accession number PRJEB51816. Amplicon Sequence Variants (ASVs) were obtained using DADA2 (v1.26)⁵⁰ and taxonomically classified as described previously⁵¹. ASVs classified as Metazoa were removed before downstream analyses to reduce the effect of in particular crustacean zooplankton on community composition. For analysis, 10 μm and 3–10 μm counts were combined, set to 1 and relative abundances calculated. Classification of Dinophyceae into majorly heterotroph and majorly autotroph taxa was done for dominant groups according to literature^{13,52,53,54,55}.

Metatranscriptomics

Metagenome and metatranscriptome sequencing were performed as described previously¹⁷. Briefly, thirty metagenomes were sequenced using PacBio Sequel II (Pacific Biosciences, Menlo Park, CA, USA) with one SMRT cell/sample while corresponding metatranscriptomes were obtained using Illumina HiSeq 3000 (~100 million reads/sample). The metagenomes were then processed to reconstruct metagenome-assembled genomes (MAGs), and mRNA reads were mapped to these genomes to identify highly expressed MAGs. Mapping and annotation were carried out using SqueezeMeta v1.3.1⁵⁶. To predict open reading frames (ORFs), FragGeneScan v1.31 with the parameters “w1” and “sanger_5” as described by Rho et al.⁵⁷ was used. The predicted ORFs were searched against various databases including GenBank r239⁵⁸, eggNOG v5.0⁵⁹, KEGG r58.0⁶⁰, and CAZy (as of 11/12/2023)⁶¹ using Diamond v0.9.24.125⁶². HMM homology searches against the Pfam 33.0 database⁶³ were conducted using HMMER3⁶⁴. The combined annotations were utilized for the manual prediction of polysaccharide utilization loci (PULs) and carbohydrate-active enzyme (CAZyme) clusters. Bowtie2⁶⁵ was employed to map mRNA reads to the MAGs, and transcripts per million (TPM) values were calculated for all MAGs in a given sample using the formula: (sum of reads successfully mapping to a MAG in the sample ×10^6)/(sum of contig lengths of the MAG × sum of reads in the sample).

Metaproteomics

Sample preparation

The metaproteomics analysis of the 0.2 µm fraction from the spring phytoplankton bloom in 2016 has been described in detail in²¹ and the free-living bacteria of the blooms from 2018 and 2020 were prepared as described previously⁶⁶. Briefly, proteins were extracted from one-eighth of a filter (Millipore Express PLUS Membrane, polyethersulfone, hydrophilic, 0.2 µm pore size, diameter 142 mm) by cutting the filter into small pieces before transfer to 15 mL low binding tubes containing 1 mL resuspension buffer 1 (50 mM Tris-HCl (pH 7.5), 0.1 mg mL^−1 chloramphenicol, 1 mM phenylmethylsulfonyl fluoride (PMSF)) and 1,5 mL resuspension buffer 2 (20 mM Tris-HCl pH 7.5, 2% SDS (w/v)). After heating (10 min at 60 °C at 1000 rpm in a thermo-mixer), 5 mL DNAse buffer (20 mM Tris-HCl pH 7.5, 0.1 mg mL^−1 MgCl₂, 1 mM PMSF, 1 μg mL^−1 DNAse I) was added, and cells lysis was carried out by ultra-sonication (amplitude 51–60%; cycle 0.5; 3× 2 min) on ice before incubation for 10 min at 37 °C at 1000 rpm. After centrifugation (10 min at 4 °C at 10,000 × g), the supernatant was collected and the pelleted filter pieces were stirred and centrifuged again for 1 min at 4 °C at 5000 × g. Pre-cooled trichloroacetic acid (20% TCA (v/v)) was added for protein precipitation to the supernatant and after inverting the tube approximately 10x, the precipitate was pelleted via centrifugation (45 min, 4 °C, 12,000 × g) and the protein pellet was washed 3× in pre-cooled (−20 °C) acetone (10 min, 4 °C, 12,000 × g) before drying at room temperature. The proteins were resuspended in 2× SDS sample loading buffer (4% SDS (w/v), 20% glycerol (w/v), 100 mM Tris-HCl pH 6.8, bromphenol blue (tip of a spatula, to add color), 3.6% 2 mercaptoethanol (v/v) (freshly added before use)), incubated for 5 min at 95 °C before vortexing and separated via SDS-PAGE (Criterion TG 4–20% Precast Midi Gel, BIO-RAD Laboratories, Inc., USA). The proteins were fixated, stained with Coomassie, and each gel lane was cut into 20 pieces⁶⁷. Gel pieces were destained 3× for 10 min with 1 mL of gel washing buffer (200 mM ammonium bicarbonate in 30% acetonitrile (v/v)) at 37 °C under vigorous shaking, dehydrated in 1 mL 100% acetonitrile (v/v) for 20 min and the supernatant was removed before drying the gel pieces in a vacuum centrifuge at 30 °C. Proteins were in-gel reduced with 100 µL 10 mM dithiothreitol in 25 mM ammonium bicarbonate buffer (1 h at 56 °C) and alkylated with 100 µL 55 mM iodoacetamide in 25 mM ammonium bicarbonate buffer (without light for 45 min at room temperature) before the supernatant was removed. The gel pieces were washed with 1 mL 25 mM ammonium bicarbonate buffer (10 min, 1000 rpm at room temperature), dehydrated with 500 µL (2018 bloom) or 800 µL (2020 bloom) 100% acetonitrile for 10 min. The supernatant was removed before gel pieces were dried in a vacuum centrifuge (20 min) and finally covered with 120 µL trypsin solution (2 µg/mL Trypsin (Promega). After incubation for 20 min at room temperature, excess trypsin solution was removed and incubated in a thermo-mixer 15 h at 37 °C without shaking. Peptides were eluted with 120 µL solvent A (water MS grade in 0,1% acetic acid (v/v)) by sonication for 15 min before protein containing supernatant was transferred into a new tube. Peptide elution was repeated with 120 µL 30% acetonitrile (v/v) by sonication for 15 min. The eluates were pooled, and eluate volume was reduced in a vacuum centrifuge to a maximum of 15 to 20 µL. The peptides were desalted via ZipTips µC18 (Merck Millipore, P10 tip size) according to the manufacturer’s protocol. The eluted samples were dried in a vacuum centrifuge and resuspended in 10 µL 0.5× Biognosys iRT standard kit in solvent A.

LC-MS/MS measurement and data analysis

Information about the LC MS/MS measurement and data analysis of the 0.2 µm fraction from the spring phytoplankton bloom in 2016 has been described in detail in ref. ²¹ and are described here for the free-living bacteria of the blooms from 2018 and 2020. An Easy-nLC1000 (Thermo Fisher Scientific, Waltham, MA, USA) was coupled to a Q Exactive mass spectrometer (Thermo Fisher Scientific) and peptides were loaded onto in-house packed capillary columns (20 cm length,75 µm inner diameter) filled with Dr. Maisch ReproSil Pur 120 C18-AQ 1.9 µm (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany) and separated using a 131 min nonlinear binary gradient from 1% to 99% solvent B (99.9% acetonitrile(v/v), 0.1% acetic acid (v/v)) in solvent A (0.1% acetic acid (v/v)) at a constant flow rate of 300 nL min^−1. The MS1 scan was recorded with a mass window of 300–1650 m/z and a resolution of 140,000 at 200 m/z. The 15 most intense precursor ions were selected for HCD fragmentation (ions with an unassigned charge or a charge of 1, 7, 8, >8 were excluded) with a normalized collision energy of NCE 27. The resulting MS/MS spectra were recorded with a resolution of 17,500 at 200 m/z. Dynamic exclusion and lock mass correction were enabled.

All MS/MS spectra were analyzed using Mascot (version 2.7.0.1; Matrix Science, London, UK) and a bloom-specific metagenome-derived database containing all protein sequences from the 18 metagenomes obtained during the spring bloom in 2018 or 15 metagenomes obtained during the spring bloom 2020 assuming the digestion enzyme trypsin. Redundant proteins were removed using cd-hit⁶⁸ with a clustering threshold of 97% identity. The non-redundant database was added by a set of common laboratory contaminants and reverse entries, amounting to 81,874,922 (bloom 2018) or 4,221,978 (bloom 2020) sequences in the final database.

For database search with Mascot⁶⁹, the following parameters were used: fragment ion mass tolerance and parent ion tolerance of 10 ppm, none missed cleavages, variable modification on methionine (oxidation), and fixed modification on cysteine (carbamidomethylation). Scaffold (version 4.11.1 (bloom 2018) or version 5.0.1 (bloom 2020); Proteome Software Inc., Portland, OR) was used to merge the search results and to validate MS/MS-based peptide and protein identifications⁷⁰. During data analysis in Scaffold, an additional X! Tandem search was performed for validation (version 2017.2.1.4; The GPM, thegpm.org; version X!Tandem Alanine)⁷¹ with default settings (fragment ion mass tolerance and parent ion tolerance of 10 ppm, carbamidomethyl on cysteine as fixed modification, Glu->pyro-Glu of the N-terminus, ammonia-loss of the N-terminus, Gln->pyro-Glu of the N-terminus and oxidation on methionine for 2018 and 2020 bloom, and additional carbamidomethyl of cysteine as variable modifications for 2018 bloom). Peptide identifications were accepted if they could be established at greater than 95% probability. Peptide probabilities from Mascot were assigned by the Peptide Prophet algorithm (bloom 2018)⁷² or the Scaffold Local FDR algorithm (bloom 2020). Peptide Probabilities from X! Tandem were assigned by the Peptide Prophet algorithm⁷² with Scaffold delta-mass correction. Protein identifications were accepted if they could be established at greater than 99% probability and contained at least two identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm⁷³. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony.

For (semi-)quantitative analysis of 2016²¹, 2018 and 2020 metaproteomic datasets, percent normalized weighted spectra (%NWS) were calculated by dividing Scaffold’s ‘Quantitative Value’ for normalized, weighted (i.e. protein size-adjusted) spectra for each protein group, by the sum of all quantitative values for the sample. Average values were calculated from three biological replicates, including ‘0’ for proteins that were not identified within a replicate. To make bacteria-specific %NWS readily comparable across all samples, all bacterial spectra were normalized to 100% (%BacNWS) using taxonomic assignment for protein groups provided by GhostKOALA v2.0⁷⁴ (genus_prokaryotes + family_eukaryotes + viruses database).

Comparative genomics

MAGs from 2010-2012, 2016²¹ (European Nucleotide Archive project accession PRJEB28156), 2018 (PRJEB38290) and 2020¹⁷ (PRJEB52999) were dereplicated using dRep⁷⁵ v3.2.0 with minimum completeness of 70% and contamination lower than 5% at 0.95 ANI (average nucleotide identity). Protein sequences for representative MAGs were predicted with Prokka⁷⁶ v1.14.6. PULs, CAZymes, SusC-like and SusD-like proteins were predicted as described previously²¹ using hmmscan v3.3.2 against dbCAN-HMMdb-V12 and diamond⁶² v2.1.1.155 against CAZyDB.07262023 provided by dbCAN⁷⁷. PULs were predicted with a sliding window of seven genes, i.e. the CAZymes and susC/D genes can only be seven genes apart for them to be considered part of the same PUL. GH13-encoding PULs were classified as „α-glucan-targeting“, whereas PULs carrying combinations of GH3, GH16, GH17, GH30 and/or GH5 enzymes were annotated as „β-glucan-targeting“. These substrate predictions were curated manually according to further PUL encoded CAZymes.

For identification of enzymes involved in α-glucan synthesis, K numbers were assigned to each sequence by GhostKOALA v2.0⁷⁴ (genus_prokaryotes + family_eukaryotes + viruses) and KofamScan⁷⁸ (ver. 01/04/2023, KEGG release 106) with an E-value ≤ 0.01. Proteins with hits for K00963, K00975, K00693, K00750, K16150, K16153, K13679, K20812, K00703, K16148, K16147, K00700 and K16149 were kept as part of the α-glucan synthesis pathway.

Strain and cultivation conditions

We used the North Sea flavobacterial strains Polaribacter sp. Hel_I_88 (isolated from seawater off Helgoland island) and Muricauda sp. MAR_2010_75 (isolated from seawater at Sylt island), as model organisms⁷⁹. For pre-cultures and polysaccharide extractions Polaribacter sp. Hel_I_88 and Muricauda sp. MAR_2010_75 were grown over night (20 °C, 200 rpm) in Marine Broth (MB 2216, Difco). Polaribacter sp. and Muricauda sp. were tested for growth on specific carbon sources in MPM medium⁸⁰ containing 0.1% (w/v) of a single poly- or monosaccharide (Glycogen from Oyster: Merck; Pullulan: Thermo Fisher Scientific; Glucose: Roth) or 0.2% (w/v) bacterial polysaccharide extract. Growth was assessed via measurement of optical density at 600 nm. Cultures for proteome analysis were carried out in 25 mL MPM medium using biological triplicates.

Proteomics of pure cultures

Cultures of Polaribacter sp. Hel_I_88 for time series sampling were grown in 100 mL batches. Cultures of Polaribacter sp. Hel_I_88 and Muricauda sp. MAR_2010_75 for extract characterization were grown in 25 mL batches. For time series sampling, 25 mL samples were sequentially filtered through 3 and 0.2 μm polycarbonate filters (Ø 47 mm, Merck) using a vacuum pump (PC 3002 VARIO, VACCUBRAND) after 16, 24 and 48 h. Filters were stored at −80 °C until further use. Protein was extracted from ¼ of each 0.2 μm filter and prepared for mass spectrometry as described for metaproteomics but using 10% 1D-SDS polyacrylamide gels.

For extract characterization, cultures were grown on either polysaccharide extract, glycogen and alginate (Sigma-Aldrich) (Polaribacter sp.) or xylan from beechwood (Sigma-Aldrich) (Muricauda sp.). After 72 h, 25 mL cultures were harvested via centrifugation at 4000 × g and stored at −80 °C until further use. Protein was extracted by resuspending the pellet in 2 mL 50 mM TEAB buffer containing 4% (w/v) SDS. Samples were incubated at 95 °C and 600 rpm for 5 min, cooled on ice and sonicated (HD/UV 2070, Bandelin, Berlin, Germany) for 5 min. Debris was removed by centrifugation (14,000 × g, 10 min) and protein concentration was determined using the Pierce BCA Protein Assay Kit (ThermoFischer Scientific). Per sample, 25 μg protein were used. Proteins were separated on a 10% 1D-SDS polyacrylamide gel at 120 V for 90 min.

Samples were measured using an easy nLCII HPLC system applying a 100 min gradient coupled to an LTQ Orbitrap Velos mass spectrometer (Resolution 30,000, Scan range 300-1 700) (Thermo Fisher Scientific Inc., Waltham, MA, USA)⁸¹. Using MaxQuant⁸², spectra were matched using a target-decoy protein sequence database with sequences and reverse sequences of Polaribacter sp. Hel_I_88 (NCBI ASM68793v1) or Muricauda sp. MAR_2010_75 (NCBI ASM74518v1) and common laboratory contaminants. A protein and peptide level FDR of 0.01 (1%) with at least two identified peptides per protein was applied. Only proteins that were detected in at least two replicates were classed as identified. Relative iBAQ (intensity based absolute quantification) values were manually calculated from automatically calculated iBAQs. Data and Results are available through the ProteomeXchange Consortium via the PRIDE partner repository (http://proteomecentral.proteomexchange.org)⁸³ with the identifier PXD043390. Statistical analysis for differential expression was performed in Perseus (2.0.1.1)⁸⁴ using one-way ANOVA followed by post-hoc Tukey’s HSD test.

Cloning, protein expression and purification

Genes coding for the proteins GH13A (P161_RS0117435), GH13B (P161_RS0117440) and GH13C (P161_RS0117455) of Polaribacter sp. Hel_I_88 (NCBI accession NZ_JHZZ01000001) were codon optimized and synthesized without their signal peptide by de novo gene synthesis (BioCat GmbH, Heidelberg, Germany). They were cloned into pET22b+ in Escherichia coli BL21 (DE3) for protein production. The susD gene (P162_RS13765) of Flavimarina sp. Hel_I_48 (NCBI accession: NZ_JPOL00000000) was amplified from genomic DNA via PCR (primers fwd: GTGTCTCGAGTTAATAACCTGGGTTTTGAGTCAGGTTT; rev: GAGAGGATCCTGAGAAT GATCTTGACGTAACCTTAGAG) and cloned into pET22b+ via restriction/ligation. Proteins were produced in 200 mL LB cultures (30 μg mL^−1 ampicillin) by induction with IPTG and incubation over night at 20 °C. Cells were harvested by centrifugation (5000 × g, 20 min), lysed using BugBuster Protein Extraction Reagent (Merck) (GH13s) or sonication on ice (3 × 2 min, 50% cycle, SusD) and centrifuged (9500 × g, 20 min) to remove debris. Proteins were purified by loading the lysate onto a prepacked 5 mL IMAC column (HisTrap HP 5 mL, Cytiva) equilibrated with IMAC Buffer A (100 mM NaCl, 20 mM Imidazole, 20 mM Tris-HCl, pH 8) using an ÄKTA Pure 25 L (Cytiva). Proteins were eluted with a step gradient of IMAC Buffer B (100 mM NaCl, 500 mM Imidazole, 20 mM Tris-HCL, pH 8). Flavimarina sp. SusD was further purified using size-exclusion chromatography (Superdex 200 Increase 10/300 GL, Cytiva) using SEC-Buffer (100 mM NaCl, 20 mM Tris-HCL, pH 7.4). The GH13 A, B & C proteins were desalted using a prepacked sepharose-based desalting column (HiPrep Desalting 26/10, Cytiva) with PBS Buffer (pH 7.4). All proteins were concentrated via spin columns (Pierce Protein Concentrator PES, 30K MWCO, 2–6 mL, Thermo Fischer).

Enzyme characterization

Activity profiles for all enzymes were generated by 3,5-dinitrosalicylic acid (DNS) reducing-end assay⁸⁵ as well as fluorophore-assisted carbohydrate electrophoresis (FACE)⁸. 25 μg purified protein were incubated with 0.5% (w/v) poly-/oligosaccharide (dextran 70, maltose, maltotriose, isomaltotriose, and maltopentose from Roth; laminarin from Laminaria digitata, α-/β-cyclodextrin from Sigma; panose and isopanose from Megazyme) for 24 h. Samples were heat-inactivated at 80 °C for 10 min and centrifuged (13,000 × g, 10 min) to remove precipitated protein.

For the reducing-end assay, samples were incubated with DNS-reagent solution (30% (w/v) Potassium sodium tartrate tetrahydrate, 10 mg mL^−1 DNS, 0.4 M NaCl) for 15 min at 95 °C and cooled to RT before measurement at 540 nm (Infinite 200 PRO M PLEX, Tecan, Männedorf, Switzerland). Values were compared against those of solutions containing only polysaccharide or only enzyme. Statistics of reducing-end assays were performed using a one-way Student’s t test with an FDR of ≥0.05.

FACE was performed with 8-aminonaphthtalene-1,3,6-trisulfonic acid (ANTS) as fluorophore. 100 μL of the reaction samples were dried in a SpeedVac (Concentrator Plus, Eppendorf) and dissolved in 4 μL 0.05 M ANTS (in DMSO, 15% (v/v) acetic acid) and 4 μL 1 M NaCNBH₃ (in DMSO). They were incubated over night at 37 °C before being loaded onto a FACE-Gel⁸⁶ and separated at 400 V for 1 h.

Thin-layer chromatography

25 μg purified protein were incubated with 0.5% (w/v) poly-/oligosaccharide in PBS for 24 h. Samples were heat-inactivated at 80 °C for 10 min and centrifuged (13,000 × g, 10 min) to remove precipitated protein. Glucose, maltose and maltotriose (all 1 mg mL^−1) in PBS were used as standard for the chromatography. The samples were analyzed as described previously⁸⁷ on silica gel plates (60 F245) with a mixture of 1-butanol, acetic acid and water (2:1:1) as solvent. Plates were sprayed with staining solution (4 g α-diphenylamine, 4 mL aniline, 200 mL acetone, 30 mL phosphoric acid 80% (v/v)) and visualized by heating above 100 °C.

Affinity gel electrophoresis

Gel electrophoresis was carried out using native 12% acrylamide-gels as described previously⁸⁸. Purified Flavimarina sp. SusD was diluted with 10 mM Tris-HCL pH 7.4 to decrease the NaCl concentration to 20 mM. Flavimarina sp. SusD and BSA, which was used as non-binding control, were loaded onto gels containing either no additive or 0.2% of the tested polysaccharide. Runs were performed at 80 V with cooled buffer on ice. Gels were stained using Coomassie Brilliant Blue.

Protein structure prediction

Structure models of Flavimarina sp. Hel_I_48 (Supplementary Dataset 5) and Muricauda sp. MAR_2010_75 SusD (Supplementary Dataset 6) were predicted using ColabFold⁸⁹ using default settings with the top-ranked structure relaxed (num_relaxed: 1). These models were used for an overlay with the structure of the experimentally characterized homolog from Bacteroides thetaiotaomicron in complex with cyclodextrin to determine differences between the SusD binding sites. Models were colored based on the AlphaFold confidence score (pLDDT) and figures were created using PyMOL (Schrödinger, New York, NY, USA)⁹⁰.

Lysate activity measurements

Samples were taken from cultures growing on 0.1% (w/v) of a single carbon source or 0.2% (w/v) intracellular enriched bacterial extract. Bacteria were harvested by centrifugation (4000 × g, 10 min) and lysed via sonication on ice in PBS (3 × 2 min, 50% cycle). Debris was removed via centrifugation (13,000 × g, 10 min). Supernatant protein concentration was measured by BSA-assay and 25 μg of protein were incubated with 0.5% of tested polysaccharide for 24 h at RT. Samples containing only polysaccharide or only extract were treated similarly as controls. Lysate activity was measured via DNS-Assay and FACE as described above. Significance of the results was determined via one-way ANOVA followed by post hoc Tukey’s HSD test.

Polysaccharide extraction

Polysaccharides were extracted from enriched intracellular fractions of Polaribacter sp. Hel_I_88. 200 mL culture were harvested via centrifugation (4000 × g, 20 min, 4 °C) and washed once with 10 mL MOPS buffer (20 mM, pH 8) before being resuspended in ddH₂O. Polysaccharide extraction was carried out according to a protocol modified from literature⁹¹. In short, attached particles were removed by centrifugation (500 × g, 10 min) and cells were lysed by sonication on ice (3 × 2 min, 50% cycle). Two more centrifugation steps (1100 × g, 30 min and 27,000 × g, 15 min) were carried out to remove unbroken cells and membrane fragments. The pellets were pooled and served as control to ensure enrichment of intracellular polysaccharide (attached fraction). Three volumes of glycine-buffer (0.2 M, pH 10.5) and two volumes of chloroform (both 4 °C) were added to the supernatant and shaken vigorously for 30 s. Phase separation was achieved by centrifugation (100 × g, 2 min) and the aqueous phase removed. The remaining organic phase was re-extracted twice with 2 volumes of glycine buffer and all aqueous phases pooled. They were centrifuged at 47,000 × g for 3 h until a gelatinous pellet remained. After resuspending the pellet in 5 mL ddH₂O, 6 volumes of ethanol (4 °C) were added to precipitate the polysaccharides over night. Precipitate was then centrifuged (14,000 × g, 1 h), resolubilized in ddH₂O, dialyzed against ddH₂O over night to remove residual salts and finally dried in a SpeedVac. Extracts were weighed and stored at −20 °C until further use.

Bacterial glycan extract characterization

To determine specific components of the bacterial polysaccharide extracts, 5 mg extract were resuspended in PBS and then were incubated with 25 μg of the characterized enzymes GH13A, GH13B, GH13C) and C. forsetii GH16⁸⁸, respectively. Samples were analyzed by reducing-end assay and FACE as described above. Mono- and oligosaccharide release was compared to samples containing either untreated extract or extract after acid hydrolysis. For acid hydrolysis, 5 mg glycan extract were boiled with 1 M HCl for 2 h, and subsequently neutralized using 1 M NaOH. Monosaccharide composition of all samples was determined via HPAEC-PAD using a Dionex CarboPac PA10 column (Thermo Fisher Scientific) and monosaccharide mixtures as standards⁹².

Determination of glucan concentrations on filters

Polysaccharide extraction was performed from 3 and 0.2 μm bloom membrane filters of the spring bloom samples and from bacterial single-cultures. Analysis of the extracts was carried out as described⁹³. In short, membrane filters were cut into small pieces and extracted using hot ddH₂O with sonication treatment and debris was removed via centrifugation (4500 × g, 15 min). α-glucan content was determined via incubation with amylase (Aspergillus oryzae, Megazyme) and amyloglucosidase (Aspergillus niger, Merck) in sodium acetate buffer (0.1 M, pH 4.5) followed by a PAHBAH-Assay⁹⁴. The measured values were corrected by the amount of water filtered, thereby taking bacterial cell numbers into account.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.