Sources of α-glucans during algal blooms
Recently, we conducted a study on the response of free-living planktonic bacteria (0.2-3 µm) to a diatom-dominated spring bloom at Helgoland Roads (54°11'N 7°54'E, German Bight). Chlorophyll a and microscopic bacterial count data across a three-month time period from the beginning of March to the end of May 2020 revealed a biphasic bloom with a first phase around the end of March until mid-April followed by a main phase from the end of April until the end of May, during which the dominant flavobacterial clades Polaribacter and Aurantivirga responded (Fig. 1A&B).
Diatoms globally produce substantial amounts of β-glucans in the form of laminarin5, but are not known to produce notable amounts of α-glucans21. To identify α-glucan sources, we analyzed 18S rRNA gene amplicon data, obtained from > 10 µm and 3–10 µm biomass size fractions (53 time points). Corresponding to microscopic biovolume data obtained in the framework of the Helgoland Roads time series22,17, 18S rRNA gene sequences confirmed the centric diatoms Dytilum bightwellii and Ceratulina pelagica as dominant algae. Importantly, algae with α-glucans such as Rhodophyta, Chlorophyta and Cryptophyceae21 were rare (Fig. 1C). However, Dinophyceae (dinoflagellates), also known to contain α-glucans21, were detected throughout the sampling period and continuously made up 7 to 47% of the eukaryotic population (> 10 µm and 3–10 µm fractions). Dinoflagellate autotrophs (e.g. Karenia spp.) and heterotrophs (e.g. Gyrodinium spp.) were both abundant during the first bloom phase (up to 27% and 20% of the eukaryotic population, respectively), whereas heterotrophic dinoflagellates dominated the main bloom phase (up to 30%), likely in response to higher bacterial cell numbers. Likewise, choanoflagellates, known to feed on bacteria, became more prominent during the main bloom phase (Fig. 1C, Tab. S1). Grazing by heterotrophic flagellates therefore likely represents a factor that promotes the release of bacterial storage α-glucan into the DOM pool. This would also corroborate an observed tight correlation between bacteria responding to blooming diatoms and the expression of their α-glucan PULs, in particular in top-expressed metagenome assembled genomes (MAGs) of the dominant responder clades Aurantivirga and Polaribacter14 (Fig. 1D).
Marine Flavobacteriia degrade different types of α-glucans
In a previous study, we sequenced 53 coastal North Sea Flavobacteriia strains, 75% of which featured α-glucan PULs, more than for any other polysaccharide19. Analysis of PUL genes coding for carbohydrate-active enzymes (CAZymes) revealed a variety of gene modules centered around one or more family 13 glycoside hydrolases (GH13) that involved also GH65, GH31 and GH97 genes alongside the characteristic susCD gene pair (Fig. S1A).
Sequence alignment of these SusD-like substrate binding proteins together with α-glucan-binding SusD proteins from bacterial MAGs of the 2020 spring bloom at Helgoland Roads17 and the well-characterized α-glucan-binding SusD from Bacteroides thetaiotaomicron23 uncovered two distinct SusD types. Corresponding AlphaFold2 structure predictions revealed a separation into two functional groups, one of which was characterized by an about 17 amino-acid-containing loop close to the binding site, similar to one described in B. thetaiotaomicron24. The second group lacked this loop along with two residues shown to be substrate-binding in B. thetaiotaomicron. The result is a more open binding site that may serve as an adaptation to a structurally distinct α-glucans (Fig. S1B). MAG-derived SusD sequences (0.2-3 µm fraction) including such with expression during the 2020 bloom clustered with both groups, indicating ecological relevance for both variants (Fig. 2). Gene composition analysis showed that PULs with an open type SusD almost exclusively coded for basic enzymes such as GH13, whereas PULs with a looped SusD comprised a wider variety of CAZymes, most notably more GH13s as well as at least one SusE-like protein. The widest variety of GH13s was found in Polaribacter strains, representing one of the most recurrent bloom-associated bacterial clades at Helgoland Roads12.
We conducted growth experiments with North Sea strains Polaribacter sp. Hel_I_88 (PUL with looped SusD) and Muricauda sp. MAR_2010_75 (PUL with open SusD). These experiments revealed a correlation between α-glucan substrate complexity and growth efficiency when α-glucan was offered as sole carbon source. While Polaribacter sp. grew well on glycogen and pullulan, which has a α-1,4- α-1,4- α-1,6 repeating unit, Muricauda sp. showed a strong preference for glycogen and grew poorly on pullulan (Fig. S1C). This demonstrates the presence of distinct α-glucans niches among marine bacteria that specialize on different types of α-glucans.
Marine bacteria contain multiple enzymes targeting α- glucans
Protein sequence alignment of all PUL-encoded GH13s within the two studied isolates showed the Polaribacter sp. enzymes to represent the majority, as roughly 70% of all isolate- and MAG-associated GH13s grouped with them. Additionally, they represented three of the GH13s encoded in highly expressed bloom-associated MAGs (Fig. 3A, Tab. S2). A notable exception was a GH13_31 with proposed α-1,6 activity. While underrepresented in the isolates, the respective gene was highly expressed during the 2020 bloom, indicating presence of this linkage type in marine bacterial α-glucans.
We heterologously expressed three GH13s encoded in the Polaribacter sp. Hel_I_88 PUL in E. coli and purified the enzymes for biochemical characterization. Via 3,5-dinitrosalicylic acid (DNS) reducing end assay and fluorophore-assisted carbohydrate electrophoresis (FACE), the enzymes GH13A (P161_RS0117435) and GH13C (P161_RS0117455) were shown to act on the α-1,4 linked glucans glycogen and pullulan. GH13A preferred the mostly α-1,4-linked glycogen, producing a dimer and smaller amounts of glucose, whereas GH13C was more active on pullulan, releasing primarily products with a degree of polymerization (dp) of three. The enzymes were inactive on α-1,6-linked dextran and β-1,3-linked laminarin. Selectivity for alpha-1,4 linkages was confirmed with malto-oligosaccharides. These oligosaccharides were degraded by both enzymes until dp2. α-1,6-linked isomalto-oligosaccharides were not hydrolyzed. Experiments with mixed-link α-1,6/α-1,4 oligosaccharides revealed minor activity on isopanose (α-1,4-α-1,6) and panose (α-1,6-α-1,4), indicating that an α-1,6-bond next to the 1,4 connected glucose monomers in -1 and + 1 subsites disables hydrolysis. Additionally, both enzymes acted on β-cyclodextrin, releasing dp1 and dp2 (GH13A) and dp1, 2 and dp3 (GH13C), respectively. Interestingly, similar activity could not be detected on α-cyclodextrin, indicating that the enzymes recognize a specific substrate diameter (Fig. 3B, S2A&C).
GH13B (P161_RS0117440) showed only minor activity on glycogen and pullulan, releasing a dimer. Of the tested oligosaccharides with only one linkage type, only such containing α-1,4-linkages were acted upon, but all activities remained minimal. However, from isopanose, GH13B released notable amounts of dp1 and dp2, clearly indicating a preference for α-1,4-bonds situated next to α-1,6 bonds. No activity could be detected with panose, and as the enzyme released only dimers from polysaccharides it can be assumed that the enzyme specifically releases isomaltose from the reducing end (Fig. 3B, S2B). These results support an adaptation of marine bacteria towards α-1,4/α-1,6 substrates.
Marine bacteria synthesize alpha glucans
α-glucans are known as major storage compounds of marine bacteria. They should therefore be formed during peak bloom phases, when excess organic carbon from algae outweighs the availability of other essential nutrients such as nitrogen, as has been shown under nitrogen limitation in vitro25. Bacteria synthesize glucose-based storage polysaccharides via proteins encoded in the glg-operon, by addition of glucose-1-phosphate or maltose-1-phosphate to ADP-glucose (via GlgA or GlgE, respectively). This results in linear glycogen, which is branched with α-1,6 linkages by the branching enzyme GlgB26. Metaproteome analysis of bacteria-dominated 0.2-3 µm filters from spring blooms at Helgoland Roads during 2016, 2018 and 2020 revealed spikes in bacterial glg-operon protein abundances. These correlated well with bloom progression as determined by chlorophyll a concentration measurements as well as laminarin- and α-glucan targeting protein abundances (Fig. S3). Except for 2018, where laminarin-degradation proteins remained comparatively low during the bloom (Fig. S3A&D), a correlation of higher abundances of laminarin-targeting Glg-proteins and α-glucan uptake proteins could be observed (Fig. S3B&C). This suggests that α-glucan synthesis is a general mechanism of bloom-associated Flavobacteriia in response to growth on laminarin. Consequently, when growing Polaribacter sp. Hel_I_88 on laminarin as sole carbon source, we detected a significant increase in α-glucan over time via specific enzymatic hydrolysis (Fig. S4A). Thus, laminarin degradation coupled with simultaneous α-glucan-synthesis could be confirmed with an isolated strain in vitro.
Proteomics revealed that proteins encoded by the glg-operon (P161_RS0109480, RS0109490, RS0109495 & RS0109500) were expressed continuously during growth on laminarin (Fig. S5A), promoting bacterial α-glucan formation. As expected, overall protein abundance was dominated by the laminarin-PUL (P161_RS0117335-P161_RS0117415) (Fig. S5B). Yet, both the SusC/D-like protein pair (P161_RS0117480/85) and a GH13 (P161_RS0117500) of the α-glucan PUL became more abundant in later growth phases for which we showed increased amounts of α-glucan present in the culture (Fig. S5C). A similar induction could not be shown for proteins of other PULs, such as the alginate PUL (P161_RS0107490- P161_RS0107540) (Fig. S5D, Tab. S3). These findings support the view that, as bacteria of the culture begin to lyse, the released organic matter was sensed, taken up and utilized. Analysis of 3 and 0.2 µm filters sampled during the 2020 Helgoland spring bloom showed that α-glucans were more abundant on 0.2 µm filters, which largely represent the free-living planktonic bacterial population. Concentrations rose to around 50 µg/L at the end of March, coinciding with the first bloom event the bacteria responded to. A spike to over 150 µg/L was observed at the beginning of May, coinciding with the main bloom phase (Fig. S4B). Taken together, these results corroborate that marine bacteria produce significant amounts of α-glucan storage polysaccharide during microalgal blooms.
Bacterial polysaccharide contains α- 1,4-glucans
Monosaccharide analysis of polysaccharide extracts from Polaribacter sp. Hel_I_88 cultures via high performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) revealed high proportions of glucose (29 mol%) (Fig. 4A). Incubation of this polysaccharide with recombinant Polaribacter sp. GH13A, GH13B and GH13C, showed visible degradation in reducing end assays (Fig. 4B), corroborating that the extract contained α-glucans. FACE-analysis of incubations with GH13A and GH13C yielded oligosaccharides of different dp, but predominantly dp2, which was also the main product formed by incubation of either enzyme with glycogen. This was supported by GH13B releasing only dp2 from the extract, indicating that the extracted polysaccharide indeed contained bacterial α-glucans (Fig. 4C).
Bacterial polysaccharide induces α-glucan PUL expression
Polaribacter sp. Hel_I_88 grew on polysaccharide extract from lysed cells as sole carbon source (Fig. S1B). Comparisons of culture lysate activity to glycogen or alginate-grown cultures showed increased activity on α-1,4-glucan-containing substrates for cultures that were grown on polysaccharide extract (Fig. 5A, Fig. S6). Corresponding proteomics revealed an induction of the entire α-glucan PUL (P161_RS0117430-P161_RS0117500) compared to alginate controls, with SusC and SusD proteins making up 0.46% and 1% of the entire proteome, respectively. Interestingly, this induction was higher than for a culture grown on glycogen as positive control, showing that the extracted polysaccharide elicited a more pronounced response (Fig. 5A, Tab. S4).
While growth on extracted polysaccharides could also be observed for Muricauda sp. MAR_2010_75 (Fig. S1B), no significant activity was visible in reducing end assays with culture lysate (Fig. 5B, Fig. S6). FACE analysis showed the degradation of α-glucan from bacterial lysates under all tested conditions, indicating a basal activity of the α-glucan PUL expression in Muricauda sp. MAR_2010_75 but no differential induction of the α-glucan degradation machinery under these conditions. Proteomics revealed the α-glucan SusC- and SusD-like proteins (FG28_RS04375, FG28_RS04380) of this strain as most abundant during growth on bacterial polysaccharide extracts, but these were also highly expressed during growth on either glycogen or xylan. This high abundance was not mirrored by the PUL’s associated CAZymes, a GH13 (FG28_RS04360) and a GH65 (FG28_RS04365), and suggests a difference in α-glucan utilization by bacteria with open-type SusD-containing α-glucan PULs (Fig. 5B, Tab. S4).