Taxonomic and functional stability overrules seasonality in polar benthic microbiomes

Abstract Coastal shelf sediments are hot spots of organic matter mineralization. They receive up to 50% of primary production, which, in higher latitudes, is strongly seasonal. Polar and temperate benthic bacterial communities, however, show a stable composition based on comparative 16S rRNA gene sequencing despite different microbial activity levels. Here, we aimed to resolve this contradiction by identifying seasonal changes at the functional level, in particular with respect to algal polysaccharide degradation genes, by combining metagenomics, metatranscriptomics, and glycan analysis in sandy surface sediments from Isfjorden, Svalbard. Gene expressions of diverse carbohydrate-active enzymes changed between winter and spring. For example, β-1,3-glucosidases (e.g. GH30, GH17, GH16) degrading laminarin, an energy storage molecule of algae, were elevated in spring, while enzymes related to α-glucan degradation were expressed in both seasons with maxima in winter (e.g. GH63, GH13_18, and GH15). Also, the expression of GH23 involved in peptidoglycan degradation was prevalent, which is in line with recycling of bacterial biomass. Sugar extractions from bulk sediments were low in concentrations during winter but higher in spring samples, with glucose constituting the largest fraction of measured monosaccharides (84% ± 14%). In porewater, glycan concentrations were ~18-fold higher than in overlying seawater (1107 ± 484 vs. 62 ± 101 μg C l−1) and were depleted in glucose. Our data indicate that microbial communities in sandy sediments digest and transform labile parts of photosynthesis-derived particulate organic matter and likely release more stable, glucose-depleted residual glycans of unknown structures, quantities, and residence times into the ocean, thus modulating the glycan composition of marine coastal waters.


RNA extraction
RNA was extracted from sediment samples from Dec 2017, Feb 2018, and May 2018 from stations 5 and 7 using RNeasy PowerSoil Total RNA Kit (QIAGEN, Hilden, Germany) with some modifications to the manufacturer's recommendation.For May 2018 sediments, RNA was extracted from station 5 from three technical replicate samples.

Modifications were as follows:
 fresh phenol/chloroform/isoamyl alcohol (pH 6.5-8.0) was added to the tube containing the beads before adding the sample.
 As only up to 2.0 g of sample can be used per tube, we subsampled each sample, performing two to five extractions per sample.In particular for winter samples, several replicates were extracted and finally combined to retrieve an amount of RNA that is sufficiently high for metatranscriptomics.
 The incubation from step 7 of the manufacturer's protocol was performed on ice and the centrifugation was done at 3000 x g.
 The centrifugation speed used in step 10 was 5000 x g.
 The last incubation (step 17) was done at -20° C for 30 minutes.
 The centrifugation (step 18) was performed at 15000 x g for 30 minutes.
 For resuspension of the nucleic acid pellet, a smaller volume between 25 to 50 µl of SR7 solution was used.
 Contaminating DNA was digested by using the Invitrogen™ TURBO DNA-free™ Kit (Thermo Fisher Scientific, Bremen, Germany) following the manufacturers' protocol.
CAZyme annotations obtained from dbCAN were accepted when two of the three integrated annotation methods (HMMER v3.3.2,DIAMOND v2.0.9.147, Hotpep version included in run_dbCAN workflow) matched [12].The annotations from Prokka and the databases (except for dbCAN) were compared using a semi-automated approach.This comparison was done using a full Damerau-Levenshtein distance [17,18] between the annotations from different tools/databases (0, identical string; 1,completely different).To account for discrepancies in the annotation string for identical functions in different databases (e.g.capitalization, extra spaces, extra letters), differences ≤25% of total string lengths were allowed.Results were sorted according to i) annotated function matching in at least two of the four databases, ii) only one database annotated the predicted gene (results marked with "*" at the end of the string) or iii), annotated function disagreed between databases (marked with "manualcheck_").Metagenomic abundance was calculated as gene counts divided by genome equivalents that were estimated using MicrobeCensus [19].
Reads classified as rRNA were used for taxonomic profiling by using the SILVAngs pipeline [https://ngs.arb-silva.de/silvangs/, release 138.1, 21] and default settings.For expression analysis, all reads not classified as rRNA or tRNA, were considered as mRNA.
Annotation of transcripts was done by mapping mRNA to predicted genes from metagenomic contigs and bins using DIAMOND blastx [v2.0.15.153, 22] at default settings.Results were filtered for the best hit using the enveomics script BlastTab.best_hit_sorted[16] and >60% identity to reference sequences and a query coverage of >70%.Values of transcripts per million (TPM) mapped reads were calculated after normalization by gene length.Differences between seasons were calculated using the average TPM values in winter (December and February) vs. spring (May) and given as log2 fold change.
Data transformation and plotting was done using R and the tidyverse packages [23,24].

Prediction of monosaccharide composition based on expression of GH genes.
For each GH family, all enzyme activities given in the CAZy database were extracted.Each enzyme activity is referred to one or several monosaccharide released/degraded.The number of different enzymes which can be referred to a specific monosaccharide is determined for each GH family and given in Supplementary Table S7.For the analysis, each enzyme activity given in the CAZy database is considered equally important.For example, for GH1, the following sugars are proposed based on 27 enzyme names: galactose (deduced from 3 enzyme names), glucose (deduced from 18 enzyme names), arabinose, fucose, glucuronic acid, rhamnose, mannose, and xylose (each deduced from 1 enzyme name).In a second step, TPM of expressed GHs were "translated" to the fraction a specific sugar account for by considering the relative contribution of this specific sugar to the GH family.For example, for GH1, release of glucose was predicted based on 18 out of 27 assignments, i.e. contributing 2/3 of total sugars.The final predicted monosaccharide composition is obtained by summing up the contributions of each sugars for each GH family.The script with detailed information is deposited on GitLab under https://gitlab.mpibremen.de/smiksch/gh_family_to_monosaccharides

Monosaccharide analysis
Freeze-dried sediment was homogenized and mixed with 5 ml Milli-Q ultrapure water per mg of sediment and incubated in a sonication bath (Bandelin Sonorex ® , Berlin, Germany) for 60 min at maximum intensity.Afterwards, samples were centrifuged at 6000 x g for 15 min at 20° C and the supernatant was preserved.Porewater and OSW were dialyzed to remove salts using ~1 kDa mesh size dialysis bags (Spectra/Por ® , Fisher Scientific, Bremen, Germany) for 24 h against Milli-Q water.Dialyzed samples were freeze-dried and re-suspended in a tenth of the original volume.Both sediment extracts and pore water and overlying sea water samples were hydrolyzed using HCl (1 M final concentration) for 24 h.Monosaccharides were quantified using High performance anion exchange chromatography (HPAEC) with pulsed amperometric detection (PAD) for details see Vidal-Melgosa et.al. [25].

Supplementary Table S4. Genomic organization of CAZymes in
Average metagenomic abundance (gene counts / genome equivalents)A BSupplementary FigureS1: Glycoside hydrolase families in the metagenomes.(A) Changes in average abundance of GH in spring metagenomes compared to winter metagenomes (calculated by abundance in spring / abundance in winter, given as log2 fold change), (B) average abundance in metagenomes (calculated by gene counts / genome equivalents estimated using MicrobeCensus) .Only GH families with > 0.1% average abundance are shown.
o b i i a B a c t e r o i d o t a D e s u l f o b a c t e r i i a

B
: Expression of key genes involved in nitrogen and sulfur cycling in Svalbard sediments.Abundances in winter and spring metatranscriptomes are shown in transcripts per million (TPM) reads.In particular, nar (nitrite reductase), dsr (dissimilatory sulfite reductase), and apr (adenylylsulfate reductase) were strongly upregulated in winter compared to spring. .Glycan concentrations in overlying seawater and porewater samples from Svalbard Isfjorden.Glycans from OSW and porewater were acid hydrolyzed and the resulting monosaccharides were measured by HPAEC-PAD analysis.Total monosaccharide concentration (sum of concentrations of all different monosaccharides) is shown.Concentrations in porewater were 18fold higher than in OSW.For concentrations of different monosaccharides refer to Supplementary TableS6.ASupplementary Figure S6.Genomic organization (A) and expression of CAZymes (B) in Colwellia binSval_st7_May.bin39.A, In two contigs, PUL-like loci, defined by at least 2 CAZymes within a sliding window of 10 genes, were detected.Additionally, several other degradative CAZymes and CBM were found in the bin.B, Expression of genes from the two PUL-like structures in the spring metatranscriptomes.Stacked bars show the sum of expression of two different genes with identical annotations.Shown open reading frames/genes were not expressed in winter metatranscriptomes except for GH103 and GH73 with low expression levels of 0.2 TPM.GH73 are peptidoglycan hydrolases and slightly higher expressed than in spring metatrascriptomes.CBM5 Glycogen synthase _ GT5 Glucosyl−3−phosphoglycerate synthase _ GT81 Lipid−A−disaccharide synthase _ GT19 UDP−3−O−acyl−N−acetylglucosamine deacetylase _ CE11 Beta−glucosidase BoGH3B _ GH3+GH30+CBM6 Glucosylglycerate phosphorylase _ GH13_18 Pullulanase _ GH13+CBM41+CBM48 Membrane−bound lytic murein transglycosylase _ GH23 Soluble lytic murein transglycosylase _ GH23 CBM50 Murein DD−endopeptidase MepM _ CBM50 Murein hydrolase activator NlpD _ CBM50 Membrane−bound lytic murein transglycosylase D .cazy.org) Bacteroidia bins.+, genes within the same open reading frame.