Dynamics of a methanol-fed marine denitrifying biofilm: 2—impact of environmental changes on the microbial community

Cultivation of the original biofilm to different culture conditions

The formulations of the artificial seawater (ASW) medium and the commercial Instant Ocean (IO) medium (Table S1), and the different conditions of the biofilm cultures were described by Payette et al. (2019). Briefly, the biomass of several carriers taken from the denitrification reactor of the Montreal Biodome was scraped, dispersed, then distributed to several vials containing twenty free carriers and 60 mL prescribed medium (Table 1; Fig. S1). The vials were incubated under anoxic conditions at 23 °C or 30 °C (Table 1) and shaken at 100 rpm (orbital shaker). In average once a week, the twenty carriers were taken, gently washed to remove the excess medium and the planktonic bacteria, then transferred into fresh anoxic medium and incubated under the same conditions (Fig. S1). The Ref300N-23C biofilm cultures (for 300 mg NO₃⁻-N/L, 23 °C) were defined as the reference biofilm cultures. These cultures were used by Payette et al. (2019) as a reference to compare results between the different culture conditions. The protocols to measure NO₃⁻ and NO₂⁻ concentrations, and to extract DNA from the biofilm cultures or the planktonic pure cultures were described in Payette et al. (2019) and Geoffroy et al. (2018).

16S rRNA gene analysis

DNA extracted from triplicate biofilm cultures was pooled before sequencing. Total DNA samples from seven biofilm cultures (Table 1; Ref300N-23C, 300N-30C, 900N-23C, 900N-30C, 0%NaCl, 0.5%NaCl and 1.0%NaCl) were sent to the sequencing service at the Research and Testing Laboratory (RTL, Lubbock, Texas, USA). A region of the 16S rRNA gene was PCR amplified using the 28F-519R primers (5′ GAGTTTGATCNTGGCTCAG 3′ and 5′ GTNTTACNGCGGCKGCTG 3′, covering the V1–V2–V3 variable regions) and subjected to pyrosequencing using a Roche 454 FLX genome sequencer system. The sequencing service (RTL) performed denoising and chimera analyses (details provided in Document S1). The high-quality reads were then processed in the RDP pipeline at the Ribosomal data project (RDP) web site (Cole et al., 2014). Reads were clustered into operational taxonomic units (OTU) using a 97% identity threshold. DNA extracted from the OB (from frozen stock) and from fresh IO biofilm cultures (Table 1) were sent to the sequencing service of Genome Quebec Innovation Center (Montreal, QC, Canada). In these cases, the 16S rRNA sequences covering the V6–V7–V8 variable regions (5′ ACACTGACGACATGGTTCTACA 3′ and 5′ TACGGTAGCAGAGACTTGGTCT 3′) were PCR amplified and sequenced by Illumina MiSeq PE250 (250 bp paired-end sequencing reactions). The reads were processed based on Peck et al. (2016). Briefly, paired-end reads were merged with minimum and maximum overlap length between the two reads of 20 and 250 bases, respectively, with 30% mismatched bp tolerance in the overlap region. The merged reads were processed using the software UPARSE (Edgar, 2013). Sequences were truncated to a uniformized length to 420 bp. Reads with a low-quality score were removed using 2.0 as the maximum expected error value. The high-quality reads were de-replicated, sorted by size and singletons were removed. The resulting reads were clustered into operational taxonomic units (OTU) with the UPARSE-OTU clustering method using a 97% identity threshold. Chimeric OTU were removed by UPARSE-REF algorithm and with the software UCHIME ran against ChimeraSlayer ‘gold’ reference database (Edgar et al., 2011). All representative sequences of the OTUs (from pyrosequencing and Illumina) were checked again for chimeras with the DECIPHER v 2.0 program (http://www2.decipher.codes/FindChimeras.html) (Wright, Yilmaz & Noguera, 2012). The affiliation of the OTUs to the most probable genus was determined by the CLASSIFIER program at the RDP web site (Document S2). 16S rRNA sequence reads were deposited in the GenBank Sequence Read Archive (SRA) under the accession number PRJNA524642. Principal component analysis of the proportion of reads associated to the bacterial profiles was performed at ClustVis web site (https://biit.cs.ut.ee/clustvis/) (Metsalu & Vilo, 2015).

Isolation of bacterial isolates

Biofilm of the Ref300C-23C biofilm cultures was scraped from the carriers and dispersed in saline solution (3% NaCl, 34.2 mM phosphate buffer pH 7.4), and serial dilutions were made and inoculated onto these agar plate media: (1) R2A medium (complex organic carbons; EMD Chemicals Inc., Gibbstown, NJ, USA), (2) Marine Agar 2216 (marine medium with yeast extract and peptone as carbon source; Becton, Dickinson and Co., Sparks, MD, USA), (3) Methylophaga medium 1403 (American Type Culture Collection [ATCC], Manassas, VA, USA) and (4) the ASW medium; these two latter media were supplemented with 1.5% agar and 0.3% v/v methanol. The isolation procedure, the taxonomic affiliation of the isolates and the measurement of their denitrifying activities were carried out as described by Geoffroy et al. (2018). The 16S rRNA gene sequences were deposited in GenBank under the accession numbers MK571459–MK571476.

Transcriptomes

Planktonic pure cultures of M. nitratireducenticrescens strains JAM1 and GP59 were performed in the Methylophaga 1403 medium and of H. nitrativorans strain NL23 in the 337a medium as described by Martineau, Mauffrey & Villemur (2015) and Mauffrey, Martineau & Villemur (2015). These cultures were carried out in triplicate with methanol (0.3%) and NO₃⁻ (21.4 mM [300 mg-N/L]) under anoxic conditions at 30 °C. The biomass was collected by centrifugation when the NO₃⁻ reduction was near completion, and total RNA was extracted as described by Mauffrey, Martineau & Villemur (2015). For the biofilm cultures, at the end of the the fifth transfer cultures, the biomass of each replicate was scraped from carriers and used to extract total RNA. The RNA samples were sent to the sequencing service for RNA sequencing (RNAseq) by Illumina (Genome Quebec Innovation Center, Montreal QC, Canada). Because of limited amount of biofilm available, total RNA from the triplicate biofilm samples were pooled before sending to the sequencing service. For the planktonic pure cultures, RNAseq was performed on each replicate. The Ribo-Zero™ rRNA Removal Kit (Meta-Bacteria; Epicentre, Madison, WI, USA) was used to deplete total RNA of the ribosomal RNA. The RNA was then treated with the TruSeq Stranded mRNA Sample Prep Kit (Illumina Inc, San Diego, CA, USA).

All computations were made on the supercomputer Briarée from the Université de Montréal, managed by Calcul Québec and Compute Canada. Raw reads were filtered to remove low quality reads using FASTX toolkit (http://hannonlab.cshl.edu/fastx_toolkit/) by discarding any reads with more than 10% nucleotides with a PHRED score <20. The resulting reads from each sample/replicate were aligned respectively to the genome of M. nitratireducenticrescens strain JAM1 (GenBank accession number CP003390.3), to the genome and plasmids of M. nitratireducenticrescens strain GP59 (CP021973.1, CP021974.1, CP021975.1) and to the genome of H. nitrativorans strain NL23 (CP006912.1) using Bowtie (v 2.2.3) with default parameters. SAMtools (v 0.1.18) and BEDtools (v 2.20.1) were used for the generation of sam and bam files, respectively. Significance for difference in the relative transcript levels of a gene (defined as transcript per million: TPM) between planktonic pure cultures and biofilm cultures was performed with the R Bioconductor NOIseq package v2.14.0 (NOIseqBio) (Tarazona et al., 2011) and run with the R software v3.2.3 (Team, 2015). Because the RNAseq from the biofilm samples were derived from one pooled RNA preparation, the “no replicate parameter” was set (pnr = 0.2, nss = 5 and v = 0.02; pseudoreplicate generated) in NOIseq as described by Tarazona et al. (2011) under the NOISeq-sim section. Briefly, NOISeq-sim assumes (quoting) ”that read counts follow a multinomial distribution, where probabilities for each gene in the multinomial distribution are the probability of a read to map to that gene”. Results from this statistical analysis showed that genes that had at least >2-fold higher transcript levels from one type of cultures to the other showed significant differences. RNAseq reads from the planktonic pure cultures and the biofilm cultures were deposited in the SRA under the accession number PRJNA525230. Annotations were based on services provided by GenBank (https://www.ncbi.nlm.nih.gov/genbank), RAST (Rapid Annotation using Subsystem Technology; http://rast.nmpdr.org) and KEGG (Kyoto Encyclopedia of Genes and Genomes; https://www.genome.jp/kegg) (Document S3).

To derive transcript reads not associated to M. nitratireducenticrescens and H. nitrativorans, reads were aligned to a concatenated sequence consisting of the three reference genomes (JAM1 + GP59 + NL23) and the two plasmids (from strain GP59). The reads that did not align were kept. Unaligned reads were de novo assembled at the National Center for Genome Analysis web site (https://galaxy.ncgas-trinity.indiana.edu) by Trinity v. 2.4.0 (Grabherr et al., 2011). These transcripts were deposited in SRA under the accession number PRJNA525230. Estimation of the transcript abundance of the de novo assembled sequences was performed by RSEM (Li & Dewey, 2011). The resulting assembled sequences were annotated at the Joint Genomic Institute (https://img.jgi.doe.gov/cgi-bin/m/main.cgi) to find open reading frames with their putative function and affiliation (GOLD Analysis Project Id: Ga0307915, Ga0307877, Ga0307760). The annotations were then verified manually for discrepancies within the assembled sequences (Documents S4–S7).

Bacterial composition of the biofilm cultures by 16S rRNA gene sequencing

As reported by Payette et al. (2019), the original biofilm (OB) collected from the Biodome denitrification reactor was used as inoculum to colonize new carriers in a series of anoxic biofilm cultures cultivated under different conditions (Table 1; Fig. S1). We selected eight of these biofilm cultures for our present analysis for the following reasons. The Ref300N-23C biofilm cultures were used as reference for comparison analysis. The 300N-30C, 900N-23C and 900N-30C biofilm cultures had higher specific denitrification rates compared to the Ref300N-23C biofilm cultures. The 0%, 0.5% and 1.0% NaCl ASW biofilm cultures and the IO biofilm cultures were chosen because of the persistence of H. nitrativorans NL23 in these cultures as opposed to the other cultures. The composition of the bacterial community of these biofilm cultures and the OB was determined by sequencing the 16S rRNA genes to assess the impact of these specific conditions on the bacterial community (Fig. 1A; Table 2). The bacterial profiles of the OB and the IO biofilm cultures were distinct to each other, and to the seven biofilm cultures cultivated with different formulations in the ASW medium (Fig. 1B). The bacterial profiles of these latter cultures were however not very different because of the high proportions of Methylophaga spp. (>85%).

In the OB, high proportions of the 16S rRNA gene sequences were related to Hyphomicrobium spp. (45.8%) followed by Oceanibaculum spp. (12.3%), Aquamicrobium spp. (11.6%); Methylophaga spp. accounted for 3.5% (Table 2). In the IO biofilm cultures, the proportion of Methylophaga spp. was 12 times higher (42.8%) than in the OB, whereas it was 5.5 times lower for Hyphomicrobium spp. (8.4%). Higher proportions of Marinicella spp. (7.2%) and Winogradskyella spp. (4.3%) along with a much lower proportion of Aquamicrobium spp. (0.44%) were observed in the IO biofilm cultures compared to the OB (Table 2).

In the biofilm cultures cultivated with the four combinations of NO${}_3^{}-$/methanol concentrations and temperatures in ASW medium containing 2.75% NaCl (Ref300N-23C, 300N-30C, 900N-23C, 900N-30C), Methylophaga spp. accounted for >90% of the 16S rRNA gene sequences followed by Marinicella spp. with proportions ranging from 1.5% to 5.0% (Fig. 1; Table 2). No sequences were found affiliated to Hyphomicrobium spp. under these conditions. In the biofilm cultures cultivated at low NaCl concentrations (0% NaCl, 0.5% NaCl, 1% NaCl), Hyphomicrobium spp. accounted for 11.8%, 6.8% and 0.25%, respectively of the 16S rRNA gene sequences (Fig. 1; Table 2). Methylophaga spp. was still the dominant genus with more than 85% of the 16S rRNA gene sequences; 16S rRNA gene sequences affiliated to Marinicella spp. were also found in significant proportions (Fig. 1; Table 2). Finally, 16S rRNA gene sequences affiliated to Stappia spp. were found in all biofilm cultures and in the OB.

The 16S rRNA gene sequences from the OB and the IO biofilm cultures that were derived by Illumina sequencing generated several thousands of reads affiliated to Hyphomicrobium spp. and Methylophaga spp. (Table 3). This tremendous amounts of sequences allowed assessing the presence of species other than H. nitrativorans and M. nitratireducenticrescens in these two biofilms. The phylogenic analyses performed on these sequences allowed regrouping the OTUs in three clusters for Hyphomicrobium spp., and also three clusters for Methylophaga spp. (Figs. S2A and S2B). The vast majority (>90%) of the 16S rRNA gene sequences associated to these OTUs were affiliated to H. nitrativorans or M. nitratireducenticrescens, respectively, in the OB and the IO biofilm cultures (Clusters 1, Table 3). A small proportion of the OTUs (clusters 2 and 3) were affiliated to other Hyphomicrobium or Methylophaga, which suggests that other members of these genera were present in these biomasses in very low proportions.

Isolation of denitrifying bacterial isolates from the biofilm cultures

The biomass of the Ref300N-23C biofilm cultures was dispersed on different nutrient agar plates to isolate denitrifying bacteria other than H. nitrativorans strain NL23 and M. nitratireducenticrescens strain GP59. Isolates affiliated to the genera Marinobacter, Pseudomonas, Paracoccus, Roseovarius, Thalassobius, Winogradskyella, Aequorivita and Exiguobacterium (Table 4) were recovered from the Marine medium 2216 plates, which contains yeast extract and peptone (Atlas, 1993). Only isolates affiliated to the genera Marinobacter and Paracoccus showed consumption of NO${}_3^{}-$ and NO₂⁻ and production of gas, suggesting that they possess the complete denitrification pathway. The three isolates affiliated to the Paracoccus spp. have identical 16S rRNA sequences with the one of Paracoccus sp. strain NL8 that was isolated from the Biodome denitrification reactor (Labbé et al., 2003). This result suggests that strain NL8 persisted in the biofilm cultures. One representative of Paracoccus isolates (GP3) could grow with methanol as sole source of carbon; Marinobacter sp. GP2 could not.

Metatranscriptomic analysis of the biofilm cultures

The metatranscriptomic approach has allowed assessing the contributions of the microbial community to the metabolic processes in the biofilm cultures. We have chosen to focus on three biofilm cultures, which were the Ref300N-23C (the reference biofilm cultures), 900N-30C (highest denitrification rates; Table 1) and 0% NaCl (persistence of H. nitrativorans strain NL23) biofilm cultures. Because the genomes of H. nitrativorans strain NL23 and M. nitratireducenticrescens strain JAM1 and strain GP59 were available, we first determined changes in the transcript levels of genes associated to these genomes between the biofilm cultures and the planktonic pure cultures of the respective strains. To assess the contribution of other microorganisms in biofilm cultures, the reads from the metatranscriptomes that did not align with the three reference genomes were used to derive de novo assembled transcripts. These transcripts were annotated for function and bacterial affiliation.

Gene expression profiles of M. nitratireducenticrescens strain GP59 in the biofilm cultures

Because >80% of the genomes of strains JAM1 and GP59 are identical, high proportions of reads from the biofilm metatranscriptomes can align to both genomes. Geoffroy et al. (2018) showed that the gene expression profiles of the common genes between both strains in planktonic pure cultures were similar. In Payette et al. (2019), the concentrations of strain GP59 and strain JAM1 in the biofilm cultures were determined by qPCR. In the three selected biofilm cultures for the metatranscriptomic analysis, the concentrations of strain GP59 (copies of nirK by ng biofilm DNA) were one to three orders of magnitude higher than those of strain JAM1 (copies of tagH by ng biofilm DNA). Because of these differences, it was assumed that most of the transcript reads associated to M. nitratireducenticrescens in the biofilm cultures were from strain GP59. Metatranscriptomic analysis in relation with strain JAM1 is described in Documents S8 and S11. The transcriptome of strain GP59 was also derived from planktonic pure cultures cultivated under anoxic conditions in the Methylophaga 1403 medium (Fig. S1). The choice of this medium was because suboptimal growth occurred with strain GP59 in ASW medium. The relative transcript levels of the corresponding genes in the biofilm cultures and the planktonic pure cultures were compared to assess changes in the metabolisms of the strain that occurred between the two environments. All quantitative changes described below of the transcript levels in the biofilm cultures are expressed relative to the transcript levels in the planktonic pure cultures.

Among all genes of strain GP59, between 11% and 21% of them had higher relative transcript levels in the biofilm cultures. At the opposite, 6 to 17% of all genes of strain GP59 were expressed at higher relative transcript levels in planktonic pure cultures (Fig. 2). Strain GP59 contains two plasmids, and most of the genes encoded by these plasmids had much lower relative transcript levels in the biofilm cultures (Fig. 2). Genes involved in the nitrogen metabolism and iron transport were globally at higher relative transcript levels in the biofilm cultures (Table 5; Figs. 2 and 3).

For the denitrification genes, narXL encoding the regulatory factors of the nar systems showed no differences between the biofilm cultures and the planktonic pure cultures in the relative transcript levels (Table 5). Small upregulation of the nar2 operon with about 3-fold increases in relative transcript levels occurred in the biofilm cultures. These levels were lower in the 900N-30C biofilm cultures for the nar1 operon and were about the same levels in the two other biofilm cultures and the planktonic pure cultures. The nor1 operon had the same relative transcript levels in the 300N-23C and 900N-30C biofilm cultures and the planktonic pure cultures, and a 3-fold decrease was noticed in these levels in the 0% NaCl biofilm cultures (Table 5). No significant changes in the expression of the nos operon occurred between both types of cultures. The relative transcript levels of nirK were 5 to 10-times higher in the 300N-23C and 900N-30C biofilm cultures, whereas these levels were similar in the 0% NaCl biofilm cultures and the planktonic pure cultures (Table 5). Higher relative transcript levels of genes involved in the ammonium transport and the assimilatory NO${}_3^{}-$/NO₂⁻ reduction pathway were observed in the biofilm cultures (Table 5; Fig. 3). Absence of nitrogen source other than NO${}_3^{}-$ in the ASW medium and presence of 37 mM NH_4⁺ in the medium used for the planktonic pure cultures (Methylophaga 1403) could explain these differences in the assimilatory pathway.

Figure 3 illustrates changes in the relative expression profiles in the biofilm cultures of major pathways in strain GP59. The relative transcript levels of mxaFJGI encoding the small and large subunits of the methanol dehydrogenase (MDH) and the cytochrome c-L increased by 2–6 folds in biofilm cultures. Two out of the four mxaF-related products (xoxF) showed 2- to 9-fold decreases in their relative transcript levels in the biofilm cultures. As observed in Methylorubrum extorquens, the genome of M. nitratireducenticrescens strain GP59 encodes three formate dehydrogenase with the same gene arrangement (Chistoserdova et al., 2004). The fdhCBAD operon that encodes the NAD-linked, Mo-formate dehydrogenase showed ca. 60-fold increases in the relative transcript levels in the biofilm cultures, whereas the two other fdh operons stayed at the same levels of those of the planktonic pure cultures. Genes encoding NAD-dependent formate dehydrogenase were upregulated in biofilms formed by Desulfovibrio vulgaris compared to planktonic cultures (Clark et al., 2012). This was also the case in biofilms formed by Staphylococcus aureus where the NAD-dependent formate dehydrogenase genes were among the highest upregulated genes (Resch et al., 2005). In both studies, this upregulation correlated with increase in formate dehydrogenase activity. Contrary to planktonic cultures, accumulation of formate could occur in cell vicinity in the biofilm that would be toxic for the cells (Resch et al., 2005). Therefore, upregulation of fdhCBAD operon could be related to detoxification. The relative transcript levels of the gene encoding the cytochrome c555 were 40–70 times higher in the biofilm cultures. Genes encoding the other cytochromes, pseudoazurins and azurin were expressed at similar levels in both types of cultures. Genes involved in the formaldehyde metabolism to formate and CO₂, glycolysis, the ribulose monophosphate pathway, the Entner Doudorof pathway, the tricarboxylic acid cycle, and the pentose pathway showed their relative transcript levels in general unchanged. Few genes in these pathways had 2- to 11-fold differences between the planktonic pure cultures and the biofilm cultures. The genome of strain GP59 encodes the major enzymes involved in the Calvin-Benson-Bassham cycle: the ribulose-bisphosphate carboxylase (Rubisco) and the phosphoribulokinase (Prk). In the 0% NaCl biofilm cultures, the relative transcript levels of the Rubisco gene operon (rbcSL) jumped by 66 times. This upregulation was less pronounced in the Ref300N-23C biofilm cultures (3-fold increase). For the prk gene, the relative transcript levels were 3 to 4 times higher in the 0% NaCl biofilm cultures. The nature of this upregulation is unknown. Except for the cytochrome c555, the relative transcript levels of genes encoding for the oxidative phosphorylation metabolism were unchanged. Several genes involved in iron transport showed higher relative transcript levels in the biofilm cultures (2- to >50-fold increases). The nature of this upregulation in the biofilm cultures is unknown, as the biofilm and planktonic pure cultures were cultivated with trace elements containing iron (Payette et al., 2019).

Gene expression profiles of H. nitrativorans strain NL23 in the 0% NaCl biofilm cultures

As with M. nitratireducenticrescens strain GP59, the transcript levels of genes associated to H. nitrativorans strain NL23 were compared between the biofilm cultures and the planktonic pure cultures to assess changes in the metabolisms that occurred between the two environments. The planktonic pure cultures of strain NL23 were cultivated under anoxic conditions in the 337a medium (Fig. S1). The choice of this medium was because strain NL23 could not grow in ASW (with 2.75% NaCl).

The overall analysis of the three metatranscriptomes confirmed the results obtained by qPCR assays (Payette et al., 2019) and the 16S rRNA gene analysis (Table 2). High number of reads (40 × 10⁶) derived from the metatranscriptome of the 0% NaCl biofilm cultures aligned with the NL23 genome, but <20,000 reads derived from the Ref300N-23C and 900N-30C metatranscriptomes did. In the 0% NaCl biofilm cultures, <10% of all NL23 genes had a higher relative transcript levels in the 0% NaCl biofilm cultures, whereas this was the case for >40% genes in planktonic pure cultures (Fig. 2). These results suggest important changes had occurred in the regulation of gene expression between the planktonic pure cultures and the biofilm cultures. Genes involved in the energy (Fig. S4) and nitrogen metabolisms (Fig. 2; Table 5) had globally higher relative transcript levels in the 0% NaCl biofilm cultures (Fig. 2).

Higher relative transcript levels (5- to 8-times) for the nap, nor and nos operons were observed in the 0% NaCl biofilm cultures (Table 5). nirK was highly upregulated in the biofilm cultures with 49-fold increase in the relative transcript levels (Table 5). The napGH operon however had a 9.4-fold decrease in the relative transcript levels in the 0% NaCl biofilm cultures. As observed with strain GP59, substantial changes in the relative transcript levels of genes involved in the ammonium transport and the assimilatory NO${}_3^{}-$/NO₂⁻ reductase were observed with 3- to 22-fold increases in these levels (Table 5, Fig. 4). These results correlate with the absence of NH_4⁺ in the ASW medium, and thus NO${}_3^{}-$ the only source of N, compared to the 337a medium used for the planktonic pure cultures, which contains 3.8 mM NH_4⁺.

Figure 4 illustrates changes in relative expression profiles of major pathways in the 0% NaCl biofilm cultures of strain NL23. The relative transcript levels of mxaFJGI increased by 2 folds in the biofilm cultures. The relative transcript levels of the mau operon (methylamine dehydrogenase) showed a 15-fold decrease in the biofilm cultures. The nature of such decrease is unknown as strain NL23 was not fed with methylamine in any of our cultures. The three xoxF genes did not show substantial changes in their transcript levels in both types of cultures. Genes involved in the formaldehyde metabolism to formate and CO₂, glycolysis, the tricarboxylic acid cycle, and the pentose pathway showed their transcript levels in general unchanged between the planktonic pure cultures and the biofilm cultures. Few genes in these pathways had 2- to 5-fold differences in their relative transcript levels. The two genes encoding the key enzymes in the serine pathway (alanine-glyoxylate transaminase, glycine hydroxymethyltransferase) had 3- to 7-fold increases in their relative transcript levels in the biofilm cultures. As Methylorubrum extorquens, the NL23 genome encodes the ethymalonyl-CoA pathway (Chistoserdova et al., 2003; Peyraud et al., 2009), which did not show changes overall in the transcript levels of the corresponding genes between the two types of cultures. Contrary to M. extorquens however, a gene encoding the isocitrate lyase is present in strain NL23 and showed a 25-fold upregulation in the biofilm cultures. The isocitrate lyase is one of the key enzymes of the glyoxylate bypass that catalyzes the transformation of isocitrate to succinate and glyoxylate. Gene encoding isocitrate lyase is also present in other available Hyphomicrobium genomes. All these results suggest that in the 0% NaCl biofilm cultures, the carbon metabolism increased in activity and that the glycine regeneration for the serine pathway by the glyoxylate was upregulated. Among genes involved in the oxidative phosphorylation, the relative transcript levels were higher (2–13 times) in the biofilm cultures with those encoding the NADH dehydrogenase, the cytochrome c reductase, with one of the cytochromes c and the F-type ATPase. Combined with increases in the relative transcript levels of the denitrification and the carbon pathways, these results suggest that increases in electron donor activities correlates with the need of electron for the nitrogen dissimilatory metabolism in the biofilm cultures. Strain NL23 possesses four types of cytochrome oxidase (aa3, bo, bd-I and cbb3) (reduction of O₂ in H₂O) that were about expressed at the same levels in both types of cultures. Figure 4 also illustrates the dynamic changes of transporters and two component systems. Several of these transporters had lower relative transcript levels in the 0% NaCl biofilm cultures. Contrary to strain GP59, genes involved in iron transport were not strongly affected in their gene expression in the 0% NaCl biofilm cultures (Figs. 2 and 4).

The composition of the active microbial community in the biofilm cultures

As mentioned above, the reads from the three metatranscriptomes that did not align with the genomes of H. nitrativorans strain NL23 and M. nitratireducenticrescens strain GP59 and strain JAM1 were de novo assembled. These reads were subsequently aligned to the de novo assembled transcripts to derive the relative levels of these transcripts in the biofilm cultures. The de novo assembled sequences were then annotated for function and affiliation. Finally, these sequences were grouped by microbial affiliation to determine the active populations in the biofilm cultures and to assess their level of involvement in these biofilm cultures (Table 6).

It was estimated that between 5 and 10% reads of the three metatranscriptomes were derived from other microorganisms than H. nitrativorans strain NL23 and M. nitratireducenticrescens strain GP59 and strain JAM1 (Document S4). The proportions of transcripts affiliated to Archaea and Eukarya accounted together for <0.1% (Table 6), which suggests very low abundance of these microorganisms in the biofilm cultures. The proportions of transcripts affiliated to viruses, phages and plasmids in the de novo assembled transcripts represented between 0.6 and 21.7% (Table 6).

Twenty-seven bacterial taxa were selected for their overall transcript levels in at least one of the three biofilm cultures (Table 6). All the taxa detected by the 16S rRNA gene sequencing are present in this list (Table 2 and Document S2). These 27 taxa represented between 22% and 35% of the de novo assembled transcripts. Among these taxa, genes encoding the four denitrification reductases were present in the de novo transcripts affiliated to Marinobacter spp., Stappia spp. and Pseudomonas spp. However, only de novo assembled transcripts affiliated to Stappia spp. showed the complete set of denitrification genes in the three biofilm cultures and organized in operons (napABC, napADFE, norCBQD, nosRZDF).

Further analysis of the expression profiles of the 27 bacterial taxa was performed to assess whether some taxa were influenced in their global metabolic activities by the specific conditions of the biofilm cultures. The overall transcript levels of the 27 bacterial taxa (Table 6) were compared between each biofilm culture by clustering analysis (Fig. 5). NaCl concentration was the main factor of clustering as two distinct clusters were derived. The low salt cluster consisting of nine bacterial taxa showed higher relative transcript levels in the 0% NaCl biofilm cultures, whereas the marine cluster of 13 bacterial taxa had higher relative transcript levels in the Ref300N-23C and 900-30C biofilm cultures. A third cluster showed five bacterial taxa with lower relative transcript levels in the 900N-30C biofilm cultures compared to the 0% NaCl and Ref300N-23C biofilm cultures. In these cases, higher temperature (30 °C vs 23 °C) and higher NO₃⁻ and methanol concentration (64.3 mM vs 21.4 mM NO₃⁻; 0.45% vs 0.15% methanol) may have negatively affected these populations.