Autotransporters Drive Biofilm Formation and Autoaggregation in the Diderm Firmicute Veillonella parvula

Veillonella parvula is an anaerobic commensal and opportunistic pathogen whose ability to adhere to surfaces or other bacteria and form biofilms is critical for it to inhabit complex human microbial communities such as the gut and oral microbiota. Although the adhesive capacity of V. parvula has been previously described, very little is known about the underlying molecular mechanisms due to a lack of genetically amenable Veillonella strains. In this study, we took advantage of a naturally transformable V. parvula isolate and newly adapted genetic tools to identify surface-exposed adhesins called autotransporters as the main molecular determinants of adhesion in this bacterium. This work therefore provides new insights on an important aspect of the V. parvula lifestyle, opening new possibilities for mechanistic studies of the contribution of biofilm formation to the biology of this major commensal of the oral-digestive tract.


INTRODUCTION
determinants. We also showed that a locus encoding a metal-dependent phosphohydrolase HD 76 domain protein is involved in biofilm formation, similarly to what was shown in the 77 prototypical monoderm Bacillus subtilis (18). Therefore, our results demonstrate that diderm 78 Firmicutes use a mixture of diderm/monoderm factors to modulate their ability to engage into 79 biofilm lifestyle, supporting the idea that monoderm and diderm molecular systems could have 80 co-evolved in these atypical Firmicutes. 81

RESULTS 83
PacBio sequencing of the genetically amenable V. parvula SKV38 strain 84 We sequenced the V. parvula SKV38 whole genome using PacBio sequencing and 85 obtained 338,310 subreads with a mean length of 9,080 bp ( Figure S1A). Assembly was 86 performed with Canu and no gaps or drops of coverage was detected based on 87 sequana_coverage output ( Figure S1B) (19). One contig of 2,146,482 bp was generated after 88 assembly closely matching the genome size (2.1422 Mbp) and GC content (38.7%, expected 89 38.6%) of the reference V. parvula DSM2008 strain (see supplementary material and methods 90 for details). PROKKA annotation of the V. parvula SKV38 genome detected 1,912 predicted 91 protein-encoding open reading frame (ORF), 12 rRNA, 49 tRNA and one tmRNA. Rapid 92 Annotation using Subsystem Technology (RAST) analysis assigned 53% of the predicted 93 proteins to known subsystems ( Figure 1A). The two-way average nucleotide identity (ANI) 94 between SKV38 and the reference strain DSM2008 is 95.43% ( Figure 1B In order to identify genes involved in biofilm formation, we performed random 101 transposon mutagenesis in V. parvula SKV38 using the pRPF215 plasmid carrying an inducible 102 transposase and a mariner-based transposon previously used to mutagenize Clostridium difficile 103 (20), a close relative of the Negativicutes. We screened 940 individual transposon mutants for 104 biofilm formation using crystal violet staining (CV) static biofilm assay in 96-well microtiter 105 plates and identified eight independent mutants with significant reduction in biofilm formation 106 ( Figure 3A). Whole genome sequencing localized the transposons in two loci putatively 107 implicated in biofilm formation ( Figure 3B). The most affected mutants correspond to 108 insertions in FNLLGLLA_00516 (seven mutants), encoding a T5SS type Vc trimeric 109 autotransporter. One transposon mutant was inserted in FNLLGLLA_01127, encoding a 110 putative HD phosphatase ( Figure 3B). 111 112 FNLLGLLA_00516 encodes a trimeric autotransporter involved in auto-aggregation 113 FNLLGLLA_00516 encodes a protein containing several domains usually identified in 114 the T5SS type Vc trimeric autotransporters. Trimeric autotransporters are OM proteins specific 115 of diderm bacteria that have been widely studied for their ability to bind to different surfaces 116 or other bacteria (21). FNLLGLLA_00516 is a homolog of V. parvula DSM2008 vpar_0464, 117 which was shown to encode a protein detected in the OM (15). FNLLGLLA_00516 was 118 annotated by PROKKA as BtaF, a trimeric autotransporter identified in Brucella suis involved 119 in adhesion to extracellular matrix and abiotic surfaces (22). Here, we renamed it Veillonella 120 trimeric autotransporter A (VtaA), as the first trimeric autotransporter involved in biofilm 121 formation identified in V. parvula SKV38. We deleted the vtaA coding sequence and showed 122 that ∆vtaA had no growth defect ( Figure S2A) but displayed a marked reduction of biofilm 123 formation in microtiter plate ( Figure 4A). Moreover, while V. parvula SKV38 cultures strongly 124 aggregated, ∆vtaA did not ( Figure 4B and S3). We constructed the strain pTet-vtaA, where the 125 chromosomal vtaA gene is placed under the control of a functional 126 tetracycline/anhydrotetracycline (aTc) inducible promoter ( Figure S4) and showed that its 127 aggregation capacity and biofilm formation directly correlated with aTc concentration ( Figure  128 4C-D), demonstrating that VtaA-mediated cell-to-cell interactions are critical for biofilm 129 formation. Whereas the microtiter plate assay corresponds to a static biofilm assay, we also 130 used continuous flow biofilm microfermentors to investigate the contribution of VtaA to 131 biofilm formation in dynamic conditions. Surprisingly, ∆vtaA formed almost six times more 132 biofilm than the WT strain in these conditions ( Figure 4E). This suggests that auto-aggregation 133 differentially contributes to biofilm formation in dynamic or static conditions. 134 135 V. parvula SKV38 encodes sixteen putative autotransporters in addition to VtaA 136 The strong biofilm phenotype displayed by the ∆vtaA mutant in microfermentor led us 137 to suspect that additional adhesins could modulate V. parvula biofilm formation capacity. 138 Indeed, searching the V. parvula SKV38 genome revealed multiple genes encoding 139 autotransporters (Table 1) Vpar_0048, respectively, and that appear to be split in SKV38 (Table 1). Among those, six 148 autotransporters plus FNLLGLLA_00035, FNLLGLLA_00036-37 and FNLLGLLA_00040-41 149 form a potential genomic cluster coding for adhesins ( Figure 1B and 5A), whereas the six others 150 are located in different areas of the genome ( Figure 1B and Figure 5B). 151

152
The cluster of trimeric autotransporters is involved in surface binding and not 153 aggregation. 154 To assess the function of the potential adhesins identified in the V. parvula SKV38 155 genome, we constructed -within the cluster of adhesin genes-independent deletion mutants for 156 the two first autotransporters (vmaA and vtaB) and a large deletion for the eight adjacent genes 157 encoding trimeric autotransporters or partial trimeric autotransporters, hereafter called ∆8 158 (∆FNLLGLLA_00036 to vtaF). We also generated independent individual mutants for each of 159 the six additional autotransporters located outside of the cluster. These mutants were all tested 160 for biofilm formation and aggregation capacities. While the previously mentioned ∆vtaA strain 161 was the only mutant involved in cell-to-cell interactions ( Figure 6A), both ∆vtaA and ∆8 led to 162 lower biofilm formation in microtiter plate ( Figure 6B and C), suggesting that the adhesins of 163 this cluster could be involved in biofilm formation independently of cell-to-cell interactions. 164 However, we observed no significant difference with the WT when evaluating biofilm capacity 165 of the ∆8 mutant in microfermentor ( Figure 6D). To determine whether this was due to flow or 166 the nature of adhesion surfaces (plastic in microtiter plate vs. glass in microfermentor), we used 167 plastic microscopy coverslip slides to grow biofilms in microfermentor. Scanning electronic 168 microscopy (SEM) showed that the WT formed a biofilm displaying filaments and protein 169 deposits that could be part of V. parvula extracellular matrix, whereas ∆vtaA formed a much 170 thicker biofilm, although without filaments and proteins ( Figure 6E). ∆8, however, only poorly 171 covered the plastic coverslip with sporadic aggregates of cells producing extracellular matrix, 172 consistent with the reduced biofilm formation observed in microtiter plate ( Figure 6E). Initial 173 adhesion assay to glass spatula showed that both vtaA and ∆8 displayed a lower percentage of 174 initial adhesion than WT, suggesting that VtaA-mediated auto-aggregation contributed to initial 175 adhesion while the adhesin cluster is directly involved in surface binding ( Figure 6F). These 176 two contributions were additive since a ∆vtaA∆8 double mutant showed a reduced initial 177 adhesion on microfermentor spatula compared to either WT, ∆vtaA or ∆8 ( Figure 6F).
In 178 addition, ∆vtaA∆8 formed 17 times less biofilm than ∆vtaA in microfermentor, indicating that 179 in the absence of VtaA, the adhesins encoded by some of these eight genes strongly promote 180 mature biofilm formation in microfermentor ( Figure 6D). 181 Taken together, these results demonstrate the differential contribution of VtaA and part 182 of the cluster of adhesin to V. parvula SKV38 adhesion and highlight the existence of potential 183 interference mechanisms between them. Large clustering of potential adhesin encoding genes observed in SKV38 strain and defined as 198 a group of two or more adhesin-coding genes immediately upstream of a conserved rRNA 199 locus, is to our knowledge a peculiar genomic character. We found no evidence of the existence 200 of this specific adhesin locus outside the Veillonella genus. We selected eight Veillonella strains 201 to study more precisely the evolution of the adhesin cluster, including SKV38 and DSM2008.

FNLLGLLA_01127 encodes an HD phosphatase that inhibits biofilm formation 211
In addition to genes encoding potential T5SS proteins, we also identified a transposon 212 mutant in FNLLGLLA_01127, encoding a protein of the HD phosphatase superfamily (Figure  213 2). The FNLLGLLA_01127 gene is homologous to YqeK, a putative phosphatase required for 214 pellicle formation and the development of biofilm in B. subtilis (18). A FNLLGLLA_01127 215 deletion mutant (∆1127) showed a moderate growth defect ( Figure S3AB) and a moderate 1.5-216 fold decrease in biofilm formation in microtiter plate after correcting for the growth defect 217 ( Figure 8A). This mutant also displayed a slightly faster aggregation rate than the WT during 218 early time points ( Figure 8B). The strongest phenotype of this mutant was detected in 219 microfermentor with a 9-fold increase in biofilm formation compared to WT ( Figure 8C), which 220 was reduced by expressing FNLLGLLA_01127 gene in trans (plasmid p1127) ( Figure 8D). 221 Scanning electronic microscopy showed that ∆1127, similarly to ∆vtaA, formed a thick layered 222 biofilm, although with fewer filaments and protein deposits compared to WT ( Figure 8E). 223 However, contrary to ∆vtaA or ∆8 mutants, ∆1127 showed no defect in initial adhesion to a 224 glass spatula ( Figure 8F). Interestingly, a ∆1127∆8 double mutant formed almost 20 times less 225 biofilm than ∆1127 in microfermentor ( Figure Table 2. V. parvula was grown in either Brain Heart 378 infusion medium (Bacto Brain Heart infusion, Difco) supplemented with 0.1 % L-cysteine and 379 0.6 % sodium DL-lactate (BHILC) or SK medium (10 g L −1 tryptone (Difco), 10 g L −1 yeast 380 extract (Difco), 0.4 g L −1 disodium phosphate, 2 g L −1 sodium chloride, and 10 mL L −1 60 % 381 w/v sodium DL-lactate, described in (17) Table S1. BHILC was constantly supplied through a peristaltic pump at 4 rpm. During the last hour, the 466 speed was increased to 10 rpm to remove planktonic bacteria. A mix of filtered 90% nitrogen/5% 467 hydrogen/5% carbon dioxide was also constantly supplied to maintain anaerobic condition. 468 After 48 hours of growth, the spatula was removed, and the biofilm was resuspended by 469 vortexing in 15 mL BHILC. We measured OD600 of the resuspended biofilms with Smart Spec 470 Plus spectrophotometer (BioRad). 471 472

Aggregation curve 473
Overnight cultures were diluted to 0.8 OD600 in Brain-heart infusion (BHI) media in semi-micro 474 spectrophotometry cuvette (Fisherbrand) and left to sediment on the bench in presence of 475 oxygen, so no growth should occur. OD600 was measured every hour in a single point of the 476 cuvette using SmartSpec spectrophotometer (BioRad). 477 478

Initial adhesion on glass 479
Glass spatula from microfermentor (described above) were dipped in overnight cultures diluted 480 to 0.5 OD600 in 15 mL Brain-Heart Infusion (BHI) media for 30 minutes to let bacteria adhere. 481 The spatulas were washed once in 15 mL BHI by submersion and the adhering bacteria were 482 resuspended in 15 mL clean BHI by vortexing. The culture used for inoculation, as well as the 483 resuspended bacteria were serially diluted and plated on SK-agar plate for colony forming unit 484 (CFU) counting.