Alternative modes of biofilm formation by plant-associated Bacillus cereus

The ability to form multicellular communities known as biofilms is a widespread adaptive behavior of bacteria. Members of the Bacillus group of bacteria have been found to form biofilms on plant roots, where they protect against pathogens and promote growth. In the case of the model bacterium Bacillus subtilis the genetic pathway controlling biofilm formation and the production of an extracellular matrix is relatively well understood. However, it is unclear whether other members of this genus utilize similar mechanisms. We determined that a plant-associated strain of Bacillus cereus (905) can form biofilms by two seemingly independent pathways. In one mode involving the formation of floating biofilms (pellicles) B. cereus 905 appears to rely on orthologs of many of the genes known to be important for B. subtilis biofilm formation. We report that B. cereus 905 also forms submerged, surface-associated biofilms and in a manner that resembles biofilm formation by the pathogen Staphylococcus aureus. This alternative mode, which does not rely on B. subtilis-like genes for pellicle formation, takes place under conditions of glucose fermentation and depends on a drop in the pH of the medium.


Introduction
Many bacteria are capable of forming multicellular communities, biofilms, which provide protection against environmental insults and facilitate symbiotic and pathogenic interactions with other organisms. The ability to form a biofilm is widespread, but the mechanisms by which community formation is achieved differs greatly between different species. In general terms, however, biofilm formation involves the coalescence of cells, often on a surface, via the production of a protective, self-produced matrix that coats the cells, sticking them together and providing a barrier against external stresses (Lemon et al. 2008). Three major components of biofilm matrix have been identified; exopolysaccharide (EPS), DNA and protein, and are used in different combinations in different bacterial species (Flemming and Wingender 2010). For example, the model Gram-positive bacterium Bacillus subtilis produces a biofilm that relies on an EPS component, EPS, and on a dedicated matrix protein, TasA, which is anchored on the cell wall and mediates cell-cell interactions (Lemon et al. 2008).
Bacillus cereus is a ubiquitous group of Gram-positive bacteria which, like B. subtilis, exhibit the ability to resist unfavorable conditions through biofilm formation and sporulation (Stenfors Arnesen et al. 2008). B. cereus is well-known as an opportunistic pathogen, particularly in connection with its frequent association with the dairy industry (Stenfors Arnesen et al. 2008). The B. cereus type strain ATCC 14579 has previously been investigated for its ability to form biofilms on surfaces, such as plastic, glass, and stainless steel (Oosthuizen et al. 2002;Shi et al. 2004;Hsueh et al. 2006;Vilain and Br€ ozel 2006;Karunakaran and Biggs 2011;Lindb€ ack et al. 2012). However, details of the genetic basis of biofilm formation remain unclear. In fact, the only matrix component that has been described to date in B. cereus is extracellular DNA (Vilain et al. 2009) although there are indications that a protein component may also be present (Vilain and Br€ ozel 2006;Vilain et al. 2009;Karunakaran and Biggs 2011).
Plant-associated strains of B. cereus have also been isolated. These are of particular interest since they could have bio-control applications for the prevention of plant disease (Xu et al. 2014). At least one potential bio-control strain, 0-9, has already been shown to colonize roots of wheat plants and to exhibit biofilm formation, which contributed to its bio-control efficacy (Xu et al. 2014). In this strain, a phosphotransferase PtsI was reported to be required for biofilm formation, root colonization, and effective bio-control. However, the components of the biofilm matrix produced by this strain were not investigated in detail.
Amongst Gram-positive bacteria matrix production is best understood in the model organism B. subtilis (L opez et al. 2010). Biofilm formation in B. subtilis begins through the receipt of specific signals by a suite of histidine kinases, KinA-E (L opez et al. 2009;McLoon et al. 2011;Chen et al. 2012). This results in the activation of a phosphorelay that culminates in the phosphorylation and activation of the master transcriptional regulator Spo0A. Phospho-Spo0A triggers a cascade of events, including the upregulation of an anti-repressor, SinI, which in turn antagonizes the DNA-binding activity of the dedicated biofilm repressor SinR. Derepression of SinR-controlled genes leads to the production of the two matrix components: EPS and the protein TasA. TasA is exported from the cell and processed to a mature form by the enzyme SipW before being anchored to the membrane through attachment to TapA (Branda et al. 2004;Romero et al. 2011). TasA monomers polymerize on the outside of the cell to form amyloid-like fibers which mediate cell-cell attachment (Romero et al. 2010).
We set out to investigate whether a plant-associated strain of B. cereus, a wheat-rhizosphere-associated strain called 905 (Wang et al. 2007), is capable of forming biofilms and, if so, how these biofilms are constructed. We wondered whether B. cereus 905 would form biofilms in a similar way to B. subtilis or if the biofilm matrix and regulatory machinery are distinct between the two species. Here, we present evidence that B. cereus 905 appears to exhibit two modes of biofilm formation. One mode is pellicle formation and appears to rely on orthologs of many of the genes known to be required for biofilm formation in B. subtilis, including two orthologs of genes for the dedicated matrix protein TasA. However, under other conditions B. cereus is able to form submerged, surfaceassociated biofilms that are not dependent on the genes required for pellicle formation. Instead, formation of these submerged biofilms is induced by low pH under conditions that result from the provision of excess glucose in the growth medium. After this work was completed Caro-Astorga et al. (2015) similarly reported a dependence of floating biofilm formation by a B. cereus strain on orthologs of TasA.

Bioinformatic analysis
To identify homologs of known B. subtilis biofilm genes in B. cereus 905 we used the cognate gene from the B. subtilis 168 genome (as indicated in Table S3) to conduct nucleotide BLAST searches of the B. cereus 905 genome sequence (complete genome sequence unpublished; for nucleotide sequences described here see Genbank KP076259-KP076281). The percentage amino acid sequence identity of the corresponding proteins was then calculated. We used the same B. subtilis 168 protein sequences to identify the closest homolog amongst all known predicted proteins in the B. cereus group in an autonomous nonredundant protein sequence database (NCBI taxid:86661). The percentage amino acid identity between this consensus sequence and the homolog from B. cereus 905 was then calculated.
To identify the putative EPS biosynthetic gene cluster in B. cereus 905 we began by using amino acid sequences encoded in the epsA-O gene cluster of B. subtilis 168 to search for orthologs encoded by B. cereus 905. This led us to identify a cluster of 21 genes in the B. cereus genome which we then compared by protein BLAST to a predicted EPS gene cluster described in the sequenced B. cereus strain ATCC 14579 (Ivanova et al. 2003).

B. cereus strains and plasmids
B. cereus strains and the plasmids used in the manipulation of these strains are described in Table S1. The B. cereus isolate 905 used in this study, which has been described previously (Wang et al. 2007), was isolated from wheat rhizosphere. This strain was assigned to the B. cereus group on the basis of its 16S rRNA gene sequence, by comparison to sequences in the NCBI database (data not shown).
B. cereus pellicle formation assay B. cereus cells were grown in LB broth at 37°C to stationary phase. For pellicle formation, 2 lL of the cells were mixed with 2 mL of MSgg broth in 24-well plates (353047; Corning, Tewksbury MA, USA). Plates were incubated at 30°C for 3 days. All images were taken using a Sony (New York, NY, USA) NEX-5 digital camera.

Construction of marked deletion mutants in B. cereus 905
The spo0A mutant strain TG001 was constructed using the following procedure. A kanamycin resistance gene (Km) was amplified by polymerase chain reaction (PCR) with Platinum Pfx DNA Polymerase (Life Technologies, Grand Island, NY, USA) from plasmid pMarA (Breton et al. 2006) using primers Km-F and Km-R (All primer sequences are shown in Table S2). The resulting fragment was digested with BamHI/PstI and cloned into pEBS (Wang et al. 2007), which had also been digested with BamHI/PstI, generating pEBSK. Regions of 967-bp upstream and 1094-bp downstream of the spo0A gene were amplified from B. cereus 905 genomic DNA using the primers spo0A-Up-F/spo0A-Up-R and spo0A-Dn-F/ spo0A-Dn-R, respectively. These fragments were digested with XbaI/BamHI and PstI/XhoI, respectively, and were cloned into the pEBSK plasmid to create pEBSKΔspo0A. A 3072-bp Up-Km-Dn fragment was amplified by PCR from pEBSKΔspo0A using primers spo0A-F and spo0A-R. The fragment was digested with BglII and MluI, and cloned into a temperature-sensitive E. coli and Bacillus shuttle vector pMAD (Arnaud et al. 2004) that had also been digested with BglII and MluI, generating pMADΔ-spo0A. The construct was confirmed by PCR and Sanger sequencing. The pMADΔspo0A plasmid was mobilized into B. cereus 905 by electroporation. Transformants were obtained after incubation for 2 days at 30°C on LB plates containing 1 lg mL À1 erythromycin and 40 lg mL À1 of X-Gal. Allelic replacement of the temperature-sensitive pMADΔspo0A in B. cereus 905 was achieved through the following procedure. Single crossover integrants were induced by growth overnight at 42°C in the presence of erythromycin. Double crossover events were promoted by two 12 h rounds of growth in liquid culture at 25°C. Erythromycin sensitive clones were isolated and mutants identified by colony PCR. All mutants were subsequently confirmed by PCR amplification and Sanger sequencing to verify the presence of the spo0A::Km allele on the chromosome (data not shown).
Replacement of other genes in the B. cereus 905 chromosome with the kanamycin resistance cassette was carried out as described for Δspo0A, with the exception that the primers shown in Table S2 were used to amplify the upstream and downstream homologous regions for each gene.

Construction of an eps unmarked deletion mutant in 905
An in-frame, unmarked deletion of the first six genes of the eps operon in the B. cereus 905 genome was constructed essentially as described in (Arnaud et al. 2004). Briefly, a construct was generated in which~900 bp regions of the B. cereus 905 genome, flanking the deletion site, were amplified by PCR and were introduced by Gibson Assembly (Gibson et al. 2009) into pMAD, linearized with NcoI and MluI. The resulting plasmid was mobilized into E. coli DH5a and transformants were selected for growth on ampicillin. The resulting construct was confirmed by PCR and Sanger sequencing. The construct was subsequently mobilized into B. cereus 905 by electroporation and transformants were selected on 1 lg mL À1 erythromycin and 40 lg mL À1 X-Gal at 30°C. Allelic replacement and mutant confirmation were carried out as described above.

Construction of complementation strains
The plasmid pGFP78 carries a constitutive promoter (F78) that was selected from a B. subtilis ISW 1214 geno- mic DNA library, for the property of driving high-level expression of green fluorescent protein (GFP) (Q. Wang, pers. comm.). The GFP open reading frame (ORF) was removed from this vector by digestion with XbaI and HindIII. The respective wild-type gene from B. cereus 905 was amplified by PCR using gene-specific primers (shown in Table S2) and was ligated to the XbaI and HindIII sites of the cut plasmid, to generate each respective complementation plasmid (as shown in Table S1). Construction of the ΔsinI complementation construct was altered from the above in that the complete 135-bp sinI gene of B. cereus 905 was amplified by PCR with primers sinI-F-C and sinI-R-C, and ligated to pHY78 digested with HindIII, to generate p78sinI. All plasmids were confirmed by PCR and Sanger sequencing. Plasmids were subsequently mobilized into the corresponding B. cereus mutants by electroporation and transformants selected with tetracycline.

B. cereus submerged biofilm assay
Submerged biofilm formation was quantified essentially as described previously (Foulston et al. 2014). B. cereus 905 was grown from a single colony in 3 mL LB overnight at 25°C with aeration. For biofilm growth, cells were diluted 1 in 1000 into fresh medium (typically filter sterilized TSB supplemented with 1% glucose (TSBG) or as specified in the text) and 200 lL was aliquoted into a Nunc TM MicroWell TM 96-Well Microplate (167008; Thermo Scientific, Tewksbury, MA, USA). Typically the starting OD 600 was recorded using a BioTek Synergy II plate reader (Bio-Tek Instruments, Winooski, VT, USA). Plates were incubated statically at 37°C for 24 h. To quantify biofilm formation, the medium in each well was carefully removed with a multichannel pipette and transferred to wells of a new 96-well microtiter plate. The biofilms were then washed once with 200 lL phosphate-buffered saline pH 7.5 (PBS) using a multichannel pipette and the wash transferred into wells of a new 96-well microtiter plate. The biofilms were then resuspended in 200 lL PBS using a multichannel pipette. The OD 600 of each fraction was then recorded using a BioTek Synergy II plate reader. To remove background absorbance the starting OD 600 was subtracted from the values for each medium sample, and the absorbance of a well containing PBS alone was subtracted from the wash and resuspended biofilm values. Replicate wells (n ≥ 4) were averaged and a standard deviation calculated. Each strain was tested independently at least three times.
For crystal violet assays biofilm cultures were grown in 12-well plates (353043; Falcon). After 24 h of growth at 37°C the medium in each well was gently removed with a pipette and each well was washed twice with 1 mL of PBS to remove unattached bacteria. The plates were stained with a 0.1% crystal violet solution (Sigma) for 30 min and then washed once with 1 mL of PBS. Plates were air dried before photographing.
To test biofilm formation of B. subtilis 3610 the strain was grown overnight in LB at 25°C before diluting 1 in 1000 into fresh medium (as indicated in figures and text) and growing in a Nunc TM MicroWell TM 96-Well Microplate (167008; Thermo Scientific) for 24 h at 37°C. Pellicles formed were photographed before being manually removed with a plastic pipette tip to visually determine whether any biomass was surface associated.

Results
Candidate biofilm genes in B. cereus 905 Genome sequences from members of the B. cereus group of bacteria encode homologs of many of the genes known to be important for biofilm formation in B. subtilis, such as the master regulator spo0A, the dedicated biofilm regulatory genes sinI and sinR, and two tasA-like genes (Pflughoeft et al. 2011;Fagerlund et al. 2014). In B. subtilis a single copy of tasA is located in an operon with sipW and tapA (yqxM) (Branda et al. 2004). By contrast B. cereus strains appear to possess two genes with predicted products with similarity to TasA (Pflughoeft et al. 2011). However, one of these two gene copies has frequently been annotated in B. cereus genome sequences as calY due to the similarity of the predicted amino acid sequences to that of camelysin, a proteolytic enzyme purified from B. cereus strain DSM 14729 (Grass et al. 2004;Pflughoeft et al. 2011). It has not yet been determined whether these proteins can also function like TasA, playing a structural role in the biofilm matrix. However, the amino acid similarity shared by the predicted proteins of the calY genes with TasA and the conserved synteny of the B. cereus tasA and calY genes with both sipW and sinI/sinR, suggests that both of these proteins are functional equivalents of B. subtilis TasA. Furthermore, in both B. anthracis and B. thuringiensis these genes are under the control of the biofilm repressor SinR as is tasA in B. subtilis (Pflughoeft et al. 2011;Fagerlund et al. 2014). Interestingly, despite the presence of two homologs of tasA we were unable to identify a homolog of tapA in the B. cereus group. This might suggest either that a TapA function is not required in B. cereus or, conceivably, that one of the two TasA homologs can functionally substitute for TapA.
Like other members of the B. cereus group, homologs of genes involved in B. subtilis biofilm formation were identified in the B. cereus 905 genome sequence (Genbank KP076259-KP076281) (Table S3). Although the identity of the predicted products of these genes between B. subtilis and B. cereus varied from 35% to 82%, the gene synteny was similar. For example, as in B. subtilis the sinI-sinR gene pair was located near the two TasA-like genes, named calY1 and calY2, and sipW (Fig. 1A). One of the tasA-like genes (calY2) occurs downstream of sipW while the other is nearby (Fig. 1A). The overall amino acid identity between TasA from B. subtilis and CalY1 and CalY2 from B. cereus 905 was about 35% and the similarity was 51%. Several amino acid motifs were conserved between the three proteins, particularly within the N-terminal half of the proteins (Fig. 1B).
Another important feature of B. subtilis biofilm formation is EPS production from the biosynthetic gene cluster epsA-O (Branda et al. 2001). We identified in the B. cer-eus 905 genome sequence, a 21-gene cluster with predicted products with many of the hallmarks of EPS biosynthetic machinery ( Fig. 1C and Table S4), such as glycosyltransferases and a tyrosine kinase (EpsAB) known to regulate EPS biosynthesis in B. subtilis ). However, the overall similarity of this region to the corresponding eps gene cluster (15 genes) in B. subtilis was relatively low and only a few genes appeared to be conserved (for example; epsC, epsN, and epsL). We also compared the B. cereus 905 gene cluster to a previously predicted capsular EPS gene cluster from the strain ATCC 14579 (BC5279-BC5263) (Ivanova et al. 2003) (Table S4). Just as with B. subtilis, however, many of the encoded proteins from the B. cereus 905 cluster were not conserved in B. cereus ATCC 14579. Only those genes likely to be involved in regulation of EPS biosynthesis such as tyrosine kinases and a LytR-type regulator shared high amino acid identity between the two strains and many genes did not appear to have an equivalent protein encoded by the other strain (Table S4). This suggests that EPS biosynthesis is not well-conserved among B. cereus strains and might reflect the use of different forms of EPS by different species and strains.

B. cereus pellicle biofilm formation
MSgg medium is commonly utilized for the formation of biofilms in B. subtilis. In MSgg medium, B. subtilis cells form floating biofilms (pellicles) at the air-liquid interface (Branda et al. 2004). B. cereus 905 similarly formed a pellicle-like structure when grown in MSgg (Fig. 2). B. cereus pellicles were not as thick or wrinkled as those made by B. subtilis but instead were thin films coating the surface of the liquid.
Are any of the genes that contribute to biofilm formation by B. subtilis important for pellicle formation by B. cereus 905? To investigate this, mutant strains of B. cereus 905 were constructed in which candidate genes were deleted from the chromosome. The deletion mutant strains were then grown alongside wild-type B. cereus 905 in MSgg medium at 30°C and visually observed for the formation of pellicles after 3 days incubation.
In B. subtilis, spo0A encodes a master transcriptional regulator of development that is required to initiate both biofilm formation and sporulation. A B. cereus 905 spo0A deletion mutant was similarly unable to make a pellicle biofilm (Fig. 2). To confirm the role of this gene as a pleiotropic regulator of development in B. cereus we also examined the ability of this strain to generate spores. By microscopy we were unable to visualize any phase-bright spores being formed by the mutant, unlike the wild-type strain in which almost all cells were able to sporulate (Fig. S1). To confirm that the phenotype of the spo0A mutant was not the result of a second-site mutation we introduced a copy of the wild-type spo0A gene into the mutant strain on a plasmid (p78spo0A). This strain was restored in the ability to form a pellicle (Fig. S2).
In B. subtilis, SinI is required to induce the derepression of biofilm matrix gene expression and is thus essential for biofilm formation. A B. cereus sinI deletion mutant was similarly unable to make a pellicle biofilm (Fig. 2). We introduced a copy of the wild-type sinI gene into the mutant strain on a plasmid (p78spo0A). This strain was restored in the ability to form a pellicle (Fig. S2).
Two homologs of tasA, the product of which is the B. subtilis matrix protein, were identified in B. cereus 905. Despite the low level of amino acid sequence identity between the predicted proteins of these genes and TasA and their previous annotation as protease-encoding genes, we wondered whether CalY1 and CalY2 were required for pellicle formation. In B. subtilis TasA is processed by the signal peptidase SipW, which is thus also essential for biofilm formation. Deletion of sipW in B. cereus inhibited pellicle formation (Fig. 2). However, deletion of calY1 or calY2 individually caused only a partial defect in pellicle formation, generating weaker pellicles that did not float on the liquid surface but could be seen at the bottom of the well (Fig. 2). These phenotypes could be complemented in trans by providing a wild-type copy of the respective gene (Fig. S2). To determine whether calY1 and calY2 have partially redundant functions in B. cereus pellicle formation we were interested to determine the phenotype of a double mutant. However, due to the challenges associated with genetically manipulating B. cereus, we instead chose to delete the entire region spanning the two genes. It should be noted that this includes another hypothetical open reading frame that lies between calY1 and calY2 (Fig. 1A). The resulting calY1-Y2 deletion mutant was completely unable to form a pellicle (Fig. 2). Despite the presence of an ORF of unknown function between the two genes, the most likely interpretation of this result, given the partial phenotypes of the single mutants, is that calY1 and calY2 display redundant functions in pellicle formation.
Finally, to assess the contribution of EPS to pellicle formation in B. cereus 905 we deleted the first six genes of the putative EPS gene cluster (Table S4 and Fig. 1C). This mutant was, however, unaffected in its ability to form pellicles (Fig. 2). This might suggest either that these six genes are unnecessary for EPS production or that the EPS product of this gene cluster is not a major component of B. cereus 905 biofilms.

Conditions promoting surface-associated B. cereus biofilm formation
We next wondered whether B. cereus can form biofilms under other growth conditions. We observed that when grown in complex media B. cereus formed submerged biofilms that were associated with the bottom surface of the wells of the microtiter plate in which they were grown, rather than floating as a pellicle. B. cereus formed measurable surface-associated biofilm in complex media such as TSB and LB but only when excess glucose was provided in the medium (Fig. 3). Approximately 20-30% of total biomass (see Experimental Procedures for details of biofilm quantification) was found to associate with the surface of tissue culture plates when cells were grown in TSB with excess glucose. However, in the absence of glucose (TSB) or in the presence of glycerol as a carbon source, biofilm biomass was reduced fivefold to~6%. B. cereus grew poorly or did not make a substantial surface-associated biofilm in the defined or minimal media tested (Fig. 3).

Submerged biofilm formation requires low pH
The requirement of B. cereus for excess glucose in order to form submerged biofilms in complex medium was reminiscent of biofilm formation by another Gram-positive bacterium, Staphylococcus aureus. S. aureus ferments excess glucose to generate mixed mild acids during biofilm growth, which subsequently decrease medium pH to around 4.5-5. This pH decrease is crucial for S. aureus biofilm formation and preventing the pH decrease by supplying a buffer prevents biofilm formation (O'Neill et al. 2008;Foulston et al. 2014). We wondered whether a similar process occurs when B. cereus is grown in the presence of excess glucose. To investigate this, B. cereus was grown under biofilm conditions and the pH of the growth medium was measured over 24 h of biofilm formation. When B. cereus was grown in TSB supplemented with 1% glucose the pH of the growth medium decreased to pH 4.5 over the 24 h period (Fig. 4A). By contrast, in the absence of glucose, although there was an initial small dip in medium pH, the final pH was 6.5. Since biofilm formation is associated with the provision of glucose, we conclude that a decrease in medium pH contributes to the ability of B. cereus to form a biofilm in TSB.
To further test our hypothesis we next used acetic acid to manually decrease the pH of TSB in the absence of glucose. When the starting pH of the growth medium was at or lower than pH 6.5 B. cereus displayed robust biofilm formation (Fig. 4B) and when measured after 24 h the final pH of the medium under these conditions was pH 5. By contrast, little to no biofilm was seen when the starting pH was pH 7 or greater even though overall cell biomass was unaffected (Fig. 4B). The pH of these cultures, when measured after 24 h, was greater than pH 6.5. Our results were recapitulated when observing biofilm formation by crystal violet staining (Fig. S3). This indicates that low pH contributes to biofilm formation and that the source of this pH drop is likely excess glucose in the medium.
We further established that a decrease in pH is required for biofilm formation by B. cereus in TSB by preventing the pH decrease with a buffer, 200 mmol/L disodium phosphate. This buffer was effective in preventing the decrease in pH observed when B. cereus was grown in TSB supplemented with glucose (Fig. 4A). Sodium phosphate buffer prevented biofilm formation by B. cereus while not affecting overall biomass (Fig. 4C). Our results were recapitulated when observing biofilm formation by crystal violet staining (Fig. S3). In toto, these results indicate that, like S. aureus, B. cereus experiences a significant decrease in medium pH during biofilm formation and that this pH decrease is crucial for the formation of a submerged biofilm.

Determining the composition of the biofilm matrix
We next investigated what might be responsible for holding cells together in the submerged B. cereus biofilm; in other words, what was the composition of the matrix? We first investigated whether protein and/or DNA are important matrix components, as they are in S. aureus submerged biofilms (Foulston et al. 2014), by preventing their accumulation outside cells with protease and nuclease treatment. When either proteinase K or DNase I were added to the cultures, from the start of the incubation, biofilm formation was inhibited, without a significant effect on growth (Fig. 5A). Proteinase K was more effective at preventing biofilm formation than DNase I, reducing the percentage of the population present in the biofilm fraction by 80%, whereas DNase I caused only a 60% reduction. This suggests that both protein and DNA play an important role in holding cells together in a biofilm and thus contribute to the matrix. Additionally, we tested the importance of EPS for submerged biofilm formation by quantifying biofilm formation by a mutant deleted for the first six genes of the eps operon (Δeps), as described above. This mutant displayed wild-type levels of submerged biofilm formation (Fig. 5B) suggesting that these genes are not essential for the formation of a submerged biofilm by B. cereus. Since homologs of genes important for B. subtilis biofilm formation were found to be essential for pellicle biofilm formation we wondered whether these genes can also contribute to the ability of B. cereus to form surfaceassociated submerged biofilms. A spo0A deletion mutant formed biofilms poorly by comparison to the wild-type strain (Fig. 5B), displaying only about 30% of the biomass of the wild-type biofilm. Biofilm formation was restored to wild-type levels by provision of the wild-type gene in trans (Fig. S4).
By contrast deletion of sipW, sinI, and calY2 caused little to no biofilm defect in B. cereus 905 (Fig. 5B). Deletion of calY1, caused a small but reproducible decrease in biofilm formation, reducing the percentage of biofilm biomass to 70% of wild-type biofilm (Fig. 5B). This strain was restored to wild-type levels of biofilm formation by provision of the wild-type gene in trans (Fig. S4). In this case, unlike for pellicle formation, CalY1 and CalY2 do not appear to act redundantly since a deletion of both genes together (ΔcalY1-Y2) recapitulated the phenotype of the single calY1 mutant (Fig. 5B).

B. subtilis 3610 does not form submerged biofilms
Since B. cereus 905 forms entirely submerged and surfaceassociated biofilm when grown in complex media in the presence of glucose we wondered whether B. subtilis 3610 would also exhibit this behavior. Under the same media conditions (TSB with 1% glucose) B. subtilis 3610 exhibited visible pellicle formation and did not appear to form any submerged surface-associated biofilm (Fig. S5). The pellicles formed under these conditions were less robust and less wrinkled than in MSgg but were hydrophobic and could easily be manually removed from the liquid surface. The ability to form a pellicle in TSB appeared to be largely independent of glucose provision and was unaffected by treatment with proteinase K or DNaseI (Fig. S5). Furthermore, the pH of the growth medium was not found to decrease substantially during biofilm growth (data not shown).

Discussion
The principal contribution of this investigation is the discovery that a plant-root-associated strain of B. cereus (905) can form two kinds of biofilms that are produced by seemingly unrelated mechanisms. One kind is pellicles, which form by a mechanism involving orthologs of genes employed by B. subtilis in pellicle formation, including sinI and the two tasA-like genes, calY1 and calY2. Strikingly, however, B. cereus 905 can also form submerged, surface-associated biofilms by a process that has several of the hallmarks of S. aureus biofilm formation.
Although it has previously been observed that B. cereus strains can form biofilms, particularly in rings around the surface-air-liquid interface, the exact mechanism of biofilm formation had not been elucidated. We found that B. cereus 905 makes visible pellicles that float at the air-liquid boundary in a similar manner to that observed for B. subtilis 3610. Furthermore, by interrogating the B. cereus 905 genome we identified homologs for many of the genes that are important for this behavior in B. subtilis. The regulatory pathway leading to matrix production appears to be conserved between B. subtilis and B. cereus since both the master regulator Spo0A and the biofilm-specific regulator SinI were required for pellicle formation in B. cereus 905. However, the exact components of the biofilm matrix appear to differ. Two homologs for the biofilm matrix protein TasA appear to be required for pellicle formation in B. cereus 905 and might act redundantly. In earlier work, proteolytic activity was attributed to the protein encoded by calY1 in B. cereus (camelysin) (Fricke et al. 2001;Grass et al. 2004). However, the shared sequence identity of these proteins with TasA, the synteny of the calY genes with sipW and sinI/sinR, and their importance for pellicle biofilm formation in B. cereus 905, suggests that they can act as matrix proteins that might be capable of exhibiting amyloid-like fiber formation (Romero et al. 2010). It was also shown that expression of both tasA homologs (calY1 and calY2) in B. cereus group members B. anthracis and B. thuringiensis is under the control of the biofilm repressor SinR, reinforcing the idea that they play a role in biofilm formation (Pflughoeft et al. 2011;Fagerlund et al. 2014). Finally, in purifying CalY1 from B. cereus the authors noted that CalY1 was prone to SDS-resistant aggregation and was tightly attached to the cell wall (Fricke et al. 2001;Grass et al. 2004). This is reminiscent of the amyloid behavior displayed by TasA from B. subtilis (Romero et al. 2010). Further investigation of the function of the CalY proteins in the matrix will focus on whether these two proteins can functionally complement the lack of tasA in B. subtilis.
By contrast, the hypothetical EPS product of the 21gene cluster identified in B. cereus did not appear to be important for pellicle formation, despite the essential nature of the equivalent product for B. subtilis biofilm formation. The EPS product made by B. cereus 905 might have another function, perhaps forming the capsule. If a second eps gene cluster exists in B. cereus we were unable to detect it by homology to the B. subtilis gene cluster. In agreement with our findings, it has been shown that unlike the tasA-sipW and calY genes of B. anthracis and B. thuringiensis which are under the control of SinI/SinR, as they are in B. subtilis, the equivalent eps gene clusters are not (Pflughoeft et al. 2011;Fagerlund et al. 2014). This might suggest that the B. cereus group of bacteria do not induce expression of eps under biofilm forming conditions and perhaps do not require EPS for biofilm formation.
As well as being able to make pellicles we also found that B. cereus can form a biofilm by a second, seemingly independent, mechanism. In complex medium supplemented with glucose B. cereus formed surface-associated submerged biofilms. These biofilms were not dependent on SinI or the TasA-like proteins, suggesting that another route to biofilm formation is important here. Although a small decrease in biofilm formation was seen in the ΔcalY1 mutant, the absence of a phenotype for the ΔsipW mutant suggests that CalY1 does not significantly contribute to matrix under these conditions, since SipW is required to process TasA to its mature form in B. subtilis matrix formation. This mode of biofilm formation was, however, dependent on the master regulator Spo0A, suggesting that it is closely tied to the physiological state of cells. Furthermore, we found that in forming these biofilms the pH of the growth medium decreased substantially. This pH decrease was dependent on provision of glucose and was required for biofilm formation. This is reminiscent of biofilm formation in the Gram-positive pathogen S. aureus. In S. aureus we suggested that the pH decrease that accompanies biofilm formation causes the accumulation of matrix protein and possibly DNA at the cell surface, which promotes the aggregation of cells (Foulston et al. 2014). We speculate that this might also be the case in B. cereus in which both a protein and DNA component of the matrix were found to be important for this form of biofilm. Efforts to test this idea directly have, however, been hampered by the relatively weak nature of these biofilms in B. cereus, making it difficult to determine the nature of the proteins in the matrix, as was possible for S. aureus; however, our genetic analysis indicates that neither of the TasA homologs CalY1 and CalY2 were essential for submerged biofilm formation.
Some undomesticated strains of B. subtilis can form thick submerged biofilms in complex media such as TSB (Bridier et al. 2011), suggesting that the ability to form submerged biofilms might be more widespread within the Bacillus group. One of these strains of B. subtilis (ND medical ) was also found to form mixed biofilms with S. aureus, which could protect S. aureus against the action of biocides and this appeared to be mediated by the matrix material made by B. subtilis ND medical (Bridier et al. 2012). This might suggest that under these conditions B. subtilis and S. aureus can share matrix material. The genetic requirements of biofilm formation in these undomesticated strains of B. subtilis were not investigated, and thus it remains unclear what matrix material is being utilized (Bridier et al. 2011(Bridier et al. , 2012. We were unable to observe such submerged, surface-associated biofilm in B. subtilis NCIB3610, which continued to form pellicle biofilms in complex media; however, this might reflect strain-specific differences or differences in the way biofilm assays were conducted. In addition to forming surface-associated biofilms the B. subtilis undomesticated strains also formed pellicles in complex media (Bridier et al. 2011). We did not find this to be the case with B. cereus 905, which exclusively made submerged biofilms in complex media.
Taken together we find that B. cereus can display at least two modes of biofilm formation dependent on environmental conditions. An interesting question that remains to be answered is whether both of these mechanisms of biofilm formation occur in nature and, if so, what role each of them plays in the biology of B. cereus. Of particular interest is which of these mechanisms is enacted in the context of the plant root and whether it is then important for conferring plant protection? Our genetic analysis of B. cereus biofilm formation sheds light on how a similar outcome, biofilm formation, can be achieved in different ways. On one hand, B. cereus 905 behaves in a similar manner to its relative B. subtilis, a harmless soil organism that also interacts with plants. On the other hand, B. cereus 905 can exhibit behaviors more in line with pathogenic bacteria, such as S. aureus. Most strikingly these two modes of behavior appear to be achieved in radically different ways and are likely operating under different environmental conditions. Figure S1. Bacillus cereus 905 Dspo0A exhibits a sporulation defect. B. cereus 905 wild type and Dspo0A::Km were grown in DSM medium at 37°C for 36 h. Cells were washed twice in PBS and spotted onto slides and covered with polylysine-treated cover slips. Slides were visualized by phase contrast microscopy with a 1009 oil objective. Figure S2. Pellicle Formation can be complemented in trans. B. cereus 905 wild-type and deletion mutants with and without complementation constructs, as described in the text, were grown in MSgg for 3 days at 30°C in a 24 well plate and images were taken. Figure S3. Crystal violet stained biofilms of Bacillus cereus 905 in growth media of different pH. B. cereus 905 was grown in the indicated growth media in wells of a 12-well tissue culture plate at 37°C for 24 h. The medium was removed and biofilms were washed twice with PBS. The plates were stained with a 0.1% crystal violet solution for 30 min and then washed once with 1 mL of PBS. Plates were air dried before photographing. Figure S4. Submerged biofilm formation can be complemented in trans. B. cereus 905 and the indicated deletion mutants with and without complementation constructs, as described in the text, were grown in TSB with 1% glucose in wells of a 96-well tissue culture plate at 37°C for 24 h. The medium was removed and the OD 600 measured (medium). The well surfaces were washed with PBS and the OD 600 of the wash measured (wash). Finally cells on the surface of the well were removed by vigorous resuspension in PBS and the OD 600 measured (biofilm). Values shown are the average of three independent experiments with error bars indicating the standard deviation. Figure S5. Pellicle Formation by Bacillus subtilis in a range of growth media. B. subtilis 3610 was grown in the indicated growth media for 24 h at 37°C in a 96-well plate before photographing pellicle formation. Table S1. Strains and plasmids used in this study. Table S2. Oligonucleotide primers used in this study. Table S3. Identification of B. cereus genes with similarity to genes known to be important for biofilm formation in B. subtilis. Table S4. A putative exopolysaccharide gene cluster in B. cereus 905.