Cyclic di-AMP Acts as an Extracellular Signal That Impacts Bacillus subtilis Biofilm Formation and Plant Attachment

ABSTRACT There is a growing appreciation for the impact that bacteria have on higher organisms. Plant roots often harbor beneficial microbes, such as the Gram-positive rhizobacterium Bacillus subtilis, that influence their growth and susceptibility to disease. The ability to form surface-attached microbial communities called biofilms is crucial for the ability of B. subtilis to adhere to and protect plant roots. In this study, strains harboring deletions of the B. subtilis genes known to synthesize and degrade the second messenger cyclic di-adenylate monophosphate (c-di-AMP) were examined for their involvement in biofilm formation and plant attachment. We found that intracellular production of c-di-AMP impacts colony biofilm architecture, biofilm gene expression, and plant attachment in B. subtilis. We also show that B. subtilis secretes c-di-AMP and that putative c-di-AMP transporters impact biofilm formation and plant root colonization. Taken together, our data describe a new role for c-di-AMP as a chemical signal that affects important cellular processes in the environmentally and agriculturally important soil bacterium B. subtilis. These results suggest that the “intracellular” signaling molecule c-di-AMP may also play a previously unappreciated role in interbacterial cell-cell communication within plant microbiomes.

whether c-di-AMP levels were lower in the DAC mutant strains and higher in the PDE mutant strains) (see Fig. S1 in the supplemental material).
In B. subtilis, colony morphology is impacted by biofilm matrix production. Thus, to determine if biofilm formation was impacted in these mutants, the colony morphology of each strain was evaluated after 48 h of growth at 30°C on MSgg medium (5 mM potassium phosphate [pH 7], 100 mM morpholinepropanesulfonic acid [MOPS; pH 7], 2 mM MgCl 2 , 700 M CaCl 2 , 50 M MnCl 2 , 50 M FeCl 3 , 1 M ZnCl 2 , 2 M thiamine, 0.5% glycerol, 0.5% glutamate) (a biofilm-inducing medium) agar plates. For comparison, a strain lacking the biofilm repressor sinR and a strain lacking all the biofilm matrix genes (epsA-O, tasA, and bslA) were used as controls for high-and low-biofilm-matrix producers, respectively. ΔcdaA and ΔcdaS exhibited small but reproducible differences in colony morphology compared with wild-type B. subtilis (Fig. 1), whereas the ΔdisA mutant exhibited a strikingly altered colony morphology on MSgg medium (Fig. 1). The PDE mutant ΔgdpP displayed a star-shaped colony morphology with large wrinkles connecting in a raised circle pattern at the center, while the PDE mutant ΔpgpH produced colonies with a flatter profile and wrinkles that were less pronounced than those seen with the wild type ( Fig. 1). Since the ΔdisA, ΔgdpP, and ΔpgpH strains exhibited the most dramatic biofilm phenotypes, we focused on these mutants in further characterizing the role that c-di-AMP plays in biofilm formation in B. subtilis.
Biofilm gene expression. To determine if disA, gdpP, and pgpH impact biofilm formation through modulation of biofilm matrix gene expression, we deleted each gene of interest in a B. subtilis strain containing a luciferase reporter for biofilm gene expression. This strain harbored the luxABCDE operon driven by the tapA promoter (P tapA -lux) integrated into the neutral sacA locus in the chromosome (30). Luminescence measurements were taken from shaking cultures of these strains grown in MSgg liquid media at 24 h. Under the conditions examined, tapA promoter activity in the ΔdisA mutant was lower than that seen with the wild type and tapA promoter activity in the ΔgdpP and ΔpgpH mutants was higher than that seen with the wild type ( Fig. 2A). These results indicate that biofilm matrix gene expression was decreased in mutant ΔdisA relative to wild-type B. subtilis and was generally increased in mutants ΔgdpP and ΔpgpH, consistent with c-di-AMP levels impacting the expression of biofilm matrix genes. Shaken liquid cultures are ideal for quantitative luminescence measurements; however, gene expression levels often differ between planktonic and biofilm-grown cells. To observe P tapA -lux in colony biofilms, these strains were spotted onto MSgg agar plates and P tapA -lux was detected after growth using chemiluminescent imaging. We found that the promoter activity was highest at the edges of the colonies in the wild type (Fig. 2B). Consistent with the liquid culture data, at the colony level, tapA promoter C-di-AMP in B. subtilis Biofilms and Plant Attachment ® activity appeared to be lower overall in the ΔdisA mutant than in the wild type ( Fig. 2B) and was higher overall in the ΔgdpP and ΔpgpH mutants than in the wild type (Fig. 2B). Since B. subtilis P tapA expression localized to different areas of the biofilm in wild-type and mutant B. subtilis colonies, we wanted to quantify the percentage of matrix-producing cells within each population. We used flow cytometry to quantify fluorescent cells in wild-type, ΔdisA, ΔpgpH, and ΔgdpP colonies containing the P tapAyfp reporter (yfp encodes yellow fluorescent protein [YFP]). We harvested biofilm colonies grown on MSgg medium for 24 h and fixed cells with paraformaldehyde. To quantify fluorescent cells, we performed gating on a sample of B. subtilis cells constitutively expressing YFP. These data show that the percentage of ΔdisA cells expressing P tapA -yfp (46%) was lower than the percentage of wild-type B. subtilis cells (69%) (Fig. 2C). The percentage of cells expressing the P tapA -yfp biofilm reporter within the ΔpgpH and ΔgdpP biofilm colonies was similar to that seen with wild-type B. subtilis (68% and 73%, respectively) (Fig. 2C). Notably, however, a greater median fluorescence intensity was observed in the ΔpgpH and ΔgdpP strains than in the wild-type strain. These data indicate that although similar percentages of cells were fluorescent in these strains, the fluorescent cells in the PDE mutants were expressing higher levels of yfp (i.e., were expressing P tapA more strongly) than the fluorescent wild-type cells (Fig. 2D). Taken together, these results imply that disA, gdpP, and pgpH are all involved in modulating biofilm formation by altering tapA biofilm gene expression.
Complementation of disA, gdpP, and pgpH. We then wanted to confirm that the observed changes in tapA promoter activity in the ΔdisA, ΔgdpP, and ΔpgpH strains were directly attributable to the disruption of these genes. To do so, we complemented each of these mutant strains with a single copy of an IPTG (isopropyl-␤-Dthiogalactopyranoside)-inducible copy of their cognate wild-type gene in the amyE site of the chromosome (31), with the expectation that (if these genes were responsible for the effects on tapA promoter activity) the complemented strains would exhibit P tapA -lux activity more similar to wild-type levels than the uncomplemented strains. Each of these strains also harbored P tapA -lux. The disA complementation strain showed a small but reproducible increase in P tapA -lux activity relative to the levels observed in the ΔdisA mutant, while the PDE complementation strains showed decreases in P tapA -lux activity relative to the corresponding deletion strain (Fig. S2). These results confirm the respective roles of these genes in c-di-AMP-mediated biofilm formation. Surfactin production. Previous studies have demonstrated that, in addition to matrix gene expression, surfactin production is relevant to biofilm architecture in B. subtilis (32,33). To determine if surfactin production was altered in the DAC and PDE mutants, we performed a drop-collapse assay using cell-free spent media obtained after growing each mutant and wild-type B. subtilis in liquid culture overnight. If surfactin is present in the spent medium, it reduces the surface tension of the liquid, allowing it to spread further when spotted onto a hard surface; adding a dye allows the spread of the spent medium to be visualized and measured. A strain harboring a deletion of srfA, the locus responsible for surfactin production, was used as a negative control. The ΔdisA mutant produced less surfactin than the wild type, similar to the ΔsrfA control, while mutants ΔgdpP and ΔpgpH both produced more surfactin than the wild type (Fig. 3). Surfactin production in these mutants therefore correlates with the observed biofilm phenotypes and tapA promoter activity.
C-di-AMP production affects plant attachment. Biofilm formation is crucial for B. subtilis attachment to plant roots (2). We therefore hypothesized that since these c-di-AMP mutants exhibited altered biofilm phenotypes, they might also impact plant attachment. To test this prediction, we examined whether the c-di-AMP mutants exhibited altered attachment to Arabidopsis thaliana roots. Six-day-old A. thaliana seedlings were added to media containing B. subtilis strains constitutively producing the fluorescent protein mTurquoise in 48-well plates, and bacterial attachment to the roots was imaged using confocal laser scanning microscopy after 24 h. In addition to the wild-type strain and the ΔdisA, ΔgdpP, and ΔpgpH mutants, we examined a biofilm matrix deletion mutant known to be unable to colonize plant roots (mutant ΔepsA-O ΔtasA ΔbslA) (2). The ΔdisA mutant displayed a severe colonization defect, similar to the results seen with the matrix-deletion control (Fig. 4), while the strains lacking either PDE gene (mutants ΔgdpP and ΔpgpH) both colonized better than the wild type (Fig. 4). We observed the same trends when bacteria were recovered from the roots and CFU were counted (Fig. S3). These results are consistent with the respective biofilm phenotypes C-di-AMP in B. subtilis Biofilms and Plant Attachment ® observed as described above and indicate that c-di-AMP signaling is important for B. subtilis plant attachment. C-di-AMP secretion contributes to B. subtilis biofilm formation. C-di-AMP has been previously demonstrated to be secreted in a variety of bacterial pathogens (27)(28)(29). To address whether B. subtilis can secrete c-di-AMP, we directly quantified extracellular concentrations of c-di-AMP using liquid chromatography-mass spectrometry (LC-MS). First, we confirmed that the ΔdisA mutant did not have a growth defect (Fig. S4). We then detected c-di-AMP in the supernatant of wild-type B. subtilis (Fig. 5), and, to a lesser extent, in that of the ΔdisA mutant grown in liquid culture, indicating that B. subtilis indeed secretes c-di-AMP.  Townsley et al. ® We then hypothesized that, if extracellular secretion and sensing of c-di-AMP were important for B. subtilis biofilm formation, the plant attachment defect of ΔdisA could be a result of its lower c-di-AMP secretion. To determine whether low extracellular levels of c-di-AMP were contributing to the inability of the ΔdisA mutant to colonize plant roots, we tested whether its attachment defect could be complemented by wild-type B. subtilis, which secretes higher levels of c-di-AMP. We performed coculture root inoculations with the ΔdisA mutant (constitutively expressing mTurquoise) with nonfluorescent wild-type cells; we mixed the cells 1:1 and inoculated plant roots as described above. Root attachment was imaged 24 h after plant inoculation. We found that the ΔdisA mutant was able to attach to plant roots when wild-type B. subtilis was present (Fig. 6). This suggests that the mutant ΔdisA plant colonization defect can be complemented by the presence of wild-type B. subtilis cells.
One trivial explanation for this effect of wild-type B. subtilis cells on the ability of mutant ΔdisA to attach to plant roots could be that cells of the biofilm-deficient ΔdisA mutant cells simply "stick" to the extracellular matrix that wild-type cells produce. To test this, we cocultured mutant ΔdisA with the non-matrix-producing ΔepsA-O strain and again examined its ability to colonize plant roots. As shown in Fig. 6, the presence of mutant ΔepsA-O also allowed mutant ΔdisA to attach to plant roots, indicating that this complementation is not affected by the ability to produce matrix. Thus, these data suggest that the production of extracellular c-di-AMP by wild-type and ΔepsA-O cells may be acting to stimulate biofilm formation in the ΔdisA cells, allowing them to colonize roots.
Identification of putative c-di-AMP transporters and their role in biofilm formation. C-di-AMP in Listeria monocytogenes is secreted through the multidrug efflux pumps MdrM and MdrT, which are controlled by the regulators MarR and TetR (20). A search of the B. subtilis genome for mdrM and mdrT homologues identified four genes that encode predicted permeases with over 30% identity to both mdrM and mdrT: ycnB, yhcA, imrB (formerly yccA), and mdtP (formerly yusP) ( Table 1). Because ycnB and yhcA shared the most similarity to the L. monocytogenes transporters, we produced strains lacking either ycnB or yhcA and compared their levels of secreted c-di-AMP to those of C-di-AMP in B. subtilis Biofilms and Plant Attachment ® the wild type to identify a possible c-di-AMP transporter. We found no significant difference between the wild-type, ΔycnB, and ΔyhcA strains in c-di-AMP levels (Fig. 7). Because these putative transporters could potentially compensate for each other, we then produced a double mutant strain lacking both ycnB and yhcA. We observed a significant decrease in the levels of secreted c-di-AMP in this double mutant strain compared to the wild type (Fig. 7). We did not observe a significant difference in intracellular levels of c-di-AMP in the ΔycnB ΔyhcA strain, suggesting that only c-di-AMP secretion (and not c-di-AMP production) is impacted in this strain (Fig. S5).
We then tested the effects that these putative c-di-AMP transporters had on biofilm formation in the context of plant roots. We cocultured a fluorescent ΔdisA strain with the transporter mutants on A. thaliana roots as described above. Similarly to the data shown in Fig. 6, the ΔdisA mutant attached to plant roots when it was cocultured with the ΔepsA-O mutant (Fig. 8). We then directly tested whether this complementation depended on these transporters by knocking them out of the ΔepsA-O strain. The ΔycnB ΔepsA-O and ΔyhcA ΔepsA-O mutants did not complement the attachment defect of ΔdisA as well as the ΔepsA-O mutant alone, and the ΔdisA mutant had a significant plant colonization defect in the presence of the ΔycnB ΔyhcA ΔepsA-O mutant (Fig. 8). The extent of ΔdisA colonization visible in these images is consistent with the quantification of mutant ΔdisA CFU recovered from the roots (Fig. S6). These results suggest that the ycnB and yhcA genes are important for the ability of ΔepsA-O cells to complement the plant attachment defect of the ΔdisA mutant and that the double mutant is unable to rescue it. These data are all consistent with a model proposing that the ycnB and yhcA genes encode c-di-AMP transporters and that their ability to secrete extracellular c-di-AMP impacts biofilm formation and plant attachment in neighboring B. subtilis cells.

DISCUSSION
Biofilm formation is important for environmental fitness and adaptation in many bacteria. Although diverse mechanisms exist for regulating biofilm formation, cyclic n.s.  di-nucleotide second messengers play a critical role in many bacteria. The intracellular signaling molecule cyclic di-guanylate monophosphate (c-di-GMP) mediates biofilm formation in a vast number of Gram-negative bacteria (34). C-di-GMP was recently discovered in B. subtilis (35,36); however, unlike its activity in the related Gram-positive bacterium Bacillus cereus (37), evidence suggests that c-di-GMP does not play a major role in biofilm formation in B. subtilis (35,36). Emerging studies, however, are indicating that c-di-AMP may be important for controlling biofilm formation in some Grampositive bacteria; increased levels of intracellular c-di-AMP stimulate biofilm formation in both Streptococcus mutans (38) and Staphylococcus aureus (39). Here we determined that altering c-di-AMP levels in B. subtilis, by deleting either the DACs that synthesize it or the PDEs that degrade it, modulates biofilm formation in B. subtilis. Few previous studies have explored the role of c-di-AMP in B. subtilis biofilm formation. One recent study reported that although there was no change in tapA and epsA expression in single mutants lacking either gdpP or pgpH, the deletion of both PDEs (which would be predicted to lead to a dramatic accumulation of c-di-AMP) downregulated the mRNA abundance of tapA and epsA in B. subtilis (40). However, transcriptome data from the double PDE mutant in this same study were inconsistent with these results: they showed an upregulation of the biofilm inducer abh and a downregulation of the biofilm repressor abrB, both of which would be predicted to increase biofilm formation. The study by Gundlach et al. was conducted using growth conditions different from ours, which could have contributed to the discrepancy between the conclusions drawn in our two studies. Our data demonstrate that increased c-di-AMP levels induce the promoter activity of the tapA operon that is required for biofilm formation in B. subtilis.

W i l d t y p e y c n B y h c A y c n B y h c A
Although our data indicate that increases in both intracellular and extracellular levels of c-di-AMP positively influence biofilm formation, we still do not know the molecular details of the mechanisms by which c-di-AMP regulates biofilm formation. One possibility is that c-di-AMP acts through alterations in the phosphorylation state of the master transcriptional regulator Spo0A. A previous study determined that the sporulation delay observed in a disA mutant is due to changes in Spo0A phosphorylation (41), although, again, the molecular details of how Spo0A is impacted by c-di-AMP remain unclear. The c-di-AMP receptors identified thus far in B. subtilis include two riboswitches that control amino acid transporter gene ydaO (renamed kimA) (42, C-di-AMP in B. subtilis Biofilms and Plant Attachment ® 43), the P II signal transducer protein encoded by darA (44), and the potassium transport protein KtrA (45). KtrA is part of one of the two main proteins associated with potassium uptake mechanisms in B. subtilis: KtrAB and KtrCD (46). When mutated, ktrC enhances biofilm formation; potassium leakage is known to induce biofilm formation in B. subtilis via the sensor histidine kinase kinC (47). Thus, integration of c-di-AMP into the potassium homeostasis network could potentially be a mechanism for impacting biofilm formation in B. subtilis. Indeed, the recently renamed YdaO protein (now KimA) has been shown to act as a potassium transporter (42). Interestingly, both ktrA and ktrC are physically located adjacent to the biofilm-relevant genes in the B. subtilis genome: ktrA is immediately downstream of bslA, while ktrC is downstream of abh and the kinC operon. Additional studies are needed to determine if these or other, yet-to-beidentified receptors are important for connecting c-di-AMP signaling to the biofilm regulatory network in B. subtilis.
We also identified two putative c-di-AMP transporters and demonstrated that B. subtilis secretes c-di-AMP and can sense and respond to extracellular c-di-AMP. These data suggest an important role for this second messenger in interbacterial communication. To our knowledge, B. subtilis is the first nonpathogenic bacterium discovered to secrete c-di-AMP, which implies that this signaling molecule may play a role in bacterial communication not only in human hosts but also in the environment. The biofilm formation and sporulation pathways in B. subtilis are controlled by many of the same regulatory elements, and it is believed that sporulation is the culmination of biofilm formation (15). A previous study was able to induce sporulation in B. subtilis by the addition of exogenous c-di-AMP (48), further corroborating our observation that B. subtilis can sense exogenous c-di-AMP and respond through the biofilm/sporulation regulatory pathway.
Our data are consistent with a model where B. subtilis secretion of c-di-AMP impacts biofilm formation and plant attachment in other B. subtilis cells. Future studies are needed to test whether B. subtilis and other bacteria can sense c-di-AMP produced by other species in the environment and to elucidate the effects that extracellular c-di-AMP production and sensing may have on bacterial community signaling and plant microbiome community structure.

MATERIALS AND METHODS
Bacterial strains and growth conditions. B. subtilis NCIB3610 was used as a wild-type strain. Escherichia coli DH5␣ and B. subtilis 168 were used for cloning. Overnight cultures were grown on Luria-Bertani (LB)-Lennox medium (10 g of tryptone, 5 g of yeast extract, 5 g of NaCl per liter) at 30°C. Biofilm assays were performed on MSgg medium (5 mM potassium phosphate [pH 7], 100 mM morpholinepropanesulfonic acid [MOPS; pH 7], 2 mM MgCl 2 , 700 M CaCl 2 , 50 M MnCl 2 , 50 M FeCl 3 , 1 M ZnCl 2 , 2 M thiamine, 0.5% glycerol, 0.5% glutamate). When needed, chloramphenicol and erythromycin-lincomycin (MLS) were used at 5 g/ml and 1 g/ml, respectively. Intracellular c-di-AMP quantification. B. subtilis colony biofilms grown on MSgg plates were scraped off, resuspended into 5 ml PBS (phosphate-buffered saline), and sonicated (amplitude ϭ 20 for 12 s with 1-s on/off pulses) to break clumps. Cultures were divided into 4.5 ml (for c-di-AMP quantification) and 500 l (for protein quantification) portions. The c-di-AMP quantification samples were centrifuged at 4,000 rpm for 20 min and resuspended in 1 ml cold extraction buffer (acetonitrile, methanol, and distilled water [dH 2 O] in a 40:40:20 ratio). Samples were snap-frozen using liquid N 2 and then incubated at 95°C for 10 min, 0.5 ml of 0.1-mm-diameter glass beads was added to samples, and a FastPrep-24 instrument (MP Biomedicals, Santa Ana, CA, USA) was used to homogenize the samples, treating them at 4 m/s for 45 s twice. Samples were then briefly centrifuged, and the supernatant was recovered and dried using a Savant SC100 SpeedVac (Thermo Fisher Scientific, Waltham, MA). Samples were resuspended in 100 l liquid LC-MS-grade H 2 O and analyzed using high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) on a Quantum Ultra triple-quadrupole mass spectrometer (Thermo Fisher Scientific, Waltham, MA) equipped with an Acquity ultraperformance LC (UPLC) separation system (Waters Corp., Milford, MA). An Acquity UPLC HSS T3 column (Waters Corp., Milford, MA) (2.1 mm by 100-mm diameter; 1.8-m particle size) was used for reverse-phase liquid chromatography. Solvent A was 10 mM ammonium formate-water, and solvent B was 10 mM ammonium formate-methanol. The injection volume was 10 l, and the flow rate for chromatography was 200 l/min. A c-di-AMP standard was prepared with purified c-di-AMP (Biolog Life Sciences, Bremen, Germany). C-di-AMP levels were normalized to total protein per milliliter of culture. Protein quantification was performed using the bicinchoninic acid (BCA) assay (Thermo Fisher Scientific, Waltham, MA) with bovine serum albumin (BSA) as standards. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with a Tukey test for multiple comparisons.
Townsley et al.