Communication determines population-level fitness under cation stress by modulating the ratio of motile to sessile B. subtilis cells

Bacterial populations frequently encounter potentially lethal environmental stress factors. Growing Bacillus subtilis populations are comprised of a mixture of “motile” and “sessile” cells but how this affects population-level fitness under stress is poorly understood. Here, we show that, unlike sessile cells, motile cells are readily killed by monovalent cations under conditions of nutrient deprivation – owing to elevated expression of the lytABC operon, which codes for a cell-wall lytic complex. Forced induction of the operon in sessile cells also causes lysis. We demonstrate that population composition is regulated by the quorum sensing regulator ComA, which can favor either the motile or the sessile state. Specifically social interactions by ComX-pheromone signaling enhance population-level fitness under stress. Our study highlights the importance of characterizing population composition and cellular properties for studies of bacterial physiology and functional genomics. Our findings open new perspectives for understanding the functions of autolysins and collective behaviors that are coordinated by chemical and electrical signals, with implications for multicellular development and biotechnology.


Introduction
Many bacterial populations frequently encounter potentially lethal environmental stress factors, and have evolved strategies that aid their survival. The Gram-positive Bacillus subtilis is an environmentally ubiquitous bacterium with a complex life style, and exhibits many beneficial traits that are exploited in diverse biotechnological applications. Studies of B. subtilis survival traits have a long history, and have contributed to a better understanding of how the species is so successful. In addition, this line of inquiry can help to identify how populations can be controlled and engineered to better serve mankind, specifically when naturally acquired resistance traits interfere with desired applications.
Vegetative B. subtilis can exist in two phenotypically distinct states: as a flagellated, motile cell that can explore its environment; and as a sessile cell that produces an extracellular matrix and can exploit locally available resources 1,2 . Motile cells have a short-rod like morphology while sessile cells tend to grow in chains of connected cells and both cell-types are capable of growing and dividing 3 The transitions between the motile and a sessile state are controlled by a complex regulatory network and can be modulated by environmental and cellular signals as well as noise 4,5 . Circumstantial evidence has also linked cell-cell communication via chemical signals to coordinate the physiologies of the different cell types during biofilm formation, via the lipopeptide surfactin 6 , although this has been recently questioned 7 .
It is tempting to equate a motile cell with a "planktonic cell" and a sessile cell with a "biofilm cell". However, there exists no planktonic or biofilm cell per se, since both biofilms and planktonic cultures are comprised of a mixture of cell types 4,5,8,9 . Thus, although biofilms are generally considered to be more stress resistant than planktonic cultures, it is unclear whether individual sessile cells are actually more robust than motile cells. In addition to that, an important implication of phenotypic diversification within isogenic populations is that population composition becomes a crucial determinant of population-level fitness under stress.
Here we show that individual sessile and motile cells exhibit different levels of stress tolerance, and that social interactions by chemical communication shift population composition which enhances population-level fitness under stress. Notably, this study was not originally designed to address these issues. Its roots lay in the serendipitous discovery that a mutant with impaired chemical communication lysed quickly in standard phosphate-buffered saline (PBS). We then employed a combination of bulk and single-cell studies to elucidate the basis for this phenomenon, which revealed unexpected links between chemical signaling and stress tolerance, and indications of a role for electrical signaling via monovalent cations. We specifically found that the quorum sensing master regulator ComA regulates population composition and favors the motile or sessile state depending on its phosphorylation status (and independent of surfactin). Motile cells are specifically sensitized to killing by monovalent cations by elevated expression the LytABC autolysin complex. This finding raises the possibility that motile subpopulations could be specifically targeted by electrical signals that induce cell death.

Starving populations of hyposocial B. subtilis are rapidly lysed by monovalent cations
By serendipity we discovered that a mutant strain with impaired chemical communication (relevant genotype: ΔcomA ΔrapCphrC ΔrapFphrF Phyperspank-PcomA-comAD55A), and for simplicity denoted as "hyposocial" strain, is highly susceptible to lysis in phosphate-buffered saline (PBS) (Fig. 1a). Suspensions of this strain in PBS, an isotonic salt solution that is commonly used for handling bacteria in the laboratory, lost 50% of their optical density (OD600nm) within an hour (T50 = 1.0 ± 0.1 h), while in Tris-buffer (Tris-HCl, pH 7.4) lysis occurred at a slower rate (T50 = 3.0 ± 1h) (Fig. 1b). To identify the stress factor(s) that cause population loss, we determined T50 values for cells suspended in Tris in the presence of different salts, osmolytes and nutrients ( Supplementary Fig.1). Cell suspensions containing monovalent cations (Na + , K + , NH4 + ) showed a rapid decline in OD, with little difference between the cations tested and regardless of the anion involved (Cl -, SO4 -2 , PO4 -3 , NO3 -); in contrast, the non-ionic osmolyte glycine betaine supplied at isoosmolar concentrations did not induce lysis. Population loss occurred in salt concentrations <150 mM and was decelerated at lower concentrations of monovalent cations. Notably, addition of divalent cations (Mg 2+ , Ca 2+ ) prevented lysis, and even halted ongoing lysis by NaCl practically immediately (Fig. 1c). The presence of small amounts of glucose also prevented lysis upon exposure to NaCl ( Supplementary Fig. 1).
Together, these results imply that, in a nutrient-depleted environment, monovalent cations cause rapid lysis of a hyposocial B. subtilis strain. and in Tris (grey curve). Data was normalized to the initial OD600. T50 denotes the exposure time required to reduce the initial OD by 50%. Data: mean ± std, nr = 6 technical replicates.
Strain: BIB1933. (c) Monovalent cations induce lysis, while addition of divalent cations prevents (and rapidly stops ongoing) lysis. Representative OD curves for cells suspended in Tris and exposed to NaCl in the absence (red), presence (orange) or upon acute addition of MgCl2. Strain: BIB2153.

ComA signaling mutants have pleiotropic effects on stress behavior and cell morphology
Cell-cell communication was not previously known to be required to protect against ionic stress. The transcription factor ComA occupies a central position in the B. subtilis communication network (Fig. 2a). ComA is regulated by several signaling systemsby the pheromone ComX, which affects its phosphorylation status via the histidine kinase ComP 20,21 , and by sequestration by Rap-modulator proteins 22,23 that are inhibited by signaling via a number of Phr peptides, respectively. Notably, there are no obvious direct or indirect gene targets of ComA that could confer resistance 24,25 . Surprisingly, signaling mutants that deregulate ComA activity showed pronounced differences in their ability to withstand lysis in PBS. A subset of mutant strains was more prone to lysis than the wild type, as evidenced by their lower T50 values, while other mutants were even more resistant than wild-type cells (Fig.   2b). In the hyposocial strain lysis occurred twice as fast (T50 = 55±5 min) as in the wild type (T50 = 101±8 min). In contrast, a mutant strain which responds sensitively to the ComX pheromone due to mutations that favor the accumulation of phosphorylated ComA (relevant genotype: ΔcomA ΔrapCphrC ΔrapFphrF Phyperspank-PcomA-comA), and for simplicity denoted as "hypersocial" strain, lysis was considerably delayed (T50 = 160±30 min). In addition to this difference in stress behavior, these two mutants differed in their morphologies (Fig. 2c). The hyposocial strain primarily contained short rod-shaped cells (that exhibited vigorous movement when immersed in liquid films; not shown), while the hypersocial strain was dominated by chains of connected cells (which were non-motile; not shown) that were reminiscent of sessile cells.

Sodium ions kill motile cells
The pleiotropic effects of deregulated ComA activity on stress behavior and cell morphology raised the possibility that the two cell types differ in their resistance traits. To investigate this, we engineered wild-type B. subtilis carrying fluorescent reporters that specifically mark motile and sessile cells by fusing the promoter of the hag gene (coding for flagellin) to YFP 4 and the promoter of the tapA gene (coding for a protein involved in extracellular matrix synthesis) to CFP 26,6 , and used fluorescence time-lapse microscopy experiments to study their stress responses at the cellular level (Fig. 3a). Wild-type populations comprised a mix of both cell types: short rods expressed YFP from the hag promoter (displayed in magenta) while cells in chains expressed CFP from the tapA promoter (displayed in cyan) (Fig. 3b, left). For simplicity, we will refer to "hag on" cells as "motile" and "tapA on" cells as "sessile", although both cell types are immobilized on Tris-based agarose pads in our experiments. Exposure of cells to NaCl selectively eliminated the motile subpopulation of cells (Fig. 3b, right, Supplementary Movie S1). Motile cells lysed rapidly and their numbers decreased exponentially at a rate of = (3.5 ± 1.0) h −1 (corresponding to a T50 = 13±4 min) resulting in the loss of more than 95% of all cells within an hour. In contrast, sessile cells did not lyse and their numbers remained stable over the same time period (Fig. 3c). We next recorded the response of each cell type to exposure to NaCl and subsequent addition of nutrients (Fig. 3d). The vast majority [99(±1)%] of motile cells underwent lysis, and only a negligible fraction survived the stress and resumed growth when nutrients became available. In contrast, the majority [50(±7)%] of sessile cells grew and another 14(±1)% were classified as "inert", i.e. these cells did not lyse, but they also failed to initiate visible growth within 30 min following nutrient addition. In a sodium-free environment little lysis of either cell type occurred (see also Fig. 3c, Inset) and 82(±5)% of motile and 94(±6)% of sessile cells were able to grow upon nutrient supplementation. Together, these data show that NaCl preferentially kills motile cells.

Differential expression of lytABC is responsible for cell-type specific lysis
The major autolysin LytC is required for cell lysis under a variety of stress conditions 10,11,19,27,28 .
LytC is a cell-wall amidase that is encoded in an operon, and expressed together with the enhancer LytB and the lipoprotein LytA. Expression is controlled by two promoters, one of which is dependent on the cell-type-specific sigma factor SigD, which is active in motile cells [29][30][31] . We therefore asked whether differential expression of lytABC contributes to cell-typespecific cation-induced lysis. Indeed, deletion of the lytABC operon largely prevented lysis (Fig.   4a). Microscopic investigations using the cell-type-specific markers revealed that the mutant retained both cell types and that lysis of motile mutant cells upon exposure to NaCl was inhibited ( Supplementary Fig. 2). We also monitored lytABC gene expression with a fluorescent promoter fusion to mCherry in a wild-type background. As expected, the distribution of mCherry fluorescence intensity obtained from motile cells was shifted to higher values relative to those in sessile cells (Fig. 4b). If differential lytABC expression was indeed responsible for cell-type-specific lysis, it should be possible to reverse the lysis pattern by directing lytABC expression to sessile cells. We therefore engineered the lytABC deletion strain to ectopically express the lytABC operon (in tandem with mCherry) under the control of the tapA promoter. Indeed, in the engineered strain, mCherry fluorescence was higher in sessile cells compared to motile cells, suggesting that the lytABC expression pattern was successfully reversed (Fig. 4c). Moreover, lysis was now preferentially observed in sessile  Micrographs of BIB2356 before (left) and after exposure to NaCl (right). Images show superposition of the YFP (magenta) and CFP (cyan) signals. Scale bar: 10 µm.

LytABC sensitizes cells to stress and mediates cation-induced cell death
We next asked whether the action of LytABC is restricted to the elimination of non-viable cells, or whether it might also sensitize cells to stress and thereby mediate stress-induced cell death.
We also wondered whether sessile cells have additional layers of protection or could be similarly sensitized to stress by raising LytABC levels. To address these questions, we engineered an IPTG-inducible lytABC operon into the lytABC deletion strain carrying the celltype specific markers and tracked the fate of each cell upon exposure to NaCl stress and the subsequent addition of nutrients (Fig. 5a) by roughly a factor of two. We thus conclude that LytABC sensitizes cells to stress and contributes to cell death in motile and, when overexpressed, also in sessile cells.

ComA has a dual function in modulating population composition
To assess how ComA affects population composition, we transformed all mutant strains with the Phag-YFP PtapA-CFP cell-type reporter. Fluorescence microscopy experiments suggested that fluorescence from both promoters provides a relatively stable marker for cell type, regardless of genotype ( Supplementary Fig. 3). Since measurements of subpopulation composition by microscopy turned out to be unreliable (due to the patchy distribution of both subpopulations on agarose pads), we used the population-averaged ratio of fluorescence as measured in a fluorescence plate reader as a proxy for population composition, r=log2(YFP/CFP). r varied with genotype (Fig. 5b). All genetic perturbations that were expected to increase the levels of ComA~P decreased r, suggesting that ComA~P favors the sessile state. In contrast, in a comA deletion strain, r was shifted to higher values than the wild type, indicating that the population had a higher proportion of motile cells. Surprisingly, the hyposocial mutant, which expresses non-phosphorylatable ComA due to an alanine substitution at its phosphorylation site D55 and which lacks RapF-and RapC-dependent ComA sequestration, showed an even higher r than the comA deletion mutant, suggesting that accumulation of unphosphorylated ComA favors the motile state.  (Fig. 5c). To distinguish between these possibilities, we analyzed effects on T50 in conjunction with the quantification of population composition by measuring log2(YFP/CFP) ratios in reporter populations (Fig. 5d). As expected, increased expression of LytABC decreased T50 values without significantly affecting population composition. In contrast, for comA and related genetic perturbations to social signaling the inferred changes in population composition with altered genotypes were highly correlated with the T50 values that were obtained from population-level lysis experiments: the higher the fraction of sessile cells in the population, the more resilient the population. Moreover, a mathematical model that assumes that these genetic mutations alter population composition without affecting the lysis properties of each cell type described the population dynamics in the complete mutant dataset very well ( Supplementary Fig. S4). The model inferred a roughly tenfold difference in the lysis rates for the two subpopulations ( m = 1.5 h −1 s = 0.1 h −1 ), which is in accordance with the single-cell experiments that show that sessile cells are much more tolerant to stress than motile cells. We thus conclude that hypersocial strains have a fitness advantage under stress due to an increased fraction of sessile cells.

Discussion
The generation of phenotypic population-level heterogeneity is an important strategy that bacteria employ as part of their adaptive repertoire to survive in nature. B. subtilis is known to diversify phenotypically under a variety of conditions, including planktonic cultures 4 and biofilms 8 . However, this factor has been -and often still is -neglected when analyzing resistance traits. A variety of stress conditions are known to result in cell lysis [10][11][12][13][14][15][16][17][18][19] . Notably, some studies have reported the "normal" cell shape of B. subtilis as "short rods" 32 while others observed cells growing in chains as being dominant in their experimental setting 15 . In many other cases, populations appeared to be comprised a mixture of both cell types, but this elicited no comment 10,18 . Our work clearly shows that motile and sessile cells have different resistance properties, and population-level resistance traits will therefore depend on population composition. Strain background, cultivation conditions and growth phase can affect population composition even in exponentially growing populations, making it difficult to compare different studies. We therefore suggest that all future studies should carefully characterize the composition and/or cell state of their populations.
The transition between motile and sessile cell states is regulated by a complex regulatory network centered around the SinR switch 1,2 , which is deeply embedded within the B. subtilis regulatory network. Thus, many genes with documented pleiotropic effects are expected to affect population-level traits, changing population composition rather than changing cellular properties. For example, the lysis-resistant isolate FJ2 exhibits cellular characteristics of bacteria locked in the sessile state, and indeed a mutation was mapped to sinR. The mutant strain did not lyse when exposed to surfactant stress 12 , cold shock 19 or ion stress 17 . On the other hand, the seminal work by Joliffe et al. 14 on cellular autolysis was performed on a spo mutant strain that was probably unable to generate sessile cellswhich presumably contributed to its invariably low stress tolerance. It will undoubtedly be a challenging but vital task to rigorously and comprehensively measure effects on cellular traits, as well as population composition on a genome-wide scale in order to clearly assign gene functions.

ComA-dependent cell-type switching may confer benefits that stabilize cooperative traits
Quorum sensing is a mode of orchestrating population-level behavior, which is commonly used in the biosphere to regulate the production of public goods. Interestingly, a growing body of evidence indicates that quorum sensing is tightly linked to stress tolerance. In several Gramnegative bacteria, quorum-sensing-deficient mutants have been shown to be vulnerable to various stress factors, including osmotic 33,34 , oxidative 35 , thermal 33 and heavy-metal stress 33 and others. In Vibrio harveyi, increased stress tolerance results from direct regulation of stressresponse pathways by a regulator that is controlled by quorum sensing 34 . Coupling of quorum sensing to stress resilience confers a private benefit to quorum-sensing-proficient cells, and thereby can contribute to stabilization of social investments and help to eliminate cheaters 33 .
Similar reasoning should also apply to B. subtilis. ComA directly controls the production of two public goods (surfactin and pectate lyase); however, a direct regulation of stress-response pathways by ComA seems unlikely 24,25 . Nevertheless, our data suggests that mutants with impaired pheromone-signaling are less fit under cation stress due to an altered population composition. Together with the observation that similarly impaired strains also fail to differentiate into competent (or K-state) cells 36 and endospores 37 respectively, this implies that an inability to communicate leads to a fitness disadvantage under a variety of stress conditions.
Our mutant data clearly show that ComA controls the composition of vegetative B. subtilis populations, and therefore must modulate switching between the motile and the sessile cell state. Phosphorylated ComA~P and non-phosphorylated ComA seem to have opposing functions and favor the sessile and motile states, respectively. How this is achieved at the molecular level requires further study. The promoters of ComA target genes have surprisingly complex architectures 24 , which might facilitate differential activation by ComA and ComA~P; and at least some target genes have (indirect, higher-order) connections to SinR, the regulator of the lifestyle switch 6,25 . Since deletion of comQXP reduced the frequency of sessile cells in our populations, ComX-based signaling seems to favor transitions to the sessile state.
However, we note that additional way(s) to favor the sessile state must exist beyond the previously proposed activation of surfactin synthesis via ComA~P 6 , since the B168 laboratory strain that was used in our study is deficient in surfactin synthesis 38 . Previous (population average) transcriptome data for comA mutants also support the notion of a ComX-dependent, but surfactin-independent transition to the sessile state, as transcripts of genes coding for matrix production were underrepresented 25 .

Elevated autolysin expression in motile cells may reflect a tradeoff
Our data show that sessile cells are more tolerant to stress but, perhaps surprisingly, these cells were not necessarily better protected from stress; instead, motile cells were more Many stress conditions, including entry into stationary phase 27 , biosurfactants 28 , cold shock 19 , sodium azide 11 and oxygen depletion 10 are known to result in LytC-dependent autolysis. Importantly, as we have shown here, LytABC not only eliminates non-viable cells, but also contributes to the induction of cell death under starvation conditions. Recent data reported by Arjes et al. shows that a lytC mutant strain has a somewhat increased viability relative to the wild type when exposed to oxygen depletion 10 . Moreover, upon closer inspection of their video material, the wild-type population seems to be made up of a mixture of motile and sessile cells.
Surviving cells predominantly grow in chains, and therefore probably represent mostly sessile cells. Others have also suggested that autolysins might contribute to killing, based on a higher viability of the lysis-resistant mutant strain FJ2 to exposure to surfactants 12 . Clearly, the function of LytABC in mediating cell death needs further study and additional mechanisms that reduce viability in response to stress conditions must exist. Yet it seems likely that motile cells are more susceptible than sessile cells to a variety of stress factors, specifically when these are encountered in conjunction with monovalent cations.

Electrical signaling may trigger the death of motile cells
The vulnerability of motile cells to monovalent cations might be profitably exploited by sessile We also observed cell-type-selective killing with potassium ions and in an undomesticated B.
subtilis strain (Supplementary Movies S5 & S6). In order for monovalent cations to kill, they probably have to act in concert with stress factors, such as starvation 14 , oxygen depletion 10 , toxins 15 or surfactants 12 , which dissipate the protonmotive force and/or disturb the membrane potential, respectively. If so, motile cells (of B. subtilis and likely also other species) could become prey for hungry B. subtilis sessile cells that elicit electrical signals to kill, lyse and feed on these cells 15,45,46 . In addition to that, hag-expressing and thus presumably motile B. subtilis cells are present at distinct locations within biofilms 8 . It is thus conceivable that cation-induced lysis of this subpopulation could contribute to the specific cell-death-based patterning that drives biofilm "morphogenesis" by projection of aerial structures 47 .

Applications
From a practical point of view, the exquisite sensitivity of subpopulations of bacteria to salt buffers such as PBS raises a strong caveat with respect to its use in experimental protocols.
Moreover, in order to improve stress tolerance of strains used in biotechnological applications lytABC is an obvious candidate for deactivation. On the other hand, cation-sensitive lysis by LytABC could be useful for some synthetic biology applications: targeting lytABC expression may allow for precise elimination of anyeven a transientcell state, as demonstrated here.
Finally, other both Gram-positive and Gram-negative bacteria have been reported to undergo autolysis in conditions similar to ours [48][49][50][51] , raising the possibility that proteins that function analogous to LytABC are widespread.

Strains
All strains were derivatives of the laboratory strain 1A700 (W168) and their genotypes are listed in Supplementary Table S1. Experiments were also performed with PS216, a recent natural isolate 52 .

Plasmid construction
All plasmids and oligonucleotides used in this study are listed in Supplementary Tables S2 and   S3, respectively. E. coli DH5α was used for cloning. Plasmids were constructed by ligationindependent cloning (LIC) and by restriction-enzyme ligation cloning (RELC), respectively. All DNA fragments were amplified from the genomic DNA of B. subtilis W168 (unless noted otherwise). Plasmids were verified by analytical restriction digestion and sequencing.
Plasmids for fluorescent promoter fusions were constructed by amplifying 100-300 bp of the promoter region located upstream of the ribosome-binding site, and inserted into pXFP_Star and pXFP_bglStar, respectively. pXFP_bglStar vectors were constructed by RELC from pXFP_Star 53  To construct the plasmid used for differential gene expression of the lytABC operon in sessile cells (EIB799), the PtapA promoter (amplified from EIB727) was fused to an optimized ribosome-binding site and the lytABC operon by PCR, and the resulting construct was inserted into pRFP_Star by LIC.

Strain construction
Strains were constructed by a one-step 57 or a two-step transformation method 58 with the ectopic integration vectors listed in Supplementary Table S3. The pLK-comK plasmid was used to transform strains with reduced competence development and the final strains were cured from the plasmid using standard protocols 59 . All reporters and expression cassettes were integrated at an ectopic locus (amyE, bglS) and present in single copy as verified by PCR. We also checked that transformations carried out with pDR111 retained an intact ldh locus 60 . Clean gene deletions were constructed using pMAD plasmids, following a protocol similar to that of Arnaud et al. 55 . At each step, we verified by appropriate PCRs that the gene had been deleted from the chromosomal locus, that the pMAD plasmid had been lost and, finally, that the gene was absent from the chromosome.

Plate Reader Assays
Population loss was monitored by absorbance measurements in 96-well plates (96 MicroWell TM , Nunc TM , Thermo Scientific) using at least n = 6 technical replicates with 200-µl sample volumes. Plates were incubated at 37°C in a preheated Epoch 2 Microplate Spectrophotometer (BioTek) under linear shaking (1 mm) at 1096 cycles per minute. The OD at 600 nm was recorded at 2-min intervals. Where applicable, the incubation was interrupted to add 10 µl of an aqueous solution of divalent cations to selected wells. Curves were normalized with respect to their initial OD. T50 was obtained by linear interpolation between ti and ti-1, where ti is defined as the time-point at which normalized OD dropped below 0.5.

Microscopy
Samples (3 μl) of cell suspension (OD = 2.0) were spotted on agarose pads (Tris-HCl solidified with 1.2% Ultrapure Agarose, dimensions: 0.9 cm diameter, 1 mm height) and stamped into 24-well SensoPlates (Greiner). Positions were manually selected to track the fate of both cell types in the same field of view (note that cell types were unevenly distributed on the same pad).
Fluorescence and bright-field images were recorded before the onset of stress. To induce cation stress, the plate was removed from the microscope and 4 µl of stress solution was added to the upper surface of the pad and imaging was resumed at the same positions. If applicable, time-lapse movies were recorded by bright-field imaging every 5 min. For cell-fate analysis at least one bright-field image was taken before adding the recovery medium (3 µl) and at the end of the indicated time period.

Image analysis
Fluorescence quantification and cell-type classification: Bright-field images were segmented using a customized software and results were manually inspected. For each cell, the mean fluorescence intensity was determined from the segmented area and the background fluorescence was subtracted 24 . Cells were classified as "motile" ("sessile") based on fluorescence thresholding. Cells that could not be unambiguously assigned (most likely as a result of a cell-type switching event) were excluded from further analysis.
Cell Fate Analysis: Cells were manually classified as "lysed" if the cell was no longer visible in the bright-field image (or visibly compromised by a dissolving cell periphery), as "growing" if cell length increased by at least 0.5 µm and as "inert" otherwise. Cells that divided during the stress period were excluded from the analysis, unless both daughter cells had the same fate.