Carbohydrate and PepO control bimodality in competence development by Streptococcus mutans

Summary

In Streptococcus mutans, the alternative sigma factor ComX controls entry into genetic competence. Competence stimulating peptide (CSP) induces bimodal expression of comX, with only a fraction of the population becoming transformable. Curiously, the bimodality of comX is affected by peptides in the growth medium and by carbohydrate source. CSP elicits bimodal expression of comX in media rich in small peptides, but CSP elicits no response in defined media lacking small peptides. In addition, growth on certain sugars increases the proportion of the population that activates comX in response to CSP. By investigating the connection between media and comX bimodality, we find evidence for two mechanisms that modulate transcriptional positive feedback in the ComRS system, where comX bimodality originates. We find that the endopeptidase PepO suppresses the ComRS feedback loop, most likely by degrading the XIP/ComS feedback signal. Deletion of pepO eliminates comX bimodality, leading to a unimodal comX response to CSP in both defined and complex media. We also find that CSP stimulates the ComRS feedback system by upregulating comR in a carbohydrate source-dependent fashion. Our data provide mechanistic insight into how S. mutans regulates bimodality and explain the puzzle of growth medium effects on competence induction by CSP.

Graphical Abstract

Abbreviated Summary

The CSP peptide induces genetic competence in only a fraction of S. mutans cells, and only in peptide-rich growth media. We show that the endopeptidase PepO is responsible for preventing a population-wide competence response. Our data indicate that PepO suppresses positive transcriptional feedback in the ComRS system, which controls comX activity, giving rise to the bimodal and medium-dependent competence. In addition we show that some carbohydrates stimulate ComRS feedback by enhancing transcription of comR. These findings reveal mechanisms behind some striking properties of the S. mutans competence pathway.

Introduction

The Gram-positive bacterium Streptococcus mutans inhabits human oral biofilms and is a primary etiologic agent of dental caries (Loesche, 1986). The ability of S. mutans to colonize the oral cavity and compete with commensal organisms is associated with the com regulon (Li et al., 2002; Suntharalingam; Cvitkovitch, 2005). The com (competence) regulon controls entry into genetic competence, a transient state during which cells are able to internalize DNA from the environment (Shanker; Federle, 2017). In S. mutans the com regulon is linked to tolerance of environmental stress, including heat, oxidative stresses, pH and carbohydrate availability (Qi et al., 2004; Ahn et al., 2005; Ahn et al., 2006; Senadheera, M. D. et al., 2007; Tremblay et al., 2009; Senadheera, D. B. et al., 2012). It is also involved in bacteriocin production and lysis, which are important in interspecies competition (Shanker; Federle, 2017), and in biofilm formation and stability (Li et al., 2002). There are many unresolved questions in the study of S. mutans competence, especially in regard to how the com regulon integrates diverse environmental cues, such as the presence of different signal peptides or nutrients or the composition of the growth medium, in order to drive different phenotypic outputs in downstream regulated genes (Hagen; Son, 2017).

The quorum sensing peptide CSP (competence stimulating peptide) (Li et al., 2002) is a primary input to the S. mutans competence pathway (Fig. 1). CSP is derived from a 46-residue precursor encoded by comC, processed to 21 residue length, then exported to the extracellular environment by the ComAB transporter. The ComC peptide is further processed by the extracellular SepM protease to yield the mature 18-residue CSP, which is understood to be the most active form of the peptide (Hossain; Biswas, 2012; Shanker; Federle, 2017). Extracellular CSP stimulates the competence pathway by interacting with the ComDE two-component signal transduction system (TCSTS): CSP bound by the transmembrane kinase ComD induces phosphorylation of ComE, which then acts as a transcriptional activator for several bacteriocin and competence-related genes, including cipB (Perry et al., 2009; Perry et al., 2009; Fontaine et al., 2015). Although the mechanism is not known, expression of cipB stimulates the ComRS system, an Rgg-type signaling system that is the immediate regulator of comX (also called sigX). ComX (or SigX) is an alternative sigma factor for late competence genes, which encode proteins for uptake and processing of DNA. Therefore, CSP drives transformability through a pathway that includes ComCDE, cipB, ComRS and ComX, with the regulatory link from cipB to ComRS being the least understood. However, a signature characteristic of this pathway (unlike the generally similar pathway in S. pneumoniae (Shanker; Federle, 2017)) is that comX responds bimodally (if at all) to CSP: when CSP is supplied in complex media containing small peptides, comX activates in only a subpopulation of cells, even though cipB activates population-wide. In defined media lacking small peptides, CSP elicits no comX response, although cipB is again activated population wide (Son, Minjun et al., 2015; Reck et al., 2015).

Fig. 1: Current model of CSP induction of competence.

CSP induction of competence in S. mutans (Shanker; Federle, 2017; Underhill et al., 2018). CSP binds the ComD transmembrane kinase, which phosphorylates ComE. ComE phosphate activates bacteriocin genes including cipB, which through an unknown mechanism stimulates the ComRS positive feedback system. In the ComRS system, ComS interacts with ComR to form a transcriptional activator for comS and comX. Thus the ComRS system has two possible states: at low ComS levels the activating complex is absent and comX remains OFF, while at high ComS levels the activating complex is sustained through positive transcriptional feedback and comX switches ON. The state of activation is thus controlled by basal ComR and ComS levels, cell-to-cell variability in ComS levels, and the rate of turnover of ComS (Son, M. et al., 2012). As a result activation of the ComRS system is heterogeneous in the population, leading to bimodal expression of a PcomX-gfp reporter in individual S. mutans. The inset at bottom shows a histogram of individual cell reporter fluorescence for cells supplied 1 μM CSP in complex growth medium (TV). Two distinct populations are visible: the overlaid (red) curves are gamma probability distribution functions corresponding to OFF (left) and ON (right) states of the ComRS/comX system. (Methods).

Bimodality arises in the ComRS system, which lies downstream of cipB (Son, M. et al., 2012). Mashburn-Warren et al. (Mashburn-Warren et al., 2010) showed that XIP (sigX-inducing peptide), comprising the C-terminal 7 residues of ComS, can induce comX if supplied exogenously to S. mutans. XIP is imported by the Opp permease and interacts with ComR to form a multimeric complex that is a transcriptional activator for both comX and comS (Mashburn-Warren et al., 2010; Son, M. et al., 2012; Fontaine et al., 2013; Underhill et al., 2018). XIP induces a population-wide (unimodal) comX response when provided as an extracellular signal in defined growth medium. However the presence of transcriptional positive feedback via comS also allows the ComRS system to operate as an intracellular positive feedback loop: endogenously produced ComS apparently interacts with ComR intracellularly to enhance transcription of comS and comX (Underhill et al., 2018). Such a feedback mechanism is sensitive to basal levels of ComR and ComS, which vary stochastically among cells (Dubnau; Losick, 2006). Consequently the positive feedback behavior is heterogeneous and only a subpopulation flips the ComRS switch by expressing comX and comS above basal levels, ultimately resulting in a bimodal distribution of ComX in the population (Fig. 1).

How CSP activates this feedback loop remains unclear. The pathway through which cipB (or other upstream elements) stimulates ComRS has not been identified. In addition, it is not understood why CSP elicits a bimodal comX response only in complex growth medium that contains small peptides. A related question is why CSP elicits no comX response in defined media that lacks small peptides. We have suggested that small peptides imported from the growth medium could strengthen ComRS feedback – and thus favor comX activation – by competing for an S. mutans enzyme that degrades endogenously produced ComS (Son, M. et al., 2012; Hagen; Son, 2017). However, no such enzyme has yet been identified. An additional question is what limits the proportion of cells that activate comX in the presence of CSP. Although the number of comX-active cells initially increases as the concentration of exogenous CSP is increased, the proportion of cells that activate comX expression saturates at an upper limit of 30–35% (Son, M. et al., 2012).

A previous study (Moye et al., 2016) identified carbon source as one of the few parameters – other than complex/defined medium - that alters the proportion of cells activating the ComRS switch, i.e. the probability of transition from comX “OFF” to “ON”. Single-cell studies found that CSP induced higher ON fractions in fructose- or trehalose-grown cultures than in glucose-grown cultures (Moye et al., 2016), with trehalose giving the strongest response. In addition, CSP-stimulated cells growing in maltose or sucrose expressed a PcomX-lacZ (comX promoter fusion) reporter at higher levels than did glucose-grown cells. In addition, growth on trehalose or sucrose led to higher transformation efficiencies than did growth on glucose. Deletion of the gene for the global carbon catabolite repression (CCR) mediator CcpA eliminated differences in CSP response between sugars (Moye et al., 2016). Moye et. al. (Moye et al., 2016) did not rule out the possibility that carbohydrate influences comDE or comC expression. However, the fact that cipB expression already saturates at moderate CSP levels, and that higher levels of CSP do not increase the fraction of responding cells (Son, M. et al., 2012), suggests that the enhanced comX response in media formulated with carbohydrates other than glucose is not due to upregulation of comCDE.

Given the link between bimodality and carbohydrate source (Moye et al., 2016), and our hypothesis that a peptidase could govern bimodality by modulating the strength of feedback in the ComRS system (Hagen; Son, 2017), we investigated whether S. mutans PepO – whose homologs are known to interact with Rgg signaling in other streptococci (Wilkening et al., 2016) – could mediate the carbohydrate effect in S. mutans by interfering with ComRS autoactivation. We studied the relationship between carbohydrate source and activation of the CSP-induced competence pathway, using fluorescent gene reporter studies of individual cells, and biochemical and transcriptional approaches. By showing a clear connection between pepO and bimodality in ComRS, and demonstrating carbohydrate-sensitive effects on ComR regulation, our studies identify two mechanisms by which S. mutans controls the proportion of cells that enter the competent state.

Results

Trehalose enhances comX, but not cipB, activation by CSP

To confirm a previous report (Moye et al., 2016) that growth in trehalose leads to a greater proportion of comX-ON cells in response to CSP than does growth in glucose, we measured the proportion of comX-ON cells as a function of the glucose and trehalose content of the growth medium. We grew a PcomX-gfp reporter strain of S. mutans in complex growth medium (TV, Methods) containing mixtures of glucose and trehalose. This was performed on a wild type and a ΔtreR mutant, which does not activate the tre operon (Baker et al., 2018) depicted in Fig. 2A. Ratios of glucose to trehalose were chosen to maintain a constant hexose concentration of 20 mM, such that 2[tre]+[glc] = 20 mM (Moye et al., 2016). We added 1 µM CSP-18 to each sample as it reached OD₆₀₀ 0.1. The fluorescent reporter activity of individual cells was measured and histogrammed as described in (Kwak et al., 2012). Growth of the two strains was checked to ensure these effects were not due to poor growth of the mutant on trehalose admixtures; the ∆treR strain grows poorly on trehalose but normally in the mixed carbohydrate media (Supplemental Figure S1). Fig. 2B shows that the presence of 0.5 to 1 mM trehalose was sufficient to increase the proportion of comX-ON cells. Repeating the experiment using the same comX reporter in a ∆treR strain showed no effect of trehalose on the proportion of comX-ON cells. As a ΔtreR strain lacks the TreR transcriptional activator and does not express treAB (Baker et al., 2018), these findings confirm that the trehalose-induced increase in the proportion of cells responding to CSP requires activation of the tre operon.

Fig. 2: Replacement of glucose by trehalose produces graded increase in percentage of cells responding to CSP.

(A) Organization of the trehalose utilization operon and surrounding genes in S. mutans UA159. (B) Effect on comX activation of replacing glucose with trehalose as the carbon source. Carbohydrate composition of the TV medium was adjusted subject to the constraint 2[tre] + [glc] = 20 mM. In each sample 1 µM CSP was added at OD₆₀₀ 0.1, and cell responses were measured by fluorescence microscopy. Data show the percentage of cells expressing comX vs. [CSP] in the wild type reporting background (blue) and a treR mutant (mutant that does not express the tre operon, red). Solid curves of corresponding colors represent n = 1 Hill function fits to the curves. (C) Median fluorescence of PcipB-gfp activity in glucose- (blue) and trehalose- (red) grown cells. At least 600 individual cells, and typically 1000 or more cells, were studied under each experimental condition shown.

The activation of cipB transcription by phosphorylated ComE is an early step in the CSP-induced competence pathway (Son, Minjun et al., 2015). We next tested whether alterations in cipB regulation, required for CSP stimulation (Perry et al., 2009), could be triggered by carbohydrate. In order to test whether carbon source affects the competence pathway by influencing cipB or comCDE activity, we studied the effect of glucose/trehalose on the CSP response of a PcipB-gfp reporter strain. Fig. 2C shows that induction of cipB transcription by CSP was not significantly different in cells growing on trehalose compared to glucose. Fitting the median (in the cell population) cipB expression level to a simple binding isotherm (Hill function with n = 1), we find indistinguishable constants K = 2.6 ± 1.4 nM for glucose and 3.0 ± 0.9 nM for trehalose. These data indicate that cipB is not differentially regulated in trehalose versus glucose. The carbohydrate effect on comX expression must arise elsewhere than in the CSP-ComDE circuit.

The carbohydrate effect on transformation efficiency is CSP-dependent

Because S. mutans maintains a low level of transformability even in the absence of CSP (Perry et al., 2009), it was necessary to test whether the carbohydrate effect on transformability occurs through the CSP-induction pathway. We compared the transformation efficiency of wild-type UA159, a pepO mutant and a comDE mutant, grown in TV that contained mixtures of glucose and trehalose, with or without CSP. Fig. 3 shows that in the absence of CSP, the efficiency of transformation of wild-type cells is only slightly lower in glucose than in trehalose (P = 0.134 by two-tailed, unpaired Student’s t test). When CSP is provided, the transformation efficiency of the wild type is significantly greater in trehalose than in glucose (P < 0.001). These data imply that the carbohydrate effect is facilitated, if not entirely generated, by the CSP induction pathway. Consistent with this interpretation, we find that the CSP dependent differences in transformability are absent in the comDE deletion, as are significant differences between the different carbohydrates.

Fig. 3: Transformation efficiency is affected by carbohydrate only in the presence of CSP.

Transformation efficiency (expressed as % transformants) in indicated genetic backgrounds in glucose (red), trehalose (green), glucose with 1 µM added CSP (blue) or trehalose with 1 µM added CSP (yellow). In each case n = 3 biological replicates were assayed and average transformation efficiency plotted, with bars representing standard deviations.

Fig. 3 also shows that deletion of pepO increases the transformation efficiency under all conditions, both in the presence and absence of CSP. In the absence of CSP the ΔpepO strain had a roughly 100-fold higher rate of transformation than the wild type in glucose (P < 0.001) and in trehalose (P = 0.01). With CSP the difference narrowed to 30-fold in glucose (P = 0.038) and 16-fold in trehalose (P = 0.064). The ΔpepO strain exhibited a higher statistical significance in the carbohydrate effect (glucose vs. trehalose) of the non-CSP transformability (P = 0.012) than UA159. Therefore, deletion of pepO generally enhances transformability, but it does not eliminate CSP-sensitivity or the effect of carbohydrate on transformability.

Carbohydrate-dependent increases in comR transcription in response to CSP

In order to investigate how certain carbohydrates could affect the ComRS bimodal switch compared to glucose, we used RT-qPCR to compare comR transcript levels (normalized to 16S rRNA) in cells grown in TV supplemented with glucose, trehalose or maltose, in the presence of 0 nM, 4 nM or 400 nM CSP. Fig. 4A shows that comS transcription was generally elevated in the presence of CSP, as expected if CSP drives competence by promoting comS transcriptional feedback (Hagen; Son, 2017; Underhill et al., 2018). The CSP enhancement of comS transcription was greater in the tested sugars than in glucose and was greater in the pepO deletion. Less expected was the finding (Fig. 4B) that addition of CSP to the wild-type strain in trehalose or maltose caused a modest but significant 2–3 fold increase (P < 0.001) in comR transcripts, which did not occur in glucose. A CSP enhancement of comR transcription was also seen in the ΔpepO strain in those carbohydrates. We note that Lemme et al. (Lemme et al., 2011) reported a similar, 1.4–1.8-fold upregulation of comR in cells that were treated with CSP in Todd-Hewitt/yeast extract medium. Curiously, we found no significant increase in comR transcripts when CSP was provided to wild-type cells growing in glucose. The finding that CSP leads to a modest upregulation of comR transcription, especially in the disaccharides tested here, suggests a mechanism by which CSP could stimulate the ComRS system and promote comX activity. As the bistable behavior of the ComRS feedback system will be sensitive to basal levels of both ComS and ComR, even a modest upregulation of comR by CSP, as occurs in trehalose and maltose, would promote positive feedback in the ComRS circuit, permitting a larger proportion of cells to enter the comX-ON state.

Fig. 4: RT-qPCR measurement of gene expression modulation by sugar and CSP.

RT-qPCR measurements of indicated transcripts normalized to 16S rRNA. Bars are grouped by strain and carbohydrate in color and correspond to 0, 4 nM and 400 nM added CSP from left to right within a color group. (A) comS expression. (B) comR expression, showing a significant change in comR transcripts in response to CSP in trehalose and maltose (P < 0.001). (C) pepO expression. (D) smu.2035 expression. In each case n = 3 biological replicates were assayed and averages plotted, with error propagated from data standard deviations plotted as the error in the mean.

Although our transformation assay (Fig. 3) demonstrated that deletion of pepO enhances com activity, Fig. 4C shows that pepO transcription was not affected by CSP in any carbohydrate tested. Therefore CSP does not stimulate comX by modulating pepO transcription. This finding is consistent with a model where CSP exerts a greater effect on comR while pepO acts through a separate mechanism to suppress competence activation.

As a test for possible effects of other genes proximal to the tre operon and pepO, we tested whether the transcription of smu.2035, which is located 143 bp downstream of pepO and transcribed in the opposite direction, was affected by CSP or carbohydrate. Fig. 4D shows that smu.2035 transcripts decreased slightly in the ΔpepO background relative to the wild type. This may be a consequence of insertion of the antibiotic resistance cassette in the pepO region (Methods). However, smu.2035 showed no particular response to CSP or carbohydrate in the wild type.

Deletion of pepO allows population-wide expression of comX

Figs. 3 and 4 show that deletion of pepO increased comS transcription and transformability. However, pepO transcription is not modulated by CSP or carbohydrate. The simplest interpretation of these data is that the endopeptidase PepO acts constitutively to limit the activation of the ComRS autofeedback loop. We tested this model by measuring the proportion of cells activating a PcomX-gfp reporter in the pepO deletion and in the wild-type background, in three different carbohydrates. Only a fraction (Fig. 5A, B) of wild-type cells became PcomX-active at saturating concentrations of CSP, where this fraction was greater in trehalose or maltose (47 ± 3 % and 38 ± 5 % responding, respectively) than in glucose (22 ± 9 % of cells responding). The deletion of pepO substantially enhanced (Fig. 5C) the level of activation observed at saturating concentrations of CSP in all three carbohydrates: in trehalose and maltose, the proportion of comX-active cells approaches 100%, while in glucose it exceeds 90% (Fig. 5D). Therefore the bimodal behavior is largely removed by the deletion of pepO and the competence pathway now responds unimodally to CSP. Within each strain, the binding parameter K (Hill function with n = 1, Supplemental Table S1) was roughly the same for all sugars (K = 11–14 nM), indicating that the sugar does not determine the CSP sensitivity threshold. However the K for the ∆pepO strain was 2–4 nM, which is roughly 2–3-fold lower than in UA159. Therefore, deletion of pepO increased overall sensitivity to CSP in all three carbohydrates, and eliminated the bimodal character of comX, allowing population-wide activation of comX. Complementation of the pepO deletion (Fig. 5E) reduced the comX-active proportion to the level of 30–40% similar to the wild type and with a similar K = 11 ± 3 nM, reverting behavior to bimodal (Fig. 5F). These data show that PepO is a major limiting factor in the activity of the ComRS feedback loop, such that pepO deletion allows population-wide comX expression in the presence of CSP.

Fig. 5: pepO deletion results in carbohydrate-dependent unimodal response to CSP.

Percentage of PcomX-gfp cells responding to CSP in TV medium supplemented with different sugars. Percentages are determined by comparing the area under the high and low-fluorescence peaks in the population distribution (see Methods). (A) Response of PcomX-gfp/UA159 in glucose (green), trehalose (blue) or maltose (red). (B) Histogram of GFP fluorescence for individual cells provided with 5 µM CSP in glucose, demonstrating bimodal response. (C) Response of PcomX-gfp ΔpepO strain in different carbohydrates, using same color code as in (A). (D) Single-cell histogram for PcomX-gfp ∆pepO strain in glucose with 5 µM CSP, showing near unimodal (>90%) activation by CSP in glucose. (E) Comparison of the above PcomX-gfp/UA159 (green) data and PcomX-gfp ΔpepO (blue) data to a PcomX-gfp ΔpepO pepO⁺ (complemented) strain (red), all grown in glucose. (F) Histogram of single-cell fluorescence of pepO complemented strain, grown in glucose with 5 µM CSP, showing restoration of the bimodal distribution seen in (B).

PepO is responsible for growth medium-dependent bimodal response to CSP

We have proposed a mechanism where the growth medium dependence of the comX response to CSP is due to intracellular XIP/ComS and small nutrient peptides from the media competing for degradation by an intracellular peptidase (Son, M. et al., 2012; Hagen; Son, 2017): In defined media that lack the peptides, the peptidase is available to degrade intracellular XIP/ComS, shutting down ComRS feedback and preventing comX activation. In complex media, the small peptides slow the degradation of XIP/ComS sufficiently to allow the ComRS feedback loop to autoactivate in some cells, leading to a population-bimodal comX response. If PepO plays the role of this hypothesized peptidase, then we would expect a pepO deletion strain to show an enhanced comX response to CSP in both complex and defined growth media. We therefore provided CSP to the ΔpepO strain in the defined medium FMC, supplemented with either glucose or trehalose. CSP normally elicits no response from comX in FMC medium (Son, M. et al., 2012). However, Fig. 6 shows that even in the absence of CSP the ΔpepO strain was bimodally activated in both trehalose-supplemented FMC and in FMC formulated with glucose as the sole carbohydrate. Further, as the CSP concentration was increased to 100–500 nM (in glucose, Fig. 6A) or to 5–10 nM (in trehalose, Fig. 6B), the comX response became unimodal (population-wide). Therefore deletion of pepO eliminated both the bimodality of the comX response to CSP and the requirement for complex growth medium.

Fig. 6: CSP-mediated activation of comX in a defined medium.

PcomX-gfp ΔpepO fluorescence response to indicated concentrations of added CSP in the defined medium FMC as measured by fluorescence microscopy in (A) glucose and (B) trehalose.

PepO degrades the ComS-derived signal XIP in vitro

We have previously shown (Underhill et al., 2018) that the bimodal response of ComRS and comX to CSP arises within an intracellular feedback loop in which endogenously produced ComS interacts with ComR to drive comS and comX transcription. The above data support the interpretation (Son, M. et al., 2012) that the growth medium dependence of CSP response is due to a peptidase, evidently PepO, which suppresses autoactivation of the ComRS system by degrading the endogenous ComS feedback signal. In order to confirm that PepO can break down ComS/XIP and prevent its interaction with ComR to drive ComRS transcriptional feedback, we tested whether recombinant PepO (rPepO) from S. mutans affected the ability of synthetic XIP to form a DNA-binding complex with ComR in vitro. Fig. 7A shows a fluorescence polarization (FP) assay in which purified ComR binds a fluorescently labeled DNA probe containing the comX promoter region in the presence of 5 µM synthetic XIP that had been incubated with 500 nM rPepO for different lengths of time. A loss of polarization was observed when XIP was incubated for more than about 20 minutes, indicating a loss of XIP-induced binding of ComR to the DNA probe. A greater loss of polarization occurred at higher rPepO concentrations. Fig. 7B shows that 2 h incubation of rPepO with XIP had little effect on the FP signal if rPepO was present at concentrations below about 30 nM, but polarization declined substantially for [rPepO] greater than about 100 nM. Fig. 7C shows the FP signal for XIP that was incubated for 5 h at the same rPepO concentrations as Fig. 7B.

Fig. 7: PepO degrades XIP in vitro.

Effect of rPepO protein on synthetic XIP. (A) Effect of XIP incubation time with rPepO on a fluorescence polarization assay for XIP/ComR binding to the comX promoter. 500 nM rPepO was added to 5 µM synthetic XIP, and fluorescence polarization measurements were then taken after the indicated incubation times, using a fluorescently-labeled DNA aptamer and 1 µM purified recombinant ComR. (B), (C) Different concentrations of rPepO added to 5 µM synthetic XIP and fluorescence polarization measurements similar to (A) taken at 2 and 5 hours of incubation respectively. (D) Fluorescence response of PcomX-gfp ΔcomS cells to addition of rPepO-treated XIP added to culture in 1:4 dilution after 5 hours of rPepO treatment.

As an additional test of rPepO degradation of XIP, we tested whether treatment with rPepO affected the ability of synthetic XIP to activate comX in a reporting strain of S. mutans. Fig. 7D shows fluorescence of a bulk culture of PcomX-gfp ΔcomS cells (incapable of producing their own ComS or XIP) that were provided with 1 µM XIP that had been incubated with rPepO for 5 hours. The reporter fluorescence of the cells shows a decline very similar to Fig. 7C as the rPepO concentration is increased, confirming that rPepO degraded the ability of the XIP to activate comX.

Discussion

Genetic competence in S. mutans is influenced by a diverse set of environmental factors, including peptide content of the medium, pH, oxidative stress and heat (Senadheera, M. Dilani et al., 2005; Ahn et al., 2006; Tremblay et al., 2009; Okinaga et al., 2010; Son, M. et al., 2012; Guo et al., 2014; De Furio et al., 2017). For S. mutans, the peptide content of the growth medium determines the proportion of cells that activate comX in response to CSP. The response ranges from zero (in defined medium, lacking small peptides) to partial (bimodal, in complex media). Only direct addition of the inducing peptide XIP, which when taken up by the Opp oligopeptide permease in defined media interacts directly with ComR to induce comX, elicits a population-wide (unimodal) comX response in wild-type S. mutans. Moye et al. (Moye et al., 2016) showed that carbon source is an additional modulator of the proportion of cells responding to CSP by activating comX. In particular, growth in the presence of the disaccharide trehalose not only increases the proportion of S. mutans expressing comX, but also increases transformation efficiency. The trehalose catabolic operon is directly upstream of, and transcribed in the same direction as, the endopeptidase pepO. In investigating the link between carbohydrate source and bimodal activation of comX, we initially hypothesized that the proximity of pepO and the treAB operon may allow induction of treAB to modify levels of the peptidase. This could lead to higher intracellular ComS, encouraging autoactivation of the bistable ComRS feedback circuit that regulates comX (Fig. 1) (Son, M. et al., 2012). Although our findings confirm the trehalose effect originally observed by Moye et al. (Moye et al., 2016) and demonstrate for the first time that PepO plays a key role in modulating bimodal competence response, the body of our data shows that carbohydrate source and PepO influence the regulatory pathway through largely independent mechanisms.

It was necessary to determine whether the trehalose effect originates with the bacteriocin genes because the trehalose operon has been linked to bacteriocin expression (Baker et al., 2018). Our data rule out a role for cipB or upstream elements such as comCDE in the effect of trehalose and other carbohydrates tested, because CSP activation of cipB was unaffected by carbohydrate source. The CSP concentration required to saturate the cipB response, and the maximal level of cipB expression, was similar in glucose and trehalose.

Figure 8A shows further evidence that carbohydrate affects the competence pathway downstream of cipB. Expression of cipB reaches maximum at a lower CSP concentration (near 5 nM CSP) than does expression of the PcomX-gfp reporter (near 100 nM CSP). Therefore, even though cipB is required to elicit the comX response in glucose (Perry et al., 2009), trehalose and maltose (Supplemental Figure S2), full activation of cipB is insufficient to induce maximum comX response. Evidently, transmission of the competence signal beyond cipB involves an additional mechanism that requires a higher threshold of CSP, such as activation of an additional gene that is subject to comDE regulation. The carbohydrate source does not affect the threshold CSP concentration (constant K in Table S1) needed to saturate either cipB or comX expression. Therefore, the carbohydrate effect on comX appears to arise in an additional mechanism unlinked to cipB, whereby CSP leads to upregulation of comR transcription, and where glucose appears to inhibit or interfere with this mechanism.

Fig. 8: Integration of carbohydrate effects into existing competence regulation models.

(A) PcipB-gfp and PcomX-gfp expression data from Figs. 2 and 5 (glucose, wild type) on a double y-axis plot showing the difference in [CSP] threshold between comX and cipB. (B) Model for the influence of carbohydrate on the comX-activating fraction of the population, in response to CSP. An unknown mechanism Z increases comR transcription under stimulation by CSP, but is repressed by glucose, such that comR is only significantly upregulated in the other carbohydrates tested. A parallel pathway Y is also postulated, through which phosphorylated ComE is able to stimulate comS via expression of cipB. The response of comS to CSP stimulation requires the autofeedback amplification of fluctuations in [ComS], which is repressed by PepO-mediated degradation, limiting the proportion of the population that activates comX.

Our data indicate that PepO affects the competence pathway in a different way than carbohydrate, most likely by degrading basally produced ComS and XIP. This action inhibits the transcriptional feedback that amplifies fluctuations in ComS levels into full activation of ComRS and induction of comX. If the effect of comCDE stimulation is to increase comS or comR transcription, then it will help to overcome the suppressive effect of PepO and permit, in at least some cells, self-activation of the feedback loop and flipping of the ComRS switch to its ON state. Consistent with this model, our transcriptional data suggest that in the tested carbohydrates (but not in glucose), CSP leads to some upregulation of comR. Although comR transcripts increase only be a modest 2- to 3-fold, bistable autofeedback systems amplify small fluctuations (Dubnau; Losick, 2006) and a small rise in ComR levels could very plausibly flip some cells from the comX-OFF to the comX-ON state.

PepO is very strongly implicated in the role of feedback inhibitor because its deletion drastically enhanced transformability and triggered population-wide expression of comX. A robust, unimodal comX response to CSP, which has not otherwise been observed in S. mutans, is then observed even in defined media, where CSP normally elicits no response from comX whatsoever (Son, M. et al., 2012) regardless of carbohydrate source (Ricomini et al., 2019). Degradation of ComS/XIP by PepO is also consistent with observations such as the degradation of the Rgg2 and three other small peptides by PepO in Streptococcus pyogenes (Wilkening et al., 2016). As PepO amino acid sequences are ~90% similar across Streptococcus spp. (Nguyen et al., 2009), the relatively non-specific degradation of small signal peptides seen in (Alves et al., 2017) is potentially a conserved property of PepO peptidase. Our in vitro data support this interpretation by showing that pre-treatment of XIP or ComS with PepO inhibits ComR binding to its cognate target, apparently by degrading exogenously added XIP. We also showed that the effect of PepO on ComR/XIP-dependent activation of comX can be reproduced in vivo where treatment of XIP with rPepO prior to addition to ComS-deficient cells in defined medium eliminates the ability of XIP to activate comX in cells.

It appears unlikely that PepO plays a direct role in the carbohydrate effect, as pepO expression was not modulated by CSP or carbohydrate source. PepO is not controlled through the ComDE circuit and instead acts independently to suppress autoactivation of ComRS. The model presented in Fig. 8B therefore proposes that PepO acts constitutively to degrade endogenously produced ComS and nutritional peptides from the medium. The model also includes a parallel, CSP-dependent pathway – denoted Z – that stimulates comR transcription in a CSP-dependent manner, but is sensitive to carbohydrate source. The efficiency with which CSP is capable of activating comR is thus controlled by the sensitivity of Z to CSP and by carbohydrate source, as our RT-qPCR data indicate in Fig. 4D.

We note that prior transcriptional studies have disagreed on whether or not CSP increases ComR levels (Lemme et al., 2011; Reck et al., 2015; Moye et al., 2016). Our data and model resolve the apparent disagreement, inasmuch as comR upregulation was not observed when CSP was provided in glucose-supplemented chemically defined medium (Reck et al., 2015), but a 1.4- to 1.8-fold upregulation was detected in THB-Y medium (Lemme et al., 2011), which contains other carbohydrates, in addition to glucose, that are presumably able to trigger the carbohydrate-sensitive pathway.

Our data still leave unanswered the question of how CSP stimulates ComRS when glucose is the sole carbohydrate source and ComR levels are unaffected by CSP. Here, it may be relevant to observe that although the pepO mutant growing in the presence of CSP and glucose activates comX far more robustly than in a wild-type genetic background (Fig. 5C and Fig. 6A), this response still requires a small (5–10 nM) concentration of CSP. Even in the absence of PepO, the ComRS feedback system is still weakly repressed, although only modest amounts of CSP are needed to overcome this repression. Similarly, comX does not respond to CSP when a cipB deletion strain grows in the disaccharides tested (Supplemental Fig. S1). All of these results indicate that cipB controls a pathway (denoted Y in Fig. 8B) that operates in all growth media to limit ComRS activation, independently of Z. It is conceivable for example that an additional peptidase (other than PepO) weakly degrades endogenously produced ComS/XIP, and that the cipB pathway acts to downregulate this peptidase. Such a model is depicted in Fig. 8B, where CSP has two parallel effects on ComRS: it downregulates the second protease (pathway Y) while also upregulating comR (pathway Z). This model predicts that in a pepO deletion strain in glucose-containing media (Z not active), the CSP threshold for the comX response will be similar to that for cipB activation – as occurs in our data (Fig. 2 and Fig. 5). However, for UA159 growing in other carbohydrates, the CSP level that is needed to induce such a pathway will be difficult to predict, owing to the two parallel routes combined with the internal positive feedback, which is present both in ComRS (Son, M. et al., 2012; Underhill et al., 2018) and in regulation of comDE by ComX (Son, Minjun et al., 2015; Reck et al., 2015).

Finally, our results allow some speculation about how the VicRKX sensory system, which is hypothesized to respond to oxidative stress (De Furio et al., 2017), links to the competence pathway. A binding site for the VicR response regulator has been identified in the pepO promoter region (Alves et al., 2017), where VicR has been demonstrated to act as a transcriptional repressor (Senadheera, D. B. et al., 2012). Thus, the vicK deletion results in higher pepO expression (Alves et al., 2017), which should give rise (under our model) to reduced transformability in these mutants. A prior study showed in fact that deletion of the vicK kinase reduces transformability, despite increasing the production of bacteriocins and comCDE mRNA (Senadheera, M. Dilani et al., 2005). A similar pattern is visible in data showing that deleting clpP raises pepO expression while concurrently lowering comR and comX expression (Kajfasz et al., 2011).

Experimental Procedures

Strains and growth conditions

S. mutans wild-type strain UA159 and mutant strains from glycerol freezer stocks were grown in BBL BHI (Becton, Dickinson and co.) at 37˚C in 5% CO₂ overnight. E. coli were grown from glycerol freezer stocks in LB at 37˚C shaking overnight. Antibiotics were used at the following concentrations where resistance is indicated in Table 1: erythromycin (10 µg ml⁻¹), kanamycin (1 mg ml⁻¹), spectinomycin (1 mg ml⁻¹), ampicillin (10 µg ml⁻¹). For all experiments, strains were washed twice by centrifugation, removal of supernatant fluids and re-suspension in phosphate buffered saline (PBS), pH 7.2. Cells were then diluted 20-fold into fresh medium and allowed to grow in the same incubator conditions until OD₆₀₀ reached 0.1. Synthetic CSP-18 (sequence SGSLSTFFRLFNRSFTQA) was purified to 98% purity and provided by NeoBioSci (Cambridge, MA, USA).

Table 1: Strains and plasmids used.

List of strains and plasmids used in this study, their relevant characteristics, and the source or reference for the material.

Strain or plasmid	Characteristics*	Source or reference
S. mutans strains
PcomX-gfp (plasmid)	UA159 harboring PcomX-gfp promoter fusion on pDL278, SpR	(Son, M. et al., 2012)
ΔpepO	pepO gene replaced with NP resistance cassette, KmR	This study.
PcomX-gfp ΔpepO	ΔpepO harboring PcomX-gfp promoter fusion on pDL278, SpR KmR	This study.
ΔpepO pepO+	ΔpepO (EmR deletion) with pepO complemented (KmR)	(Alves et al., 2017)
PcomX-gfp ΔpepO pepO+	ΔpepO pepO+ harboring PcomX-gfp promoter fusion on pDL278, SpR KmR EmR	This study.
PcipB-gfp	UA159 harboring PcipB-gfp promoter fusion on pDL278, SpR	(Son, Minjun et al., 2015)
ΔcomDE	comDE replaced with NP resistance cassette, SpR	(Shields et al., 2017)
E. coli strains
BL21(DE3)	Used for recombinant protein expression	New England Biolabs, MA
Plasmids
pDL278	E. coli – Streptococcus shuttle vector, SpR	(LeBlanc et al., 1992)
pIB184	Shuttle expression plasmid with the P23 constitutive promoter, EmR	(Biswas et al., 2008)
pET45b(+)his-comRUA159	pET45b(+) derivative containing the translational fusion PT7lac-6xhis-comRUA159, Apr	(Underhill et al., 2018)

^†Em = erythromycin, Sp = spectinomycin, Km = kanamycin, Ap = ampicillin

Construction of pepO deletion mutant

The pepO gene was replaced by a non-polar kanamycin cassette in S. mutans strain UA159 by homologous recombination. PCR primers (Table 2) with ends containing a BamHI recognition site were used to amplify the flanking regions of the gene. Ends were digested with BamHI to ligate the flanking region product to the kanamycin cassette. The resulting linear DNA was transformed into UA159 in which competence was induced by XIP in the defined medium FMC (Terleckyj et al., 1975). The transformants were confirmed by PCR and Sanger sequencing to ensure that the pepO gene was deleted and the sequences flanking pepO that were used for the recombination event were intact. Previously described fluorescent protein reporter fusion constructs (Son, M. et al., 2012) were transformed as needed to generate strains for use in experiments.

Table 2: Oligonucleotides used.

Primers used for this study, including qPCR primers. Restriction enzyme sites are underlined.

Purpose	Sequence (restriction enzyme sites underlined)
pepO deletion upstream	ATGCCAGACAGTAGCGATTTTAGCA
BamHI - pepO deletion upstream	ATGCGGATCCGTGGTTTATCATCAGGAATGAC
BamHI - pepO deletion downstream	ATGCGGATCCTCCACCAAGAATTTGCGGTTA
pepO deletion downstream	ATGCTATCGGCGCTAAGGTCACTAT
pepO deletion sequencing 1	ATGCTGCTGAAACAGTTGAGCTAGA
pepO deletion sequencing 2	ATGCTCATCCTTGATAAACATCCTGTTCTAA
pepO deletion sequencing 3	ATGCTGTTACAATAGAAAGCGT
pepO deletion sequencing 4	ATGCAAACTAGATATCTACCAAATAATAACA
16S fwd qPCR	CCTACGGGAGGCAGCAGTAG
16S rev qPCR	CAACAGAGCTTTACGATCCGAAA
comR fwd qPCR	TATTACGAAGGCCAACCTAT
comR rev qPCR	TTCTTCTTCAGGCAAATCAT
comS fwd qPCR	TCAAAAAGAAAGGAGAATAACA
comS rev qPCR	TCATCGGAGATAAGGGCTGT
pepO fwd qPCR	TTGCTTCCAACATAGCCAC
pepO rev qPCR	TTACCTTTGTCATTACCTTCAGC
Smu.2035 fwd qPCR	TTCTGATATCCATGACGCTT
Smu.2035 rev qPCR	AGAGCCGTTGTATCTGAATA

Development of buffered TV medium

The pH of tryptone-vitamin (TV) medium was found to be approximately 6.7. Due to the influence of pH on the com regulon (Guo et al., 2014; Son, M. et al., 2015; Son, M. et al., 2015), a medium buffered at a pH close to 7 was desired. A phosphate buffer (170 mM KH₂PO₄, 720 mM K₂HPO₄) was diluted into TV medium 100-, 80-, 50- or 10-fold and cells from overnight cultures were diluted 20-fold into the differently buffered media. The resulting cell suspensions plus an unbuffered control were put into a well plate (Falcon 24 well plate, Corning inc.) and growth was monitored by measuring the OD₆₀₀ every 5 minutes in a Biotek Synergy 2 plate reader (Biotek Instruments, inc.). It was found that the 80-fold dilution was the highest concentration of buffer that did not inhibit growth of S. mutans (Supplemental Figure S3) so medium buffered in this was used for all carbohydrate experiments involving TV. The initial pH of the buffered TV was 7.2, which is known to be permissive for competence signaling by CSP (Guo et al., 2014; Son, M. et al., 2015).

Single cell experiments

For single-cell experiments involving planktonic growth in different carbohydrates in the presence of CSP, a chromosomally integrated PcomX-gfp reporter strain was used. Cells were diluted 20-fold from overnight cultures into buffered TV supplemented with the desired carbohydrate(s) at final concentrations of 20 mM for monosaccharides and 10 mM for disaccharides. When the cells reached an OD₆₀₀ of 0.1, 1 µM CSP-18 was added and the cultures were incubated for 2 h. At this point, cultures were gently sonicated using a Fisher Scientific FB120 sonic dismembrator probe to break up long chains and pipetted onto a glass coverslip. Phase contrast and fluorescence imaging were performed at 60x magnification on a Nikon TE2000U phase contrast microscope followed by image analysis, as described previously (Kwak et al., 2012). We typically measured about 1000 cells, or as many as 3000 cells, under each experimental condition. However only 200–300 cells were studied at the highest CSP concentrations in Figs. 5 and 6, where growth is strongly inhibited by the activation of the competence pathway. The percentage of cells deemed to be expressing comX was determined using a two-distribution fit to the bimodal data (see below).

Concurrent monitoring of growth and fluorescence

Well plate experiments were carried out in a Falcon 96 well plate with clear bottom and black side (Corning, Inc.). 200 µL of culture was pipetted into each well and covered with mineral oil to prevent evaporation and oxygen diffusion. Fluorescence and OD₆₀₀ were read in a Biotek Synergy 2 plate reader (Biotek Instruments).

Transformation efficiency assay

Cells were prepared for each transformation efficiency assay from overnight cultures as for the CSP experiments described above. DNA used for transformation was plasmid pIB184, carrying an erythromycin (Em) resistance marker, added at a concentration of 600 ng ml^-1. At OD₆₀₀ of 0.1, DNA was added to the cells and tubes were mixed by inversion; at this point CSP was added where used. After 4 hours of exposure to DNA, cells without CSP (or comDE mutants in all cases, as these do not respond to CSP (Perry et al., 2009)) were diluted and plated on BHI supplemented with 10 µg ml⁻¹ erythromycin. A no-DNA control from each sugar was plated on BHI-erythromycin in order to verify that no spontaneously resistant variants arose during the incubation. Total viable cell counts were obtained by plating the diluted cultures onto BHI agar with no antibiotics. Efficiency was calculated by dividing the number of transformants per microliter plated by the total viable cell count per microliter plated, correcting for dilution and concentration factors appropriately. Statistical significance was assessed using the two-tailed, unpaired Student’s t test.

Calculating the comX-active proportion of cells

The double peaked (bimodal) distribution of PcomX activity under CSP stimulation was found empirically to be well characterized as a combination of two well-separated gamma distribution functions (Friedman et al., 2006; Taniguchi et al., 2010; Taniguchi et al., 2010) (as in Figure 1), corresponding to comX-ON and OFF cells respectively. For cells responding bimodally to CSP, the histogram P(F) of PcomX fluorescence F can then be fit to obtain a parameter λ (0 ≤ λ ≤ 1) equal to the proportion of cells in which comX is active:

$$P(F)=(1-\lambda )H(F|a,b)+\lambda G(F|c,d)$$

Here H is the gamma distribution with shape parameters a and b, describing the OFF cells, and G is the gamma distribution with parameters c and d, describing the ON cells. In experiments when PcomX activity was nearly unimodal (single peaked), this fit did not determine all parameters robustly, and so the comX-active proportion λ was found by a cutoff method, based on counting the proportion of cells whose fluorescence F exceeds a cutoff value F_c. We chose the cutoff F_c by identifying the value that, for a comparable dataset where PcomX activity is strongly bimodal, minimizes the probability that a simple cutoff wrongly assigns a cell to either the ON or OFF distribution. That is, the fluorescence cutoff F_c was chosen to minimize

$$P(error)= {\int }_{F_c}^{\infty }HdF+ {\int }_0^{F_c}GdF$$

Hill functions were fit to data exhibiting saturating behavior using a Hill function model with n set to 1 and hence two free parameters, the saturating height and K. Fits were performed by the chi squared nonlinear fitting method and error in parameters estimated by a 1000 iteration bootstrap for each data set.

In vitro degradation of XIP by PepO

Purified recombinant PepO (rPepO) was a kind gift of Livia Alves and Dr. Jacqueline Abranches of the University of Florida College of Dentistry. Purified protein was frozen in PBS pH 7.2 with 10% glycerol. The concentration of rPepO in solution was estimated from its absorbance at 280 nm. XIP was treated with rPepO by diluting XIP from 10 mM phosphate buffer at pH 7.0 (5.8 mM K₂HPO₄, 4.2 mM KH₂PO₄) with indicated concentrations of rPepO. The mixtures were then incubated at 37˚C for 5 hours before addition to cells.

To detect how much XIP was left, the rPepO-XIP mixtures were added to PcomX-gfp ΔcomS plasmid-based reporter cells at OD₆₀₀ 0.1 in a ratio of 1:4 (such that the XIP mixture was 5-fold diluted into the cell culture) in a 96-well plate and OD₆₀₀ and green fluorescence was recorded every 5 min. The GFP fluorescence was normalized to OD₆₀₀ and averaged over the same time period for each sample corresponding to the time between onset and decline of the fluorescence peak. The data were plotted using the standard deviation of F/OD₆₀₀ values over this time as error bars.

Fluorescence polarization using rPepO-treated XIP and ComS

Native ComR was purified by previously reported methods (Underhill et al., 2018). Briefly, E. coli BL21 (DE3) cells containing a plasmid harboring N-terminally tagged 6x His-ComR were lysed and the ComR purified by Ni-NTA affinity chromatography. The histidine tag was cleaved using EKMax enterokinase (Invitrogen), and the resulting ComR dialyzed into PBS pH 7.4. Concentration was estimated using the Pierce bicinchonic acid (BCA) assay (Thermo Scientific). The purified protein was used with the same Bodipy-FL-X labeled fluorescent DNA aptamer and binding assay buffer as described elsewhere (Underhill et al., 2018) to assess the ability of rPepO-treated XIP to induce ComR binding of the PcomX DNA region.

XIP (5 µM) was treated with the indicated concentrations of rPepO in phosphate buffer. After 2 and 5 hours, 40 µL of the XIP-rPepO solutions were pipetted into a well containing 160 µL of binding buffer, 1 nM fluorescent DNA and sufficient ComR to obtain 1 µM final concentration in the 200 µL volume. Fluorescence polarization was then measured by a Biotek Synergy 2 plate reader (Biotek Instruments inc.) in a 96-well black-bottomed, black-side assay plate in polarization mode using a 485 nm excitation filter and 528 nm emission filter. The same experiment was then performed using 5 µM XIP and 500 nM rPepO and taking polarization measurements at the indicated time points.

Reverse transcriptase-quantitative PCR (RT-qPCR)

Cells for RT-qPCR were grown to OD₆₀₀ = 0.1 in TV supplemented with the indicated sugar from 20-fold dilution from washed overnight cultures. At this point, CSP was added to samples and the cultures were incubated for 2 h before centrifugation and removal of the TV. One ml of TRIzol reagent (ThermoFisher) was then added, the pellet was resuspended, and cells were mechanically lysed in a bead beater in the presence of 100 µm glass beads. Extraction of RNA was then performed following the TRIzol phenol:chloroform method (Chomczynski, 1993) and the resulting RNA sample treated with the Turbo DNA-free™ kit from ThermoFisher.

RNA concentration was estimated using absorbance at 260 nm and 1 μg was reverse transcribed using iScript reverse transcription mix (Biorad) containing random primers. Resulting cDNA was diluted 50-fold in water and used as the basis for qPCR reactions. qPCR was performed using iTaq™ SYBR Green Supermix (Biorad) in a Biorad CFX Connect thermal cycler. Transcript counts for genes of interest were normalized to a count of 16S rRNA for the same volume of sample. Three biological replicates of each condition were grown and three independent technical replicates assayed for each of these. Resulting values of transcript count divided by 16S rRNA count were averaged to compute reported values. Standard deviations of mRNA counts in technical replicates were propagated forward to the calculated quotient and the computed error used to represent error bars.

CSP activity in defined medium

To examine the influence of CSP on comX expression in a pepO mutant in defined medium, FMC, was supplemented with the indicated sugar (20 mM for glucose, 10 mM for trehalose) (Terleckyj et al., 1975). The experiment was otherwise performed exactly as the single cell experiments in TV medium.