|Home | About | Journals | Submit | Contact Us | Français|
In response to limiting nutrient sources and cell density signals, Bacillus subtilis can differentiate and form highly resistant endospores. Initiation of spore development is governed by the master regulator Spo0A, which is activated by phosphorylation via a multicomponent phosphorelay. Interestingly, only part of a clonal population will enter this developmental pathway, a phenomenon known as sporulation bistability or sporulation heterogeneity. How sporulation heterogeneity is established is largely unknown. To investigate the origins of sporulation heterogeneity, we constructed promoter-green fluorescent protein (GFP) fusions to the main phosphorelay genes and perturbed their expression levels. Using time-lapse fluorescence microscopy and flow cytometry, we showed that expression of the phosphorelay genes is distributed in a unimodal manner. However, single-cell trajectories revealed that phosphorelay gene expression is highly dynamic or “heterochronic” between individual cells and that stochasticity of phosphorelay gene transcription might be an important regulatory mechanism for sporulation heterogeneity. Furthermore, we showed that artificial induction or depletion of the phosphorelay phosphate flow results in loss of sporulation heterogeneity. Our data suggest that sporulation heterogeneity originates from highly dynamic and variable gene activity of the phosphorelay components, resulting in large cell-to-cell variability with regard to phosphate input into the system. These transcriptional and posttranslational differences in phosphorelay activity appear to be sufficient to generate a heterogeneous sporulation signal without the need of the positive-feedback loop established by the sigma factor SigH.
When nutrient sources are dwindling, the Gram-positive bacterium Bacillus subtilis can utilize a number of adaptive phenotypes such as the secretion of proteases and the development of genetic competence (for a recent review, see reference 49). The most sophisticated survival strategy that B. subtilis employs is the formation of a highly resistant endospore (34). Initiation of spore formation is governed by a complex multicomponent phosphorelay (Fig. (Fig.1)1) including the primary kinases KinA and KinB and two intermediate phosphotransferases, Spo0B and Spo0F (33). The phosphorelay activates Spo0A, the master sporulation regulator of sporulation, by phosphorylation (4), upon which phosphorylated Spo0A (Spo0A~P) can directly regulate more than 100 genes (29). This induces a chain of events that takes several hours to complete and involves the activation of a set of alternative sigma factors that give unidirectionality to the cascade, culminating in the formation of the endospore (10).
Interestingly, within isogenic populations of B. subtilis grown under identical conditions, not all cells enter this timely and costly developmental pathway. This type of phenotypic bifurcation is known as sporulation bistability or sporulation heterogeneity (9, 38). How populations of genetically identical cells bifurcate into phenotypically distinct subpopulations in the same environment is an important question for developmental biology. Competence development is another example of phenotypic bistability in B. subtilis (38). Positive feedback of the master competence regulator ComK on its own expression has been shown to be essential and sufficient for competence bistability (28, 37). For the initiation of sporulation, the origins of its heterogeneity seem to be more complex: although Spo0A~P also activates its own transcription, this positive-feedback loop is not essential for heterogeneous initiation of sporulation (50). Several lines of investigation indicate that the activity of the phosphorelay (the phosphorelay phosphate charge) determines the fraction of cells that commit to spore formation (8, 14, 15, 46, 50). However, whether sporulation heterogeneity originates from cell-to-cell variation in the phosphate charge and/or expression levels of the phosphorelay components is unknown. Another major question is whether sporulation heterogeneity originates from one or several (and if so, which) genes/proteins in the system. To address these questions, we constructed promoter-green fluorescent protein (GFP) fusions of the genes coding for the core phosphorelay components and PsigH so that their expression could be followed in real time at the single-cell level. The expression of each individual phosphorelay gene was correlated with sporulation, and the impact of phosphorelay gene expression on sporulation heterogeneity was tested by systematically perturbing the system using knockouts and overproduction strains.
With time-lapse microscopy and flow cytometry experiments, we show that expression of the phosphorelay genes is distributed not bimodally but unimodally within the population. However, transcription of several phosphorelay genes appears to be dynamic and variable in time (heterochronic), and we found a correlation between the temporal gene activation and whether a cell sporulates or not. In addition, we provide data suggesting that these characteristics of phosphorelay gene expression result in cell-to-cell variability of the phosphorelay phosphate flow and thus sporulation heterogeneity. Finally, our results indicate that the fluctuations in the complex sporulation signal transduction cascade are sufficient for establishing a heterogeneous sporulation signal, without the need for specific positive-feedback loops.
Oligonucleotides, plasmids and bacterial strains are listed in Tables Tables1,1, ,2,2, and and3,3, respectively. Bacillus subtilis 168 trp+ was grown in TY, TLM, or 15% CDM at 37°C or 30°C (see below). Escherichia coli DH5α and MC1061 were used as hosts for cloning and grown in TY at 37°C. When required, the growth media were supplemented with antibiotics at the following concentrations: 5 μg ml−1 of chloramphenicol, 5 μg ml−1 of kanamycin, 100 μg ml−1 of spectinomycin, 4 μg ml−1 of erythromycin (B. subtilis), 150 μg ml−1 of erythromycin (E. coli), and 100 μg ml−1 of ampicillin (E. coli). Agar at 1.5% was included for solid medium.
Procedures for DNA purification, restriction, ligation, agarose gel electrophoresis, and transformation of E. coli were carried out as described before (36). Restriction enzymes were obtained from Roche (Mannheim, Germany), and all PCRs were performed with Phusion (NEB, United Kingdom), unless stated otherwise. Oligonucleotides were purchased from Biolegio (Nijmegen, Netherlands). All constructs were sequence verified. B. subtilis was transformed as described before (19).
Overnight cultures were inoculated from −80°C stocks and grown at 30°C in TLM [62 mM K2HPO4, 44 mM KH2PO4, 15 mM (NH4)2SO4, 6.5 mM sodium citrate, 0.8 mM MgSO4, 0.02% Casamino Acids, 27.8 mM glucose, 0.1 mM and l-tryptophan, with the pH set to 7 using a KOH solution]. The following morning, the cells were diluted 1:10 in 15% prewarmed CDM [62 mM K2HPO4, 44 mM KH2PO4, 15 mM (NH4)2SO4, 6.5 mM sodium citrate, 0.8 mM MgSO4, 2.2 mM glucose, 2.1 mM l-glutamic acid, 6 μM l-tryptophan, 7.5 μM MnCl2, 0.15× metal mix (45)] and grown for 4 h at 30°C. Single cells (rediluted in 15% CDM) were loaded on 15% CDM supplemented with 1.5% high-resolution agarose (Sigma) (for detailed information, see reference (50) and, if relevant, 100 μM isopropyl-β-d-1-thiogalactopyranoside (IPTG) or 0.5% xylose (for induction of transcription from the Pspank or Pxyl promoter, respectively). The outgrowth of a single cell into a microcolony monolayer was achieved by growing the cells in an environmental chamber at 30°C. The following equipment/settings were used for time-lapse microscopy (provided by DeltaVision, United Kingdom): IX71 microscope (Olympus), CoolSNAP HQ2 camera (Princeton Instruments), 300-W xenon light source, 60× bright-field objective (1.25 numerical aperture [NA]), GFP filter set (Chroma; excitation at 470/40 nm and emission at 525/50 nm), mCherry filter set (Chroma; excitation at 572/35 nm and emission at 632/60 nm). Snapshots for movies were taken at intervals of 8 or 12 min using 10% APLLC white LED light and a 0.05-s exposure for bright-field pictures, 10% xenon light and a 0.5-s exposure for GFP detection (unless indicated otherwise), and 32% Xenon light and a 0.8-s exposure for mCherry detection. Raw data were stored using softWoRx 3.6.0 (Applied Presicion) and adjusted for publication using Adobe Photoshop 5.0 (Adobe), ImageJ (http://rsbweb.nih.gov/ij/), and CorelDRAW X3 (Corel Corporation).
Cultures were inoculated from −80°C stocks and grown overnight at 37°C in TY medium supplemented with chloramphenicol. After dilution to an optical density at 600 nm (OD600) of ~0.1, cells were grown for 2 h and transcription of the inducible genes was induced with 100 μM IPTG and/or 0.5% xylose at the beginning of the exponential phase. Samples for flow analysis were taken every hour, and GFP levels of at least 20,000 cells were measured with a Coulter Epics XL-MCL flow cytometer (Beckman Coulter, Mijdrecht, Netherlands) using an argon laser (488 nm). Raw data were taken using EXPO32 software (Beckman Coulter). WinMDI 2.9 was used for data analysis (http://en.bio-soft.net/other/WinMDI.html).
To construct the promoter-GFP fusion plasmids, approximately 600 to 700 bp of the corresponding promoter regions, including all known regulatory sequences (according to http://dbtbs.hgc.jp/), was amplified by PCR using B. subtilis 168 chromosomal DNA as a template. For this study we only examined transcriptional regulation, and to obtain maximal sensitivity we replaced the native ribosome binding site (RBS) in all cases with a strong RBS using ATG as a start codon (47, 51). The PCR fragments were inserted in front of the gfpmut1 gene present on plasmid pSG1151 (25) via the HindIII and EcoRI or HindIII and PstI restriction sites.
To construct pmCherry-kinA_sacA, approximately 600 bp of the kinA promoter region, including all known regulatory sequences (according to http://dbtbs.hgc.jp/), was amplified by PCR using B. subtilis 168 chromosomal DNA as a template. The PCR fragment was inserted in front of the mCherry gene present on plasmid pJWV012 (48) via the KpnI and NheI restriction sites.
Plasmid pDR110 (a kind gift of David Rudner) carrying the tight IPTG-inducible spank promoter was used as vector for the IPTG-inducible constructs (Table (Table2).2). Genes coding for the proteins to be overproduced were amplified by PCR using B. subtilis 168 chromosomal DNA as a template, and strong translation signals were inserted in the primer for all constructs. PCR fragments were cleaved with NheI and SalI and ligated into the corresponding sites of pDR110. To construct a xylose-inducible kinA overexpression plasmid, kinA was amplified by PCR as described above, whereby SpeI and BamHI were used as integration sites for ligation of the PCR product into pAX01.
The promoter-GFP fusion plasmids were inserted into the native locus on the B. subtilis 168 trp+ chromosome by Campbell-type integration, leaving the wild-type background of the corresponding genes intact. Colonies were selected for chloramphenicol resistance, and correct integration was confirmed by PCR using a specific forward primer for each strain on the genome located upstream of the promoter region and pGFP-seq-R (5′GTTGGCCATGGAACAGGTAG3′) as a reverse primer for all constructs.
pmCherry-kinA_sacA was inserted into the sacA locus on the B. subtilis 168 trp+ chromosome of IDJ008, leaving the wild-type background of kinA intact. Colonies were selected for kanamycin resistance, and correct integration was checked by PCR.
The IPTG-inducible overexpression plasmids were inserted into the B. subtilis 168 trp+ chromosome by double crossover into the nonessential amyE locus. Colonies were selected on spectinomycin plates, and correct integration of the overproduction constructs was confirmed by lack of a halo on TY plates containing 1% starch. The xylose-inducible pAX01-kinA construct was inserted into the B. subtilis 168 trp+ chromosome by double crossover into the nonessential lacA locus. Colonies were selected on erythromycin plates, and correct integration of the overproduction constructs was confirmed by PCR.
Replacement mutants were obtained as follows. About 2,000 bp upstream (including the start codon) and downstream (excluding the stop codon) of the corresponding genes was amplified using B. subtilis 168 chromosomal DNA as a template. Corresponding primers are listed in Table Table1.1. The neo cassette was taken from pBEST501 (21) using PstI and XbaI or XbaI and EcoRI, respectively. The three fragments (upstream region, neo cassette, and downstream region) were ligated and directly transformed to B. subtilis strain JWV002 (48), which was used as an in-between host, before transformation to the target strain B. subtilis 168 trp+. Colonies were selected for kanamycin resistance, and correct chromosomal integration was verified by PCR.
Combinations of the strains were made by transforming a promoter-GFP strain with pmCherry-kinA_sacA or chromosomal DNA from the IPTG-inducible strains and/or knockouts. Transformants were selected for the antibiotic resistance marker, and correct integration was verified by PCR and/or lack of a halo on TY plates containing 1% starch.
The major sporulation kinases KinA and KinB autophosphorylate and transfer the phosphate to the phosphotransferase Spo0F (4). Spo0B receives the phosphate from Spo0F~P, and Spo0B~P subsequently phosphorylates Spo0A (Fig. (Fig.1).1). In response to phosphorylation, Spo0A~P dimerizes and is able to bind to its recognition sequence, the so-called 0A box (26). Sporulation-specific promoters, such as the spoIIA promoter, are activated by high concentrations of Spo0A~P (14). Using bulk population-wide LacZ assays, it has been shown that transcription of the phosphorelay genes increases during the early stages of sporulation (see, e.g., reference 1). However, it is unknown how the phosphorelay genes are regulated at the single-cell level. Previously it was shown that not all cells within clonal B. subtilis populations activate the spoIIA promoter and consequently form spores (8, 46) (Fig. (Fig.2H2H and Fig. Fig.3H;3H; see Movie S8 in the supplemental material). The simplest explanation for the bimodal expression pattern of spoIIA is that the heterogeneity is already present earlier than expression of spoIIA, at the level of the individual phosphorelay components. To test this hypothesis, we constructed promoter-GFP fusions of the core phosphorelay components and PsigH and examined their GFP expression at the single-cell level using time-lapse microscopy. Individual cells were grown directly under the microscope on top of a thin layer of agarose in such a way that single cells grew into sporulating microcolonies (see Materials and Methods). Bright-field and fluorescence images were acquired at periodic intervals to generate time-lapse movies which contain information encompassing the complete cell division history of individual cells, their GFP expression status, and the final cell fate decision (see Materials and Methods and reference 50). We focused on sporulation heterogeneity only. Other heterogeneous phenotypes (e.g., competence development or protease production) that might be displayed were not examined. For each promoter-GFP fusion, at least two independent experiments were performed, and at least two developing microcolonies were analyzed per experiment (thus there were minimally four movies per strain). Time-lapse microscopy revealed that the expression levels of kinA, kinB, spo0F, spo0A, and sigH (also known as spo0H) increased with time until spore formation occurred and that the activity of all genes exhibited a large cell-to-cell variability (see Movies S1 to S8 in the supplemental material). Single-cell analyses of the movies showed a unimodal, broad, long-tailed Gaussian distribution in gene expression for kinA, spo0F, and sigH (Fig. (Fig.2;2; see Fig. S1 in the supplemental material). When all cells from a single microcolony (80 to 120 cells) were analyzed, the fluorescence distribution for the reporter strains of kinB and spo0A appeared to be almost bimodal (Fig. (Fig.2).2). However, when the data for corresponding time points from multiple microcolonies are pooled, a broad unimodal distribution becomes apparent (Fig. (Fig.2,2, insets; see Fig. S1 in the supplemental material). Thus, we conclude that also in the case of kinB and spo0A, expression is unimodal. Interestingly, GFP from the spo0B promoter was expressed with a relatively small cell-to-cell variability (note the low GFP expression levels, close to the background fluorescence of the wild type), and in contrast to that of the other phosphorelay genes, spo0B expression did not seem to increase during microcolony development (Fig. (Fig.2F;2F; see Fig. S1F and Movie S6 in the supplemental material).
Taken together, these data indicate that bimodal expression of spoIIA is not the result of bimodal expression of one of the phosphorelay components.
Although the phosphorelay genes did not exhibit a bimodal expression pattern, their variability in expression levels might cause sporulation heterogeneity if the timing of their induction differs between cells. To test this, we used GFP reporters to follow the expression of genes encoding components of the phosphorelay in a number of lineages that resulted in either sporulating or nonsporulating cells as judged by bright-field microscopy after 20 h during microcolony development (see Movies S1 to S8 in the supplemental material). At least two independent experiments were performed, and a minimum of four sporulating and four nonsporulating lineages of each promoter-GFP strain (IDJ001 to IDJ007) were analyzed per independent experiment. Note that it is not possible to directly compare the timings of trajectories from different experiments, since not only does the time of adaptation to the agarose slide vary between individual cells, but also the time between the preparation of the microscopic slide and the start of the movie differs for each experiment. Nevertheless, the timings of GFP expression from trajectories of sporulating cells can be compared by approximating the time between a specific expression trend and actual spore formation. Single-cell trajectories of the phosphorelay reporters showed that cells that formed spores early during microcolony development showed higher expression of kinA, spo0F, and spo0A than their nonsporulating siblings at the same time point within the same microcolony (Fig. (Fig.3,3, red lines; see Fig. S2 and S3 [left panel] in the supplemental material). In some cases nonsporulating cells also showed a peak of spo0F expression simultaneously with sporulating cells, but never earlier (see Fig. S2 and S3 in the supplemental material). The peak in expression occurred approximately 250 min before the endospore became visible (indicated by the last red point plotted in each graph). Subsequently, the GFP signals expressed from PsigH, PkinA, Pspo0F, and Pspo0A dropped dramatically, probably because of activation of the late sporulation-specific sigma factors and/or deactivation of sigma A (13, 20, 27).
Interestingly, cells that did not form spores during the time course of the experiment often showed similar (kinA and spo0A) or even stronger (sigH and spo0F) promoter activity of the phosphorelay genes at later stages of microcolony development than their sporulating siblings within the same microcolony (Fig. (Fig.3,3, black lines; see Fig. S2 and S3 [right panel] in the supplemental material). It cannot be excluded that some cells which did not sporulate during the experiment would ultimately have formed spores later (compare, e.g., red lines in Fig. S3G in the supplemental material). In all experiments performed, no particular differences could be found in the kinB expression patterns of sporulating and nonsporulating cells. This might be due to a supporting, but not primary, function of KinB to feed phosphate into the system. However, expression of kinB as well as sigH increases with time, suggesting that they might be sensitive to very low levels of Spo0A~P and that their regulation differs from that of other phosphorelay genes. The other peculiar phosphorelay component is Spo0B. As mentioned above, its expression is low, and the cell-to-cell differences are too small to reveal variations in expression patterns of sporulating and nonsporulating cells (Fig. (Fig.3F;3F; see Fig. S2F and S3F in the supplemental material).
These data support the hypothesis that the timing of phosphorelay gene expression, or at least of kinA, spo0F, and spo0A expression, is important in determining whether a cell will initiate sporulation or not.
As shown in Fig. Fig.33 and in Fig. S2 and S3 in the supplemental material, early activation of at least kinA, spo0F, and spo0A seems to be important for successful spore formation. Strikingly, however, most nonsporulating cells exhibited relatively high phosphorelay transcription levels during late microcolony development but did not develop spores during the experimental time course. Thus, the question remains why these cells do not initiate sporulation despite relatively high levels of the phosphorelay components such as KinA. The most likely explanation for this paradoxical observation is that although the level of kinases is sufficient, the net activity of the phosphorelay is too low to obtain high levels of Spo0A~P. Phosphate groups can be drained from the phosphorelay by a number of phosphatases, including the major sporulation phosphatase RapA (33). Recently, it was shown that the levels of RapA are high in a subpopulation of cells during microcolony development (3). To assess the transcription levels of rapA and kinA simultaneously in single cells, we performed time-lapse microscopy experiments using a double-labeled strain in which PrapA was fused to the GFP gene and PkinA was fused to the mCherry gene. Information on the background fluorescence can be found in Fig. S4 in the supplemental material. Single-cell trajectories of the double-labeled strain IDJ039 show that mCherry transcription from PkinA is more or less similar in both sporulating and nonsporulating cells, whereas GFP transcription from PrapA is relatively low in sporulating cells, compared to high RapA transcription levels in cells that did not sporulate within the experimental time frame (Fig. (Fig.4A).4A). The peak of mCherry expression from PkinA is not as pronounced and is shifted to the right in time compared to the corresponding fusion to GFP shown in Fig. Fig.3.3. These differences might be observed due to the limited brightness and longer maturation time of mCherry compared to GFP. To ensure that the difference in rapA expression is valid for all cells, all 61 cells from the microcolony were analyzed at a time point at which the first spore is not yet visible (786 min). This analysis shows that indeed a significant difference (P < 0.01 by Student's t test) in PrapA-GFP expression levels exists between spore formers and nonsporulating cells (22 ± 15 versus 75 ± 28 arbitrary GFP units, respectively) (Fig. (Fig.4B4B).
Together, these data suggest that increased phosphatase expression prevents cells from sporulation, even if transcription of the phosphorelay genes is high.
While the heterochronicity of kinA, spo0F, and spo0A gene expression is correlated with the initiation of sporulation (Fig. (Fig.22 and and3;3; see Fig. S1, S2, and S3 in the supplemental material), sporulation heterogeneity also seems to originate from cell-to-cell differences in the phosphate flow of the phosphorelay (8, 46) (Fig. (Fig.4).4). To test whether perturbations in the phosphate input have an effect on sporulation heterogeneity, we constructed strains with knockout or overproduction mutations of the genes coding for the sporulation kinases KinA and KinB. The corresponding mutations were introduced in a strain harboring the PspoIIA-GFP reporter, and cells were monitored by fluorescence time-lapse microscopy and/or flow cytometry. As shown in Fig. Fig.5A,5A, spores in either a kinA (IDJ035) or kinB (IDJ036) single mutant were still readily formed. While fewer cells formed spores in either a kinA or kinB mutant background, the effect of the single mutation in our genetic background (168 trp+) was not as dramatic as the effect of a kinA mutation on sporulation in the JH642 background (8).
Besides KinA and KinB, three other phosphorelay kinases (KinC, KinD, and KinE) that are not essential for spore formation have been identified (22, 44). To validate that KinA and KinB are the primary sporulation kinases in our experimental setup, a kinA kinB double mutant was constructed and analyzed by time-lapse microscopy. Consistent with previous reports (31, 39), spore formation was completely blocked in the kinA kinB double mutant (IDJ037) (Fig. (Fig.5A5A).
If cell-to-cell differences in phosphate input within the system result in bimodal spoIIA expression, overproduction of KinA or KinB might abolish sporulation heterogeneity. It was previously shown that artificial induction of kinA or kinB expression induces spore formation during exponential growth (15), indicating either that the kinases autophosphorylate readily and do not require a specific stationary-phase signal to switch to their kinase mode or that the basal level of kinase activity is sufficient to initiate sporulation under overproduction conditions. To investigate the effects of KinA or KinB overproduction on PspoIIA activation, we cloned the genes coding for these kinases under the control of the strong IPTG-inducible spank promoter (see Materials and Methods) and transformed the corresponding DNA to strain IDJ007 (PspoIIA-GFP), resulting in strains IDJ021 and IDJ022. To ensure that other stationary-phase signals leading to spoIIA activation were absent, cells were grown in rich TY medium and supplemented at early exponential phase with 100 μM IPTG to induce kinA or kinB expression. SDS-PAGE analysis of cell lysates clearly showed KinA overproduction at this IPTG concentration (data not shown). After 5 h of induction, cells were analyzed by flow cytometry. As shown in Fig. Fig.5B,5B, overproduction of KinA or KinB resulted in a shift of the GFP levels to the right, indicating that more cells activated the spoIIA promoter than in noninduced cultures. Interestingly, after 5 h of induction, not all cells have activated spoIIA, and a broad, almost bimodal spoIIA expression pattern is observed (Fig. (Fig.5B).5B). This indicates that Spo0A~P does not reach the levels required to initiate sporulation in all cells, implying that the phosphorelay activity is not saturated under these KinA-overproducing conditions.
To confirm these results and directly assess the effect of KinA and KinB overproduction on sporulation heterogeneity in a clonal population originating from a single cell, the kinase overproduction strains were grown on agarose slides containing 100 μM IPTG and monitored by time-lapse microscopy. Under these conditions, overproduction of KinA or KinB resulted in premature spore formation (see Movies S11, S12, and S13 in the supplemental material). These results are consistent with the flow cytometry data showing an increased percentage of cells expressing high levels of PspoIIA-GFP compared to the control. Importantly, upon KinA overproduction, microcolonies in which all cells formed spores (100% spores) were regularly found and sporulation heterogeneity was completely abolished (Fig. (Fig.5C;5C; see Movie S11 in the supplemental material). In general, sporulation efficiency upon KinA overproduction was 92% ± 9% (mean ± standard deviation; data from six microcolonies, with 187 cells counted). Also in the case of KinB overproduction, most cells formed spores (82% ± 7%; data from seven microcolonies, with 356 cells counted), in sharp contrast to the case for wild-type microcolonies, where only approximately 14% (±10%; data from six microcolonies, with 630 cells counted) of the cells formed spores (Fig. (Fig.5C;5C; see Movies S10 and S12 in the supplemental material). It is interesting to note that the total cell count (spores plus vegetative cells) per microcolony decreases with a decrease in sporulation heterogeneity (or an increase in sporulation efficiency). This can partly be explained by a block of vegetative cell division during sporulation (2), but nevertheless this indicates that sporulation heterogeneity gives a numerical (reproductive) advantage compared to homogenous sporulating populations.
Taken together, these data corroborate previous results showing the importance of the kinases in the initiation of sporulation (8, 15). Importantly, our results indicate that the phosphate input (through autophosphorylation of KinA or KinB), is a major determinant in the “decision” to sporulate or not.
The activity of the phosphorelay is determined by a mix of environmental and cell cycle cues (33). Thus, the question remains how these external signals are propagated through the network to initiate sporulation; how do high kinase levels manage to abolish sporulation heterogeneity? To test this, we examined the expression of PkinA-GFP, PkinB-GFP, Pspo0F-GFP, Pspo0B-GFP, Pspo0A-GFP, and PsigH-GFP under conditions in which the phosphorelay was artificially charged (KinA overproduction) at a stage in which normally no stationary-phase signals are present (induction in TY medium at early exponential phase). Flow cytometry analysis showed that transcription of kinA, kinB, spo0F, spo0A, and sigH was upregulated by overproduction of KinA (Fig. (Fig.6).6). This indicates that the phosphate charge of the phosphorelay directly affects the transcription of the phosphorelay, which might be important for signal propagation. spo0B seems to be the exception, as its transcription is not under (indirect) positive control of KinA (Fig. (Fig.6E).6E). This result might suggest that Spo0B is a limiting factor for the initiation of sporulation, but our experimental data show that this is unlikely (see Discussion and Fig. S5 in the supplemental material). Interestingly, at least for sigH, kinB, spo0F, and perhaps spo0A, a bimodal expression pattern arises upon KinA overproduction, again suggesting that upon KinA overproduction the phosphorelay is not fully saturated in all cells (Fig. (Fig.66).
Our data and previous work from other labs suggest that overproduction of KinA leads to an increase in the phosphate charge of the phosphorelay. In addition, it has been shown that overproduction of the Spo0E phosphatase reduces Spo0A~P levels and thus sporulation (30, 32) and that deletion of spo0E causes more cells to initiate sporulation (46). To further support the hypothesis that sporulation heterogeneity is regulated through the phosphate flow of the phosphorelay, we constructed a strain in which spo0E is under the control of Pspank (IDJ014). Genomic DNA of this strain was used to construct strain IDJ027, which also carries a xylose-inducible kinA construct (see Materials and Methods). Flow cytometry experiments were performed as described above for Fig. Fig.5B,5B, using 0.5% xylose for induction of kinA expression and 1 mM IPTG for induction of spo0E expression. Consistent with the data obtained with the IPTG-inducible kinA strain IDJ011, induction of kinA expression from the xylose-inducible promoter (Fig. (Fig.7A,7A, green line) resulted in a significant increase in GFP expression from the spoIIA promoter compared to that in the noninduced culture (Fig. (Fig.7A,7A, red line). Upon overproduction of Spo0E, GFP expression from PspoIIA was significantly reduced (blue line). Since Spo0E dephosphorylates Spo0A~P (30), these data support the idea that phosphate availability in the phosphorelay is the major determinant responsible for sporulation heterogeneity. Overproduction of both the kinase KinA and the phosphatase Spo0E (gray line) clarifies this even more, since the GFP expression pattern strongly resembles that of the noninduced control (red line). This indicates that the result of the increased KinA levels is diminished by the increased Spo0E levels. The same strain (IDJ027) was also used to examine the effect of decreased phosphate availability on sporulation heterogeneity by microscopy under the conditions used for time-lapse microscopy (Fig. 7B to E). The cells were grown on a 15% CDM agar pad supplemented with 0.5% xylose for kinA overexpression and 1 mM IPTG for spo0E overexpression, and snapshots of the microcolonies were taken after 24 h. As expected, sporulation heterogeneity was observed for the noninduced control (Fig. (Fig.7B)7B) and heterogeneity was lost upon KinA overproduction (100% spore formation) (Fig. (Fig.7D).7D). Conversely, upon overproduction of Spo0E, only a few cells reached the Spo0A~P levels required for spore formation, resulting in a significant reduction in sporulation efficiency (Fig. (Fig.7C).7C). Consistent with the flow cytometry data, spore formation during Spo0E overproduction was restored by KinA overproduction (Fig. (Fig.7E7E).
The alternative sporulation sigma factor SigH activates transcription of kinA, spo0F, and spo0A, and indirectly its own transcription, but not that of spo0B (14, 35, 40). Thus, SigH constitutes a major positive-feedback loop in the sporulation network and might be important in determining whether a cell will sporulate or not. We showed that overproduction of KinA caused activation of sigH (Fig. (Fig.6A),6A), suggesting that SigH plays a pivotal role in the propagation of the sporulation signal. To further explore the role of SigH, we deleted sigH and assessed the effect on heterogeneous spoIIA expression in microcolonies (IDJ034). Interestingly, while the spoIIA promoter is under SigH control (52), cells mutated in sigH still exhibited heterogeneous GFP expression (Fig. (Fig.8A;8A; see Fig. S6 in the supplemental material). However, after 24 h of microcolony development, PspoIIA-GFP expression in nonsporulating cells mutant in sigH was at least 2.5-fold lower than that in wild-type cells (Fig. (Fig.8A).8A). Most likely the difference is even greater at earlier stages of microcolony development, when wild-type cells show the highest spoIIA expression. To test whether the heterogeneity was caused by Spo0A, we performed similar experiments with a PspoIIE-GFP reporter strain in the sigH mutant background. spoIIE is also under the control of Spo0A but is not dependent on SigH for its expression (53). Although GFP expression from the spoIIE promoter in the sigH mutant background was even closer to the background fluorescence than GFP expression from PspoIIA in this mutant, heterogeneity could still be observed (data not shown). Thus, both experiments indicate that heterochronic phosphorelay gene expression (in conjunction with the phosphorelay phosphate charge) is sufficient to create a heterogeneous sporulation signal and that the positive-feedback loop by SigH is probably required only as an amplifier to ensure that phosphate levels can reach the threshold required for actual spore formation. Nevertheless, we cannot exclude that SigH might be essential for other factors required for sporulation that were not tested with our studies.
If SigH merely amplifies the heterogeneous sporulation signal, an artificial increase of the phosphorelay phosphate charge might bypass the requirement of SigH for spore formation. To test this hypothesis, we overproduced KinA in a sigH mutant background (IDJ038). In this experimental setup, the activity of the phosphorelay is artificially charged in the absence of the SigH feedback loop. As shown in Fig. Fig.8A,8A, under these conditions spores were also not formed, indicating that in the absence of the positive SigH feedback loop, Spo0A~P cannot reach levels high enough to initiate sporulation even if the phosphorelay is artificially charged.
If KinA overproduction activates initiation of sporulation only in the presence of an intact SigH feedback loop, the temporal separation between charging the phosphorelay and SigH expression might be important for sporulation heterogeneity. To test this, kinA was placed under the control of the xylose-inducible promoter Pxyl and sigH under the control of Pspank. Both constructs were stably integrated in the chromosome of the PspoIIA-GFP reporter strain, and the resulting strain (IDJ026) was grown in the presence or absence of xylose and/or IPTG, respectively. Flow cytometry analysis of exponentially growing cells showed that PspoIIA-GFP was not increased when sigH expression was artificially induced (in the presence of 1 mM IPTG) (Fig. (Fig.8B,8B, red line). KinA overproduction (induced by 0.5% xylose) activated PspoIIA-GFP, as expected (blue line). When both KinA and SigH were overproduced simultaneously, PspoIIA-GFP was activated to a similar degree (green line) as under conditions in which only kinA was overexpressed. Strikingly, when KinA was overproduced first and SigH 1 h later, even more cells activated expression of the spoIIA operon (gray line). This might reflect the temporal order of processes within the system prior to heterogeneous initiation of sporulation. Upon external signal accumulation, the first process being operated is signal integration by buildup of the phosphate charge, followed by an increase in expression of the alternative sigma factor SigH, which in turn facilitates expression of spo0A and the sporulation-specific genes that require high levels of Spo0A~P.
Clonal populations of Bacillus subtilis bifurcate into sporulating and nonsporulating subpopulations. This phenotypic bistability, or sporulation heterogeneity, is proposed to have evolved as a “bet-hedging” strategy to ensure that part of the population is always prepared for uncertain future and rapidly changing environmental conditions (50). In contrast to competence development, the direct positive feedback on the main regulator Spo0A is not essential for sporulation heterogeneity (50). Since Spo0A phosphorylation and initiation of sporulation are controlled by a multicomponent phosphorelay, we hypothesized that sporulation heterogeneity can arise at two distinct levels: (i) variability in the expression of phosphorelay genes and/or (ii) variability in the overall net activity of the phosphorelay. Using single-cell analysis tools, we tested these predictions. In a previous study it was shown that there is a large cell-to-cell variability in low levels of spo0A expression during exponential growth (5). We now show that the other phosphorelay genes also show a large variability in gene expression (Fig. (Fig.22 and Fig. Fig.3;3; see Fig. S1, S2, and S3 in the supplemental material). The simplest explanation for sporulation heterogeneity would be that at least one of the genes involved in activation of sporulation-specific genes is bimodally expressed. Surprisingly, all phosphorelay genes were expressed unimodally across the population (Fig. (Fig.2;2; see S1 in the supplemental material).
In line with previous genetic evidence (7, 16, 35, 41-43), histograms based on expression data of developing microcolonies showed that transcription of all phosphorelay genes, except for spo0B, increases as cells enter stationary growth (Fig. (Fig.2;2; see Fig. S1 in the supplemental material). If this temporal increase of transcription does not occur simultaneously in all cells, this could constitute another source for sporulation heterogeneity. In fact, single-cell expression traces showed that sporeformers in general activate transcription of kinA, spo0F, and spo0A earlier than nonsporulating cells within the same microcolony (Fig. (Fig.3;3; see Fig. S2 and S3 in the supplemental material). At this point we cannot distinguish whether the early increase in kinA expression is caused by early activation of Spo0A or by KinA itself. This is currently under investigation in our laboratory.
Interestingly, the subpopulation of nonsporulating cells demonstrated an interesting pulsating gene expression pattern during microcolony development, which resembles recently reported fluctuations of rapA and sda gene expression (3, 48). It is tempting to speculate that cells undergo cell cycle-dependent regulation of the activity of Spo0A~P as a “bet-hedging” strategy to ensure that cells will sporulate only when the conditions are favorable again to bifurcate into sporulating and nonsporulating populations (50). Furthermore, cell cycle regulation of Spo0A~P is important to ensure correct chromosome copy number during sporulation (48).
Why do the nonsporulating cells show a relatively high level of transcription of the phosphorelay genes but do not initiate sporulation? To initiate sporulation, cells require high levels of Spo0A~P (14). Thus, although the phosphorelay gene expression of the nonsporulating subpopulation was shown to be high at late stages of microcolony development (Fig. (Fig.3;3; see Fig. S2 and S3 in the supplemental material), the net phosphate charge of the phosphorelay might be too low to initiate sporulation. In this respect it is interesting to note that transcription of the genes encoding the RapA and Spo0E phosphatases are activated by Spo0A~P as well (14, 32). Concomitantly with the phosphorelay genes, the expression of these negative regulators of phosphorelay activity is elevated, and Spo0A~P levels could be kept low although phosphorelay gene expression is relatively high. Overproduction of Spo0E resulted in a decrease in spore formation, a phenotype that could be bypassed by simultaneous overproduction of KinA (Fig. (Fig.7).7). In fact, using a double-labeled strain, we showed that nonsporulating cells with relatively high levels of kinA expression have significantly higher rapA transcription levels than their sporulating siblings (Fig. (Fig.4).4). RapA levels might, for instance, be influenced by growth rate, cell densities, and the nutritional status of the cell (3). This suggests that all cells have the predisposition to initiate sporulation but that only those able to reach a sufficiently high phosphorelay phosphate charge can enter the pathway.
It has been suggested that the phosphorelay kinases need a (biochemical) signal to initiate autophosphorylation, although the nature of this signal still remains elusive (24). Transcriptional activation (overproduction) of kinA or kinB is sufficient to increase Spo0A~P levels and initiate spore formation (15). Here we show that overproduction of kinA or kinB results in elevated levels of phosphorelay gene expression (Fig. (Fig.55 and and6),6), even under conditions in which no late-stationary-phase signals are present. Furthermore, KinA overproduction itself can abolish sporulation heterogeneity in microcolonies originating from a single cell, resulting in 100% sporulating cells (Fig. (Fig.5C).5C). However, flow cytometry shows that overproduction of KinA does not abolish heterogeneity per se (Fig. (Fig.5B)5B) but that it seems to reduce the time for individual cells within the heterogeneous population to commit to sporulation. These data indicate that autophosphorylation of the kinases occurs readily, possibly without the need of an external signal, in line with previous reports on the phosphorelay and other two-component systems in which overproduction of the signal kinase leads to signal-independent activation of the cognate response regulator (23). Whether the kinases are fully in their kinase mode during overproduction or whether the basal level of the kinase activity is sufficient to initiate sporulation under overproduction conditions is as yet unknown.
As mentioned before, phenotypic bistability can originate from positive feedback within the gene network. A major positive-feedback loop within the sporulation network is represented by the alternative sigma factor SigH. SigH was shown to activate transcription of kinA, spo0F, and spo0A (35). Furthermore, transcription of sigH is repressed by the unstable transcriptional repressor protein AbrB, whose transcription is repressed by low levels of Spo0A~P (14, 40). Thus, indirect transcriptional activation of sigH by Spo0A~P results in elevated levels of Spo0A~P, constituting a feedback loop (Fig. (Fig.9).9). Our data suggest that once the phosphorelay reaches a certain threshold phosphate charge, the SigH feedback loop kicks in, leading to even larger amounts of KinA and KinB (Fig. (Fig.66 and and8).8). This causes a rapid increase of Spo0A~P, ultimately resulting in activation of the unidirectional sporulation program. It is tempting to speculate that this built-in time delay requires a complex network topology with many feedback loops, as it is present within the phosphorelay (Fig. (Fig.9).9). Although the heterogeneous sporulation signal is established without the need for SigH, this positive-feedback loop might provide the system with more robustness.
Spo0B is special in the phosphorelay, as it is not under feedback control and might act to keep the system in the signal integration regimen (3). Initially, our data suggested that Spo0B might be a limiting factor for sporulation, since its expression was shown to be low and did not increase with time (Fig. (Fig.2F2F and and3F;3F; see Fig. S1F, S2F, and S3F and Movie S6 in the supplemental material). These data are consistent with data from Ferrari et al., who estimated that only 50 to 100 Spo0B molecules are present in the cells (12). However, overproduction of Spo0B or Spo0F resulted in fewer cells expressing PspoIIA-GFP (see Fig. S5 in the supplemental material). A likely explanation for this phenotype is that the ratio of nonphosphorylated to phosphorylated phosphotransferase has increased. Since all the reactions in the phosphorelay are reversible (11) (Fig. (Fig.1),1), the effective concentration of the phosphate-bound phosphotransferase is reduced upon overproduction. This might reduce the speed of phosphotransfer to its cognate partner, subsequently resulting in reduced levels of Spo0A~P. This indicates that Spo0B is not a limiting factor for sporulation and that minor fluctuations in the levels of any of the phosphorelay components might rapidly be equilibrated between the components within the phosphorelay, making the system robust to posttranscriptional noise. On the other hand, the broad distribution of expression levels of some of the phosphorelay genes, such as kinA, spo0F, and spo0A, indicates that there is a large cell-to-cell variability in Spo0A~P levels which possibly originates from stochastic fluctuations that arise during the process of transcription.
In conclusion, our data strongly support a model in which sporulation heterogeneity originates from intercellular differences in phosphorelay phosphate availability regulated at both the transcriptional and posttranslational levels. While the Spo0A~P level increases over time, a large cell-to-cell variability in Spo0A~P exists, which may be caused by the cell cycle state, the metabolic activity of the cell, and/or noise in phosphorelay gene transcription. Here we use the term heterochronicity, since our data suggest that there is not only cell-to-cell variability in phosphorelay gene expression but also temporal variation between the development of cells. Thus, sporulation heterogeneity seems to originate from a combination of asynchrony and stochastic influences, such as transcriptional noise, but might also include specific metabolites or signaling molecules. Stochasticity of transcription might be an important regulatory mechanism for heterogeneity in phosphorelay gene expression, especially variations in kinase gene expression. Furthermore, our data suggest that successful entry into sporulation might require that all phosphorelay components reach a certain threshold concentration at the same time. If one or more components of the phosphorelay are below this threshold, sporulation cannot be initiated. This might also offer an explanation for the few cells that show a relatively early peak of spo0F transcription but do not go on to sporulate.
Not all B. subtilis cells sporulate at the same time, but with increasing time more cells sporulate. Whether cells initiate sporulation later might depend on the nutrients released by lysed cells, for instance, via the cannibalism route, but it also depends on the cell cycle state of the cell (17, 48). Here, we have followed sporulation only for a period of approximately 24 h. It would be interesting to see whether cells that have not sporulated during this period would divide again later and bifurcate into sporulating and nonsporulating cells. It is tempting to speculate that heterochronicity is an important feature of B. subtilis sporulation to allow for an optimal ratio of sporulating and vegetative cells.
We thank the anonymous referees for useful suggestions and one referee for pointing out the term “heterochronicity.” We thank David Rudner for the kind gift of plasmid pDR110 and Leendert Hamoen for critically reading the manuscript.
I.D.J. was supported by BaCell-SysMO and NWO. J.-W.V. was supported by startup funds from the University of Groningen and by a Marie-Curie Reintegration grant. This project was carried out within the research program of the Kluyver Centre for Genomics of Industrial Fermentation, which is part of the Netherlands Genomics Initiative/Netherlands organization for Scientific Research.
Published ahead of print on 12 February 2010.
†Supplemental material for this article may be found at http://jb.asm.org/.