Direct cell–cell contact activates SigM to express the ESX-4 secretion system in Mycobacterium smegmatis

Significance

A conjugation model of mycobacterial interaction recently revealed that intercellular communication occurs between donors and recipients. This communication links two ESAT-6 (ESX) (type VII) secretion systems that are both required for conjugation. Functionally distinct ESX secretion systems are found in all mycobacteria, and they serve important virulence functions in pathogenic mycobacteria. We demonstrate that SigM, an extracytoplasmic transcription factor, activates ESX-4. Direct donor-recipient cell contact triggers the recipient cell to release membrane-sequestered SigM, which rapidly induces an ESX-4-focused regulon. The conservation of SigM and ESX-4 throughout mycobacteria suggests that this interaction-response network is intact and active in pathogens. Contact-dependent responses similar to those identified in our model system may therefore also mediate communal processes within infectious mycobacterial populations.

Abstract

Conjugal cell–cell contact between strains of Mycobacterium smegmatis induces the esxUT transcript, which encodes the putative primary substrates of the ESAT-6 secretion system 4 (ESX-4) secretion system. This recipient response was required for conjugal transfer of chromosomal DNA from the donor strain. Here we show that the extracytoplasmic σ factor, SigM, is a cell contact-dependent activator of ESX-4 expression and is required for conjugal transfer of DNA in the recipient strain. The SigM regulon includes genes outside the seven-gene core esx4 locus that we show are also required for conjugation, and we show that some of these SigM-induced proteins likely function through ESX-4. A fluorescent reporter revealed that SigM is specifically activated in recipient cells in direct contact with donor cells. Coculture RNA-seq experiments indicated that SigM regulon induction occurred early and before transconjugants are detected. This work supports a model wherein donor contact with the recipient cell surface inactivates the transmembrane anti-SigM, thereby releasing SigM. Free SigM induces an extended ESX-4 secretion system, resulting in changes that facilitate chromosomal transfer. The contact-dependent inactivation of an extracytoplasmic σ-factor that tightly controls ESX-4 activity suggests a mechanism dedicated to detect, and appropriately respond to, external stimuli from mycobacteria.

Conjugation is a form of horizontal gene transfer in which DNA is transported from a donor into a recipient bacterium (1). Conjugative plasmids (and integrative conjugal elements) use conjugation to efficiently disseminate through a bacterial population. Plasmids that are self-transmissible encode all of the activities required to facilitate their transfer from a plasmid-containing donor into a plasmid-naïve recipient, activities that include mating-pair formation, DNA transfer, and replication (2, 3). Coordinated regulation of conjugal transfer genes is critical, as efficient conjugation systems must identify compatible partners and culminate with the stable maintenance of the transferred mobilized DNA in the resulting transconjugant (4). This coordination is exemplified by some conjugative systems that encode pheromones to regulate donor and recipient functions (5–7).

Mycobacterial conjugation, in which many unlinked chromosomal fragments are cotransferred from a donor strain into a genetically distinct recipient, is known as distributive conjugal transfer (DCT) (8, 9). DCT is mechanistically distinct from oriT-mediated transfer and does not involve plasmids; thus, the functional and regulatory genes for conjugation are all chromosomally encoded (10). Previous transposon mutagenesis studies in donor and recipient strains of the Mycobacterium smegmatis DCT model system established two tenets of mycobacterial DCT. First, transposon insertions showed that the ESAT-6 secretion system 1 (ESX-1) is intimately involved in the process (11–13). In the recipient, transposon insertions in the esx1 locus prevented DNA transfer (11), while insertions in the orthologous donor esx1 locus increased DNA transfer efficiencies (12). Second, a pattern emerging from transposon screens suggests that mycobacterial conjugation is a recipient-driven process, contrasting with traditional donor-driven conjugative mobile elements (11, 14). With the caveat that the screened arrayed transposon libraries represented only a subset of their respective genomes, only a few insertion mutations in the donor strain reduced DNA transfer (15), while many were identified in the recipient strain that prevented DNA transfer. A linkage analysis of transconjugants for donor activity identified a mating identity (mid) locus that switched the conjugal identity from recipient to donor. Subsequent back-crossing delimited mid to a six-gene region within esx1, further implicating a central but complex role for ESX-1 in DCT (9).

Characterization of DCT in mycobacteria led to the initial identification of contact-dependent intercellular communication (14). Cell–cell contact between conjugal mating M. smegmatis pairs elicits strain-specific transcriptional responses. One particular response was shown to be required for DCT and involved another ESX system. ESX-4 is the progenitor of the other four paralogous ESX secretion systems encoded in mycobacterial chromosomes, but its function was not known (16–18). The largest transcriptional response to coculture was the induction of recipient esxUT mRNA, which encodes a small heterodimer that is the putative primary secretion substrate of ESX-4. Mutations in the recipient esx4 locus that likely block ESX-4 secretion activity are also defective in DCT. Taken together, these data demonstrated that donor and recipient coculture induces ESX-4 secretion in the recipient strain, and that this induction is critical for DCT.

The recipient esxUT induction response to donor coculture represents an identified target of intercellular communication networks in mycobacteria. The cell–cell interaction during coculture generates an unknown signal that is transmitted from the recipient cell surface to the genes required for DCT. The transmittal mechanism inducing esxUT is the logical place to begin to trace the interaction network of DCT.

Here, we show that the extracytoplasmic function (ECF) σ factor, SigM, is required in the recipient strain for conjugation. We define the SigM regulon in M. smegmatis by ectopic SigM expression, followed by RNA-seq to map the transcription units that enable DCT. To reproduce a biologically relevant response context, we conducted our analysis in the DCT recipient strain of M. smegmatis, MKD8. We found that SigM expression induced other genes that encode the core ESX-4 secretion apparatus. Annotations for many of the 21 directly regulated SigM genes include ESX-4 secretion, cell wall hydrolase, and nuclease activities, which could conceivably support the process of DNA transfer. Additional coculture RNA-seq at time points that precede detectable conjugation showed robust SigM-regulon induction. This is consistent with SigM directing a rapid transcriptional response to cell–cell contact at an early stage of the DNA transfer process. We developed an EGFP reporter of esxUT transcription to provide single-cell reporting of SigM activation, and visualized recipient cells responding to the stimulus of direct donor contact. This activated subpopulation identifies preconjugal recipients and reveals the establishment of phenotypic heterogeneity within mixed mycobacterial populations. The findings presented here show that the ECF SigM transmits signals initiated by donor contact to activate an expanded ESX-4 secretion apparatus in the DCT recipient strain of M. smegmatis. The conservation of SigM and ESX-4 suggests that a similar interaction response network is active in diverse mycobacterial populations.

Results

SigM and DCT.

Our published data showed that the strong induction of esxUT in the recipient strain was required for conjugation (14). Previous independent microarray studies of SigM ectopic expression in Mycobacterium tuberculosis found esxU and esxT to be the most highly induced genes (19–21), although a link to ESX-4 has not been explored. We hypothesized that if the SigM–esxUT regulatory link was conserved, SigM would be required for conjugal transfer of DNA into the M. smegmatis recipient. We tested this by creating knockouts of the orthologous sigM locus in the recipient strain (MKD8) of M. smegmatis. MKD8_6931 and MKD8_6932 comprise the two-gene locus, and encode SigM and anti-SigM, respectively (Fig. 1A). We then assessed this knockout in a standard DCT conjugal transfer assay. Deletion of sigM reduced the transfer efficiencies to below the level of detection (Fig. 1C). Therefore, SigM is required in the recipient strain for DCT.

Fig. 1.

SigM is required in the recipient for conjugation. (A) The M. smegmatis MKD8 sigM genomic locus includes MKD8_6931 and MKD8_6932, respectively encoding SigM and anti-SigM. (B) The indicated integrating plasmids used to complement deletions of the sigM locus. Wild-type and nonsense mutant SigM or anti-SigM derivatives were each expressed from the native promoter. SigM was additionally expressed constitutively from a promoter that also enabled Atc-control when TetR was coexpressed. (C) Conjugal DNA transfer as measured by marker transfer from a wild-type donor into MKD8 recipient strain derivatives. Wild-type, ΔsigM, Δanti-sigM, and full locus deletions were tested. Complementation by constitutive SigM overexpression (up arrows) or nonsense mutants (asterisks) is indicated below.

In M. smegmatis, the predicted ORF for anti-sigM overlaps the upstream sigM gene by 50 nt. Therefore, complete removal of the sigM coding sequence will also affect anti-sigM. To ensure an unambiguous interpretation, we first deleted the entire chromosomal sigM/anti-sigM locus. We then complemented the full deletion by cloning a fragment that included both genes and the native promoter (Fig. 1B). The complemented version restored the conjugal DNA transfer efficiency to near wild-type levels (Fig. 1C). Complementation plasmids were also created to include engineered nonsense mutations that specifically truncated either sigM or anti-sigM. The sigM nonsense mutant failed to complement, while the anti-sigM mutant (expressing wild-type sigM) complemented the full deletion of the chromosomal locus (Fig. 1). This result demonstrated that DCT in the recipient strain required SigM. Recipients with the truncated mutation of anti-sigM were not apparently constitutively activated for DCT, as transfer efficiencies were similar to wild-type recipients. This suggests that SigM activation is not rate-limiting in mycobacterial conjugation.

We had previously shown that esxUT of the conjugal donor strain, mc²155, was not responsive to coculture and esx4 genes were not needed for DCT (14). Similarly, donor sigM mutations had no effect on conjugal DNA transfer efficiencies (SI Appendix, Table S1). Because the SigM requirement for DCT is recipient-specific, a corollary hypothesis is that SigM activation is both necessary and sufficient for recipient proficiency in conjugation. We hypothesized that donor function in conjugation is to activate recipient SigM, which then induces a transcriptional program to acquire DNA. We tested this in two ways. First, we constitutively expressed SigM in the donor and then determined whether it could now act as a recipient when cocultured with a wild-type donor. In an analogous approach, we tested whether an anti-sigM mutant derivative (untethered SigM) of MKD8 could acquire DNA from a wild-type recipient. Transfer was not detected in either approach. Thus, while free SigM is necessary for DCT in the recipient, the network it activates is not sufficient to define recipient activity. Therefore, directed activation of SigM does not bypass the requirement for DCT-compatible mating pairs, determined by genes in the esx1 locus to be either donor or recipient (9).

We used ectopic overexpression of SigM to maximally induce the SigM regulon in the recipient strain. This strain also complemented the genetic deletion (Fig. 1C). This ectopic expression approach was also used to assess cross-species complementation. The M. tuberculosis sigM gene was introduced into the MKD8 ΔsigM/anti-sigM strain and tested for complementation of DCT. Constitutive expression of M. tuberculosis SigM complemented the conjugal phenotype (Fig. 1C). This indicated that the M. tuberculosis SigM transactivated all of the genes necessary for DCT in the M. smegmatis conjugal recipient. Therefore, SigM’s cognate binding site and its interaction with RNA polymerase are conserved in fast- and slow-growing mycobacteria.

SigM and Extracellular Structure.

Overexpression of SigM generated visible phenotypic effects. The donor mc²155 strain forms a robust, pellicle biofilm in standing culture, but the recipient MKD8 strain does not (Fig. 2A). Constitutive overexpression of SigM in MKD8 induced a biofilm-like pellicle in this strain (Fig. 2A). Colony morphology also changed. The colony morphology for wild-type MKD8 is symmetric, smooth, and shiny. Overexpression of SigM shifted the colony morphology to an asymmetric, corrugated, and dry appearance, grossly similar to the mc²155 conjugal donor strain (Fig. 2B). We hypothesized that SigM modifies colony morphology by activating the ESX-4 secretion apparatus. In a strain with a deletion of the seven-gene esx4 locus, we found that SigM overexpression did not affect colony morphology. This indicated that ESX-4 mediates SigM’s effect on cell wall composition. It also indicated that SigM induces all of the genes necessary to produce a functional ESX-4 secretion apparatus, as well as the substrates needed to elicit the observed change in morphology.

Fig. 2.

Constitutive SigM expression changes colony and biofilm phenotype. (A) MKD8 becomes biofilm-proficient upon constitutive expression of SigM. (B) High-level constitutive expression of SigM alters the colony morphology of MKD8. This effect is abrogated by high-level constitutive coexpression of anti-SigM, and requires ESX-4. Colonies are shown at ∼0.5× magnification.

SigM Regulon in MKD8.

We used an inducible ectopic expression strain to identify the SigM regulon. An integrating plasmid expressing the tet repressor (TetR) was introduced into the sigM/anti-sigM deletion strain. A second plasmid encoding SigM under the control of a TetR-regulated promoter, or an empty-vector control, was also introduced. Replicate monocultures were grown under standard conditions and induced with Atc for 30 min. Total RNA was harvested for library construction and deep sequencing. Replicate SigM-induced cultures were compared with induced empty-vector controls. The RNA-seq reads were mapped to the MKD8 genome (accession no. CP027541). Twelve transcripts were identified, which were highly responsive to SigM expression and encoded 21 genes (Table 1 and SI Appendix, Fig. S1). These highly induced transcripts appear to constitute the direct regulon of SigM, as all had promoters with matches to a MEME-identified consensus −35 and −10 SigM binding site (Fig. 3 and Table 1).

Table 1.

SigM regulon: 21 genes (each row) listed in the order as they appear on 12 mRNAs

Transcript	Gene	Initiation	Stop	Fold-change	Annotation	AA	DCT	CC
1	0914	1090202	1089330	20	DUF559 family	290	NR
2	1535	1679161	1677851	45	EccD4	441	Req*	✓
2	1533	1677851	1676526	24	MycP4	436	Req	✓
3	1536	1679284	1682919	218	EccC4	1,211	Req*	✓
3	1537	1682904	1684010	951	Rv3446c family	368	Req*	✓
4	1538	1684083	1684394	2,098	EsxU	103	Req	✓
4	1539	1684420	1684716	3,181	EsxT	98	Req	✓
5	15690	1700522	1699671	273	DUF559 family	283	NR
6	1560	1704074	1704382	>10	WXG100/EspC family	102	Req	✓
6	15750	1704383	1705762	459	Hypothetical	459		✓
7	16105	1736523	1736562	>258	Hypothetical	11
8	1687	1839232	1838942	603	WXG100/EspC family	96		✓
8	1686	1838935	1837868	408	NlpC-fam	355	Req
8	1685	1837871	1837704	251	DUF2580 family	55	NR
9	21385	2278367	2278296		Hypothetical	24
9	2151	2278183	2276822	150	Transposase IstA protein	453
9	21370	2276825	2276655	116	Transposition helper prot	56
10	53960	5701789	5701520	23	Pseudo	89
10	53950	5701789	5701520	37	DUF559 family	146
11	5844	6021032	6021931	340	DUF559 family	299	NR	✓
12	6925	7097900	7097403	30	Pullulanase	165

Columns: Transcript induced by SigM (numbers correspond to Fig. 3 and Fig. S1), gene numbers (four-digit numbers derived from mc²155 ortholog number and five-digit numbers identify unique MKD8 genome number), SigM induction fold-change, length in amino acids (AA), required (Req) or not required (NR) for conjugation. Req gene mutants produce no transconjugants (<1 × 10⁻⁹), whereas NR genes exhibit wild-type chromosomal marker transfer efficiency (∼5 × 10⁻⁵). Detected in DCT coculture (CC).

^*As reported previously (14).

Fig. 3.

Sequences of the M. smegmatis SigM regulon promoters define SigM −35 and −10 binding sites. A MEME-identified logo (39) appears above, and individual promoter sequences are aligned below. The predicted translation initiation codon (ATG or GTG) for the adjacent gene is reverse-shaded near the right side of the alignment. Note that the MKD8_1535 promoter likely starts transcription at a GTG that initiates a two-residue ORF, which in turn is translationally coupled to the annotated initiation codon of MKD8_1535. Two SigM binding motifs were discernable by inspection, but not identified by standard MEME parameters. Genomic positions of all SigM target loci are listed in Table 1 and shown in SI Appendix, Fig. S1, with transcripts 2–4 mapping within esx4, and peri-esx4 transcripts 5–8 mapping near the locus.

The esx4 Locus.

The esxUT transcript was the most highly induced transcript in the M. smegmatis SigM regulon (MKD8_1538 and MKD8_1539) (Fig. 4, transcript 4, Table 1). This is consistent with studies in M. tuberculosis that found esxUT to be most responsive to SigM overexpression by microarray analysis (19–21). Our previous M. smegmatis conjugal coculture study also identified esxUT as highly induced (14). Two other SigM-induced bicistronic transcripts mapped to the esx4 locus (Fig. 4, transcripts 2 and 3, and Table 1). All ESX secretion systems include a dedicated FtsK-SpoIII ATPase, designated EccC in the current nomenclature scheme (22). EccC₄ is encoded by MKD8_1536, which is the 5′ gene of a transcript that also encodes MKD8_1537. The latter gene is conserved in all esx4 loci but is uniquely found only in esx4, not in other paralogous chromosomal esx loci. The transcript encoding these two esx4 genes in MKD8 was strongly up-regulated upon SigM overexpression (Fig. 4 and Table 1). SigM also induced another bicistronic mRNA, divergently transcribed relative to eccC₄. This transcript encodes the putative inner membrane ESX channel protein, EccD₄, and an ESX-associated mycosin-like protein, MycP₄. Our data show that six of the seven genes in the annotated esx4 locus are direct targets of SigM transcriptional activation. A putative fourth esx4 transcript that encodes MKD8_1534/eccB₄ was not SigM-induced and does not have an identifiable SigM cognate binding site upstream. The previously reported microarray data in M. tuberculosis are mostly in agreement with our observations, although the array studies did not detect induction of all esx4 genes. This presumably reflects the reduced resolution and sensitivity of microarray data compared with RNA-seq. Taken together, these data indicate that the ESX-4 secretion system is a primary target of SigM in mycobacteria.

Fig. 4.

Core esx4 locus induction by SigM. A schematic of the esx4 locus and its RNA-seq profiles are shown following a 30-min SigM induction period (green peaks in middle track) or in a ΔsigM strain (purple peaks in lower track). Scales were customized for each induced mRNA for profile accuracy, with maximum read depth for each window shown bracketed at the right; read depths in adjacent genes that exceed the maximum appear red. Read peaks above the midline map to the positive genomic strand, while those below correspond to the negative strand. SigM responsive promoters are shown using the same reverse-shaded numbers shown in Fig. 3. Blue boxes in the top track represent annotated genes, and the yellow block arrows indicate SigM-induced genes. Adjacent esx4 core genes are labeled within their respective blue box.

Previously, we deleted and tested three of the genes in the esx4 core locus: eccD₄, eccC₄, and esx4-specific MKD8_1537. All were shown to be required for DCT, leading to the generalization that all esx4 genes are needed for a functional ESX-4 system (14). Here, we deleted the remaining annotated core esx4 genes: eccB₄, mycP₄, and esxUT. In support of our previous conclusion, we found that each mutant was defective for DCT, consistent with an absolute requirement for ESX-4 function in DCT (Table 1 and SI Appendix, Table S1).

An Expanded Functional esx4 Locus.

In addition to inducing genes in the annotated esx4 locus, SigM induced additional transcripts that mapped in close proximity to the core locus (Table 1 and SI Appendix, Figs. S1 and S2), consistent with a genetic linkage. We noted that proteins encoded by peri-esx4 had predicted annotations that were consistent with ESX secretion substrates (Table 1), including two genes (MKD8_1560 and MKD8_1687) encoding putative WXG100 homologs (23). These colocalized and coregulated genes suggested that the encoded gene products function through ESX-4. We reasoned that the SigM regulon defines a functional ESX-4 locus that extends beyond the annotated core locus. This led us to hypothesize that some of these SigM regulon genes are also required for DCT.

We generated precise deletions of a subset of SigM target genes in MKD8 and tested these for recipient activity in standard DCT mating assays with a wild-type donor. While MKD8_15690 (the 1,569th annotated gene in the MKD8 genome; the five digits indicate that there is no clear ortholog in the reference mc²155 genome) and MKD8_1685 had no effect on DCT recipient activity, MKD8_1560 and MKD8_1686 were each required for DCT (Table 1 and SI Appendix, Fig. S2). Complementation experiments verified that the targeted genes were required for DCT in these deletion strains. We posit that the MKD8_1560 and MKD8_1686 deletions phenocopy mutations in the nearby annotated esx4 locus because their gene products, which have annotations consistent with secretion substrates, also function through ESX-4 during DCT. Our results indicate that definitions of a fully functional ESX-4 secretion system will extend beyond the relatively diminutive seven-gene esx4 core locus, and that the coregulated genes of the SigM regulon encode candidates for that expansion.

Cell Wall Hydrolases.

MKD8_1686 was required for DCT. The annotation predicts that it encodes a secreted cell wall-associated glycoside hydrolase belonging to the NlpC family (Table 1 and SI Appendix, Fig. S2). A peptidoglycan endopeptidase could function in cell wall remodeling. While there are homologs encoded in the M. tuberculosis genome, there are no clear orthologs. Another putative hydrolase, however, was identified in the SigM regulons of both species. MKD8_6925 (Rv3906c in M. tuberculosis) encodes a putative pullulanase-like polysaccharide debranching hydrolase. Mycobacterial orthologs of this gene are well conserved (also in Mycobacterium marinum and Mycobacterium abscessus) (SI Appendix, Fig. S3 and Table S2), but have not been characterized, likely because this gene is only expressed under the same restrictive conditions that have similarly hindered the study of ESX-4. Mycobacterial MKD8_6925 orthologs have conserved SigM binding sites in their promoters (SI Appendix, Fig. S3), suggesting that the activating stimulus for SigM to induce this hydrolase is encountered by many mycobacterial species.

Unannotated Small Proteins.

Approximately one-quarter of all mycobacterial transcripts begin with an AUG or GUG (collectively, RUG), which is the sole requirement for leaderless translation initiation (24). Strikingly, this hallmark of leaderless translation initiation is seen in 11 of the 12 SigM-induced transcripts (Fig. 3). The annotated initiation codon for MKD8_1535/eccD₄ is likely coupled by an RUGA start/stop motif to an ultrasmall (2 amino acid) leaderless ORF (yellow highlighted GTG in Fig. 3). Two leaderless RUGs initiate unannotated small ORFs: MKD8_16105 and MKD8_21385 (gene numbers ending in 5 indicate that they were overlooked by computational annotation prediction; see transcripts 7 and 9 in SI Appendix, Fig. S1). MKD8_16105 encodes a small, 11-amino acid protein and maps between the annotated MKD8_16100 and MKD8_16110 genes, near the other peri-esx4 SigM-induced genes (SI Appendix, Figs. S2 and S4). MKD8_21385 is an unannotated 5′ ORF of a transcript that also encodes MKD8_2151 and MKD8_21370. This leaderless ORF encodes a small protein of 24 amino acids (SI Appendix, Fig. S4). A function for either of these small proteins in the SigM-induced phenotype has not been investigated, but these examples further reinforce the gap in our understanding of the role of small proteins in the biology of mycobacteria (24).

Contact Dependence and Population Heterogeneity.

We reported previously that the conjugal communication activation of esxUT was contact-dependent: separation of donor and recipient cells by a porous membrane prevented activation of esxUT (14). The observed contact dependence presented the possibility that the recipient transcriptional response was restricted to those in direct contact with a donor cell. To test this hypothesis, we developed a fluorescent reporter of the esxUT transcriptional response to allow individual cell observations of intercellular signaling through the SigM response network. A translational fusion including the MKD8 esxU promoter, 5′ UTR, Shine Dalgarno sequence, and initiating methionine was cloned in-frame with an EGFP reporter gene that had been codon-optimized for mycobacteria (25) (Fig. 5A). This construct was electroporated into MKD8 cells that also expressed codon-optimized mCherry from a constitutive promoter, integrated at the L5 attB site (position 4,829,450 in MKD8). This reporter strain of MKD8 generated uniformly red fluorescence, but contact with a donor strain (constitutive blue) revealed brightly green fluorescing individual MKD8 cells (Fig. 5B). SigM activation of the recipient esxU-EGFP occurred only in those recipients with direct and extensive contact with the conjugal donor, and did not spread to other recipients nearby. This is consistent with direct contact being a logical physical prerequisite for the conjugal transfer of DNA between cells. The heterogeneity of the response also distinguishes the intercellular communication to esx4 in DCT from uniform population-wide responses typical of quorum-sensing systems.

Fig. 5.

Fluorescent reporter shows SigM activation at points of extensive donor cell contact. (A) Schematic of fluorescent protein expression in cocultured DCT strains. The MKD8 recipient strain has two integrated plasmids. One expresses mCherry from a constitutive promoter. The other integrating plasmid expresses EGFP from the esxU promoter. SigM activation will confer green fluorescence to an otherwise red cell. The mc²155 donor strain constitutively expresses enhanced blue fluorescent protein (EBFP2). (B) Confocal fluorescence microscopy image of fluorescent cocultured donor (blue) and recipients (red). Individual recipients expressing EGFP indicate SigM induction at points of donor and recipient contact. Note that not all points of contact elicit a recipient response. Imaged at 63× magnification, and an area of productive contacts is enlarged (2.6×) for clarity.

Identifying SigM Signatures in DCT Coculture.

DCT is not detected before 16 h of coculture, and reaches peak efficiencies between 18 and 24 h (10). Accordingly, our previous work assessed transcriptional responses at 24 h, and identified esxUT as a major transcriptional response in the recipient strain (14). The posttranslational activation of ECF σ-factors presents the possibility that SigM activation of its regulon is a rapid-response mechanism to the inciting contact stimulus. We hypothesized that SigM activation is an early event leading to conjugal activity. We tested this by repeating the RNA-seq coculture analysis following 6, 12, and 24 h of coculture. We focused on the responses of the MKD8 recipient strain. Genes in the esx4 core locus were readily identified among the genes that were induced at just 6 h of coculture (Fig. 6, green stars). Several other genes in the SigM regulon were also induced (Fig. 6, orange spots). Genes with insufficient expression or having no strain-specific SNPs are not displayed. The observed coculture initiation of induction of the SigM regulon was diminished by 24 h (Fig. 6, Right). This rapid SigM response mechanism is active at time points well ahead of detectable DCT events. At these pre-DNA transfer time points, the SigM regulon is a major contributor to the recipient response to donor contact. We note that the individual cell transcriptional response to donor contact must be very high, as the subset of activated recipient cells visualized in Fig. 5 are averaged over the total population in Fig. 6.

Fig. 6.

Time course shows that the SigM regulon is a rapid and major contributor to the recipient transcriptional response to DCT coculture. Recipient transcripts identified by informative SNPs from cocultured RNA pools were compared with their levels in time-matched recipient monocultures. Changes in annotated genes with adjusted P values less than 1.0 were plotted. Adjusted P value significance elevates responding genes in the volcano plot. Adjusted P values less than e⁻²⁰ are binned at that maximum stringency (red dashed line). The degree of induction shifts genes to the right (log₂ scale). Genes in the esx4 core locus are shown as green stars and labels, and noncore SigM-regulon genes as orange dots with black labels.

Discussion

Bacteria must detect and appropriately respond to other bacteria in their immediate environment. Recently, we demonstrated that conjugal donors and recipients altered transcriptional networks upon coculture with each other. Specifically, we identified the induction of esxUT, a transcript that encodes putative secretion substrates of the ESX-4 secretion system, in the recipient strain (14). Here, we use this recipient esxUT response as a biomarker of intercellular communication to begin to elucidate the signaling network. ECFs are held inactive at the inner membrane by their dedicated transmembrane anti–σ-factor (26). In response to a stimulus, the anti-σ releases the tethered σ-factor, freeing it to activate transcription of its regulon. Overexpression of SigM in M. tuberculosis had previously been found to induce the orthologous esxUT genes (19–21). We show that cell–cell contact in M. smegmatis is an activating stimulus for SigM, and that SigM and some of its transcriptional targets are required for DNA uptake in the conjugal recipient. Our data indicate that SigM-induced genes support ESX-4 function and contribute activities required for DNA transfer between mycobacteria.

SigM induced orthologous genes in both M. tuberculosis and M. smegmatis, suggesting that a conserved regulon generates a conserved response to the initiating SigM stimulus (SI Appendix, Table S2). M. tuberculosis SigM functionally complemented a sigM deletion in the conjugal M. smegmatis recipient (Fig. 1C). In addition, SigM binding sites were found upstream of orthologs of the M. smegmatis SigM-regulated genes in other mycobacterial species, including M. abscessus and M. marinum, supporting conservation of the SigM-controlled ESX-4 response throughout mycobacteria (SI Appendix, Fig. S3). The genes most highly induced by SigM were esxU and esxT. The higher resolution and sensitivity provided by RNA-seq identified other esx4 genes as transcriptional targets of SigM. Our data also show that the profound effect of SigM on colony morphology required an intact esx4 locus. Collectively, these findings identified the ESX-4 secretion system as a primary transcriptional target of SigM in mycobacteria.

One of the remarkable features of mycobacteria is their unusually complex cell envelope. In contrast to most Gram-positive bacteria, mycobacteria are surrounded by a lipid-rich envelope that contains an outer (myco)membrane composed of mycolic acid-lipids (27, 28). Thus, their cell wall structure more closely resembles those of Gram-negative diderms, and it is thought to constitute an additional barrier against antimycobacterial agents (29). This specialized cell envelope requires specialized secretion systems. ESX secretion systems have been consistently associated with mycobacterial cell envelope synthesis, permeability, and related characteristics, such as colony morphology (30, 31). The SigM regulon includes putative hydrolases, MKD8_1686 and MKD8_6925, which may be secreted by ESX-4. The enzymatic activities and specificities of these proteins, whether they are secreted by ESX-4, and whether they remodel the cell wall of the activated recipient or the cell wall of the contacting donor, are all important questions. Some alterations in cell wall composition or structure may also be a secondary response to ESX-4 activation. The PDIM system in M. tuberculosis was observed to be down-regulated upon SigM expression, consistent with a feedback mechanism that reduced PDIM lipid composition of the cell wall (20). This secondary PDIM response is suggestive of compensatory responses to SigM-initiated cell wall remodeling in mycobacteria.

Four of the 12 SigM target mRNAs in MKD8 transcribe single genes that encode homologous hypothetical proteins: MKD8_0914, MKD8_15690, MKD8_53960-53950 (frame-shifted split pseudogenes in MKD8), and MKD8_5844. These belong to a superfamily of proteins conserved in mycobacteria that are predicted to include a C-terminal DUF559 nuclease domain (SI Appendix, Fig. S5). While the annotations associated with this superfamily suggest a possible endonuclease activity by inference with homologs in other bacteria, this has not been assessed for mycobacterial members. If these mycobacterial homologs possess nuclease activity, we speculate that this component of a SigM response could function in acquiring donor chromosome fragments during conjugation, in providing eDNA to support biofilm formation, or in modulating the host response (32, 33). The potential for functional redundancy within this large family could complicate analyses of their contributions to the SigM regulon. Deletion of individual SigM-induced DUF559 genes had no effect on DCT efficiencies (Table 1).

Some differences exist between ectopic SigM and contact-initiated transcriptional responses. Of note, gene MKD8_1534 encodes EccB₄, and is the only exception to SigM’s direct activation of the full esx4 core locus. The eccB₄ gene lacks a discernible SigM binding motif in its promoter, and was not transcriptionally activated by SigM in a 30-min Atc induction period. However, in coculture, eccB₄ shows transcriptional increases comparable to the other members of the core locus (Fig. 6), indicating that it is coregulated with the rest of the locus in a conjugal context and time frame. A parallel communication pathway may activate eccB₄, but we speculate that it is more likely to be indirectly activated by a product of the primary SigM regulon. The esx4-dependence of the SigM-induced colony morphotype switch supports this speculation. Expression of the SigM regulon during hours of coculture, compared with 30 min of Atc induction, could amplify disparities in the relative synthesis and degradation rates that determine the steady-state abundance of SigM regulon transcript levels. Our esxU-EGFP analysis showed a high level of SigM induction in only a small fraction of recipients. The response of a small subpopulation reduces the sensitivity of population-averaged quantitative assays such as RNA-seq. A heterogeneous signal thus diluted could obscure biologically important responses of SigM regulon mRNAs that are not as highly transcribed or as stable as eccC₄ or esxUT.

While the five paralogous chromosomally encoded ESX secretion systems likely operate similarly, phenotypes that arise from single ESX mutations indicate they have nonredundant functions. ESX-4 is the widely distributed and conserved progenitor of the ESX secretion system family (16). The annotated esx4 locus consists of seven genes in most mycobacteria, with an eighth (eccE₄) found in M. abscessus (34). The esx4 locus has been solely defined by gene homologs it shares with the other paralogous esx loci (22). With DCT as a functional assay of ESX-4 activity, we can more fully identify the gene products that have ESX-4 roles. Our data show that deletion of genes in the esx4 core locus, as well as other SigM-induced genes near esx4, are required for conjugation. The most parsimonious explanation is that they all contribute to ESX-4 function. We propose that the presently annotated seven-gene esx4 locus will be expanded to include other SigM-coregulated genes required for ESX-4 function and DCT. We also suggest that the satellite genes of an expanded functional esx4 locus were once clustered nearer to the core locus, but have drifted apart over evolutionary time-scales of this ESX progenitor. In support of this, the M. tuberculosis ortholog of MKD8_1560, Rv3440c, is located just four genes away from the annotated esx4 locus (Rv3444c-Rv3450c) in that species.

M. tuberculosis is the defining member of a pathogenic cluster of mycobacteria, termed the Mycobacterium tuberculosis complex (MTBC). Most members of the MTBC are host-adapted clonal siblings of M. tuberculosis that show little evidence of horizontal gene transfer or genome mosaicism consistent with DCT (35). Mycobacterium canettii is a more distantly related member of the MTBC distinguished by its smooth colony morphology and by genetic heterogeneity. Genomic comparisons between M. canettii strains revealed mosaicism strikingly similar to that generated by experimental DCT in M. smegmatis (36). DCT has been experimentally demonstrated between strains of M. canettii (37), showing that the requisite stimulus and response pathways are present in this pathogenic species. One evolutionary model posits that a historic DCT event between M. canettii strains produced a particularly effective pathogenic transconjugant that was the founder of the extant clonal members of the MTBC (38). Based on our DCT studies with M. smegmatis, we would predict that the stimulus and response pathways mediated by ESX-1, ESX-4, and SigM play similar roles in conjugal M. canettii isolates.

Even though we describe a role for ESX-4 in the process of conjugation, how it enables DNA transfer is unclear. It may provide a function specific to conjugation (e.g., establishes mating-pair formation), or it may provide a more general role (e.g., modifies the recipient cell wall to facilitate passage of DNA). These possibilities are consistent with macroscopic observations of augmented biofilm formation and of rough colony morphology upon ESX-4 induction. Recently published work in M. abscessus found that ESX-4 subverts lysophagosomal maturation in macrophage and amoeba models of infection (34). Notably, an insertion in the M. abscessus sigM gene phenocopied mutations in esx4, further supporting our hypothesis that SigM’s role in activating esx4 is conserved across mycobacterial species. We hypothesize that some secretion substrates will confer activities conserved throughout the genus, while others may have species-specific roles. Characterizing the ESX-4 secretome, as well as the molecular composition and structural changes of the modified cell wall, will elucidate ESX-4’s fundamental role in mycobacterial biology that has ensured its evolutionary conservation.

Whatever ESX-4’s role, the secretion apparatus clearly requires SigM for expression. Therefore, the extracytoplasmic domain of anti-SigM represents the sensor for the stimulus that ultimately drives the ESX-4 secretion system response. In conjugating mycobacteria, direct contact with the conjugal donor initiates that cascade. An emerging model posits that the donor cell surface triggers anti-SigM on the periplasmic side of the recipient inner cell membrane to release the SigM bound to its cytoplasmic domain (Fig. 7). The intercellular communication may be indirect, as the periplasmic surface of the inner cell membrane is not likely to be exposed to a neighboring donor cell surface. In the context of conjugation, donor contact initiates SigM-induction of ESX-4, resulting in the recipient cell taking up donor chromosomal DNA. SigM-directed transcriptional activation likely leads to the secretion of recipient ESX-4 substrates, and a reasonable possibility is that these secreted effector proteins act on, or communicate back to, the contacting donor. We envision that physical contact initiates reciprocal communication exchanges. The conservation of SigM and ESX-4 across mycobacteria suggests this stimulus and response network is both active and evolutionarily advantageous.

Fig. 7.

SigM identifies a central link in the interaction network between conjugal strains of mycobacteria. Previous data showed that ESX-1 determines the mating identity of the donor and recipient strains, possibly by decorating their cell surface with mating recognition ligands and receptors (red octagon, 1). Previous data also showed that ESX-1 was upstream of the ESX-4 response in the recipient, but how it is required for SigM activation is unknown (red octagon, 2). Contact incites an unknown stimulus for the anti-sigM (red octagon, 3). The present data connects contact at the cell surface to a biologically appropriate response through SigM (yellow diamond, 4), which induces target genes (circled M and arrow). ESX-4 expression alters colony morphology and is required for donor DNA uptake; we speculate that SigM regulon-encoded activities include hydrolases and nucleases, which could have roles in those respective phenotypes (red octagon, 5).

Methods

Details of cell lines, plasmid constructs, and molecular genetic techniques (for mutagenesis, conjugation fluorescent imaging, and RNA analyses) are all provided in SI Appendix.