General secretion signal for the mycobacterial type VII secretion pathway

Abstract

Mycobacterial pathogens use specialized type VII secretion (T7S) systems to transport crucial virulence factors across their unusual cell envelope into infected host cells. These virulence factors lack classical secretion signals and the mechanism of substrate recognition is not well understood. Here we demonstrate that the model T7S substrates PE25/PPE41, which form a heterodimer, are targeted to the T7S pathway ESX-5 by a signal located in the C terminus of PE25. Site-directed mutagenesis of residues within this C terminus resulted in the identification of a highly conserved motif, i.e., YxxxD/E, which is required for secretion. This motif was also essential for the secretion of LipY, another ESX-5 substrate. Pathogenic mycobacteria have several different T7S systems and we identified a PE protein that is secreted by the ESX-1 system, which allowed us to compare substrate recognition of these two T7S systems. Surprisingly, this ESX-1 substrate contained a C-terminal signal functionally equivalent to that of PE25. Exchange of these C-terminal secretion signals between the PE proteins restored secretion, but each PE protein remained secreted via its own ESX secretion system, indicating that an additional signal(s) provides system specificity. Remarkably, the YxxxD/E motif was also present in and required for efficient secretion of the ESX-1 substrates CFP-10 and EspB. Therefore, our data show that the YxxxD/E motif is a general secretion signal that is present in all known mycobacterial T7S substrates or substrate complexes.

Most pathogenic mycobacteria, such as Mycobacterium tuberculosis, the etiological agent of human tuberculosis, are facultative intracellular pathogens that primarily infect host macrophages (1). Similar to other intracellular pathogens, mycobacteria secrete proteins that manipulate host cellular processes, thereby creating an environment favorable for intracellular survival and replication (2). Gram-negative bacterial pathogens often use specialized secretion systems, such as the type III and type IV secretion systems, to secrete virulence factors directly into host cells (2). Mycobacteria have a highly unusual cell envelope (3) and therefore need an accompanying specialized secretion system. Such a system has recently been identified in mycobacteria and other bacteria with similar cell envelope organization and is now known as the ESX or type VII secretion (T7S) pathway (4). This pathway is responsible for secretion of multiple proteins that lack classical signal peptides in mycobacteria (5–8).

Mycobacterial genomes contain up to five paralogous ESX loci, named ESX-1 to ESX-5 (9, 10). The archetypical T7S system, ESX-1, is responsible for the secretion of several proteins, including the ESX-1–encoded T-cell antigens ESAT-6 (early secreted antigenic target of 6 kDa) and CFP-10 (culture filtrate protein of 10 kDa) (7, 11–15). ESX-1 and its substrates play an important, albeit not yet fully understood role in M. tuberculosis virulence. For instance, the primary cause of attenuation of the Mycobacterium bovis bacille Calmette–Guérin vaccine strain is attributed to a deletion of nine genes from the ESX-1 locus (7, 16, 17). Consistently, deletion of ESX-1 genes from the genomes of M. tuberculosis or M. bovis results in diminished virulence (18, 19). Of the other ESX loci, both ESX-3, which is involved in iron acquisition, and ESX-5 have been demonstrated to encode functional secretion systems (20–22).

Four of the ESX clusters contain members of the pe and ppe gene families (9, 10), both of which are unique for mycobacteria and some closely related species. Intriguingly, whereas nonpathogenic mycobacteria generally only carry a small number of pe/ppe genes, these gene families are highly expanded in the genomes of M. tuberculosis and the fish pathogen Mycobacterium marinum (23, 24). The pe and ppe genes are named after Pro-Glu (PE) and Pro-Pro-Glu (PPE) motifs, located in the conserved ∼110 (PE) and ∼180 (PPE) amino acid N-terminal domains of their gene products. Sometimes, these N-terminal domains constitute all of the PE/PPE proteins, but they can also be fused to highly variable C-terminal domains (23). Although they are associated with mycobacterial virulence, the function of the PE and PPE proteins is largely unknown (25). Many PE and PPE proteins are exported to the bacterial cell surface or the culture supernatant (20, 21, 26–29). The vast majority of these intriguing proteins are exported by the most recently evolved T7S system, i.e., ESX-5 (21). Like other T7S substrates, PE and PPE proteins lack classical signal peptides (4). Here, we investigate which sequences are required for export of PE and PPE proteins. We find that PE proteins share a short conserved secretion signal at their C termini. However, this signal does not determine ESX system specificity. Furthermore, we show that the PE secretion signal is shared by and required for secretion of the ESX-1 substrates CFP-10 and EspB. Therefore, we conclude that mycobacterial T7S substrates and substrate complexes are targeted for secretion by dual signals: (i) a general T7S signal and (ii) a yet unknown system specificity signal.

Results

PE25 and PPE41 Are Secreted as a Dimer by ESX-5 in M. marinum.

Previously, PPE41 was shown to be exported by ESX-5, both in its native host M. tuberculosis (30) and in M. marinum upon heterologous expression (20). The PPE41 gene forms an operon with PE25 (31) and both proteins form a heterodimer when expressed in Escherichia coli (32). To investigate whether PE25 is also an ESX-5 substrate, we generated a PE25–PPE41 operon in which the encoded PE25 protein was labeled with an HA epitope at its C terminus. This construct was introduced in a WT strain of M. marinum and its corresponding ESX-5 mutant strain, and secretion was analyzed. Immunoblot analysis showed that both PE25-HA and PPE41 were secreted in the WT strain, but hardly in the ESX-5 mutant (Fig. 1A), similar to what has been described for PPE41 (20, 21). This secretion defect was not accompanied by an increase in cellular levels of PE25 and PPE41. This phenomenon is often observed for ESX-1 and ESX-5 substrates and indicates that the stability of the substrates depends on the presence of the T7S systems in the bacterial cell (5, 11, 13, 33). PPE41 was detected in two forms, a 25-kDa band, which represents full-length PPE41, and a smaller form of ∼19 kDa, which probably represents degraded PPE41 (20). As previously observed, only full-length PPE41 was secreted (20). The cytosolic control protein GroEL2 was not detected in the culture supernatant fractions, confirming the integrity of the cells. Together, these results show that PE25 is also an ESX-5 substrate.

Fig. 1.

PE25 and PPE41 are secreted as a dimer by ESX-5. (A) Immunoblot analysis of PE25-HA, PPE41, and the intracellular control GroEL2 in cell pellet (P) and culture supernatant (S) fractions of M. marinum WT and its ESX-5 mutant (ESX5::Tn). Equivalent OD units are loaded. Full-length PPE41 is marked by an arrow (←). (B) Immunoblot showing PE25-HA and its interacting partner PPE41 from M. marinum culture filtrate, after immunoprecipitation (IP) on beads coated with an anti-HA mAb. Monomeric PE25-HA and PPE41 are marked by asterisks (*). Detection of the ESX-5 substrate EsxN and the Sec-dependent Apa protein exclusively in the flow-through (FT) fraction confirms the specificity of the interaction. Both PE25-HA and PPE41 were also detected at higher molecular weights than their predicted masses, which could be due to complex formation or residual detergent in these fractions. W, washing fraction. (A and B) PE25-HA was detected with anti-HA mAb and PPE41 with polyclonal antiserum.

Previously, it was shown that PPE41 requires coexpression of PE25 for stable production and secretion (20). Similarly, PE25 is only expressed and secreted if a plasmid containing the entire operon is introduced in M. marinum, whereas it is not detectable if the PE25-encoding gene is introduced alone. These observations imply that PE25 and PPE41 form a complex before or during translocation. To examine whether these two proteins are secreted as a complex, we purified HA-tagged PE25 from culture supernatants with anti-HA agarose. In these experiments, PPE41 was copurified, whereas two control proteins, the ESX-5 substrate EsxN and the general secretory (Sec) pathway-dependent Apa protein, were not (Fig. 1B). In conclusion, these data indicate that PE25 and PPE41 are likely secreted as a heterodimer.

C Terminus of PE25 Is Required for Secretion of PE25/PPE41.

The N terminus of PE25 and the C termini of both PE25 and PPE41 are absent from the crystal structure of the dimer (32). Some of these flexible tails contain several highly conserved amino acids. To investigate the role of these regions in secretion, we monitored the effect of a series of small N- and C-terminal deletions in both proteins in M. marinum. This analysis showed that deletions in the N termini of both proteins or in the C terminus of PPE41 did not affect secretion (Fig. 2A, lanes 3 and 4 and Fig. S1A). However, deletion of 7 or 15 amino acids from the extreme C terminus of the 99-aa-long PE25 protein completely abolished secretion of both PE25 and PPE41 (Fig. 2A, lanes 5–10). Because both truncated PE25 and PPE41 were identified in the cell pellet we reasoned that dimer formation and protein stability were not significantly affected by the C-terminal deletions of PE25. To confirm this finding, we expressed these mutated operons also in E. coli (Fig. S2A, lanes 3 and 4) and purified PE25 again using anti-HA agarose. This experiment showed that PPE41 could still be copurified with the truncated forms of PE25 (Fig. S2B, lanes 2 and 3), indicating that the interaction between these proteins was indeed not affected by the deletions. Next, we also determined whether the HA tag, which is located adjacent to the deleted C-terminal region, had an affect on PE25/PPE41 secretion. To this end, we produced C-terminal deletions of PE25 in a WT PE25-PPE41 operon. Because similar secretion defects were observed when deletions of 7 or 15 amino acids from the PE25 C terminus were generated in a construct lacking the HA tag, the HA tag does not interfere with secretion (Fig. S3). To examine whether these results would be similar in the native host species, we also introduced the plasmids with the operon encoding HA-labeled WT PE25 or PE25 lacking the last 7 or 15 residues in M. tuberculosis and analyzed secretion (Fig. S4, lanes 1–8). M. tuberculosis expresses and secretes endogenous PPE41 (30), and consequently PPE41 was detected in the supernatant of strains with and without plasmids. However, introduction of the WT operon on a plasmid resulted in elevated PPE41 levels in the culture supernatant, whereas in strains harboring plasmids with truncated forms of PE25, the PPE41 levels in the culture supernatant were reduced to background levels. Consistently, detection of plasmid-encoded PE25 with a mAb recognizing the HA epitope confirmed that full-length PE25 with an HA tag was efficiently secreted into the medium, whereas the truncated forms remained in the cell pellet. Therefore, it can be concluded that the extreme C terminus of PE25 is required for secretion of both PE25 and PPE41 in various mycobacterial species.

Fig. 2.

Identification of a motif required for secretion via ESX-5. (A, D, and F) Immunoblot analysis visualizing the effects of (A) N- or C-terminal truncations in PE25-HA and PPE41 or point mutations in (D) PE25-HA and (F) C-terminally HA-labeled LipY_tub (LipY_tub-HA), on protein translocation in WT M. marinum. PE25 and LipY_tub were detected with HA antiserum, PPE41 with a polyclonal Ab, and GroEL2 was included as control for cell lysis. Equal OD units of fractions containing cell pellets (P) and culture supernatants (S), and in (F) Genapol-treated pellets (Gp) and Genapol surface extracts (Gs), are shown. In F, bands representing full-length (*) and processed (◀) LipY_tub-HA are marked. (B) Multiple sequence alignment visualized as a sequence logo to emphasize conserved residues within the C termini of PE proteins. (C and E) Mutation analysis of the C termini of (C) PE25-HA and (E) the PE domain of LipY_tub-HA, with altered residues (one-letter code) followed by the amino acid into which they were mutated. Mutated residues are bold.

Highly Conserved Motif in the Flexible C Terminus of PE25 Is Critical for Secretion of PE25/PPE41.

The results of our deletion analysis suggested that the last 7 amino acids of PE25 are important for ESX-5 secretion. To identify which of these residues were important for secretion, an alignment was generated with protein sequences of other ESX-5–dependent PE proteins (Fig. 2B). This alignment showed that the sequence conservation was in fact quite low in this region. On the basis of this alignment we decided to mutate four conserved or distinctive residues: Asn-94, Ile-95, Lys-96, and Phe-98 (Fig. S1B). However, individual substitutions of these amino acids in PE25 did not impair secretion (Fig. S1C).

The sequence alignment did reveal the presence of two highly conserved residues that were located a little further upstream, i.e., a tyrosine located 12 amino acids from the C terminus of PE25 and a negatively charged residue (Asp or Glu) at 8 amino acids from the C terminus (Fig. 2B). To address whether these residues play a role in secretion, they were independently replaced by Ala residues (Fig. 2C). Strikingly, upon introduction in M. marinum, each substitution completely abolished secretion of both PE25 and PPE41 (Fig. 2D, lanes 3–6). Similar results were obtained upon introduction of these constructs in the native host M. tuberculosis (Fig. S4, lanes 9–12).

We noticed that the Tyr and Asp/Glu residues are invariably spaced by 3 amino acids of relatively low conservation (Fig. 2B). Therefore, we hypothesized that the spacing between these residues could be important for secretion. To test this hypothesis, a construct was generated in which Ala-90 of PE25 was deleted, leaving only two residues between Tyr-87 and Glu-91 (Fig. 2C). In addition, a second construct was created in which an extra Ala was introduced (Fig. 2C). Both mutations completely abrogated secretion of PE25 and PPE41 (Fig. 2D, lanes 7–10). Importantly, the mutated versions of PE25, and PPE41, were detected in the pellet fractions of both M. marinum (Fig. 2D, lanes 3, 5, 7, and 9) and E. coli (Fig. S2A, lanes 5–8). As both PE25 and PPE41 are unstable in the absence of their partners (20, 32), this observation suggests that the heterodimer remained intact when the four critical point mutations were introduced. Moreover, successful coimmunoprecipitation of PPE41 with PE25_Y87A-HA from E. coli lysates again confirmed that the mutation did not affect the complex stability (Fig. S2B, lane 4). Taken together, these results show that PE25/PPE41 secretion via ESX-5 requires a YxxxD/E motif within the flexible C terminus of PE25.

ESX-5 Secretion of M. tuberculosis LipY Depends on a Similar Conserved Motif.

To investigate whether this secretion motif is also required for other ESX-5 substrates, we analyzed M. tuberculosis LipY (LipY_tub). LipY_tub is a lipase that is secreted via ESX-5, and the only PE protein with a known function (33, 34). Although LipY_tub contains a large C-terminal domain that exhibits lipase activity, it also has a complete N-terminal PE domain, including the conserved YxxxD/E motif (Fig. 2E). This PE domain is proteolytically removed upon transport (33). To determine whether the YxxxD/E motif is required for secretion, Y88A and E92A point mutations were generated. Both mutations abrogated translocation to the bacterial surface and concomitant processing of LipY_tub in WT M. marinum (Fig. 2F). These results confirm that the YxxxD/E motif forms a conserved ESX-5 secretion signal.

YxxxD/E Motif Is also Required for Secretion via ESX-1.

In pathogenic mycobacteria, several different ESX systems exist and we were interested in identifying the secretion signal for these systems as well. Recently, we identified a putative ESX-1–dependent PPE substrate by proteomic analysis, i.e., M. marinum PPE68_1 (hereafter referred to as MmPPE68_1) (15). MmPPE68_1 is adjacent to PE35 (hereafter referred to as MmPE35) in a two-gene operon (Mmar_0185–Mmar_0186). To investigate whether these proteins are indeed ESX-1 substrates, this operon was placed under control of the hsp60 promoter and MmPE35 was modified to express a C-terminal HA tag. Western blot analysis demonstrated that MmPE35-HA is secreted when expressed in WT M. marinum and its ESX-5 mutant, but not in an ESX-1 mutant strain (Fig. 3A). In the ESX-1 mutant the detection levels of MmPE35 were lower than in the WT and ESX-5 mutant strains. As discussed above for PE25/PPE41, this is likely due to MmPE35’s instability in the absence of a functional ESX-1 system (5, 11, 13, 33). These results show that MmPE35 is indeed secreted via ESX-1.

Fig. 3.

Secretion of MmPE35 via ESX-1 requires C-terminal Tyr and Asp residues. (A and B) Immunoblot analysis of MmPE35-HA, detected with the HA mAb, and intracellular GroEL2 in (A) M. marinum WT, ESX-1 mutant (ESX1::Tn), and ESX-5 mutant (ESX5::Tn) strains expressing WT MmPE35-HA and MmPPE68_1 and (B) WT M. marinum expressing WT, truncated, or mutated forms of MmPE35-HA and MmPPE68_1. Equivalent OD units of cell pellets (P) and culture supernatants (S) are shown. (C) Alignment of C-terminal sequences of PE25 and MmPE35. The residues required for PE25 secretion are (partly) conserved in MmPE35.

Next, deletions were generated in MmPE35 and/or MmPPE68_1 to determine which domains are required for PE/PPE secretion via ESX-1. This analysis showed that removal of the extended C-terminal domain of MmPPE68_1 (MmPPE68_1_Δ191C) had no effect on secretion of MmPE35-HA in M. marinum (Fig. 3B, lanes 5 and 6). However, deletion of 15 residues from the C terminus of MmPE35 abrogated secretion, either when expressed with full-length or with C-terminally truncated MmPPE68_1 (Fig. 3B, lanes 3, 4, 7, and 8).

Interestingly, MmPE35 has, similar to PE25, a YxxxD/E motif near its C terminus (Fig. 3C). Single-amino-acid substitutions of Tyr-86 and Asp-90 for Ala residues severely diminished (Y86A) or completely abrogated (D90A) secretion of MmPE35 (Fig. 3B, lanes 9–12). Thus, similar to the ESX-5 substrates PE25/PPE41 and LipY_tub, the ESX-1 substrate MmPE35 requires the YxxxD/E motif for secretion.

C Termini of PE25 and MmPE35 Are Functionally Equivalent.

The observation that a conserved motif is required for secretion of both ESX-1 and ESX-5 substrates raises the question of whether this signal determines system specificity. To investigate this question, hybrid proteins were produced in which the last 8, 12, or 15 residues of MmPE35 were exchanged for the corresponding parts of PE25. Subsequently, these chimera were expressed in the M. marinum WT, ESX-1, and ESX-5 mutant strains, and secretion was analyzed. Curiously, although all MmPE35 C-terminal swaps were efficiently secreted, they were still exported by ESX-1 (results for all three swaps are exemplified with the 15-amino acid substitution in Fig. 4A). To confirm these data, we also produced the reciprocal mutation, in which the last 15 residues of PE25 were substituted for the last 15 residues of MmPE35. Indeed, this fusion was also efficiently secreted, but again the modification did not alter the secretion route of this ESX-5 substrate (Fig. 4B). These data show that, whereas the presence of the YxxxD/E motif-containing C terminus is absolutely required for secretion, this part does not determine the system specificity.

Fig. 4.

The C termini of PE25 and MmPE35 are functionally equivalent. The last 15 residues of MmPE35 were replaced by the equivalent part of PE25 (MmPE35_C15(PE25)-HA), and vice versa (PE25_C15(MmPE35)-HA). (A and B) Fractions containing equivalent OD units of cell pellets (P) and culture supernatants (S) of M. marinum WT, ESX-1 mutant (ESX1::Tn), and ESX-5 mutant (ESX5::Tn) strains, expressing (A) MmPE35_C15(PE25)-HA and (B) PE25_C15(MmPE35)-HA were separated by SDS/PAGE and analyzed by immunoblotting for the presence of HA-tagged chimera using the HA mAb, PPE41 using polyclonal antiserum, and GroEL2 as a control for cell lysis.

To identify the signal that is required for ESX secretion system specificity, a number of larger swaps were produced. However, these swaps were undetectable on Western blot, likely as a result of unstable protein complexes. The only stable complex was observed when the flexible N terminus and the first α-helix of PE25 (in total 44 amino acids) were exchanged for the corresponding part of MmPE35. However, this complex was only detected in the cell pellet fraction and not secreted by M. marinum (Fig. S5). This means that the second signal cannot be easily identified and could be a conformational signal involving different parts of the complex. Future experiments will be directed to identify the domain that determines specificity for T7S systems in mycobacteria.

YxxxD/E Is a General T7S Motif That Can Be Used to Predict T7S Substrates.

Previously, the ESX-1 substrates ESAT-6 and CFP-10 were shown to be secreted as a heterodimer, depending on a signal located in the C-terminal seven residues of CFP-10 (35). Surprisingly, whereas the last seven residues of CFP-10 are unique and could not be used to identify other substrates (35), this protein also has a YxxxD/E motif near the C terminus that aligns perfectly with the PE secretion motif (Fig. 5A). We hypothesized that this motif is critical for secretion and set out to test this hypothesis by using a C-terminally HA-tagged CFP-10 protein and a derivative in which Tyr-83 was replaced by an Ala residue. Upon introduction in M. marinum, secretion was markedly reduced in the Y83A mutant (Fig. 5B), indicating that the YxxxD/E motif is indeed part of and extending the previously identified CFP-10 secretion signal.

Fig. 5.

A general T7S signal. (A) Alignment of the secretion motif for a number of T7S substrates. The conserved Y and D/E residues are highlighted in red. (B–C) Immunoblot analysis of M. marinum WT expressing HA-labeled WT and mutated versions of (B) CFP-10 and (C) EspB. Fractions containing cell pellets (P) and 10-fold more OD units of culture supernatants (S) are shown. CFP-10 and EspB were detected with the HA mAb, and GroEL2 was included as lysis control. Full-length CFP-10-HA and HA-EspB are marked by an asterisk (*), and C-terminally processed forms of HA-EspB by an arrowhead (◀). (D) Model for ESX secretion, in which T7S substrates are recognized by dual signals: (i) a conserved secretion signal located approximately at position 80–95 in PE/CFP-10/EspB (red box), and (ii) a yet unidentified signal that determines system specificity. IM, inner membrane; ML, mycolic acid containing layer; lipase domain, C-terminal domain of LipY.

Interestingly, another unrelated ESX-1 substrate, i.e., EspC (12), also contains a C-terminal YxxxD/E motif (Fig. 5A). The YxxxD/E motif thus appears to be shared by all known classes of T7S substrates, except for PPE and ESAT-6–like proteins that are secreted as a complex with a motif-containing substrate. Therefore, we reasoned that the YxxxD/E motif might be used to predict putative unknown T7S substrates. Notably, the overall sequence conservation is very low between the known T7S substrate families, and homology-based searches resulted in hits in PE and CFP-10 proteins but failed to identify other sequences. Therefore, we developed a method that could identify proteins that contain an amino acid pattern, without taking into account the remaining primary amino acid sequence. This search pattern was derived from a multiple alignment of the amino acid sequences of all PE and CFP-10 proteins of M. tuberculosis H37Rv that contain the YxxxD/E motif (Fig. S6). Furthermore, in PE and CFP-10 proteins, the YxxxD/E motif is found approximately at position 80–95, in a flexible region that follows a typical helix-turn-helix structure. Therefore, only those proteins that contained the pattern at a similar position, in a region preceded by a predicted helix-turn-helix secondary structure, were selected. Using this approach, all putative proteins of M. tuberculosis H37Rv were screened. This resulted in the identification of, apart from all of the PE and CFP-10 proteins in Fig. S6, a relatively small number of other proteins (Table S1). Interestingly, one of the identified proteins was encoded by Rv3440c, a gene located near the ESX-4 locus that is coregulated with ESX-4 genes by SigM (36), indicating that this protein is a putative ESX-4 substrate. Other proteins belong to the Pfam esxAB clan (37) and are therefore also putative secretion substrates. Another likely substrate is ESX-1–encoded EspJ, which was previously identified in M. tuberculosis cell envelope extracts (38). Finally, we also identified a T7S motif in EspB, one of the larger proteins secreted by ESX-1 (14). It is unknown whether EspB is recognized directly by ESX-1 or whether it is cosecreted with a small carrier protein as described for other T7S substrates (4). To verify the role of the secretion motif in one of these putative substrates, an N-terminally HA-labeled version of M. marinum EspB and a corresponding Y81A mutant were generated. Strikingly, whereas secretion and concomitant C-terminal processing (14) was observed for WT EspB in M. marinum, the Y81A mutant was predominantly located in the bacterial cell pellet (Fig. 5C).

Together these results show that the YxxxD/E motif is required for secretion not only of PE proteins, but also of CFP-10 and EspB. Therefore, we conclude that this motif constitutes a general secretion signal, which can be used to predict unknown T7S substrates.

Discussion

Most members of the large PE and PPE protein families are transported across the mycobacterial cell wall by the specialized T7S system ESX-5 (20, 21). Similar to the substrates of other Sec-independent secretion systems (39, 40), most PE and PPE proteins lack recognizable signal peptides. We have recently shown that the conserved PE and PPE domains are required for targeting PE and PPE proteins to ESX-5 (33). However, within these domains no consensus sequence has been identified that defines a secretion signal. In this study, we show that secretion of two independent ESX-5 substrates, i.e., PE25/PPE41 and LipY_tub, is dependent on a conserved motif (YxxxD/E) located at the end of the PE domain. The Tyr and Asp/Glu residues are among the most highly conserved amino acids in PE proteins, implying a conserved functional role. Notably, a deletion of the last seven residues of PE25, which did not affect the identified secretion motif, also abolished secretion. However, because individual amino acid substitutions or replacement for the HA sequence failed to define residues that were important for secretion, the precise role of this part of the protein remains unclear. Although introduction of structure-altering Pro residues had no effect on secretion, it is possible that the secondary structure is of importance or that this region is involved in the proper exposure of the YxxxD/E motif.

Although the majority of PE/PPE proteins are substrates of ESX-5 (20, 21), we show here that M. marinum PE35 is secreted via ESX-1. This PE protein also required its C-terminal YxxxD/E motif for secretion. Interestingly, the MmPE35 C terminus was found to be functionally equivalent to that of PE25. This finding indicates that although they are substrates of different ESX systems, PE25 and MmPE35 share a conserved C-terminal signal. As PE25 and MmPE35 are specifically secreted by ESX-5 and ESX-1, respectively, we would like to propose that PE proteins are targeted for secretion by dual signals: a general C-terminal ESX signal, and a yet unknown signal that determines system specificity (Fig. 5D). This is not without precedent, because dual targeting and translocation signals have been described for substrates of other specialized secretion systems, e.g., the type I and type IV systems (41, 42).

Inspection of the amino acid sequences of known ESX-1 and ESX-5 substrates does not reveal any obvious conserved motifs that might determine system specificity. It is thus possible that system specificity results from a conformational signal, which could be located in another part of the PE protein. Indeed, analysis of the secretion of another ESX-5 substrate, PE_PGRS33, showed that this protein requires the first 30 residues of its PE domain for secretion (43). It is also possible that the specificity signal is located in the PPE partner protein or that it is formed by a combination of parts of the PE/PPE complex. Although many PE and PPE proteins are encoded by single genes on the genome, bioinformatic analysis proposed that they might have cognate interaction partners that are encoded elsewhere in the mycobacterial genome (44). In this work, attempts to identify the specificity signal by swapping large parts of the PE and/or PPE proteins of PE25/PPE41 and MmPE35/MmPPE68_1 were unsuccessful due to lack of secretion or instability of the protein chimeras. As PE/PPE proteins appear to have preferred partners (44), this was likely a result of disrupted interactions between the cognate PE/PPE pairs.

Both the PE/PPE and the CFP-10/ESAT-6 proteins form heterodimeric complexes that are composed of elongated bundles of α-helices (32, 45). These structural similarities point to a functional and/or evolutionary link between these major mycobacterial protein families. Here we identify an additional link between the different classes of T7S substrates, as we show that the PE secretion motif is shared by members of the CFP-10 protein family. Furthermore, the observation that the YxxxD/E motif is required for the secretion of CFP-10 and EspB suggests that this motif not only constitutes a general PE secretion signal, but is in fact a signal conserved in all classes of mycobacterial T7S substrates or substrate complexes (Fig. 5D). Notably, a similar motif can be detected in the CFP-10 homologs of other high G+C Gram-positive bacteria with T7SSs, although the Tyr is often replaced by another aromatic amino acid, i.e., Phe (Fig. S7). This finding suggests that the motif identified in this study might represent a general secretion signal for the T7S pathway.

Using the YxxxD/E motif, we identified a number of potential unknown T7S substrates in M. tuberculosis, with the notion that proteins that are secreted as part of a complex, such as PPE and ESAT-6, are of course not identified with this method. Interestingly, the gene coding for one of the identified proteins, Rv3440c, is encoded adjacent to the ESX-4 locus, and this gene and several ESX-4 genes appear to belong to the same SigM regulon (36). Future investigations will reveal whether Rv3440c and our additional candidates are indeed true T7S substrates.

Materials and Methods

Experimental procedures including bacterial strains and growth conditions, plasmid construction, protein sample preparation, SDS/PAGE and immunoblotting procedures, and immunoprecipitation are described in SI Materials and Methods. This section also describes the bioinformatics analysis used to predict putative T7S substrates. Plasmids and oligonucleotide sequences used for plasmid construction are listed in Table S2.