C-Terminal Signal Sequence Promotes Virulence Factor Secretion in Mycobacterium tuberculosis

See allHide authors and affiliations

Science  15 Sep 2006:
Vol. 313, Issue 5793, pp. 1632-1636
DOI: 10.1126/science.1131167


Mycobacterium tuberculosis uses the ESX-1/Snm system [early secreted antigen 6 kilodaltons (ESAT-6) system 1/secretion in mycobacteria] to deliver virulence factors into host macrophages during infection. Despite its essential role in virulence, the mechanism of ESX-1 secretion is unclear. We found that the unstructured C terminus of the CFP-10 substrate was recognized by Rv3871, a cytosolic component of the ESX-1 system that itself interacts with the membrane protein Rv3870. Point mutations in the signal that abolished binding of CFP-10 to Rv3871 prevented secretion of the CFP-10 (culture filtrate protein, 10 kilodaltons)/ESAT-6 virulence factor complex. Attachment of the signal to yeast ubiquitin was sufficient for secretion from M. tuberculosis cells, demonstrating that this ESX-1 signal is portable.

Proteins are sorted for translocation across cellular membranes through recognition of signal sequences (1, 2). In prokaryotes, most proteins are secreted through the general secretion pathway, which recognizes N-terminal signal peptides (3). Additionally, Gram-negative pathogenic bacteria use specialized secretion machines to secrete virulence determinants during infection (4, 5).

Mycobacterium tuberculosis does not have recognizable homologs of these specialized secretion systems. Instead, the ESX-1 system (ESAT-6 system-1) is required for controlling host-cell response to infection (68). ESX-1 is encoded by genes in the RD1 (region of difference 1) locus of the genome that is missing in the M. bovis Bacille Calmette-Guérin vaccine strain (9, 10). This system includes a multitransmembrane protein, Rv3877 (Snm4), and two putative SpoIIIE/FtsK adenosine triphosphatase (ATPase) family members, Rv3870 (Snm 1) and Rv3871 (Snm2). These three proteins are required for secretion of two virulence factors, ESAT-6 and CFP-10 (8). ESAT-6 (product of the esxA gene) and CFP-10 (product of the esxB gene) interact to form a 1:1 dimer (11, 12), and the stability of these proteins is interdependent in vivo. CFP-10, but not ESAT-6, interacts with the C-terminal domain of Rv3871, a cytosolic component of the ESX-1 system (8). Although the secretion of ESAT-6 and CFP-10 is critical for M. tuberculosis virulence, the molecular mechanisms of ESX-1 substrate selection and secretion are unclear.

To define targeting sequences responsible for directing secretion by ESX-1, we probed a series of N- and C-terminal deletions in M. tuberculosis CFP-10 for interaction with ESAT-6 and Rv3871 (13). Deletion of 25 amino acids from either the N or C terminus of CFP-10 abrogated the interaction of CFP-10 with ESAT-6 (Fig. 1A). In contrast, the last 25 amino acids of CFP-10 were necessary and sufficient for interaction with Rv3871 (Fig. 1B). Proteins containing deletions of the N-terminal 25 and 50 amino acids of CFP-10 retained Rv3871 binding activity, albeit at lower levels than full-length protein. These deletions interrupted domains of the folded protein (Fig. 1F), and the partial structures may have interfered with presentation of the interaction domain in the two-hybrid system.

Fig. 1.

The C terminus of CFP-10 is necessary and sufficient for Rv3871 interaction. Two-hybrid analysis of CFP-10 bait deletions with ESAT-6 (A) or Rv3871 (B) preys. (C) CFP-10 alanine scan mutants were tested for interaction with either ESAT-6 (top) or Rv3871 (bottom). (D) The last three to seven amino acids of CFP-10 were tested for interaction with Rv3871. For (A) to (D), β-galactosidase activity is shown; error bars represent standard deviation. (E) The interaction of HA-Rv3871 (amino acids 248 to 591) with CFP-10 by in vitro pulldown assays followed by CFP-10 immunoblot analysis. Equal amounts of CFP-10 and CFP-10Δ7CT were loaded into the pulldown (lanes 1 and 2), and CFP-10 and CFP-10Δ7CT were added to beads alone (lanes 4 and 5) or to Rv3871-bound beads (lanes 6 and 7). (F) ESAT-6 (gray) and CFP-10 (blue) solution structure (11) modeled by Protein Explorer (24) with the N and C termini of CFP-10 labeled.

Because the C-terminal domain of CFP-10 was required for interaction with both ESAT-6 and Rv3871, we identified individual residues required for the interaction of CFP-10 specifically with Rv3871 (Fig. 1C and fig. S1). Mutations in four of the seven C-terminal amino acids of CFP-10 (L94A, M98A, G99A, and F100A) abolished interaction with Rv3871 (Fig. 1C and fig. S1) but not with ESAT-6, demonstrating that these interactions are separable (Fig. 1C). Further deletion within the C-terminal tail of CFP-10 revealed that the last seven amino acids were sufficient for Rv3871 interaction (Fig. 1D). N-terminal deletions within the last seven amino acids of CFP-10 abrogated Rv3871 binding, demonstrating that L94 is critical for this interaction (Fig. 1D).

To independently test the interaction between CFP-10 and Rv3871, we performed in vitro pulldown experiments (Fig. 1E). HA-Rv3871 fusion protein bound to agarose beads was incubated with lysates from Escherichia coli cells expressing CFP-10 or CFP-10 lacking the C-terminal seven amino acids (CFP-10Δ7CT) in the presence of ESAT-6. Under these conditions, CFP-10 did not bind to Rv3871-coated beads. Indeed, interactions of secreted proteins with targeting proteins are typically weak and transient (14). Upon addition of protein cross-linker, however, the interaction was stabilized. Although low levels of CFP-10 were cross-linked nonspecifically to antibody-coated beads in the absence of Rv3871, CFP-10 interacted specifically with Rv3871 (Fig. 1E). In contrast, deletion of the C-terminal seven amino acids reduced binding to background levels (Fig. 1E). Thus, the last seven amino acids of CFP-10 are critical for interacting directly with Rv3871.

The published CFP-10/ESAT-6 solution structure shows that the C-terminal 14 amino acids of CFP-10 form an unstructured tail that does not interact with ESAT-6 (11) (Fig. 1F). Because most of the residues required for CFP-10 interaction with Rv3871 map to this unstructured C terminus (Fig. 1F), we hypothesized that targeting information for both ESAT-6 and CFP-10 could lie in the last seven amino acids of CFP-10.

To test the role of the CFP-10/Rv3871 interaction in secretion in vivo, we introduced mutant forms of esxB into the DesxB M. tuberculosis strain. Despite bearing an in-frame deletion in the esxB gene, neither CFP-10 nor ESAT-6 was detectable in the cell lysate (Fig. 2A), consistent with previous results (8, 15). Expression of wild-type CFP-10 and ESAT-6 in the DesxB strain restored production and secretion of these proteins into the culture supernatant (Fig. 2A), whereas expression of CFP-10 lacking the terminal 25 amino acids (CFP-10Δ25CT) did not, presumably because of a lack of interaction with ESAT-6 (Figs. 1A and 2A). In contrast, expression of CFP-10 lacking the seven terminal amino acids (CFP-10Δ7CT) resulted in stable production of both ESAT-6 and CFP-10. Thus, the carboxy-terminal seven amino acids of CFP-10 are not required for protein stability or for interaction of CFP-10 with ESAT-6. The CFP-10Δ7CT and ESAT-6 proteins, although stable, were not secreted into the culture supernatant (Fig. 2A), establishing that these residues target both CFP-10 and ESAT-6 for secretion in vivo.

Fig. 2.

The last seven amino acids of CFP-10 are necessary and sufficient for secretion from M. tuberculosis in vivo. (A) Immunoblot analysis of cell pellets (P) and supernatants (S) from wild-type (WT) M. tuberculosis (lanes 1 and 2), or DesxB strains bearing either the pMH406 complementation plasmid (COMP) (lanes 5 and 6), or plasmids expressing ESAT-6 and CFP-10Δ7CT (lanes 7 and 8) or CFP-10Δ25CT (lanes 9 and 10). GroEL was a control for autolysis. (B) Immunoblot analysis from DesxB M. tuberculosis harboring plasmids expressing ESAT-6 and either CFP-10 S96A (lanes 1 and 2), CFP-10 M98A (lanes 3 and 4), or CFP-10 F100A (lanes 5 and 6). (C) Immunoblot analysis of pellets and supernatants from DesxB M. tuberculosis strains harboring either pMH406 (vector) (lanes 1 and 4), or plasmids expressing yeast ubiquitin (Ubi) (lanes 2 and 5), or yeast ubiquitin tagged at its C terminus with LSSQMGF (Ubi-LSSQMGF) (lanes 3 and 6).

We generated single-point mutations in the last seven amino acids of CFP-10 and tested these mutant proteins for stability and secretion. Stable expression of the CFP-10 S96A, CFP-10 M98A, or CFP-10 F100A mutant protein was detectable in DesxB cell lysates, consistent with the dispensability of these residues for interaction with ESAT-6 in vivo. CFP-10 S96A, which could interact with Rv3871 (Fig. 1C), also allowed secretion of ESAT-6 into the culture supernatant from DesxB M. tuberculosis cells (Fig. 2B). In contrast, neither CFP-10 M98A nor CFP-10 F100A was secreted into the culture supernatant (Fig. 2B). Like CFP-10Δ7CT, these mutant CFP-10 proteins could not promote the secretion of ESAT-6 into the culture filtrate (Fig. 2B), demonstrating that these two residues are required for targeting both CFP-10 and ESAT-6 for secretion in vivo.

Because the last seven amino acids of CFP-10 were necessary for CFP-10/ESAT-6 secretion, we tested whether these residues are sufficient for secretion. We expressed yeast ubiquitin or the last seven amino acids of CFP-10 fused to the C terminus of ubiquitin in the DesxB M. tuberculosis strain and monitored secretion into the culture supernatant by immunoblot analysis. Native ubiquitin was found exclusively in the cell pellets (Fig. 2C). In contrast, addition of the CFP-10 signal sequence led to secretion of ubiquitin from the cells (Fig. 2C). Quantitative immunoblot analysis revealed that the ratio of secreted to cell-associated protein was lower by a factor of about 10 for the ubiquitin fusion protein than for CFP-10/ESAT-6, indicating that other features of CFP-10 or the CFP-10/ESAT-6 complex probably facilitate transport of these substrates by ESX-1. However, the last seven amino acids of CFP-10 can constitute a portable signal sequence sufficient to direct the secretion of a heterologous protein.

Ten CFP-10 paralogs are encoded by the M. tuberculosis genome as a result of numerous duplications of the esxA/esxB operon (fig. S3A). At least three CFP-10 paralogs (EsxG, EsxW, and EsxJ, K, or M) and five ESAT-6 paralogs (EsxH, EsxL, EsxN, EsxO, and EsxR) have been identified in the culture supernatants of M. tuberculosis (1618) and are candidate virulence factors. It is unclear whether these paralogs are secreted by the ESX-1 system or by distinct secretory systems (19). Alignment of the amino acid sequences of these proteins (Fig. 3A) revealed that five CFP-10 paralogs (EsxJ, EsxK, EsxP, EsxM, and EsxW) end in “QILSS,” which is similar to the “QALSS” found near the C terminus of CFP-10, but lack the terminal “MGF” required for CFP-10 secretion. EsxG ends in “ASTYTGF,” substituting the critical L94 and M98 residues of CFP-10 with alanine and threonine, respectively. We would predict that these paralogs and their associated ESAT-6 paralogs would be secreted independently of ESX-1.

Fig. 3.

CFP-10 paralogs are secreted from ESX-1 mutant strains. (A) Clustal alignments (25, 26) of the C termini of secreted CFP-10 paralogs, and CFP-10 homologs, colored according to the BLOSUM62 substitution matrix. The potential signal sequence region is underlined in red. (B) Example signature ion regions from CFP-10 and EsxG (Rv0287) peptides from proteins identified in culture supernatants of M. tuberculosis (13). Peaks at 114 Da, 116 Da, and 117 Da represent peptides found in WT, DesxB and DRv3877 culture supernatants. (C) Model of the M. tuberculosis ESX-1 secretion system. (mAg), mycolyl-arabinogalactan, (PG), peptidoglycan, (CM), cytoplasmic membrane.

To determine this, we measured the levels of these proteins in culture supernatants from wild-type, DRv3877, and DesxB M. tuberculosis strains using quantitative mass spectrometry (13, 20). Peptides from CFP-10 and ESAT-6, three ESAT-6 paralogs (EsxN, EsxL, and EsxO), and at least three CFP-10 paralogs (EsxG, EsxP or K, and EsxJ or W) (Fig. 3B, Table 1, and fig. S2) were present in wild-type supernatants, consistent with previous findings (1618). Although peptides from CFP-10 and ESAT-6 were absent from supernatants from DesxB and DRv3877 deletion M. tuberculosis strains (Fig. 3B), each of the other Esx proteins was secreted to wild-type levels (Fig. 3B, Table 1, and fig. S2). Thus, the C-terminal signal sequence possibly allows the ESX-1 system to discriminate between CFP-10 and its paralogs. Interestingly, five of these paralog pairs are embedded within loci with synteny to the ESX-1 region (21), raising the possibility that other ESX-1–like secretion systems function to secrete these Esx paralogs.

Table 1.

ESAT-6 and CFP-10 paralogs identified in the culture filtrate by quantitative liquid chromatography–tandem mass spectrometry (LC-MS/MS).

Protein nameRv No.Confidence of ID% Sequence coverage >99% conf. peptidesView inlineRatio 116/114 ΔesxB/WTView inlineRatio 117/114 ΔRv3877/WTView inlineC-terminal peptide present?
EsxA Rv3874 >99.9% 93.70% ≈0 ≈0 yes
EsxB Rv3875 >99.9% 92.00% ≈0 ≈0 yes
EsxN Rv1793 >99.9% 98.90% 1.02:1 1.1:1 yes
EsxL Rv1198 >99.9% 98.90% 0.92:1 0.92:1 yes
EsxO Rv2346c >99.9% 98.90% 0.99:1 1.01:1 yes
EsxPView inline Rv2347c >99.9% 93.90% 1.00:1 1.00:1 yes
EsxKView inline Rv1197 >99.9% 93.90% 1.00:1 1.00:1 yes
EsxJView inline Rv1038c >99.9% 93.90% 1.1:1 0.99:1 yes
EsxWView inline Rv3620c >99.9% 93.90% 1.1:1 0.99:1 yes
EsxG Rv0287 >99.9% 73.20% 1.00:1 1.00:1 yes
  • View inline* Sequence coverage = total amino acids identified/total in database, including signal sequences and N-terminal methionine. Only reported for those peptides with MS/MS confidence scores >99%.

  • View inline Ratio normalized for lysis and mixing error. 95% confidence interval was <±0.2 for each Esx peptide.

  • View inline‡,§ Insufficient MS/MS or protein sequence evidence exists to differentiate these isoforms.

  • Alignment of CFP-10 protein sequences from other bacterial species that are secreted by homologous ESX-1 systems (68, 15, 22, 23) reveals a notable stretch of conservation at the C terminus (Fig. 3A and fig. S3B). Because these residues do not play a structural role in the folded CFP-10/ESAT-6 dimer, the conservation at the C terminus suggests strongly that the mechanism of ESX-1 C-terminal signal sequence recognition has been conserved during bacterial evolution.

    We propose the following model for the targeting and secretion of substrates by the ESX-1 secretion system in M. tuberculosis (Fig. 3C). ESAT-6 and CFP-10 fold and form a stable dimer in the cytoplasm before targeting. The C-terminal domain of CFP-10 is recognized by Rv3871, targeting both ESAT-6 and CFP-10. Substrate-bound Rv3871 interacts with Rv3870, a membrane-bound component of ESX-1, thus linking the cytosolic component of the system with membrane components. Because Rv3870 and Rv3871 are both members of the SpoIIIE/FtsK ATPase family, these proteins probably perform the work necessary to secrete ESX-1 substrates. This is reminiscent of Type IV secretion systems in Gram-negative bacteria, in which a membrane-bound SpoIIIE/FtsK–like ATPase functions to recognize an unstructured C-terminal signal sequence and direct the secreted substrate to the cytoplasmic membrane (4).

    Supporting Online Material

    Materials and Methods

    SOM Text

    Figs. S1 to S3

    Tables S1 and S2

    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article