Conserved Pro-Glu (PE) and Pro-Pro-Glu (PPE) Protein Domains Target LipY Lipases of Pathogenic Mycobacteria to the Cell Surface via the ESX-5 Pathway

Conserved Pro-Glu (PE) and Pro-Pro-Glu (PPE) Protein Domains Target LipY Lipases of Pathogenic Mycobacteria to the Cell Surface via the ESX-5 Pathway. The type VII secretion system ESX-5 is a major pathway for export of PE and PPE proteins in pathogenic mycobacteria. These mycobacteria-specific protein families are characterized by conserved N-terminal domains of 100 and 180 amino acids, which contain the proline-glutamic acid (PE) and proline-pro-line-glutamic acid (PPE) motifs after which they are named. Here we investigated secretion of the triacylglycerol lipase LipY, which in fast-growing mycobacteria contains a signal sequence, but in slow-growing species appears to have replaced the signal peptide with a PE or PPE domain. Selected LipY homologues were expressed in wild-type Mycobacterium marinum and its corresponding ESX-5 mutant, and localization of the proteins was investigated by immunoblotting and electron microscopy. Our study shows that Mycobacterium

Mycobacteria such as Mycobacterium tuberculosis, the etiological agent of tuberculosis, have a highly unusual and complex cell envelope (1). To secrete virulence factors across the cell envelope, these bacteria use specialized protein secretion sys-tems, known as ESX or type VII secretion systems (2)(3)(4)(5)(6)(7)(8). Mycobacterial genomes contain up to five genetic loci coding for type VII secretion systems, named ESX-1 to ESX-5 (9,10). The most well studied system, ESX-1, is responsible for the secretion of 10 substrates, including the important T-cell antigens ESAT-6 and CFP-10, and is required for full virulence of M. tuberculosis (7,(11)(12)(13)(14)(15)(16)(17). Phylogenetic analyses and comparative genomics suggest that the five ESX clusters have evolved by gene duplication and that ESX-5 is the result of the most recent duplication event (9). Interestingly, ESX-5 is restricted to a group of mycobacterial species known as the slow-growing mycobacteria, which include all major pathogens, such as M. tuberculosis, Mycobacterium leprae, and Mycobacterium ulcerans, and the fish pathogen Mycobacterium marinum (9). Four of the ESX loci contain also PE and PPE genes (named after the conserved Pro-Glu (PE) and Pro-Pro-Glu (PPE) motifs near the N termini of their respective gene products) (9,10), and the appearance of ESX-5 predates the huge expansion of these gene families in slowgrowing mycobacteria (18). Intriguingly, although fast-growing, non-pathogenic mycobacteria encode only a small number of PE and PPE proteins, pathogens such as M. marinum and M. tuberculosis dedicate nearly 10% of the coding potential of their genomes to members of the PE and PPE gene families (19). Although the precise function of these proteins is largely unknown, members of both families are important for mycobacterial virulence (20 -24). Moreover, various PE and PPE proteins are located on the cell surface, where they can interact with the host during infection (21,(25)(26)(27)(28). However, because the PE and PPE proteins lack detectable secretion signals, the route of translocation across the cell envelope remained unclear until members of the two protein families were recently shown to be secreted by the ESX-5 system in M. marinum (3,29).
PE and PPE proteins are characterized by unrelated, conserved N-terminal domains 2 of ϳ100 (PE) and ϳ180 (PPE) amino acids (aa) (19), and they can be divided into different subfamilies on the basis of their C-terminal domains (19,30,31). To date, little is known about the function of the various domains of the PE and PPE proteins. In fact, the M. tuberculosis PE protein LipY (hereafter referred to as LipY tub ) is the only PE protein for which a function has been characterized (32). M. tuberculosis LipY tub is involved in the degradation of triacylglycerols (TAGs) and is the major active lipase under nutrientdeprived conditions. This led to the hypothesis that LipY tub plays a role in fatty acid metabolism during the dormancy and reactivation stages of the M. tuberculosis infection cycle (32). Although the lipolytic activity of LipY tub is expressed by the C-terminal part of the protein (32), the function of its N-terminal PE domain is less clear. It has been shown that the PE domain of LipY tub has an inhibiting effect on the lipase activity (33). Furthermore, cell wall and cell surface localization of LipY was shown to occur independently of the presence of the PE domain (33). Similar results were obtained for M. marinum LipY (hereafter referred to as LipY mar ), which contains a PPE domain instead of a PE domain. These results contradict those obtained in Refs. 26 and 27, which showed that localization of heterologously expressed PE_PGRS33 (member of the polymorphic GC-rich repetitive sequence subfamily of PE proteins) in the cell wall of Mycobacterium smegmatis is dependent on the PE domain.
In this study, we show that, like many other PE and PPE proteins, the M. tuberculosis and M. marinum LipY homologues are secreted by ESX-5. Using a combination of molecular techniques and analysis of the lipolytic activity of LipY, we show that ESX-5-mediated secretion of the LipY homologues is dependent on the respective PE and PPE domains.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-M. marinum wild-type strains E11 (34) and M vu (35) and their respective ESX-5 mutants 7C1 (29) and Mx2 (3) and Mycobacterium gilvum PYR-CGK (ATCC 700033) (36) were grown at 30°C with shaking at 90 rpm, whereas Mycobacterium bovis bacille Calmette-Guérin (BCG) Copenhagen (37), M. bovis BCG Tice (38), and M. tuberculosis mc 2 6020 (39) were grown standing at 37°C. M. smegmatis mc 2 155 (40) was grown at 37°C with shaking at 90 rpm. All mycobacterial strains were grown in Middlebrook 7H9 broth (Difco-BD Biosciences), supplemented with Middlebrook ADC and 0.05% Tween 80 (Sigma-Aldrich). In secretion experiments, mycobacteria were grown to mid-logarithmic phase, at which they were washed to remove BSA (part of the ADC supplement) and subsequently grown in Middlebrook 7H9 supplemented with 0.2% dextrose and 0.05% Tween 80 for another 24 h. If cultures were grown for electroporation, glycine was added to a final concentration of 2.5 mg/ml during the early exponential phase to increase electroporation efficiency (3). Transformants were selected on Middlebrook 7H10 agar plates and supplemented with Middlebrook OADC and appropriate antibiotics. Escherichia coli DH5␣ was grown at 37°C in LB medium (Difco-BD Biosciences) for cloning experiments. Hygromycin was used at a concentration of 50 g/ml for mycobacteria and 100 g/ml for E. coli, and kanamycin was used at a concentration of 25 g/ml.
Plasmid Construction-All DNA manipulations were carried out according to standard protocols. Pfu DNA polymerase, restriction enzymes, and T4 DNA ligase were purchased from Fermentas. Chromosomal DNA from M. marinum E11 and M. gilvum PYR-CGK was used as template in PCR reactions. Plasmids and primers are described in supplemental Tables S1 and S2, respectively.
Cloning of Rv3097c (gene encoding LipY tub ), modified to express a C-terminal HA tag, in the E. coli-mycobacterial shuttle vector pMV10 -25 to generate pAL36 is described in Ref. 26. pMV10 -25 contains a minimal pAL5000 origin of replication, which lacks orf5 that is required for initiating stable replication (41). Therefore, a 1749-bp fragment containing the hsp60 promoter sequence, the Rv3097c gene, and the in-frame sequence encoding an HA tag followed by a stop codon was isolated from pAL36 by digestion with HindIII and XbaI. The fragment was ligated into the pSMT3 vector (42), which contains a full pAL5000 origin of replication, digested with HindIII and XbaI, to generate pSMT3::LipY tub . To construct pSMT3::LipY tub -⌬PE, a gene fragment encoding LipY tub without the first 125 aa was amplified from pAL36, cloned in pMV10 -25 to generate pAL45, and subsequently transferred to pSMT3 by digestion with HindIII and XbaI.
For cloning of sequences encoding LipY mar and LipY gilv or truncated forms of LipY tub and LipY mar , forward primers were designed to contain an NheI site immediately upstream of the start codon, and reverse primers contained an HA tag encoding sequence in-frame with the coding sequence, a stop codon, and a BamHI or BglII site. PCR fragments were inserted into pSMT3::LipY tub , digested with NheI and BamHI, thereby replacing the Rv3097c gene and HA-encoding sequence.
Plasmids for expression of fusions of LipY tub and LipY gilv were generated as follows. To replace the PE domain of LipY tub with the signal peptide of LipY gilv , a 145-bp fragment encoding the signal sequence was amplified from M. gilvum chromosomal DNA and digested with NheI and EcoRI. Then, a 1045-bp fragment coding for the linker and lipase domains of LipY tub and the HA tag was isolated from pSMT3::LipY tub by digestion with EcoRI and XbaI. The two fragments were ligated into pSMT3, digested with NheI and XbaI, to generate pSMT3::SP gilv -LipY tub . For the construct in which the signal peptide of LipY gilv was replaced by the PE domain of LipY tub , a 309-bp fragment encoding the LipY tub PE domain was isolated from pSMT3::LipY tub by digestion with NheI and EcoRI. In addition, a 1390-bp fragment encoding the linker and lipase domains of LipY gilv and the HA sequence was amplified from M. gilvum chromosomal DNA and digested with EcoRI and BglII. These two fragments were ligated into a 5691-bp NheI-BamHI fragment of pSMT3::LipY tub , replacing the Rv3097c gene, resulting in the pSMT3::PE tub -LipY gilv vector. For the fusion of the PE and linker domains of LipY tub to the lipase domain of LipY gilv , a nested PCR approach was taken to amplify fragments from the pSMT3::LipY tub and pSMT3::LipY gilv vectors. After digestion with NheI and BamHI, the resulting 1436-bp fragment was inserted into pSMT3::LipY tub , as described above, to generate pSMT3::PE/linker tub -LipY gilv .
Because M. tuberculosis mc 2 6020 carries a hygromycin resistance marker (39), LipY tub was placed under control of the hsp60 promoter in the pBH10 vector (3,43). A 1355-bp fragment containing Rv3097c, the in-frame HA sequence, and the stop codon was cut out from pSMT3::LipY tub with BspEI and BamHI, and ligated into pBH10 digested with the same enzymes. The constructs were verified by sequencing.
Site-directed Mutagenesis of the LipY tub Cleavage Site-Derivatives of pSMT3::LipY tub containing point mutations were generated using the QuikChange site-directed mutagenesis kit (Stratagene). Because pSMT3::LipY tub was too large for the Pfu Turbo DNA polymerase to amplify, the smaller pAL36 was used as template. The 1749-bp HindIII-XbaI fragments containing the hsp60 promoter and the mutated gene were transferred into pSMT3 as described above. The mutations were verified by sequencing.
N-terminal Sequencing-To identify the LipY tub processing site, we determined the N-terminal amino acid sequence of the cleaved form of the protein. Surface-exposed proteins were extracted with Genapol as described above and separated by SDS-PAGE. Proteins were transferred to a PVDF membrane (Millipore) and stained with 0.1% Coomassie Brilliant Blue R-250 (Bio-Rad) in 1% acetic acid, 40% methanol. After destaining with 50% methanol, the band corresponding to processed LipY tub was cut out from the membrane, and the sequence of the first seven amino acids was determined by Edman degradation (Eurosequence, Groningen, The Netherlands).
Immunogold Electron Microscopy (EM)-Fixation of bacterial cells and immunogold labeling of whole cells were done as described in Ref. 17. The HA.11 mouse monoclonal (Covance) was used to visualize the HA-tagged LipY homologues.
Lipase Activity Assay-The lipolytic activity of surface-exposed LipY tub was analyzed with a modified method of the assay described in Ref. 33. Intact M. marinum cells grown in the dark (to prevent production of yellow pigment that might interfere with the absorbance measurements) were centrifuged, and an equivalent of 1 OD unit was resuspended in Tris-HCl buffer (100 mM, pH 8.0). p-Nitrophenyl stearate was added from a 20 mM stock solution prepared in isopropyl alcohol to reach a final concentration of 0.5 mM in 200 l, and the mixture was vortexed and incubated for 15 min at 37°C. Thereafter the bacteria were centrifuged, and the supernatant was transferred to a 96-well plate before measuring the production of p-nitrophenol with a spectrophotometer at 405 nm. Reactions were carried out in triplicate.
Bioinformatic Analyses-Nucleotide and protein sequences were retrieved from the MarinoList, TubercuList, and National Center for Biotechnology Information (NCBI) databases. Linear genomic comparisons were performed with the Artemis Comparison Tool software release 9 (44), with a cutoff value of 100 bp.
Cells Lines and Culture Conditions-For LipY macrophage infection experiments, the human acute pro-monocytic leukemia THP-1 cell line was used. Cells were cultured in RPMI 1640 with Glutamax-1 medium (Invitrogen) supplemented with 10% FCS, streptomycin, and penicillin and differentiated into macrophage-like cells in the presence of 10 ng/ml phorbol 12-myristate 13-acetate (Sigma-Aldrich).
LipY Processing by Mycobacteria in Macrophages-For infection, THP-1 cells were used between passage 13 and 20. Cells were seeded at 10 ϫ 10 6 per flask in 75-cm 2 diameter flasks. Mononuclear cells (3-5 ϫ 10 7 ) were infected with M. tuberculosis or M. bovis BCG (multiplicity of infection 10) and incubated for 3 h at 37°C and 5% CO 2 . The supernatant was removed, and the infected cells were washed three times with medium to remove extracellular bacteria. Subsequently, the cells were incubated in fresh medium with 10% FCS at 37°C and 5% CO 2 for the indicated time periods. Enrichment of mycobacterial proteins was performed as follows. Briefly, infected THP-1 cells were collected and washed twice with PBS. Cellular membranes were solubilized by treatment with 1% Triton X-100 for 15 min at room temperature. Intact mycobacteria were sedimented at 25,000 ϫ g for 30 min and subsequently disrupted by sonication. Proteins were separated by SDS-PAGE and visualized by immunoblotting using the HA.11 mouse monoclonal (Covance).

Homologues of LipY Have Evolved Different N-terminal
Domains-M. tuberculosis LipY consists of three domains: a typical PE domain at the N terminus followed by a linker domain of unknown function and a C-terminal domain containing the triacylglycerol lipase motif (for a schematic view, see Fig. 1A). Nearly identical orthologues of LipY tub are found in various other species of the M. tuberculosis complex, such as M. bovis. However, with the exception of a pseudogene in M. leprae (32), only one orthologue has been reported in a mycobacterial species outside the M. tuberculosis complex (33). This gene is present in M. marinum and codes for a protein with a different domain organization. Remarkably, although the linker and lipase domains of LipY mar are 65% identical to LipY tub , the N-terminal PE domain is replaced by a PPE domain (Fig. 1A).
We performed a BLAST search to identify additional homologues of LipY in mycobacterial species. Two lipY copies were found in Mycobacterium kansasii. Interestingly, one of these has a PE domain (NCBI accession number ZP_04747303), whereas the other has a PPE domain (NCBI accession number ZP_04747310). The latter variant is probably not functional because its lipase domain is disrupted by a stop codon so that this PPE-LipY protein lacks the histidine residue of the Ser-Asp-His catalytic triad. The two M. kansasii lipY genes are located six genes apart in a region of conserved synteny with the sequences surrounding the lipY genes in M. tuberculosis and M. marinum (Fig. 1B). Short stretches of non-homologous DNA immediately up-and downstream of the lipY genes indicate that some insertions and/or deletions have occurred in this region. When the BLAST analysis was extended to the fastgrowing mycobacteria, two additional putative LipY proteins were identified in M. gilvum (hereafter referred to as LipY gilv ) and Mycobacterium vanbaalenii that shared 44 and 43% identity, respectively, with the C-terminal half of the linker domain and the lipase domain of LipY tub (NCBI accession numbers YP_001135683 and YP_952650). Strikingly, the N termini of these sequences do not share homology with PE or PPE domains but instead appear to contain classical signal peptides (Fig. 1A), as indicated by SignalP analysis.
To gain insight in the evolutionary relationships between the LipY homologues, a phylogenetic tree rooted to the Nocardia farcinica lipase Nfa23150 (NCBI accession number YP_ 118526) was constructed based on the conserved part of the lipase domains, using the MEGA4 software (45) (not shown). The tree topology is similar to that of the mycobacterial species tree, suggesting that the original lipY gene was modified to contain PE/PPE domains. The alignment tree also suggests that PE-LipY evolved before PPE-LipY. This implies that the signal sequence was initially replaced by a PE domain. Subsequently, PE-LipY was (partially) duplicated and modified to contain an N-terminal PPE domain.
LipY tub and LipY mar Are Substrates of ESX-5 in M. marinum-Because several PE and PPE proteins have been shown to be secreted by the ESX-5 secretion system (3,29), we hypothesized that the LipY homologues with PE/PPE domains are also ESX-5 substrates. To investigate this, the genes encoding LipY tub and LipY mar were each cloned under control of the hsp60 promoter and modified to express a C-terminal HA tag (Fig. 1A). The two constructs were subsequently introduced into wild-type M. marinum strain E11 and its ESX-5 mutant derivative 7C1, and the localization of the LipY-HA fusions was analyzed. Surface-exposed proteins were extracted by treating intact cells with the mild detergent Genapol X-080 (17,26). For the wildtype strain, a minor fraction of both LipY tub and LipY mar was detected in the culture supernatant (Fig. 2, A and B, lane 4), whereas substantial amounts of both proteins were detected in the surface extracts (Fig. 2, A and B, lane 3). GroEL2 was absent from the surface extracts and culture supernatants, confirming the integrity of the cells. Furthermore, ESAT-6 was included in the analysis as a positive control for secretion, and this protein was normally secreted in all conditions. LipY tub and LipY mar were undetectable in the surface extracts and culture supernatants of the ESX-5 mutant (Fig. 2, A and B, lanes 7 and 8), which confirms our hypothesis that these two proteins are substrates of ESX-5. In addition, the LipY protein levels were markedly lower in the ESX-5 mutant as compared with in the wild-type lysates. Similar results have previously been obtained for the ESX-5 substrates PPE41 and PE_PGRS (3,29) and indicate that when secretion is blocked, stability of the ESX-5 substrates is affected. The ESX-5-dependent secretion and surface localization of LipY tub was confirmed in M. marinum strain M vu and its ESX-5 mutant Mx2 (not shown).
In the total cell lysate of the wild-type strain expressing LipY tub , two bands are detected (Fig. 2A, lane 1). The upper 40-kDa band likely represents the full-length protein, which has a calculated molecular mass of 45.9 kDa, whereas the lower band at ϳ28 kDa probably represents an N-terminally pro-cessed or degraded form. Only the smaller form is detectable in the surface extract and in the culture supernatant ( Fig. 2A,  lanes 3 and 4), whereas full-length LipY tub remains in the Genapol-treated pellet and thus appears to be intracellular ( Fig. 2A, lane 2). In contrast, LipY tub was detected as a single, 40-kDa band in the total cell lysate of the ESX-5 mutant, indicating that processing only occurs when LipY tub is secreted (Fig. 2A, lane 5). For LipY mar , three bands are detected, a minor band of 56 kDa probably representing fulllength LipY mar and two processed or degraded forms at 39 and 36 kDa (Fig. 2B, lane 1). As for LipY tub , only the fulllength band is detected in the ESX-5 mutant strain (Fig. 2B,  lane 5). However, unlike for LipY tub , all three forms are represented in the surface extract of the wild-type strain (Fig.  2B, lane 3).
To confirm the surface exposure of LipY tub and LipY mar , the localization of the proteins was analyzed by immunogold EM on whole cells. Wild-type cells expressing LipY tub were efficiently labeled (Fig. 2E), whereas the ESX-5 mutant showed little or no labeling (Fig. 2F), consistent with the immunoblot results ( Fig. 2A). Wild-type M. marinum expressing LipY mar also showed surface labeling (Fig. 2G), although to a lesser extent than LipY tub . Again, the ESX-5 mutant showed little or no labeling (Fig. 2H), confirming that LipY mar is also secreted by ESX-5. M. marinum harboring the empty plasmid showed no reaction with the HA antiserum in immunoblots or whole cell immunogold EM (not shown).
To investigate whether surface-exposed LipY tub is functional, bacterial cells were incubated with p-nitrophenyl stearate, after which the bacteria were removed by centrifugation. Lipolytic activity resulted in the release of yellow p-nitrophenol in the supernatants, which could be measured spectrophotometrically. The wild-type strain expressing LipY tub showed strongly increased activity as compared with the wild-type strain harboring the empty plasmid pSMT3, confirming that LipY tub is secreted to the cell surface where it is enzymatically active (Fig. 3). Interestingly, the empty wild-type strain also exhibited lipolytic activity, which was completely abolished in the ESX-5 mutant strain, with or without lipY-containing plasmids. This activity could be due to chromosomally encoded LipY mar .
LipY gilv Is Not Secreted in M. marinum or M. smegmatis-The Sec-dependent secretion pathway in mycobacteria is, as in other bacteria, responsible for translocation of proteins with signal sequences across the inner membrane (4). Because LipY gilv is predicted by the SignalP program to have a signal peptide, we set out to determine its subcellular localization. To this end, we amplified and cloned the lipY gilv gene with a C-terminal HA tag under control of the hsp60 promoter (Fig. 1A). Upon introduction into M. marinum wild-type E11 and its ESX-5 mutant, no bands could be detected in the culture supernatant or Genapol extract fractions (Fig. 2C, lanes 3 and 4 and lanes 7 and  8). In the whole cell lysate, several bands were detected of which the largest is ϳ65 kDa (Fig. 2C, lane 1). Because LipY gilv has a theoretical molecular mass of 51 kDa, this could indicate that the protein is post-translationally modified. However, because a similar band was observed upon expression of lipY gilv in E. coli (results not shown), this hypothesis is not likely to be correct. In agreement with the results obtained by immunoblotting, no labeling was observed of whole cells in immunogold EM (Fig. 2, I and J). We attempted to repeat the experiments in M. gilvum to rule out that LipY gilv is secreted by its natural host, but despite trying a variety of experimental settings, we failed to transform M. gilvum. Therefore, secretion was analyzed in another fast-growing species, i.e. M. smegmatis. Similar to M. marinum, several bands were observed in the cell pellet (Fig.  2D, lane 1), but also, in this host, LipY gilv is neither secreted nor surface-exposed (Fig. 2D, lanes 3 and 4). In conclusion, LipY gilv is neither exposed on the bacterial surface nor secreted to the extracellular environment.
The PE and PPE Domains Are Required for Secretion of LipY tub and LipY mar via ESX-5-To investigate the role of the conserved PE and PPE domains in ESX-5 secretion, N-terminally truncated forms of LipY tub and LipY mar lacking the first 125 and 176 aa, respectively (Fig. 1A), were studied. Both mutant proteins were stably produced in the cell lysates of the wild-type and mutant strains (Fig. 4, A and B, lanes 1 and 5). However, the absence of bands in the Genapol extracts and culture supernatants (Fig. 4, A and B, lanes 3 and 4 and lanes 7 and 8) indicated that these truncated proteins were not secreted. Detection of ESAT-6 in culture supernatant fractions confirmed that the lack of secretion was specific for the LipY homologues. To exclude the possibility that LipY tub -⌬PE and LipY mar -⌬PPE were still secreted but could not be recognized by the antiserum, we analyzed the lipolytic activity of M. marinum wild-type E11 and its ESX-5 mutant expressing LipY tub -⌬PE. As shown in Fig. 3, E11 expressing LipY tub -⌬PE exhibits the same lipase activity as the wild-type cells. The ESX-5 mutant expressing LipY tub -⌬PE did not show any activity. Because LipY tub and LipY mar lacking their N-terminal PE or PPE domains have been shown to have higher lipase activity than the corresponding full-length proteins (33), these results confirm that our N-terminally truncated LipY homologues are indeed not secreted.
To test whether the signal peptide of LipY gilv could replace the PE domain and facilitate secretion of LipY tub to the cell surface, we exchanged the PE domain (1-101 aa) of LipY tub for FIGURE 3. Lipolytic activity of cell surface-exposed LipY tub in M. marinum. Lipid-hydrolyzing activity was determined by measuring the release of p-nitrophenol upon incubation of intact cells of M. marinum wild-type E11 and its corresponding ESX-5 mutant 7C1 (ESX-5::Tn) overexpressing LipY tub and LipY tub -⌬PE, with p-nitrophenyl stearate. E11 and 7C1 without plasmids or carrying the empty vector pSMT3 were included as control strains, so increases in activity can be specifically attributed to lipase exposed on the bacterial cell surface. The bars indicate means, and the error bars indicate standard deviations of two experiments, both performed in triplicate. Abs, absorbance.
the signal peptide encoding sequence (1-45 aa) of LipY gilv (Fig.  1A). This chimera was, however, neither secreted nor surfaceexposed upon expression in M. marinum (Fig. 4C, lanes 3 and 4  and lanes 7 and 8). Finally, we also tested whether the PE and PPE domains of LipY tub and LipY mar without the linker and lipase domains could be secreted. To this end, we cloned the PE (1-99 aa) and PPE (1-176 aa) domains of LipY tub and LipY mar and modified them to express C-terminal HA tags (Fig. 1A) under the control of the hsp60 promoter. However, upon introduction into M. marinum, neither PE LipYtub nor PPE LipYmar could be detected in immunoblots (results not shown), suggesting that they were not stable. In conclusion, these results show that LipY tub and LipY mar require their respective N-terminal PE and PPE domains for translocation to the cell surface by the ESX-5 secretion machinery.
LipY gilv Fused to a PE Domain Is Not Secreted-To test whether the PE domain can target unrelated proteins for ESX-5 secretion, we replaced the signal peptide of LipY gilv (1-45 aa) with an in-frame N-terminal fusion of the PE domain of LipY tub (1-103 aa) to the linker and lipase domains of LipY gilv (46 -494 aa) (Fig. 1A). However, upon introduction into M. marinum, this chimeric protein was neither surface-exposed nor secreted (Fig. 4D, lanes 3 and 4). This led us to believe that (part of) the linker domain following the PE domain could be critical. Therefore, a second fusion was constructed in which the signal peptide and the linker domain of LipY gilv (1-236 aa) were replaced by the PE domain and the linker domain of LipY tub (1-205 aa) (Fig. 1A). However, this chimera also failed to be secreted (Fig.  4D, lanes 7 and 8). The first chimeric protein showed the aberrant molecular weight features of the LipY gilv protein, whereas the second construct migrated at the expected molecular weight (Fig. 4D, lanes 1 and 5), indicating that the linker domain is responsible for the multiple bands observed with LipY gilv in SDS-PAGE, of which some are larger than expected.
These results indicate that heterologous proteins cannot easily be secreted via the ESX-5 pathway. Failure to secrete even a closely related fusion sequence (44% identity on amino acid level) suggests that structure or folding kinetics of substrates may be of importance for ESX-5-dependent translocation.
Processing of Secreted LipY tub -As discussed above, secreted and surface-exposed LipY tub is detected in a processed form ( Fig. 2A). This could mean either that LipY tub is processed during transport or that secreted LipY tub is sensitive to extracellular proteases. To gain more insight into the mechanism of processing, cell lysates of M. marinum wild-type strain E11 and its ESX-5 mutant 7C1 overexpressing LipY tub were incubated for 1, 2, 4, and 24 h at 30°C. No additional processing or degradation of LipY tub was observed in either strain (Fig. 5A). These results suggest that LipY is not a protease-sensitive protein, which indicates that processing of LipY tub is a specific maturation event.
Next, processed LipY tub was isolated from the Genapol supernatant (Fig. 5B), and the N-terminal sequence was determined by Edman degradation. The results showed that LipY tub is cleaved between Gly-149 and Ala-150 (Fig. 5C). The cleavage position indicates that the PE domain and part of the linker domain are removed, whereas the intact lipase domain is retained. The processed protein has a predicted mass of 30.3 kDa, which correlates well with the observed protein band with an apparent molecular mass of 28 kDa.
To investigate the nature of the recognition sequence of the enzyme responsible for processing of LipY tub , we introduced point mutations in residues surrounding the cleavage site (Fig.  5C). Most strikingly, substitution of the alanine at position 150 for an aspartic acid residue resulted in a complete disappearance of processed LipY tub , whereas a slightly larger cleavage product was visible (Fig. 5D, lane 11). The S148D and G149D mutations resulted in reduced production of the 28-kDa band and accumulation of a slightly larger product, similar to that resulting from the A150D mutation (Fig. 5D, lanes 7 and 15). Although the processing was affected, none of these mutations affected the surface localization of LipY tub (Fig. 5D, lanes 7, 11, and 15), nor did they significantly alter the lipase activity (results not shown). Three additional point mutations, P147A, F153A, and D154A, had no effect on LipY tub processing (not shown).
Taken together, these results suggest that the entire PE domain and the first 50 amino acids of the linker domain of LipY tub are removed in a maturation event linked to secretion. Processing depends on the SGA motif encoded by residues 148 -150, but if those residues are mutated, cleavage occurs at another adjacent site, and secretion and lipase activity are unaffected.
Expression and Processing of LipY tub in M. tuberculosis and M. bovis BCG-PE and PPE proteins have previously been detected on the surface of M. bovis BCG and M. tuberculosis (21,25). To investigate whether LipY tub is also surface-exposed in members of the M. tuberculosis complex, we expressed HAtagged LipY tub in M. bovis BCG Copenhagen, M. bovis BCG Tice, and the M. tuberculosis double auxotroph mutant mc 2 6020. Although we observed secretion of endogenous PPE41, a protein that has previously been shown to depend on ESX-5 (3), LipY tub was not was detected in the Genapol-extracted fraction or the culture supernatant (Fig. 6A, lanes 3 and  4, for results with M. tuberculosis mc 2 6020; data for M. bovis BCG strains are not shown). Moreover, unlike in M. marinum, LipY tub appeared to remain unprocessed (Fig. 6A, lane 1). Consistently, no labeling was observed in whole cell immunogold EM (not shown). FIGURE 5. The N terminus is proteolytically removed from secreted LipY tub . A and D, processing of wt and mutated forms of LipY tub was analyzed by immunoblotting using HA antiserum. GroEL2 was detected as a loading control in A and as a control for bacterial lysis in D. A, no additional processing or degradation was observed when cell lysates of M. marinum wt E11 or its ESX-5 mutant 7C1 (ESX-5::Tn) expressing LipY tub were incubated for 1, 2, 4, or 24 h after sonication. B, Coomassie Brilliant Blue-stained SDS-PAGE gel with Genapol surface extracts of wt M. marinum E11 or E11 expressing LipY tub . Proteins were transferred to a PVDF membrane, and the band representing processed LipY tub (marked with an arrow in the gel) was analyzed by Edman degradation. C, LipY tub is processed in the linker domain between residues Gly-149 and Ala-150. Residues altered by site-directed mutagenesis that affected processing are labeled with asterisks, whereas those that had no effect are marked with triangles. D, The S148D, G149D, and A150D point mutations resulted in (partial) production of a surface-exposed slightly larger processed form of LipY tub , marked with an asterisk, when expressed in M. marinum strain M vu . Equal OD units of cell pellets (P Ϫ), Genapol-treated cell pellet (P ϩ), and Genapol surface extract (S ϩ), and 2-fold more culture supernatant (S Ϫ) are shown. Because the expression of M. tuberculosis lipY is highly induced upon infection of host cells (46), we hypothesized that the lack of secretion in culture could be due to the fact that additional proteins required for LipY secretion are missing. To test this possibility, we infected human THP-1 cells, differentiated to macrophage-like cells in the presence of phorbol 12-myristate 13-acetate, with M. tuberculosis mc 2 6020 and M. bovis BCG Copenhagen expressing HA-tagged LipY tub . At 1, 2, 3, and 6 days after infection, the THP-1 cells were lysed by Triton X-100 treatment and centrifuged to collect mycobacteria. Interestingly, these samples showed the same processing pattern for LipY tub as was observed for this protein upon expression in M. marinum (see Fig. 6B, lane 2, for results from day 2 after infection with M. tuberculosis mc 2 6020; data for M. bovis BCG are not shown). Together these results indicate that processing of LipY tub occurs both in M. marinum and in members of the M. tuberculosis complex, albeit indicating that in the latter species, the secretion and/or processing of LipY is induced during infection.

DISCUSSION
Comparative genomics and phylogenetic analyses indicated that the evolution of the type VII secretion system ESX-5 predated the expansion of the PE and PPE gene families in slowgrowing mycobacteria (18). These analyses also showed that the recently evolved PE_PGRS and PPE_MPTR (major polymorphic tandem repeat subfamily of PPE proteins) subfamilies have probably originated from PE and PPE genes within the ESX-5 cluster, which resulted in the hypothesis that ESX-5 has a functional link with the recent expansion of PE and PPE proteins. Recently, we showed that ESX-5 is indeed responsible for the secretion of various members of the PE_PGRS and PPE_MPTR protein subfamilies to the cell surface and culture supernatant of M. marinum (3,29). In this work, we report the identification of two novel substrates of the M. marinum ESX-5 system, the LipY homologues of M. tuberculosis and M. marinum, which contain a PE and PPE domain, respectively. Interestingly, the analysis of LipY allows a further expansion of the original hypothesis. Additional lipY orthologues were identified in fast-growing mycobacterial species, namely M. gilvum and M. vanbaalenii. These lipY homologues have putative N-terminal signal sequences instead of a PE or PPE domain. LipY gilv was not transported to the cell surface, indicating that the PE or PPE domain is required for surface localization. However, it must be acknowledged that the localization of LipY gilv could not be studied in its native host due to problems with transformation. We would like to propose that LipY homologues in slow-growing mycobacteria have evolved to secreted proteins by substituting the original signal peptide for PE or PPE domains and thereby have turned into substrates of the ESX-5 system (Fig. 7). A remarkable snapshot of this evolution process can be observed in the genome of M. kansasii, which has two copies of lipY, one containing a PE domain and one with a PPE domain.
N-terminally truncated forms of LipY tub and LipY mar were not transported across the cell envelope, showing that the presence of a PE or PPE domain is required for secretion via ESX-5. Unfortunately, our attempt to investigate whether PE and PPE domains are sufficient for secretion failed because PE LipYtub and PPE LipYmar were unstable when expressed alone. However, in another study, it was shown that the PE domain of PE_PGRS33 can be secreted on its own by ESX-5. 3 In M. marinum, surface-exposed LipY tub and LipY mar were processed, which resulted for LipY tub in the removal of the PE domain and first half of the linker domain. No further processing or degradation was observed of LipY tub after extended incubation of M. marinum lysates, indicating that LipY cleavage results from a specific maturation process, linked to ESX-5 translocation (Fig. 7). Similar patterns of N-terminal processing have previously been observed for PE_PGRS and PPE_MPTR proteins secreted by ESX-5 (29), although the responsible protease was not identified. This could mean that processing of ESX-5 substrates is a common process, similar to the removal of a signal sequence. Interestingly, processing has also been observed for the ESX-1 substrate EspB (16), which was recently shown to be cleaved by the ESX-1-encoded protease MycP1 (47). To further investigate the nature of the LipY tub cleavage site, we altered the neighboring residues using site-directed mutagenesis. Although introducing point mutations could block processing at the original cleavage site, the mutations always resulted in an alternative processing pattern, as is also observed for signal sequence removal (48,49). Besides the removal of the secretion domain, processing could also have a functional implication. M. tuberculosis utilizes fatty acids as the principal energy source during the latent stage of infection (50 -52), and it has been shown that it stores fatty acids in the form of TAGs (51). Dormancy leads to high up-regulation of LipY tub , whereas a ⌬lipY tub mutant has a severely compromised ability to degrade stored TAGs upon starvation (32). This indicates that LipY is involved in degradation of intracellular TAGs. In this study, we show that surface-exposed LipY tub can also hydrolyze extracellular lipids. LipY may therefore also be involved in fatty acid acquisition by hydrolyzing host lipids during infection. Mishra et al. (33) showed that the lipase activity of Our data support a model in which the PE or PPE domain of the LipY homologues functions as a secretion signal that is recognized by the ESX-5 secretion system (1). After recognition, LipY is translocated across the mycobacterial cell envelope by the ESX-5 system (2), and during or after translocation the PE or PPE domain is proteolytically removed (3), possibly by the mycosin associated with the ESX-5 system. Maturation only occurs in strains with a functional ESX-5 system, and no further processing is observed during extended incubation of lysed bacteria. After secretion, processed LipY is loosely associated with the bacterial surface, where it can hydrolyze TAGs (4). IM, inner membrane; MOM, mycobacterial outer membrane.
both LipY tub and LipY mar is significantly higher in the absence of the PE or PPE domains. Therefore, processing of surfaceexposed LipY will result in increased lipase activity and more efficient hydrolysis of host lipids.
Intriguingly, LipY tub was found to be neither surface-exposed nor processed when expressed in cultures of M. tuberculosis and M. bovis BCG. However, infection of THP-1 macrophages with M. tuberculosis or M. bovis BCG did result in LipY tub processing, suggesting that the pathway required for secretion and/or processing in M. tuberculosis complex members is activated upon infection of host cells. Endogenous M. tuberculosis lipY is induced upon infection of macrophages (46), and it is conceivable that additional proteins required for LipY secretion and processing are also up-regulated during an infection. These proteins are not the ESX-5 system itself, which seems to be constitutively produced. Instead, it could for instance be a specific chaperone required for secretion or because proteins secreted via type VII secretion systems are often secreted as dimers, another ESX-5 substrate (3,53).
Our data showing that the PE/PPE domain of LipY is essential for secretion conflict with the results presented in Ref. 33, where it was shown that localization of LipY tub and LipY mar in the cell wall of M. smegmatis and M. bovis BCG and on the bacterial surface of M. bovis BCG is independent of these domains. Moreover, our analysis in M. tuberculosis and M. bovis BCG showed that LipY is not surface-exposed when grown in culture. Although we cannot explain these differences, we have analyzed the LipY secretion pathway in more detail using ESX-5 mutants. Immunoblotting and whole cell immunogold labeling EM showed a clear lack of surface localization of both LipY tub and LipY mar in the ESX-5 mutant, demonstrating that these are indeed ESX-5 substrates. This was again confirmed in our whole cell lipase assay, in which the p-nitrophenyl stearate hydrolysis observed with LipY tub in wild-type M. marinum was absent in the ESX-5 mutant strain.
Some PE and PPE proteins are encoded in operons (18). One such couple, consisting of PE25 and PPE41, forms a complex (54) and is secreted by ESX-5 as a heterodimer. 4 However, computational predictions suggest that many other PE and PPE proteins, including several that are encoded alone in the genome, also have cognate interaction partners (55). This raises the hypothesis that PE and PPE proteins are targeted to ESX-5 in pairs. Experiments are underway in our laboratory to investigate how PE and PPE domains may interact in targeting to ESX-5 and to identify the residues within the PE and PPE domains that constitute the ESX-5 secretion signal.