Understanding specificity of the mycosin proteases in ESX/type VII secretion by structural and functional analysis

Abstract

Mycobacteria use specialized ESX secretion systems to transport proteins across their cell membranes in order to manipulate their environment. In pathogenic Mycobacterium tuberculosis there are five paralogous ESX secretion systems, named ESX-1 through ESX-5. Each system includes a substilisin-like protease (mycosin or MycP) as a core component essential for secretion. Here we report crystal structures of MycP₁ and MycP₃, the mycosins expressed by the ESX-1 and ESX-3 systems, respectively. In both mycosins the putative propeptide wraps around the catalytic domain and does not occlude the active site. The extensive contacts between the putative propeptide and catalytic domain, which include a disulfide bond, suggest that the N-terminal extension is an integral part of the active mycosin. The catalytic residues of MycP₁ and MycP₃ are located in a deep active site groove in contrast with an exposed active site in majority of subtilisins. We show that MycP₁ specifically cleaves ESX-1 secretion-associated protein B (EspB) in vitro at residues Ala358 and Ala386. We also systematically characterize the specificity of MycP₁ using peptide libraries, and show that it has evolved a narrow specificity relative to other subtilisins. Finally, comparison of the MycP₁ and MycP₃ structures suggest that both enzymes have stringent and different specificity profiles that result from the structurally distinct active site pockets, which could explain the system specific functioning of these proteases.

1. Introduction

Tuberculosis (TB) is a devastating infectious disease caused by Mycobacterium tuberculosis that resulted in more than 8.7 million new cases and 1.4 million deaths in 2011 (Floyd, 2012). M. tuberculosis possesses five related secretion pathways designated ESX-1 through ESX-5 that are involved in the transport of multiple proteins across the complex cell envelop (Abdallah et al., 2007; Stoop et al., 2012). Although the ESX secretion pathways mediate secretion of proteins across two membranes, they are unlike the known secretion pathways for Gram-negative bacteria and, as such, are also referred to as type VII secretion (Abdallah et al., 2007). The importance of ESX-1 secretion for virulence has been established through comparison of mycobacterial strains containing the full complement of ESX secretion genes versus avirulent strains with deleted regions, such as the Mycobacterium bovis BCG vaccine strain (Guinn et al., 2004; Majlessi et al., 2005; Pym et al., 2002). The ESX-1 secretion apparatus is involved in early bacterial replication in macrophages and contributes to virulence by allowing escape of mycobacteria from the phagosome into the cytosol of infected macrophages (Houben et al., 2012a; Simeone et al., 2012). In contrast, the ESX-3 system is essential for mycobacteria, possibly because of its role in iron acquisition (Serafini et al., 2009; Siegrist et al., 2009). The ESX-5 system of M. tuberculosis is responsible for secretion of multiple proteins of the PE and PPE families that modulate cell wall integrity and macrophage responses (Abdallah et al., 2009; Bottai et al., 2012; Houben et al., 2012a). The functions of ESX-2 and ESX-4 systems in mycobacteria are still completely unknown.

A comparison of five ESX loci present in mycobacteria and ESX clusters identified in closely related species, such as Nocardia farcinica and Streptomyces sp. allowed identification of a set of genes that likely encode the core components of the ESX secretion systems (Guinn et al., 2004; Houben et al., 2012b). Interestingly, all known ESX clusters encode their own subtilisin-like proteases, named mycosins or MycP₁ through MycP₅ (Brown et al., 2000). Mycosins are membrane proteins with extracytoplasmic serine protease domains (Dave et al., 2002). The functional role of MycP family proteases in the secretion process is poorly understood, but because MycP₃ is essential for in vitro growth of M. tuberculosis H37Rv (Griffin et al., 2011; Sassetti et al., 2003), we can infer that these mycosins can not substitute for each other. A substrate has only been identified for MycP₁ protease, which is the ESX-1 secretion-associated protein B (EspB), a secreted substrate of ESX-1, in vitro and in vivo (Ohol et al., 2010). EspB is a glycine-rich protein essential for ESX-1 secretion and required for virulence and growth of bacteria in macrophages (Xu et al., 2007). Importantly, the other ESX systems do not seem to secrete an obvious EspB homologue. Furthermore, mycosins seem to have a dual function in ESX secretion, as deletion of the mycP1 gene results in a lack of secretion, however inactivation of MycP₁ catalytic activity leads to increased secretion of ESX-1 substrates (Ohol et al., 2010). Importantly, the catalytically inactivated mutant is attenuated in mice compared to wild-type M. tuberculosis (Ohol et al., 2010), suggesting that MycP₁ inhibitors might have efficacy as anti-tuberculosis antibiotics.

To better understand the system-specific role of mycosins in the ESX secretion systems, we determined the crystal structures of two mycosins, Mycobacterium thermoresistibile MycP_1mth and Mycobacterium smegmatis MycP_3msm. We also carried out a detailed characterization of MycP_1mth substrate recognition motifs. In addition, we compared our findings with the recently reported structure of M. smegmatis MycP_1msm (Solomonson et al., 2013). Our structural and biochemical analyses of MycP₁ and MycP₃ provide insights into the molecular recognition mechanisms used by this family of enzymes. These insights are important for understanding the diverse roles of individual ESX secretion systems, and provide a starting point for rational design of mycosin-targeted therapeutics.

2. Materials and methods

2.1. Cloning, expression and purification of MycP_1mth, MycP_1msm, MycP_3msm and EspB_mtu

The DNA fragment corresponding to the soluble domain of MycP_1mth (residues 24–393) was PCR-amplified from genomic DNA of M. thermoresistibile ATCC 19527 and cloned into pET-22b vector (EMD Millipore) for periplasmic expression with an N-terminal pelB signal sequence and a C-terminal tobacco etch virus (TEV) protease cleavage site followed by His₆-tag. Alternatively, the same DNA fragment was cloned into pET-21d vector (EMD Millipore) for cytoplasmic expression with a C-terminal TEV protease cleavage site and His₆-tag. Initial crystallization screening and mass-spectrometry analysis of disulfide bonds were performed with MycP_1mth produced in the periplasm. The crystal growth optimization and kinetic analysis was performed with MycP_1mth produced in the cytoplasm. The gene fragment corresponding to the soluble domain of MycP_1msm (residues 24–396) was amplified from genomic DNA of M. smegmatis mc² 155 (ATCC Catalog No. 700084) and cloned into a modified pETDuet-1 vector (EMD Millipore) to encode an N-terminal His₆-tag and a TEV protease cleavage site. The DNA fragment encoding the soluble domain of MycP_3msm (residues 26–403) was cloned into a modified pET-22b vector to include pelB signal sequence followed by a His₆-tag and a TEV protease cleavage site. The DNA fragment encompassing full-length EspB_mtu and EspL_mtu genes was amplified from genomic DNA of M. tuberculosis H37Rv and cloned into a modified pET-28b vector (EMD Millipore) with an N-terminal His₆-tag and a TEV protease cleavage site. The catalytically inactive mycosin variants MycP_1mth S334A, and MycP_3msm S341A were generated using the QuickChange (Stratagene) site-directed mutagenesis protocol. The plasmid to produce the N-terminal EspB_mtu fragment (residues 1–358) was constructed by introducing a stop codon in position 359. All plasmids were sequenced to verify absence of inadvertent mutations.

All proteins were expressed in Rosetta (DE3) E. coli cells after induction with 0.5 mM IPTG for 4 h at 18 °C (mycosins) or at 24 °C (EspB_mtu). Cells were harvested by centrifugation and resuspended in buffer containing 20 mM Tris-HCl pH 8.4, 300 mM NaCl, and 20 mM imidazole. The resuspended cells were lysed using EmulsiFlex-C5 (Avestin) and proteins were purified via a Ni-NTA column (Qiagen). Following the cleavage of His₆-tag by TEV protease, proteins were purified on size-exclusion Superdex200 column (GE Biosciences) in buffer containing 20 mM HEPES pH 7.5 and 100 mM NaCl.

2.2. Identification of MycP₁ disulfide bonds by mass-spectrometry

MycP_1mth samples were subjected to digestion with trypsin in solution. 6 μL of 10 ng μl⁻¹ trypsin in 40 mM (NH₄)HCO₃ with 9% acetonitrile solution was added to a 50 μl sample MycP_1mth (~10 μg ml⁻¹ in 50 mM (NH₄)HCO₃) and incubated at 37°C for 16 h. Two sets of digested samples were prepared. One set was pre-treated with dithiothreitol and iodoacetamide before trypsin digest, while the other one was not. The digested solution was subjected to LC-MS/MS analysis as previously described (Kuo et al., 2012). LC-MS/MS analysis was performed using an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific) coupled with an Eksigent Nanoflex cHiPLC system (Eksigent) through a nanoelectrospray ionization source. The LC-MS/MS data were subjected to database searches for protein identification using Proteome Discoverer software V. 1.3 (Thermo Fisher Scientific) with a local MASCOT search engine. A custom database containing the protein of interest was used.

2.3. Cloning of MycP_1mm-HA and MycP_1mth-HA and expression in M. marinum

The DNA fragments corresponding to MycP_1mm and MycP_1mth were PCR-amplified from M. marinum E11 and M. thermoresistibile ATCC 19527 genomic DNA, respectively, using a 5′ primer containing an NheI restriction site and a 3′ primer that contains a BamHI restriction site and a HA-encoding sequence. After cloning of the PCR products into the pJet1.2 vector (Fermentas), the fragments were digested by NheI and BamHI and ligated into the E. coli-mycobacterial shuttle vector pSMT3::LipY_tub (Daleke et al., 2011), which was cut with the same enzymes to remove the lipY_tub gene. The resulting vectors, named pSMT3::MycP_1mm-HA and pSMT3::MycP_1mth-HA, were introduced in M. marinum E11 by electroporation. The various M. marinum strains were grown at 30°C in Middlebrook 7H9-broth (Difco) supplemented with 0.2% glycerol, 10% oleic acid-albumin-dextrose-catalase (OADC) (BBC) and 0.05% Tween 80. Expression and subcellular localization of the HA-tagged proteins in M. marinum were analyzed by subcellular fractionation followed by SDS-PAGE and immunoblotting as previously described (Houben et al., 2012b; van der Woude et al., 2013). Briefly, bacteria were grown until an OD₆₀₀ of ~1, harvested by centrifugation and resuspended in PBS, 250 mM sucrose. The cells were lysed in a One-Shot cell disruptor and cellular debris was removed by centrifugation 3 times for 5 min at 3,000 × g. Subsequently, the cytosolic fraction was separated from the cell envelope fraction by centrifugation at 100,000 × g.

2.4. Crystallization and data collection

The initial crystal hits were found using JCSG Core Suites I–IV (Qiagen) (Lesley and Wilson, 2005). The thin rod-shaped MycP_1mth crystals were obtained by vapor diffusion method mixing protein concentrated to 4.5 mg ml⁻¹ and 0.2 M Zn acetate, 0.1 M imidazole pH 8.0, 25% 1,2-propanediol, 10% glycerol. Crystals were harvested directly from crystallization solution and flash-cooled in liquid N₂ prior to data collection. The optimized MycP_3msm crystals were produced using protein at concentration 10–14 mg ml⁻¹ and 0.2 M MgCl₂, 0.1 M Tris-HCl pH 8.4, 20% PEG8000, 0.01 M NiCl₂. Crystals were cryoprotected in crystallization solution supplemented with 10% ethylene glycol and vitrified in liquid N₂. Data were collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. All data were processed and scaled using XDS and XSCALE (Kabsch, 2010).

2.5. Structure determination and refinement

The crystals of MycP_1mth belong to space group P2₁2₁2₁ with one molecule in the asymmetric unit. The structure was solved by molecular replacement using BALBES (Long et al., 2008) and thermitase, a subtilisin-fold protease from Thermoactinomyces vulgaris (PDB 1THM), as a search model (Teplyakov et al., 1990). Following extensive manual rebuilding using Coot (Emsley et al., 2010) the model was completed and solvent molecules were added using ARP/wARP and AutoSolve module in PHENIX (Adams et al., 2010; Langer et al., 2008). The structure was refined to 2.0 Å resolution with R_work 0.169 and R_free 0.206 using REFMAC5 (Murshudov et al., 2011) and applying twelve translation, libration and screw-rotation displacement (TLS) groups determined by the TLSMD server (Painter and Merritt, 2006). The final model includes residues 24–391 of MycP_1mth, seven Zn²⁺ ions identified by the presence of positive peaks in anomalous difference map, and other solvent molecules.

The crystals of MycP_3msm belong to space group P1 with one molecule in the asymmetric unit. The structure was solved by molecular replacement using Phaser (McCoy et al., 2007) and MycP_1mth structure without the N-terminal extension (residues 67–391) as a search model. The model was rebuilt using ARP/wARP and Coot. The structure was refined to 1.5 Å resolution with R_work 0.151 and R_free 0.185 using REFMAC5 and two TLS groups defined by the TLSMD server. The final model includes residues 26–399 of MycP_3msm with the loop residues 241–245 missing due to disorder.

Data collection and refinement statistics are listed in Table 1. The quality of final models was assessed using Coot and the MolProbity server (Chen et al., 2010). The structural figures were generated using PyMol (www.pymol.org) and Chimera (Pettersen et al., 2004). The coordinates and structure factors were deposited to the Protein Data Bank with accession codes 4HVL (MycP_1mth) and 4KG7 (MycP_3msm).

Table 1.

Data collection and refinement statistics.

	MycP1mth (PDB 4HVL)	MycP3msm (PDB 4KG7)
Data collection
Space group	P212121	P1
Cell dimensions
a, b, c (Å)	61.95, 73.08, 75.42	42.82, 46.95, 47.54
α β γ(°)	90, 90, 90	77.27, 68.35, 74.47
Resolution (Å)	47.9–2.00 (2.11–2.00)1	43.8–1.50 (1.58–1.50)
Rsym	0.117 (0.684)	0.048 (0.411)
I/Iσ	14.2 (2.2)	15.6 (2.3)
Completeness (%)	99.7 (98.0)	91.4 (65.4)
Multiplicity	6.7 (4.1)	2.6 (2.5)
Refinement
Resolution (Å)	47.9–2.00	43.8–1.50
No. reflections (total/free)	23667/1208	48069/2432
Rwork/Rfree	0.169/0.206	0.151/0.185
No. atoms
Protein	2653	2648
Ligand/ion	31	1
Water	209	533
B-factors
Protein	23.5	13.3
Ligand/ion	24.0	17.8
Water	31.1	27.0
Wilson B	28.5	19.2
R.m.s. deviations
Bond lengths (Å)	0.010	0.010
Bond angles (°)	1.310	1.410
Ramachandran distribution (%)2
Favored	97.6	96.4
Outliers	0.0	0.0

¹Values in parentheses are for the highest-resolution shell.

²Calculated using the MolProbity server (http://molprobity.biochem.duke.edu).

2.6. In vitro assay of MycP_1mth activity

EspB at a concentration of 0.15 mg ml⁻¹ was incubated with 0.015 mg ml⁻¹ MycP_1mth or 0.015 MycP_3msm at 37°C in reaction buffer (20 mM HEPES pH 7.5, 100 mM NaCl, 2 mM CaCl₂) for the indicated time periods. At each time point a 10 μL aliquot of the mixture was mixed with SDS loading buffer and heated at 95 °C for 5 min. The aliquots were analyzed on 15% SDS-PAGE gel. EspB_mtu-C was separated on SDS-PAGE and transferred to PVDF membrane (Pierce) using standard protocols. The appropriate bands were excised and submitted to 5 cycles of N-terminal sequencing using the Edman degradation method. The sequencing was performed on an Applied Biosystems 494 Procise HTS sequencer, Tufts University Core Facility.

For quenched fluorescent peptide assays MycP_1mth, MycP_1msm or MycP_3msm were concentrated in the reaction buffer (50 mM HEPES pH 7.5, 100 mM NaCl). The quenched fluorescent substrate Abz-AVKAASLGK(Dnp)-OH was obtained from Genscript Inc. Concentrated substrate stock solutions were made by dissolving the substrate in dimethylformamide (DMF) to a final concentration of 20 mM. Prior to starting the assay 0.5 mM stocks of the substrate were made by adding the reaction buffer. The concentration was confirmed using a Nanodrop spectrophotometer to measure the absorbance of the Dnp quencher at 360 nm (E = 17,400 M⁻¹ cm⁻¹ at 360 nm) (Carsten and Eisen, 1953). 200 μl volumes of substrate solutions in the reaction buffer at the concentrations indicated were warmed to 37 °C in a 96 well opaque black plate. The protease was added to all wells simultaneously to a final concentration of 2 μM. Fluorescence measurements were read starting immediately and every 30 sec for 1 hr in a Spectramax Gemini XPS plate reader with settings: excitation = 319 nm, emission = 419 nm.

The fluorescence increase per cleavage event was measured by reading the fluorescence of control Abz-AVKAASLGK(Dnp)-OH solutions that were completely degraded by trypsin in parallel with MycP digestions. V_max of MycP_1mth or MycP_1msm at each substrate concentration was then calculated by measuring the initial slope of the data between 10 min and 60 min. V_max versus the substrate concentration was then plotted and kinetic constants calculated using Graphpad Prism software.

For immobilized fluorescent peptide library digestions cellulose discs with individual peptides covalently attached were purchased from JPT Peptide Technologies GmbH (Berlin, Germany). The discs contained ~54 nmol of peptide each and were provided in 96-well format. The discs were washed with methanol and reaction buffer (50 mM HEPES pH 7.5, 100 mM NaCl) according to the manufacturer’s instructions. The fluorescence of the final wash was checked to ensure spontaneous desorption from the discs was not significant. After washing the discs were submerged in 150 μL of reaction buffer containing 4 μM MycP_1mth and incubated for 2 hr at 37 °C. After 2 hrs the wells were thoroughly mixed by pipetting up and down then 75 μL was aspirated from each well and transferred to a second 96-well plate. The fluorescence of each well was measured using an M5 Spectramax platereader with settings: excitation = 319 nm, emission = 419 nm.

2.7. Circular dichroism (CD) spectroscopy

Full length EspB_mtu and EspB_mtu-N were concentrated to 20 μM in 50 mM HEPES, 100 mM NaCl. EspB_mtu-C was obtained by digesting EspB_mtu with MycP_1mth overnight at 20 °C. EspB_mtu-C was then purified from the digestion mixture by size exclusion chromatography (SEC) using a Superdex 200 column. A single 0.5 mL fraction with two protein fragments migrating at ~10 kDa and ~11 kDa was concentrated to a final concentration of 20 μM. Ellipticity measurements were obtained using a 0.1 cm quartz cuvette maintained at 37 °C in a Jasco J-810 spectropolarimeter.

3. Results

3.1. MycP₁ mycosin forms two disulfide bonds

Based on homology with other subtilisins it has been suggested that mycosins are translated as a preproprotein with a secretory signal peptide (residues 1–21 for MycP_1mtu), a putative propeptide (residues 22–63) (Dave et al., 2002), a catalytic domain (residues 64–394) and a predicted C-terminal transmembrane region (Brown et al., 2000) (residues 419–439) (Supplementary Fig. S1). TOPCONS analysis (Bernsel et al., 2009) of the mycosin cellular localization predicts that the protease domain is located in the periplasmic space between the inner membrane and outer mycomembrane of mycobacteria. To provide a suitable folding environment and obtain soluble mycosin variants, we substituted the native signal peptide of MycP_1mtu with the signal peptide from Erwinia carotovora PelB and truncated the predicted transmembrane region. Our efforts to produce soluble mycosin proteases were initially hampered by poor expression and/or lack of solubility of M. tuberculosis mycosins expressed heterologously in Rosetta(DE3) cells. However, we successfully expressed and purified homologous mycosins from M. thermoresistibile (MycP_1mth), and M. smegmatis (MycP_1msm), as well as MycP₃ from M. smegmatis (MycP_3msm), using the same methods.

All mycosins have four conserved Cys residues present in their sequences (Supplementary Fig. S1), which prompted us to investigate the presence of potential disulfide bonds. Using mass spectrometry methods to analyze tryptic digest peptides of reduced and non-reduced MycP_1mth samples, we identified two disulfide bonds, one formed between residues Cys51-Cys120 and the other between Cys206-Cys244 (Supplementary Fig. S2). Surprisingly, Cys51 is located in the putative propeptide region, which raised doubts concerning the proteolytic removal of this predicted propeptide.

To investigate whether the N-terminal extensions of mycosins are processed in vivo, a C-terminally HA-tagged version of MycP₁ from M. thermoresistibile and M. marinum (MycP_1mth and MycP_1mm, respectively) were expressed in M. marinum (Fig. 1A). Although the expression of MycP_1mth was poor, we could demonstrate that both MycP₁ proteins showed an apparent molecular size of ~52 kDa (calculated molecular mass of MycP_1mm without signal peptide is ~43.3 kDa and without putative propeptide ~38.9 kDa), indicating that the N-terminal extension remains attached to the protease domains. Furthermore, MycP_1mm-HA almost exclusively localized to the cell envelope fraction (Fig. 1B), in accordance with the presence of a predicted transmembrane region at its C-terminus.

Fig. 1. Expression of HA-tagged MycP_1mm and MycP_1mth in M. marinum.

Immunoblot analysis with anti-HA antibodies of (A) total cell lysates prepared from M. marinum expressing MycP_1mm-HA (lane mm) and MycP_1mth-HA (lane mth) and (B) cytosol (lane cyt) and cell envelope (lane CE) fractions prepared from M. marinum expressing empty vector (lane −) and MycP_1mm-HA (lane +). The HA-tagged proteins are indicated by an asterisk.

In addition, the presence of the N-terminal extension in MycP_3mm and MycP_5mm was observed by carefully analyzing previously obtained mass spectrometry results of total cell envelope fractions of M. marinum (van der Woude et al., 2013), while no peptide corresponding to the N-terminus of MycP_1mm was identified by tandem MS analysis (Supplementary Fig. S3). Moreover the N-terminal extension of MycP_3mtu of M. tuberculosis was also previously identified in a proteomics study (Malen et al., 2007). Together, these findings suggest that the N-terminal extensions of these two different mycosins remain attached to the protease domains in vivo.

3.2. MycP_1mth and MycP_3msm have a subtilisin-like fold

In order to gain further insight into the basis for MycP specificity, we determined the crystal structures of MycP_1mth and MycP_3msm (Fig. 2A). Structures were refined to 2.0 Å (MycP_1mth) and 1.5 Å (MycP_3msm) resolution (see Materials and Methods and Table 1 for structure solution details). Both MycP_1mth and MycP_3msm contain a typical subtilisin fold, including all the secondary structure elements of the extended core (Siezen et al., 1991). These include a 7-stranded β sheet core surrounded by 8 α helices (Fig. 2A and Supplementary Fig. S4). The structures of MycP_1mth and MycP_3msm superimpose with RMSD 1.7 Å and 48% sequence identity.

Fig. 2. Crystal structures of MycP_1mth and MycP_3msm.

A. The structures are shown in the same orientation with α helices and β strands of core subtilisin domain in green, the insertion loops in purple and the N-terminal extension in blue. The catalytic triad residues (yellow) are shown as sticks. Bound ions are omitted for clarity. The disordered amino acid residues in loop 2 of MycP_3msm are indicated by a purple dashed line.

B. Structures of MycP_1mth and MycP_3msm and subtilisin Carlsberg (PDB 3UNX) (Fisher et al., 2012) are shown in surface representation with catalytic triad residues highlighted in yellow.

A Dali search (Holm and Rosenstrom, 2010) using the structure of MycP_1mth indicated that subtilisin Carlsberg (PDB 3UNX) of Bacillus licheniformis is the closest structural homolog with an overall RMSD of 2.0 Å (Fisher et al., 2012). However, relative to subtilisin Carlsberg, MycP_1mth and MycP_3msm have three sequence insertions consisting of residues 160–176 (loop 1, MycP_1mth numbering), 241–262 (loop 2), and 319–330 (loop 3) (Figs 2 and 3). These loops pack within the N-terminal extensions of MycP_1mth (residues 24–65) and MycP_3msm that wrap around the perimeter of the molecule, and all three insertion loops cluster around the active site pocket. These loops and the N-terminal extension together contribute to three 2-stranded anti-parallel β sheet elements in addition to the core subtilisin fold. These new structural elements surround the active site creating a relatively deep active site pocket (Figs 2B and 4). The two disulfide bonds, which are present in both MycP_1mth and MycP_3msm structures, likely contribute to the stability of these extended loops.

Fig. 3. Structure based sequence alignment of MycP_1mth, MycP_3msm and subtilisin Carlsberg.

Secondary structure elements of MycP_1mth and subtilisin Carlsberg are displayed above and below the sequences, respectively. The signal protease cleavage site is indicated by vertical arrow. The N-terminal extension is indicated by a blue line. The mycosin insertion loops are indicated by purple lines. The catalytic triad residues and conserved Asn residue coordinating oxyanion hole are highlighted in blue. The conserved Cys residues that form disulfide bonds in mycosins are highlighted in yellow.

Fig. 4. The mycosin-specific N-terminal extension and loop insertions.

MycP_1mth and MycP_3msm structures are displayed as ribbons with semitransparent gray surface around the core subtilsin domain. The catalytic triad residues are highlighted in yellow. Loop insertions are highlighted in purple. The N-terminal extension amino acid residues are shown as blue sticks surrounded by σA-weighted 2F_O–F_C electron density map contoured at 1 σ.

3.3. A unique N-terminal extension in mycosins does not occlude the active site

Subtilase propeptides normally prevent substrate access to the active site until degradation of the propeptide occurs (Gallagher et al., 1995; Shinde and Inouye, 1995). However, we found that the N-terminal extensions of MycP_1mth and MycP_3msm follow a meandering path around the outer surface of the protease domain without obscuring the active sites (Fig. 4). Thus the catalytic residues are exposed at the bottom of the active site pocket. Similarly, the structure of a mycosin from M. smegmatis revealed an enzyme with intact propeptide without any obvious obstruction of the active site (Solomonson et al., 2013). Together these data suggest that, in contrast to other subtilisins, the N-terminal extension of MycP_1mth and MycP_3msm is part of the mature enzyme and not a cleavable propeptide.

Analysis of the N-terminal extension/MycP_1mth interface using the PISA server (Krissinel and Henrick, 2007) found 4200 Ų total buried surface area, 6 salt bridges and 26 hydrogen bonds. Similarly, the interface between the N-terminal extension and protease domain in MycP_3msm buries 4100 Ų, and contains 1 salt bridge and 30 hydrogen bonds. These interactions together with the presence of disulfide bonds between residues Cys51-Cys120 (MycP_1mth numbering) indicate the integral nature of the N-terminal extension in the mycosin fold and further support a suggested role in fold stabilization (Solomonson et al., 2013).

3.4. Catalytic triad residues are in active conformation

In MycP_1mth the catalytic triad consisting of Asp92, His123, and Ser334 adopts the conformation seen in other subtilases with the exception of Ser334 (Fig. 5A). During catalysis the hydroxyl group of the catalytic serine must point upwards toward the substrate, but in MycP_1mth it is turned sideways and lies about 1.8 Å out of place. This slightly aberrant positioning was also reported for the catalytic serine of MycP_1msm (Solomonson et al., 2013). Conformational rearrangement of the serine side chain probably occurs upon substrate binding. Small side chain rearrangements are common in protein-peptide interactions (London et al., 2010) therefore the observed conformation of Ser334 should not be considered an obstacle to hydrolysis. The catalytic triad of MycP_3msm consisting of Asp95, His126 and Ser341 adopts an almost identical configuration (Fig. 5B). Interestingly, Ser341 appears in a double conformation, the first corresponding to the typical subtilisin-like orientation and the second corresponding to the orientation of Ser334 in MycP_1mth. Again, we expect that the presence of a substrate at the active site will stabilize the solvent-facing rotamer of Ser341 resulting in the correct orientation for nucleophilic attack on the peptide bond.

Fig. 5. Catalytic triad and binding grooves of MycP_1mth, MycP_3msm and subtilisin Carlsberg/eglin complex.

A. The surface of MycP_1mth is shown in gray with residues of the catalytic triad and Asn239 highlighted in yellow. Zinc ion is shown as a pink sphere. The substrate recognition subsites and important residues are labeled explicitly. Distances between catalytic triad side chains are shown as dashed lines. Residues 242, and 320–329 of MycP_1mth (loop 3) were not displayed for clarity.

B. MycP_3msm is represented in the same way as (A). Residues 326–335 of MycP_3msm (loop 3) were omitted for clarity.

C. Subtilisin Carlsberg/eglin complex (PDB 2SEC) (McPhalen and James, 1988) represented as in (A). Only eglin residues 53–61 are shown as cartoon with side chains at P4–P1′ positions shown as sticks. The glycine/serine rich linker strands (residues 99–103 and 126–130) that make backbone interactions with eglin are shown as cartoons. Subtilisin side chains close enough to interact with eglin are shown as sticks within the surface and labeled explicitly.

The catalytic triad histidine in the MycP_1mth structure has an unusual backbone conformation with φ–136.7° ψ–66.5° for His123 in MycP_1mth, similar to the conformation in MycP_1msm structure (Solomonson et al., 2013). The typical backbone conformation of the catalytic histidine residue is different in MycP_3msm (φ–73.5° ψ–50.3° for His126) and other subtilisins, for example φ–61.4° ψ–48.2° for His64 in subtilisin Carlsberg (PDB 3UNX) (Fisher et al., 2012). The functional consequences of these differences in MycP_1mth are unclear, because the distances between the catalytic triad side chains in the mycosins are similar relative to other subtilisins. For example Asp → His = 2.6 Å and His → Ser = 3.0 Å in MycP_1mth and Asp → His = 2.7 Å and His → Ser = 2.7 Å in subtilisin Carlsberg (PDB ID 2SEC) (Fig. 5C). Most likely there is no functional consequence since all the functional groups participating in the serine protease catalytic mechanism are in place. Furthermore, it is not uncommon for main-chain scaffolds supporting serine protease catalytic triads to vary widely (Dodson and Wlodawer, 1998).

3.5. The MycP_1mth and MycP_3msm substrate binding sites suggest unique substrate specificity

Subtilase substrate specificity is typically determined by interactions with substrate residues designated P6–P3′ with residues P4–P1 having particular importance (Gron et al., 1992; Siezen et al., 1991). Inhibitor complex structures of subtilisin Carlsberg and subtilisin BPN indicate that residues P4–P1 follow a conserved surface groove and make anti-parallel sheet like interactions between two β sheet elements (Bode et al., 1987; McPhalen and James, 1988; McPhalen et al., 1985; Siezen and Leunissen, 1997). These structural elements correspond to MycP_1mth residues 158–162 (β4/β5 linker) and 202–205 (β6/β7 linker) (Fig. 3). In MycP_1mth and MycP_3msm, as well as MycP_1msm (Solomonson et al., 2013), these strands are separated by ~15 Å. This wide groove makes it impossible for a single strand to make sheet-like interactions as expected in classic subtilisin/substrate interactions. Furthermore, in MycP_1mth and MycP_3msm the canonical binding groove is completely blocked near S4 by loop 1 (residues 160–176 for MycP_1mth and 162–175 for MycP_3msm), which extends helix α4 by one turn in MycP_1mth and by two turns in MycP_3msm. This insertion effectively walls off the distal end of the binding groove in both MycP_1mth and MycP_3msm, albeit more dramatically in MycP_3msm (Fig. 5).

Additional differences in MycP_1mth and MycP_3msm substrate binding surfaces compared to classic subtilisins occur downstream of the active site at substrate positions P1′–P3′. Typically substrates exit abruptly after brief backbone interactions with the β strand corresponding to β13 strand of MycP_1mth (residues 325–332). This strand is roughly level with the active site and the rest of the binding groove in subtilisin Carlsberg, but in MycP_1mth and MycP_3msm it is tilted 90° and overhangs the catalytic triad. Therefore, the substrates must exit at a 90° angle from the active site or else bend to one side of the β13 strand. Either option results in a dramatically bent substrate exit groove that is likely to result in an extended downstream binding groove relative to classic subtilisins. The downstream groove probably mediates interaction with the large aromatic substrate residues that are favored at positions P2′ and P3′ (as discussed in chapter 3.9).

3.6. MycP_1mth and MycP_3msm structures show system specific differences

The substrate binding groove of MycP_1mth is similar to that of MycP_1msm (78% sequence identity for full-length MycP_1mth and MycP_1msm) (Solomonson et al., 2013). All of the residues highlighted in Fig. 5A are identical in the two proteins with the minor exception of His155 (MycP_1mth) which is exchanged for Gln155 (MycP_1msm). In contrast, the substrate binding grooves of the MycP_1mth and MycP_3msm structures reveal significant differences that probably result in ESX system specific substrate specificities. First, the overall surface electrostatic potential inside the binding groove of MycP_1mth is more negative compared to MycP_3msm (Fig. 6). Second, the insertion loop 2, which is located beside the substrate binding groove, is 5 amino acids longer in MycP_3msm relative to MycP_1mth (residues 241–245 in MycP_3msm). In MycP_3msm, these residues are disordered, but they appear to occupy the S4 site within the binding groove. Moreover, loop 1 moves 9 Å towards the P4 site in MycP_3msm relative to MycP_1mth (Fig 2). The combined effect of loop 1 and 2 is to truncate the putative binding groove in MycP_3msm, which may decrease the number of residues it recognizes upstream of its cleavage site. Finally, there is a significant difference between MycP_1mth and MycP_3msm at site S2. This site is relatively deep and polar in MycP_1mth. The same site in MycP_3msm is shallow, because it is filled by nonpolar Leu202 (Fig. 5B). Thus, the structures of MycP_1mth and MycP_3msm contain unique architecture that indicates they have evolved to recognize different specific sequences.

Fig. 6. Active site grooves of MycP_1mth and MycP_3msm have distinct surface properties.

The structures of MycP_1mth and MycP_3msm are shown as surfaces colored according to the solvent accessible electrostatic potential. The electrostatic surface potential was calculated using the PDB2PQR server, APBS and APBS/PyMol plugin (London et al., 2010; McLaughlin et al., 2007).

3.7. MycP_1mth cleaves EspB_mtu at Ala358 in vitro

Although most ESX-1 substrates seem to be secreted as full-length proteins, it has been reported that EspB, is processed during secretion in M. tuberculosis, M. smegmatis, and M. marinum (Gao et al., 2004; McLaughlin et al., 2007; Ohol et al., 2010; Xu et al., 2007). However, the exact location of the cleavage site has been the subject of conflicting reports (Ohol et al., 2010; Solomonson et al., 2013). We incubated MycP_1mth with EspB_mtu then isolated the C-terminal EspB_mtu fragments and determined their molecular mass using mass-spectrometry (Fig. 7). The major C-terminal cleavage product had a mass of 9,678 corresponding to residues 359–460, whereas the minor C-terminal fragment had a mass of 7,360 corresponding to residues 387–460. Incubation of EspB_mtu with MycP_1msm resulted in the same cleavage products, although with a slower kinetics (data not shown). N-terminal sequencing of these fragments further confirmed that Ala358-Ser359 is the primary cleavage site and Ala386-Gly387 is the secondary cleavage site in agreement with the data reported for MycP_1msm (Solomonson et al., 2013).

Fig. 7. EspB_mtu is processed by MycP_1mth in vitro.

A. EspB_mtu samples were incubated with MycP_1mth or catalytically inactivated variant MycP_1mth S334A for indicated time periods and analyzed by SDS-PAGE. The cleavage leads to accumulation of the N-terminal fragment (EspB_mtu-N) and two C-terminal fragments (EspB_mtu-C) indicated by asterisks.

B. A schematic diagram of the domain structure of EspB_mtu with predicted PE and PPE domains. The C-terminal homology region is indicated by an orange bar (Supplementary Fig. S6). The primary and secondary MycP₁ cleavage sites are located in the low complexity region of EspB_mtu.

C. The sequence conservation of the primary MycP₁ cleavage site is displayed as a sequence logo (http://weblogo.berkeley.edu) (Crooks et al., 2004) based on the sequence alignment of EspB homologs shown in (D).

D. Sequence alignment of EspB homologs from the ESX-1 clusters with the primary and secondary MycP₁ cleavage sites indicated by vertical arrows. The full-length alignment is shown in Supplementary Fig. S6.

In contrast to the MycP₁ proteins, EspB homologs have a lower sequence identity, especially in a low complexity region in their C-terminal parts (Supplementary Fig. S5). However, residues around the identified primary and secondary cleavage sites appear to be conserved in EspB homologs (Figs 8C and 8D). In each EspB protein the conserved residues are consistent with the known specificity of subtilases with an aromatic/hydrophobic residue at position P4, Arg/Lys at P3, Pro/Ala in P2, and Ala or another small residue in P1, (Gron et al., 1992) (using the commonly accepted nomenclature (Schechter and Berger, 1967)).

Fig. 8. Comparison of the relative secondary structure in EspB_mtu, EspB_mtu-N and EspB_mtu-C.

A. Disorder predictions for EspB_mtu with location of the predicted PE and PPE domains indicated by rectangles.

B. CD spectra of EspB_mtu (black long dashed line), EspB_mtu-N (blue short dashed line) and EspB_mtu-C (red solid line). Data are expressed as molar ellipticity. EspB_mtu-N (residues 1–358) was produced recombinantly and EspB_mtu-C was obtained as a mixture of two proteolytic fragments (residues 359–460 and 387–460) as described in Methods.

3.8. The C-terminal EspB_mtu fragment lacks secondary structure

According to the HHpred protein homology prediction algorithm the N-terminal fragment of EspB contains two regions that show similarity with PE and PPE domains, respectively (Supplementary Fig. S6) (Soding et al., 2005). Downstream of the identified MycP_1mth cleavage site (residues 359–460) the sequence of EspB_mtu contains 28% glycine residues with few large hydrophobic or aromatic residues. In silico bioinformatics tools including PONDR-FIT (Xue et al., 2010) predict a disordered C-terminal region (Fig. 8A). We tested the hypothesis that EspB_mtu contains a disordered C-terminal region by circular dichroism (CD). CD spectra clearly show that more secondary structure is present in full-length EspB_mtu relative to the N-terminal EspB_mtu fragment (residues 1–358). However, the CD spectrum of the C-terminal fragments of EspB_mtu (residues 359–460 and 387–460) is characteristic of a protein without secondary structure (Fig. 8B). Taken together, these results indicate that the C-terminal domain of EspB_mtu is at least partly structured in the context of the full-length protein, but fails to maintain any secondary structure by itself. This lack of secondary structure may indicate the presence of unstructured loop(s) in the C-terminal part of EspB_mtu that serve to expose the cleavage site sequence to MycP_1mth.

3.9. Specific residues near cleaved bond are necessary for MycP_1mth recognition

To further investigate the activity of MycP_1mth and MycP_1msm we designed a fluorescent resonance energy transfer (FRET) substrate consisting of EspB_mtu residues 354–361 with an N-terminal o-aminobenzoic acid (Abz) fluorescent group and an additional C-terminal lysine residue derivatized with 2,4-dinitrophenyl (Dnp) quencher acceptor (Abz-AVKAA|SLGK(Dnp)-OH). Both MycP_1mth and MycP_1msm hydrolyzed this substrate and the accumulation of fluorescent product was used to follow proteolytic activity (Fig. 9). Mass spectrometry analysis confirmed that the cleavage occurs at the residue corresponding to Ala358. Kinetic analysis using this peptide showed that the substrate associates with MycP_1mth with a K_m = 60 μM, and a V_max = 6.0 nmol min⁻¹ mg⁻¹, indicating relatively slow catalysis by MycP_1mth. Similarly, kinetic analysis using MycP_1msm to degrade the same substrate yielded a K_m = 86 μM, and a V_max = 1.8 nmol min⁻¹ mg⁻¹. These data suggest that the substrate residues at positions P5–P3′ are sufficient for substrate recognition by MycP₁ proteases. Notably, MycP_3msm protease was completely inactive against this model substrate (data not shown).

Fig. 9. Comparison of the kinetics of MycP_1mth and MycP_1msm in quenched fluorescent peptide assay.

Rate of substrate (Abz-AVKAA|SLGK(Dnp)-OH peptide) cleavage by MycP_1mth and MycP_1msm is plotted as a function of substrate concentration. Kinetic parameters were K_m = 60 mM, and V_max = 6.0 nmol min⁻¹ mg⁻¹ (MycP_1mth) and K_m = 86 mM, and V_max = 1.78 nmol min⁻¹ mg⁻¹ (MycP_1msm).

To determine the sequence specificity requirements for MycP₁ in more detail we tested activity of the mycosins against a peptide library composed of 8-residue N-terminal Abz labeled peptides based on the substrate peptide (Abz-AVKAASLG). Peptides were synthesized with positions P5–P3′ systematically changed to every other amino acid and attached to a cellulose matrix. Digestion with MycP_1mth revealed that MycP_1mth is highly selective in regard to which amino acids it can tolerate at certain positions (Fig. 10). In the P4 position valine is preferred, all other residues significantly abrogate activity. In P3 a basic residue such as lysine or arginine is strictly required consistent with the 100% conservation of Lys356 in EspB (Fig. 7C). Finally, P1 position is selective towards Ala, Met, or Pro residues.

Fig. 10.

Single site substitutional analysis based on the MycP₁ substrate Abz-AVKAASLG using ProteaseSpot^™ system. Bars represent the fluorescent signal from peptide fragments released from the cellulose membrane over 2 hours. Fluorescence is directly proportional to the activity of MycP_1mth against each peptide. The relative activity of MycP_1mth against peptides was normalized such that activity versus EspB_mtu peptide sequence (Abz-AVKAASLG) = 1.0.

Interestingly, acidic residues were not tolerated at any of the sites tested. This probably reflects electrostatic repulsion from the overall negative charge inside the MycP_1mth binding site (Fig. 6). The observed recognition pattern is consistent with the properties of the substrate binding groove of MycP_1mth. A model of peptide in the active site of MycP_1msm has been proposed (Solomonson et al., 2013), however the preference for a hydrophobic residue in position P4, both in the EspB protein family and in our analysis, is not explained by the model. It should be noted, however, that an altered peptide sequence with an additional Gly residue between P3 and P4 positions of the cleavage site was used for docking (Solomonson et al., 2013).

4. Discussion

Mycobacteria, a group of Gram-positive bacteria, have a very complex cell envelope composed of a typical phospholipid bilayer plasma membrane, followed by a periplasmic space and an outer mycomembrane, containing unique lipids and glycolipids (Brennan, 2003; Hoffmann et al., 2008; Zuber et al., 2008). Therefore, the proteins secreted by mycobacteria have to cross not only the plasma membrane, but also the second highly hydrophobic outer membrane. As a result, mycobacteria developed specialized ESX, or type VII, secretion systems, which transport proteins necessary for virulence in M. tuberculosis. The M. tuberculosis genome encodes five such secretion systems, named ESX-1 through ESX-5. The secreted proteins of the ESX system belong to the EsxAB clan (http://pfam.sanger.ac.uk/clan/EsxAB) including WXG100 and PE/PPE family members. The structures of different members of this clan show that they are usually heterodimers composed of a bundle of four α-helices and one of the monomers has a conserved secretion motif adjacent to the second α-helix (Daleke et al., 2012; Strong et al., 2006). The only known mycosin substrate, EspB, also belongs to the EsxAB clan and contains a similar predicted PE/PPE fold and a secretion motif (Supplementary Fig. S6).

One of the most intriguing aspects of the ESX secretion system is the presence of membrane associated subtilisin-like proteases, or mycosins, in all ESX clusters. The only functionally characterized mycosin — MycP₁ — is essential for ESX-1 secretion and was shown to require its protease activity for efficient infection (Ohol et al., 2010). A link between EspB cleavage and virulence was lacking until it has been recently reported that the N-terminal PE/PPE domain of EspB specifically binds phosphatidic acid (PA) and phosphatidylserine (PS), but only in the absence of the C-terminal region downstream of the MycP₁ cleavage site (Chen et al., 2013). Chen et al. show that EspB secretion leads to increased M. tuberculosis virulence in a mouse model of infection, and argue that processed EspB promotes bacterial growth within macrophages by disrupting phospholipid mediated signaling (Chen et al., 2013). Therefore, MycP₁ hydrolysis may serve to activate EspB as it is transported into the host cell.

Another ESX secreted protein, LipY, is also known to undergo proteolytic activation upon secretion. LipY promotes virulence by breaking down triacylglycerol in a reaction that may aid in fatty acid acquisition by the bacilli. Hydrolysis of LipY after Gly149 removes the PE domain necessary for secretion via ESX-5 (Daleke et al., 2011) and also activates the enzymatic activity of LipY (Mishra et al., 2008). There is not currently any evidence linking MycP₅ to LipY processing, and the sequence around Gly149 does not bear an obvious resemblance to the MycP₁ recognition motif described here. However, since there is significant variation in the specificity of mycosins from one system to another it is still possible that MycP₅ is responsible. Apart from EspB and LipY, the majority of ESX secreted cargo proteins do not have assigned functions. As more information is learned the mycosins may be appreciated as a means of regulating both the quantity and the functional maturity of secreted proteins.

In this work we determined the crystal structures of M. thermoresistibile MycP_1mth and M. smegmatis MycP_3msm. We found that MycP_1mth does not contain an autoinhibitory propeptide, similar to the recently reported structure of MycP₁ from Mycobacterium smegmatis (Solomonson et al., 2013). Indeed, our structures extend this observation to include MycP₁ from a second species (MycP_1mth) and from a different genetic cluster (MycP_3msm). In agreement with these data, our in vivo experiments indicate that the N-terminal extension of MycP₁, MycP₃ and MycP₅ are not cleaved in their homologous host. These analyses further suggest that the mycosins may not require the N-terminal extension removal for activation.

A characteristic feature of the mycosins is a relatively deep active site cleft. In contrast, the majority of subtilisins have a shallow solvent-exposed substrate binding groove, which contributes to low selectivity of many subtilisin-like proteases and dictates the requirement for the propeptide acting an intramolecular inhibitor. Thus, the three loop insertions surrounding the active site may act to increase the specificity of the mycosins toward proteins that meet certain size and conformational requirements. The relatively steep active site pocket of the mycosins suggests that substrates adopt a less extended conformation when inserted into the binding pocket. It is tempting to speculate that substrates may loop back over themselves within the binding pocket. In this scenario the conserved glycine/serine rich regions of EspB (Fig. 7D) would provide a fitting substitute for one of the glycine/serine rich strands seen interacting with substrate in the binding grooves of subtilisin Carlsberg and subtilisin Novo (McPhalen and James, 1988).

We found that MycP_1mth and MycP_1msm selectively recognize and cleave the C-terminal region of EspB_mtu at Ala358 with a secondary recognition site at Ala386. Our experiments with fluorescent peptides based on EspB_mtu sequence indicate that eight amino acids are sufficient to be recognized and cleaved by MycP_1mth. Moreover, substitution of the amino acids at all eight positions revealed strict specificity at several positions indicating that these enzymes have evolved to hydrolyze only specific sequences. Finally, the MycP_1mth peptide substrates identified here are not hydrolyzed by MycP_3msm suggesting differences in substrate recognition motifs. This disparate substrate specificity is consistent with the different shape and surface chemistries of MycP_1mth and MycP_3msm active site grooves (Figs 5 and 7). The effort to identify MycP₃ substrates is ongoing.

EspB is the only ESX substrate with a function linked to proteolytic maturation, and it is unique to the ESX-1 system. This raises the question as to what other secreted proteins might be substrates for the remaining mycosin paralogs. CD spectra analysis and in silico predictions suggest that EspB has an α-helical N-terminal domain and a partly disordered C-terminal domain. The cleavage sites of MycP_1mth and MycP_3msm are located in the intervening domain predicted as disordered. This suggests that mycosins may act to remove autoinhibition of cargo proteins by separating ordered domains with flexible loop regions. In addition, mycosins might promote transport through the ESX machinery by removing PE/PPE domains used as ESX secretory signal sequences.

The results presented here are initial stages in our understanding of the mycosin function in ESX secretion at the molecular level. Information presented here on the sequence recognition motif of MycP₁ as well as the unique structural features of MycP₃ relative to MycP₁, may enable the design of specific competitive inhibitors aimed at disrupting proteolytic activation of ESX secreted proteins and generation of new therapeutics against M. tuberculosis.

5. Accession numbers

The coordinates and structure factors were deposited to the Protein Data Bank with accession codes 4HVL (MycP_1mth) and 4KG7 (MycP_3msm).