Interactome Analysis Identifies MSMEI_3879 as a Substrate of Mycolicibacterium smegmatis ClpC1

ABSTRACT The prevalence of drug-resistant Mycobacterium tuberculosis infections has prompted extensive efforts to exploit new drug targets in this globally important pathogen. ClpC1, the unfoldase component of the essential ClpC1P1P2 protease, has emerged as one particularly promising antibacterial target. However, efforts to identify and characterize compounds that impinge on ClpC1 activity are constrained by our limited knowledge of Clp protease function and regulation. To expand our understanding of ClpC1 physiology, we employed a coimmunoprecipitation and mass spectrometry workflow to identify proteins that interact with ClpC1 in Mycolicibacterium smegmatis, a surrogate for M. tuberculosis. We identify a diverse panel of interaction partners, many of which coimmunoprecipitate with both the regulatory N-terminal domain and the ATPase core of ClpC1. Notably, our interactome analysis establishes MSMEI_3879, a truncated gene product unique to M. smegmatis, as a novel proteolytic substrate. Degradation of MSMEI_3879 by ClpC1P1P2 in vitro requires exposure of its N-terminal sequence, reinforcing the idea that ClpC1 selectively recognizes disordered motifs on substrates. Fluorescent substrates incorporating MSMEI_3879 may be useful in screening for novel ClpC1-targeting antibiotics to help address the challenge of M. tuberculosis drug resistance. IMPORTANCE Drug-resistant tuberculosis infections are a major challenge to global public health. Much effort has been invested in identifying new drug targets in the causative pathogen, Mycobacterium tuberculosis. One such target is the ClpC1 unfoldase. Compounds have been identified that kill M. tuberculosis by disrupting ClpC1 activity, yet the physiological function of ClpC1 in cells has remained poorly defined. Here, we identify interaction partners of ClpC1 in a model mycobacterium. By building a broader understanding of the role of this prospective drug target, we can more effectively develop compounds that inhibit its essential cellular activities.

tagged ClpC1 can substitute for endogenous ClpC1 in M. smegmatis, albeit with a growth defect ( Fig. S1C to H), indicating that this construct retains at least partial ability to support proteolysis in cooperation with ClpP1P2.
ClpC1 constructs, or an empty vector control, were expressed in M. smegmatis MC 2 155. ClpC1 and interaction partners were coimmunoprecipitated from lysates in the presence of 10 mM ATP using anti-FLAG beads and separated by SDS-PAGE (Fig. 2). Samples were prepared by in-gel trypsin digestion and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) using higher-energy collisional dissociation (HCD) fragmentation. Proteomic analysis was performed in the Proteome Discoverer software suite against the M. smegmatis MC 2 155 proteome using Sequest HT. Proteins observed in at least two of three biological replicates were considered for further analysis. Five proteins were observed in the empty vector control (Fig. 3A), presumably due to nonspecific interactions with beads. Occurrences of these proteins were ignored in all other samples. In total, 163 and 93 proteins were identified in ClpC1 WT and ClpC1 EQ data sets, respectively ( Fig. 3A; see Tables S1 and S2 in the supplemental material). We expected substrates to bind more stably to ClpC1 EQ than to ClpC1 WT because the wild-type enzyme can presumably unfold substrates in the pulldown conditions. Within the ClpC1 EQ data set, 28 proteins matched these criteria and were absent from the ClpC1 WT data ( Fig. 3B and Table 1). These include cell wall synthesis protein Wag31, succinate dehydrogenase iron-sulfur protein SdhB, aconitate hydratase A (AcnA), and ATP synthase epsilon chain AtpC. There were 65 proteins observed in both the ClpC1 WT and ClpC1 EQ data sets ( Fig. 3A and B). A semiquantitative comparison indicates that 20 of these were enriched in ClpC1 EQ over ClpC1 WT by at least a ratio of 2:1 in average normalized score ( Fig. 3C; Table S2). However, only three proteins were enriched significantly (with a P value of ,0.05): the DNA repair and stress response protein RecA, electron transport chain component ubiquinol-cytochrome c reductase cytochrome b subunit (MSMEG_4263), and a MarR family protein transcriptional regulator (MSMEG_2538) (Fig. 3B). The co-occurrence of these in ClpC1 WT and ClpC1 EQ data sets suggests that these interactions occur stably regardless of ATPase and unfolding activity, as would be expected for adaptors, regulators, or other non-substrate interactors.
We additionally cross-referenced the 191 proteins observed across ClpC1 WT and ClpC1 EQ data sets with M. tuberculosis Clp protease interaction partners identified elsewhere by bacterial two-hybrid screening (23) or by LC-MS/MS upon knockdown of ClpP1 and ClpP2 (21). Thirty M. smegmatis proteins in our full-length ClpC1 interactome have M. tuberculosis homologs identified in these prior studies (see Table 3 below). Subcomponents of ClpC1 interact with specific cellular proteins. The ClpC1 NTD is a discrete folded module thought to interact with substrates and adaptors, such as the N-end rule adaptor ClpS (22), but is dispensable for hexamer formation, ATP hydrolysis, and unfolding activities. We hypothesized that the NTD and the D1/D2 ATPase core of ClpC1 can bind to different interaction partners. To discriminate between interactions with the NTD and ClpC1 core, we created two truncated ClpC1 constructs: one consisting of only the NTD (ClpC1 NTD ) and another lacking the NTD but possessing the full D1 and D2 core (ClpC1 CORE ) (Fig. 1B). We repeated our pulldown and proteomics workflow with these constructs.
We observed 243 proteins and 116 proteins that coimmunoprecipitate with ClpC1 NTD and ClpC1 CORE , respectively (Tables S1 and S2). Comparing data sets, we found 42 proteins that interact with all core-bearing ClpC1 constructs (ClpC1 WT , ClpC1 EQ , and ClpC1 CORE ) (Fig. 4A). Notably, 72 out of 116 proteins (62%) that interact with the ClpC1 CORE were found in at least one other core-bearing data set, which provides additional confidence that these are bona fide interaction partners ( Fig. 4A; Table S2). Of these, only 15 occurred exclusively in core-containing data sets, while the remaining 57 were also found in the ClpC1 NTD data set. Appearance in both ClpC1 NTD and ClpC1 CORE data sets may indicate that these proteins bind nonspecifically. Alternatively, they may genuinely interact with both the NTD and core, for which there is precedence in other organisms. For example, the Bacillus subtilis proteolytic adaptor MecA contacts both the ClpC NTD and M domain (63).
Forty-two proteins were found to interact with all NTD-bearing constructs (ClpC1 WT , Similarly, as shown on the right only 7 proteins were observed interacting with the core-containing constructs and not in the NTD data sets. ClpC1 EQ , and ClpC1 NTD ), although only 7 of these were exclusive to NTD-bearing constructs and absent from ClpC1 CORE . Interestingly, only 84 proteins (35%) identified in the ClpC1 NTD data set were observed in at least one of the other NTD-bearing ClpC1 WT or ClpC1 EQ data sets, suggesting differences in how proteins interact with an isolated monomeric NTD and an NTD in the context of a ClpC1 hexamer.
Our analysis identified few proteins that consistently and uniquely interact with NTD-containing or core-containing ClpC1 constructs. Only 7 proteins coimmunoprecipitated with all of the NTD-bearing constructs and were absent from the ClpC1 CORE data set. Seven separate proteins interacted with all of the core-bearing constructs, but not ClpC1 NTD (Fig. 4B). Surprisingly, 35 proteins were found in all four data sets, suggesting either that some ClpC1-interacting proteins bind to both NTD and core regions or that truncated constructs are able to coassemble with endogenous ClpC1. Overall, our data suggest that many physiological ClpC1 interaction partners exist and that these may bind to multiple regions on the unfoldase.
Mycobacterial ClpC1 has a diverse and far-reaching interactome. To assess whether specific classes of proteins are preferentially targeted by M. smegmatis ClpC1, we performed Gene Ontology (GO) annotation analysis on the interactome using the Blast2GO software suite (64). GO terms were applied to M. smegmatis (strain ATCC 700084/MC 2 155) proteins based on the closest homologs present in the Swiss-Prot/UniProt database. We found that ClpC1 potentially regulates proteins with a diverse range of cellular functions ( Fig. 5 and Table 2). Of particular interest were protein annotation groups that were highly represented in the full-length ClpC1 interactome. The most represented group of proteins were those associated with nucleic acid binding and metabolism, including transcriptional regulators (Hup, Rho, RpoC, Rne, RpoA, Pnp, IleS, RpoB, TopA, DnaX, SigA and transcriptional regulators belonging to the MarR, PadR, GntR, XRE, Crp/Fnr, TetR families, among other proteins). Additionally, proteins involved in redox reactions were identified, such as GabD2, IspG, Pks, DapB, among others. Another well-represented group were proteins involved in cellular transport (including SecA1, SecF, Ffh, LprG, TatA). Some of these proteins are transmembrane, but have cytoplasmic regions through which proteolytic regulation might occur (65,66). In addition to nucleic acid metabolism, observed proteins were involved in other forms of metabolism, such as carbohydrate metabolism (GlmS, Eno, SucB, GlpK, SdhA, SdhB, AcnA, Mqo, Kgd), electron transport chain (Ndh, AtpA, AtpC, AtpD, AtpFH), amino acid metabolism (HemL, SerA, GabT, HisG, CarB, GlnA, MSMEG_3973, MSMEG_0688, MSMEI_3879), and lipid metabolism (fatty acid synthase MSMEG_4757, AcpM, acyl coenzyme A [acyl-CoA] synthases MSMEG_0599 and MSMEG_3767). Taken together, these observations suggest that ClpC1 plays a central role in the framework of mycobacterial cellular physiology by potentially regulating metabolic processes and/or ensuring quality control of metabolic enzymes (31). Functional categorization of cellular proteins interacting with ClpC1 WT and/or ClpC1 EQ is shown. Also shown is the distribution of these proteins by Gene Ontology (GO) annotation based mostly on biological process. Functional classification was performed using Blast2GO and was based on annotation of the closest homologs.    Proteins involved in protein quality control, such as protein folding and unfolding, proteome turnover, and protein homeostasis in general, were also important constituents of the interactome (Fig. 5 and Table 2). These included chaperone proteins GroL1, GroL2, DnaJ1, DnaJ2, DnaK, and Tig and Clp proteins ClpX and ClpC2 (MSMEG_2792). ClpC1 is thus named because of the existence of the orthologous ClpC2 which possesses homology to the ClpC1 N-terminal domain, but lacks AAA1 ATPase modules (22,67). As ClpC1 is itself involved in protein homeostasis and protein fate in general, it is not surprising to see its interaction with these proteins. Also, as noted above, components of the ribosome and other proteins involved in translation were abundant as well, making up ;13% of the total data set (Fig. 5). It is possible that nascent ClpC1 stays in contact with the ribosome during folding, either directly or indirectly through other elements of folding machinery. Alternatively, ribosomal proteins are commonly observed as contaminants in cell-based mass spectrometry experiments (68).
Taken together, we observe a functionally diverse set of cellular proteins that interact with M. smegmatis ClpC1, revealing a broad diversity of putative substrates and interaction partners of the ClpC1P1P2 protease. This buttresses just how far-reaching its regulatory effects on cellular proteins could be and may help account for the essentiality of these proteases in mycobacteria (19,(27)(28)(29)(30)(31).
Physiochemical analysis of termini of interacting proteins. Some Clp protease substrates are recognized by short terminal degron sequences of various lengths and  (13). M. tuberculosis ClpC1 recognizes Hsp20 via the C-terminal sequence TQAQRIAITK (21) and PanD via the C-terminal sequence NAGELLDPRLGVG (20). In addition, substrates displaying some hydrophobic N-terminal residues are delivered to ClpC1 by the adaptor ClpS as part of the N-end rule proteolytic pathway (23,69). Because the sequence determinants required for recognition by ClpC1 are not well defined, we compared terminal sequences of proteins pulled down by full-length ClpC1 (ClpC1 WT or ClpC1 EQ ) to the terminal sequences found across the entire M. smegmatis proteome (Fig. 6). The average charge and hydrophobicity of the first and last 10 amino acids (aa) of protein sequences varied widely. We saw no statistically significant differences in these parameters for residues at the N terminus. However, C-terminal regions of interactome proteins were, on average, less positively charged (closer to neutral) and more hydrophilic than equivalent regions from the full proteome. These differences were statistically significant, although the magnitude of the shift was much smaller than the standard deviation (SD) in either data set. The apparent preference for hydrophilic C-terminal character is surprising, as ATP-dependent proteases are often thought to recognize exposed hydrophobic regions as markers for protein misfolding (8). This analysis suggests either a slight preference for uncharged polar termini or that polar termini are more exposed to solvent and more available for ClpC1 binding. Additionally, the discrepancy between expected and observed physicochemical parameters might be explained by a preponderance of ClpC1interacting proteins that are not proteolytic substrates, and thus may not be subject to the same recognition trends as the substrates.
ClpC1P1P2 recognizes MSMEI_3879 as a proteolytic substrate. We selected several hits from our data set that (i) were expected to be soluble cytosolic proteins, (ii) had M. tuberculosis homologs identified in prior screens for Clp protease interaction partners (Table 3) (21,23), and (iii) could be readily expressed and purified from E. coli. These included chaperones DnaJ1 and DnaJ2, transcriptional regulators GntR and XRE, magnesium chelatase, and the apparent pseudogene product MSMEI_3879. Five of these (DnaJ1, GntR, XRE, Mg chelatase, MSMEI_3879) produced clear binding curves to ClpC1 by microscale thermophoresis, with K app values in the low micromolar range (Fig. S2), which validates the ability of our overall method to identify bona fide interaction partners. We tested whether these hits were recognized as substrates by ClpC1P1P2 in vitro and found that only MSMEI_3879 was degraded, with ;75% of the protein hydrolyzed within 40 min (Fig. 7A and B). These in vitro results provide preliminary evidence of the substrate status of MSMEI_3879.
The MSMEI_3879 locus encodes a 446-aa product with homology to ATP-hydrolyzing hydantoinase/oxoprolinase enzymes in the hydantoinase A family, which catalyze ringopening reactions on lactam substrates (70-72). Phylogenetic analysis (Fig. S3)  cluster of orthologs with .90% sequence identity to MSMEI_3879 in several related species, including Mycolicibacterium fortuitum, Mycolicibacterium peregrinum, and Mycolicibacterium brisbanense. More distantly related homologs occur across Mycobacteriaceae (73). The closest homolog in M. tuberculosis is OplA, which appears to belong to a separate subclass of enzymes, sharing only 34% sequence identity with MSMEI_3879 and incorporating a C-terminal fusion with a hydantoinase B enzyme.
Notably   (74) predictions of MSMEI_3879 and M. fortuitum AcxA, which share 93% sequence identity and possesses an intact N terminus, suggests that truncation removes a set of surface helices in MSMEI_3879 (Fig. S5). Loss of these helices is predicted to expose hydrophobic residues on two small structural elements that project outward from the body of the protein. In spite of this, MSMEI_3879 expressed well in E. coli, was straightforward to purify and remained stable throughout purification without precipitation or aggregation. Many Clp protease substrates are recognized by short terminal degron sequences (13,20,21,23). To test whether ClpC1 recognizes MSMEI_3879 by a particular terminus, we engineered green fluorescent protein (GFP) constructs with MSMEI_3879 fused to either the N or C terminus and assayed proteolysis by ClpC1P1P2. Michaelis-Menten analysis of the resulting degradation rates showed that the construct with MSMEI_3879 at the N terminus ( MSMEI_3879 GFP) is degraded with a k cat of %0.26 substrate Á min 21 Á enzyme 21 and a K m of %2 mM (Fig. 7C). For comparison, under the same assay conditions, GFP with an M. smegmatis ssrA tag (GFP ssrA ) was degraded faster (k cat % 0.4 substrate Á min 21 Á enzyme 21 ) but with a much higher K m of %16 mM (Fig. 7D), similar to the K m reported for degradation of GFP ssrA by M. tuberculosis ClpC1 (23). A construct carrying MSMEI_3879 at the C terminus (GFP MSMEI_3879 ) was degraded at a substantially lower rate (k cat % 0.09 substrate Á min 21 Á enzyme 21 ) and higher K m (;19 mM) than MSMEI_3879 GFP (Fig. 7E). The 9-fold lower K m of MSMEI_3879 GFP suggests that the N terminus of MSMEI_3879 contributes to efficient degradation by ClpC1, but is not the sole determinant of recognition. Neither GFP fusion was degraded by ClpXP1P2 (Fig. S1B), demonstrating that MSMEI_3879 recognition is specific to ClpC1. We also assessed whether sequences at the beginning or end of MSMEI_3879 function as simple degrons by adding its first 13 residues to the beginning of GFP or its final 13 residues to the end of GFP. Neither construct was degraded with as low a K m or high a k cat as MSMEI_3879 GFP. While we cannot rule out that longer sequences would be recognized more robustly, our results suggest that efficient recognition involves multivalent contacts on the folded MSMEI_3879 module, likely through the hydrophobic regions exposed by the truncation (Fig. S5). We tested this by assessing proteolysis of "full-length" MSMEI_3879, incorporating a restored N terminus created by removing the nucleotide insertion that causes a frameshift. As expected, this construct was not degraded by ClpC1P1P2 in vitro (Fig. S6), confirming that the truncation leads to recognition by ClpC1. Given the lower K m for degradation of MSMEI_3879 compared to GFP ssrA , we tested whether untagged MSMEI_3879 can compete with GFP ssrA degradation and found that equimolar MSMEI_3879 effectively blocks GFP ssrA proteolysis (Fig. 7F). Inhibition was a specific feature of MSMEI_3879, as DnaJ1, a non-substrate protein from our interactome data set (Fig. 7A), had no effect on degradation (Fig. 7F). Taken together, these data indicate MSMEI_3879 is a novel and robustly degraded ClpC1P1P2 substrate and that its N-terminal sequence contributes to its recognition.
Chemoinhibition of ClpC1 blocks MSMEI_3879 degradation. We sought to assess whether MSMEI_3879 constructs would have utility as reporters for small molecule disruption of ClpC1. We used two well-characterized mycobacterial ClpC1 dysregulators, ecumicin (ECU) and rufomycin (RUF), which bind the ClpC1 NTD and dysregulate the ability of ClpC1 to degrade protein substrates (75)(76)(77). We assayed proteolysis of MSMEI_3879 GFP and GFP ssrA by ClpC1P1P2 in vitro in the presence and absence of 10 mM of each compound. Compared to the untreated control, both ECU and RUF reduced the rate of MSMEI_3879 GFP degradation by about 60% (Fig. 8A). Inhibition of GFP ssrA degradation was also observed, but with different magnitude for each compound: ECU inhibited by ;80%, while RUF inhibited by ;45% (Fig. 8B). Taken together, these results demonstrate that known ClpC1 dysregulators inhibit proteolysis of MSMEI_3879 by ClpC1P1P2.

DISCUSSION
It is now well established that Clp protease components, including ClpC1, are essential for mycobacterial viability (19,(27)(28)(29)(30)(31), and these enzymes have consequently emerged as promising antibiotic targets. Here, we sought to better understand the physiological roles of ClpC1P1P2 by identifying partners that interact with full-length and truncated ClpC1 constructs. In total, we found 370 unique cellular proteins that interact with one or more ClpC1 construct. The diversity of the identified ClpC1 interactome is consistent with a multifaceted role for this enzyme in mycobacterial physiology. Moreover, our data align with recent explorations of the M. tuberculosis ClpC1 interactome/degradome (21,23), as 30 proteins found here possess orthologs identified in those screens. Several factors likely limit our ability to capture the full breadth of cellular interaction partners. Indeed, some known partners were absent from our data set, including substrates WhiB1 and CarD (30) and the proteolytic adaptor ClpS (23). Our approach was generally biased toward abundant and stably bound partners. Proteolytic substrates-which are actively degraded by the protease and interact with ClpC1 only while ATP is abundant-may be more challenging to detect due to dissociation during capture and wash steps. Clp proteases themselves are known to assemble dynamically and disassemble when substrate or ATP is exhausted (13), which likely occurs during sample preparation. Finally, overexpression presumably creates a stoichiometric excess of ClpC1 FLAG over ClpP1P2, which may bias capture toward certain classes of partners.
While ClpC1 likely participates in multiple proteolytic programs in mycobacteria, we note that not all interacting proteins identified here are substrates. While five of six candidates tested directly bind ClpC1 in vitro (see Fig. S2 in the supplemental material), only one, MSMEI_3879, proved to be a bona fide substrate (Fig. 7). Non-substrate interaction partners may play a role in regulating ClpC1 activity, may selectively interact with protease-deficient higher-order ClpC1 oligomers (25,78,79), or may simply bind nonspecifically under our capture conditions. It is also notable that capture experiments utilizing ClpC1 EQ , a construct carrying Walker B mutations expected to stabilize interactions with proteolytic substrates (56,(60)(61)(62), yielded fewer identified interaction partners (93 proteins) than experiments with ClpC1 WT (163 proteins) (Fig. 3, Table 1, and Table 2; see Table S2 in the supplemental material). This suggests that many of these interaction partners are not engaged as the substrates and instead adopt different modes of interaction that are dependent on nucleotide binding, ATP hydrolysis, and perhaps oligomeric state. A diverse set of dynamic interaction partners would plausibly allow mycobacterial cells to nimbly modulate ClpC1 activity across different growth conditions by altering the oligomeric state of ClpC1, changing substrate preferences, or tuning the stability of the entire Clp protease.
ClpC1 is composed of a globular N-terminal domain that binds to substrates and adaptors and an ATPase core consisting of two AAA1 modules that carry out chemomechanical substrate unfolding and translocation (8,10,(80)(81)(82). Surprisingly, our studies identified few interaction partners that bind exclusively to either the NTD or the ATPase core. This suggests that many ClpC1-interacting proteins make multivalent binding to both components. Alternatively, this could result from truncated constructs (ClpC1 NTD and ClpC1 CORE ) assembling with endogenous ClpC1 in the cell. Regardless of the mechanism involved, this illustrates that ClpC1 acts through the cooperation of its constituent parts.
Importantly, our interactome analysis led us to identify MSMEI_3879 as a bona fide ClpC1P1P2 substrate (Fig. 7). This protein joins a short list of verified mycobacterial Clp protease substrates (20,21,23,24). MSMEI_3879 is degraded relatively efficiently: a GFP construct carrying an N-terminal MSMEI_3879 fusion was degraded with a K m ;7-fold lower than that of the model substrate GFP ssrA (13,17). As expected for this lower K m , degradation of GFP ssrA was effectively blocked by an equimolar amount of MSMEI_3879 (Fig. 7F). Our experiments indicate that recognition of MSMEI_3879 by ClpC1 likely involves elements exposed as a result of its N-terminal truncation, due to a unique frameshift mutation ( Fig.  S4 and S5). Indeed, ClpC1 has been proposed to selectively recognize some substrates via hydrophobic regions and disordered termini (21). It is unclear whether MSMEI_3879 carries out any significant enzymatic function, or whether its proteolysis by ClpC1 plays any role in M. smegmatis physiology. Regardless, MSMEI_3879-based substrates may serve as useful tools for probing ClpC1P1P2 activity and dysregulation. As a proof of concept, we show that known ClpC1-targeting antimicrobials inhibit MSMEI_3879 GFP proteolysis in vitro (Fig. 8).
Similar substrates could serve as the basis for cell-based screens to identify novel ClpC1-targeting antibiotics.
In vivo substrate trapping. Plasmids encoding M. smegmatis ClpC1 constructs were electroporated into M. smegmatis (ATCC 700084/MC 2 155) at 5 kV in a MicroPulser (Bio-Rad). Liquid starter cultures were made in Middlebrook broth base (HiMedia) supplemented with 0.2% (vol/vol) glycerol (Fisher Scientific), 0.2% (wt/vol) glucose (TCI), and 0.05% (vol/vol) Tween 80 and then (with orbital shaking) grown for 60 h at 37°C. At a starting A 600 of 0.05, starter cultures were subcultured into 200 mL of fresh medium. The cultures were then grown at 37°C until the mid-log phase (A 600 % 0.6 to 1.0), at which point expression was induced by 28 mM « -caprolactam (Sigma-Aldrich) for 20 h. All constructs continued to grow similarly over the course of expression. For cell harvesting, centrifugation was carried out at 9,000 Â g for 20 min at 4°C and pellets were resuspended in 5 mL of lysis buffer (25 mM HEPES, 10 mM magnesium chloride, 200 mM potassium chloride, and 0.1 mM EDTA, supplemented with 10 mM ATP [pH 7.5]). Cell lysis was done in a microfluidizer (Microfluidics), and subsequent clarification of lysates was performed at 16,000 Â g for 30 min at 4°C. Cell supernatants were stored at 280°C. A Bradford assay (Bio-Rad) was performed to estimate total protein content.
Coimmunoprecipitation. Forty microliters of EZView Red anti-FLAG M2-affinity gel beads (Sigma-Aldrich) was equilibrated and washed twice in 0.5 mL lysis buffer by centrifugation at 8,200 Â g for 30 s. To pull down the expressed 3ÂFLAG-tagged M. smegmatis ClpC1 constructs and interacting proteins, the equilibrated beads were incubated with 1 mL of lysate for 1 h at 4°C with gentle agitation. After incubation, the bead-lysate slurry was spun at 8,200 Â g for 30 s and the bead pellet was subsequently washed with lysis buffer (containing 10 mM ATP). To elute, the washed beads were then mixed with 20 mL of 2Â Laemmli sample buffer (10% glycerol, 4% SDS, 167 mM Tris-HCl, 0.02% bromophenol blue) in the absence of reducing agent and boiled for 5 min. Samples were then vortexed briefly and centrifuged, and the supernatant containing the eluate was stored at 280°C prior to further analysis. As a negative control, lysates of cells containing empty pNIT vector were processed by the same workflow.
Mass spectrometry sample preparation. Proteins in the eluate were analyzed on a 6 to 15% SDS-PAGE gradient gel. Each lane was diced, and gel pieces were put into 1.5-mL microcentrifuge tubes. The samples were prepared by adapting standard protocols for in-gel mass spectrometry sample preparation (84). Incubation with 10 mM dithiothreitol (DTT) was carried out for 45 min at 55°C in order to reduce disulfide bonds. Afterwards, carbamidomethylation of cysteines was with 55 mM iodoacetamide for 30 min at room temperature (in the dark). Gel pieces were washed with gel wash buffer (25 mM ammonium bicarbonate and 50% acetonitrile) and dehydrated with 100% acetonitrile prior to drying in a SpeedVac vacuum centrifuge (Thermo). Trypsin digestion was carried out in 25 mM ammonium bicarbonate containing 10 mg mL 21 MS-grade trypsin protease (Pierce), and the mixture was incubated overnight at 37°C. Peptides were subsequently extracted by a two-step process, which was repeated twice: incubation of gel pieces with 5% formic acid (Sigma-Aldrich), followed by 100% acetonitrile. To remove residual salts, the peptide samples were desalted using a HyperSep C 18 column (Thermo Fisher Scientific) according to a previously described procedure (84,85). C 18 column-eluted samples were dried by vacuum centrifugation and frozen prior to mass spectrometry.
LC-MS/MS. Liquid chromatography tandem mass spectrometry (LC-MS/MS) analysis was performed using a Q Exactive Orbitrap interfaced with Ultimate 3000 Nano-LC system (Thermo Fisher Scientific). Trypsin-digested samples were loaded on an Acclaim PepMap rapid-separation liquid chromatography (RSLC) column (75-mm by 15-cm nanoViper) using an autosampler. Analysis of samples was done using a 150-min gradient running from 2% to 95% buffer B (0.1% formic acid in acetonitrile) in buffer A (0.1% formic acid in water) at a flow rate of 0.3 mL min 21 . MS data acquisition was done using a data-dependent top10 method, with the most abundant precursor ions from the survey scan chosen for higher-energy collisional dissociation (HCD) fragmentation using stepped normalized collision energies of 28, 30, and 35 eV. Survey scans were acquired at a resolution of 70,000 at m/z 200 on the Q Exactive. LS-MS/MS data were collected in independent biological triplicates.
MS data analysis. For proteomic analysis, extraction of raw data was performed in the Proteome Discoverer software suite (version 1.4; Thermo Fisher Scientific). The raw data were searched against Mycolicibacterium smegmatis (strain ATCC 700084/MC 2 155) UniProt Reference Proteome (Proteome ID UP000000757) using Sequest HT (University of Washington and Thermo Fisher Scientific). Iodoacetamide-mediated cysteine carbamidomethylation was set as a static modification. Precursor mass tolerance was set at 10 ppm, while allowing for fragment ion mass deviation of 0.6 Da for the HCD data, and full trypsinization with a maximum of two missed cleavages. Peptide-spectrum match (PSM) validation was done using Percolator, with false-discovery rates (FDRs) of 1% and 5% for stringent and relaxed validation, respectively. Gene Ontology (GO) annotation analysis on the data sets was performed using the Blast2GO software suite (64). Some GO annotation terms were binned into a finite number of final terms.
CRISPRi assays. Integrative plasmid PLJR962 (Addgene) (90), which encodes dCas9 machinery driven by an anhydrotetracycline (aTc)-responsive promoter for CRISPR interference (CRISPRi) in M. smegmatis, was modified by Gibson Assembly to contain either a nontargeting (GAGACGATTAATGCGTCTCG) or clpC1-targeting (ATGAGCGCGTCGTCGTCGCCGAA) small guide RNA (sgRNA). Versions of these plasmids were constructed by Gibson Assembly to contain a secondary clpC1 (or clpC1 FLAG , clpC1 EQ,FLAG ) locus downstream of the CRISPRi machinery. The plasmid-borne copy of clpC1 was modified to incorporate strategic codon substitutions to the sgRNA-targeted region that prevent sgRNA binding and thus escape transcriptional knockdown. Plasmids were transformed into M. smegmatis, and growth was monitored on Middlebrook agar plates at 37°C at various levels of aTc induction.
Data availability. The mass spectrometry data from this work have been submitted to the ProteomeXchange Consortium via the PRIDE partner repository (91) and assigned the identifier PXD030385.

ACKNOWLEDGMENTS
We gratefully thank R. Neunuebel and V. Parashar for the use of instrumentation. We thank Papa-Nii Asare-Okai for technical support. We also thank S. Cho and S. Franzblau of the Institute for Tuberculosis Research at the University of Illinois at Chicago for their gift of ecumicin and rufomycin. P.C.B. was supported by T32GM133395. K.R.S. was supported by NIH NIGMS award P20GM104316. The UD Department of Chemistry and Biochemistry Mass Spectrometry core was additionally supported by NIH NIGMS award P30GM110758-02. This content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.