Proximity‐dependent biotin identification links cholesterol catabolism with branched‐chain amino acid degradation in Mycobacterium smegmatis

Cholesterol is a crucial component in Mycobacterium tuberculosis virulence as it is required for phagocytosis of mycobacteria by macrophages. In addition, the tubercle bacilli can grow using cholesterol as the sole carbon source. Thus, cholesterol catabolism represents a valuable target for the development of new antitubercular drugs. However, the molecular partners of cholesterol catabolism remain elusive in mycobacteria. Here, we focused on HsaC and HsaD, enzymes involved in two consecutive steps of cholesterol ring degradation and identified putative partners, using a BirA‐based proximity‐dependent biotin identification (BioID) approach in Mycobacterium smegmatis. In rich medium, the fusion protein BirA‐HsaD was able to fish the endogenous cognate HsaC, thus validating this approach to study protein–protein interactions and to infer metabolic channeling of cholesterol ring degradation. In chemically defined medium, both HsaC and HsaD interacted with four proteins, BkdA, BkdB, BkdC, and MSMEG_1634. BkdA, BkdB, and BkdC are enzymes that participate in the degradation of branched‐chain amino acids. As cholesterol and branched‐chain amino acid catabolism both generate propionyl‐CoA, which is a toxic metabolite for mycobacteria, this interconnection suggests a compartmentalization to avoid dissemination of propionyl‐CoA into the mycobacterial cytosol. Moreover, the BioID approach allowed us to decipher the interactome of MSMEG_1634 and MSMEG_6518, two proteins of unknown function, which are proximal to the enzymes involved in cholesterol and branched‐chain amino acid catabolism. In conclusion, BioID is a powerful tool to characterize protein–protein interactions and to decipher the interconnections between different metabolic pathways, thereby facilitating the identification of new mycobacterial targets.


| INTRODUCTION
Mycobacterium tuberculosis (Mtb) is the etiological agent of tuberculosis (TB), which causes 1.6 million deaths per year. There is an alarming increase of TB cases caused by Mtb strains, which are (multi)resistant to available antibiotics. Therefore, there is an urgent need to identify new targets and to develop new anti-TB agents. Mtb is capable of infecting, surviving, and multiplying within the alveolar macrophages of the infected host, and cholesterol is required for the phagocytosis of mycobacteria by macrophages, 1 as mycobacteria enter phagocytes through cholesterol-enriched membrane microdomains. 2 Moreover, during the intramacrophagic stage, access to nutrients is very limited for mycobacteria. Host cholesterol is an essential carbon and energy source for mycobacterial growth, 3 and cholesterol metabolism is required to maintain optimal Mtb infection in animal models. [4][5][6][7] Thus, cholesterol catabolism plays a major role in both the establishment and the development of Mtb infection.
Cholesterol is composed of a ring and a side chain, whose degradation is ensured by a distinct set of enzymes (for review). [8][9][10][11] Furthermore, cholesterol degradation generates potentially toxic metabolic intermediates for mycobacteria, such as propionyl-CoA, which requires a very tight regulation of this degradation pathway. 12 Recently, several studies have shown that chemical compounds, able to block or perturb cholesterol utilization, inhibit Mtb growth. [13][14][15] Thus, cholesterol catabolism may represent a valuable target for the development of new antitubercular drugs.
We have recently shown that proximity-dependent biotin identification (BioID) is a powerful tool to detect and characterize protein-protein interactions (PPI) in mycobacteria by defining the heparin-binding hemagglutinin HbhA interactome. 16 Briefly, after generating a fusion protein between the protein of interest and the Escherichia coli biotin ligase (BirA), proximity labelling allows the identification of proteins that strongly, weakly, or transiently interact or are closely associated (within 10-20 nm) with the fusion protein, 17 allowing to define its proxisome (or proximal interactome). 18 In addition, this technology can help to identify multi-partner complexes directly in the organism of interest, preserving the intact subcellular structures, the presence of enzyme cofactors and posttranslational modifications (for review). [19][20][21][22] In this study, we used the BioID approach in saprophytic Mycobacterium smegmatis mc 2 155 in order to characterize the proxisome of two consecutive enzymes, HsaC (MSMEG_6036) and HsaD (MSMEG_6037). In Mtb, HsaC (Rv3568c) and HsaD (Rv3569c) are required for growth on cholesterol, 23 as they are both involved in cholesterol ring catabolism, 8,10 and HsaD was shown to be required for mycobacterial survival within macrophages. 24 In addition, it was shown that in the absence of HsaC, mycobacterial survival is impaired in mice and guinea pigs. 7 We show here that in rich medium, such as lysogeny broth (LB), the fusion protein BirA-HsaD is able to interact with endogenous HsaC in M. smegmatis, thus validating the BioID approach. In a chemically defined medium, such as 7H9 supplemented with glycerol, both HsaC and HsaD were closely associated with BkdA, BkdB, and BkdC, which are involved in branched-chain amino acid (BCAA) degradation, such as leucine, isoleucine, and valine. In addition, MSMEG_1634 interacted with both HsaC and HsaD. By defining the MSMEG_1634 proxisome, encompassing proteins involved in BCAA degradation and MSMEG_6518, the BioID technology allowed us to characterize mycobacterial proteins of unknown function. Thus, BioID proves to be useful for the characterization of metabolic pathways, leading to the identification of an interconnection between cholesterol ring catabolism and BCAA degradation when M. smegmatis is grown in 7H9.

| Bacterial strains and culture conditions
Escherichia coli TOP10 (Invitrogen) used for the cloning steps was cultured in LB medium (MP Biomedicals) at 37°C supplemented with ampicillin (100 μg/mL) or kanamycin (25 μg/mL), when required. M. smegmatis mc 2 155 and the recombinant strains constructed in this study (Table S1) were grown in Middlebrook 7H9 broth containing glycerol (without OADC) or on 7H11 agar plates supplemented with glycerol and OADC enrichment (Becton Dickinson), containing kanamycin (25 μg/mL), when required.
To construct the vectors for production of HsaC with a N-terminal 6His-tag, the gene was PCR amplified from M. smegmatis mc 2 155 genomic DNA using primers pM-V361_6His_hsaC_dir and pMV361_6His_hsaC_rev (Table S2). The corresponding amplicon was digested by HindIII and inserted into pMV361 digested by PvuII and HindIII to yield pMV361_6His_hsaC.
To generate the vectors expressing the fusion between MSMEG_1634 and birA, the gene was PCR amplified from M. smegmatis mc 2 155 genomic DNA using primers pcDNA_birA_MS1634_dir and pcDNA_bi-rA_MS1634_rev, or pcDNA_MS1634_birA_dir and pcD-NA_MS1634_birA_rev (Table S2). For N-terminal fusions with BirA, the MSMEG_1634 amplicon was digested by HindIII and EcoRV and inserted into pcDNA3.1 mycBi-oID (Addgene) 25 digested with the same enzymes to yield pcDNA3.1_birA_MSMEG_1634. For C-terminal fusions with BirA, the MSMEG_1634 amplicon was digested by EcoRI and KspAI and inserted into pcDNA3.1 MCS-BirA(R118G)-HA (Addgene) 25 digested with the same enzymes to yield pcDNA3.1_MSMEG_1634_birA. The birA_MSMEG_1634 fragment was amplified from pcD-NA3.1_birA_MSMEG_1634 using primers pMV361_birA_ XXX_dir and pcDNA_birA_MS1634_rev (Table S2), digested by HindIII and inserted into pMV361, previously digested by HindIII and PvuII, to yield pMV361_birA_MS-MEG_1634. The MSMEG_1634_birA fragment was obtained after digestion of pcDNA3.1_MSMEG_1634_birA by BstBI and PmeI and inserted into pMV361, digested by BstBI and KspAI, to yield pMV361_MSMEG_1634_birA. All the constructions were checked by DNA sequencing.

| BioID protocol with M. smegmatis cultures
A total quantity of 50 mL of medium were inoculated at OD 600nm = 0.025 from 3 days saturated cultures of M. smegmatis and grown overnight (16 h) at 37°C under agitation. To avoid any depletion during exponential phase of growth, biotin was added at a final concentration of 200 μM and cultures were grown for additional 24 h. After 40 h of growth, cultures were centrifuged at 4000 g for 10 min and then rinsed twice with phosphate-buffered saline (PBS). Each pellet was resuspended in 1 mL of PBS per 300 mg of bacteria. A total quantity of 1 mL of each culture was transferred into Lysing Matrix B tubes (MP Biomedicals) and the tubes were subjected to FastPrep (MP Biomedicals) treatment (6 cycles at maximum speed for 1 min, 1 min of incubation on ice). After quick spin, 700 μL of lysates was transferred to new Eppendorf tubes and centrifuged at 16 500 g for 1 h at 4°C. Dynabeads MyOne Streptavidin C1 (ThermoFisher Scientific) were rinsed three times with PBS, and 100 μL of beads was incubated at 4°C with each cleared lysate using a rotating wheel. After overnight incubation, beads were washed three times with PBS containing 0.1% bovine serum albumin and two times with PBS (to decrease the bovine serum albumin signal during mass spectrometry analysis). Finally, enriched biotinylated proteins were eluted using Laemmli buffer.

| In vivo pulldown in M. smegmatis
Mycobacterium smegmatis containing empty pMV361 or producing HsaC with a N-terminal 6His-tagged was grown in 100 mL of 7H9 until OD600 reaches ~1. Cultures were split in two, centrifuged at 3000 g, and rinsed twice with PBS containing 0.05% Tween 80. The pellets were resuspended in 1 mL of PBS, transferred into Lysing Matrix B tubes (MP Biomedicals), and lysed using a FastPrep (MP Biomedicals) treatment (6 cycles at maximum speed for 1 min, 1 min of incubation on ice). Lysates were harvested and SDS was added to a final concentration of 1%. The samples were further incubated for 30 min at room temperature under gentle rotation. Cell debris were eliminated by centrifugation at 5000 g for 15 min at 4°C. The supernatants were then incubated with 16 μL of PureCube INDIGO Ni-MagBeads (Cube Biotech, dilution 1:4) for 1 h at room temperature under gentle rotation. After incubation, the beads were rinsed two times with PBS, two times with PBS containing 50 mM imidazole, and the proteins were finally eluted with 100 μL PBS containing 300 mM imidazole.

| Western blot analysis
Ten microliters of total protein extract was resolved by SDS-PAGE using 12% acrylamide gels and then electrotransferred onto a nitrocellulose membrane. The membrane was saturated with Tris-buffered saline (TBS), pH 7.5, containing 0.05% Tween 80 and 5% milk and probed overnight at 4°C with anti-HA monoclonal antibody (Sigma-Aldrich, dilution 1:1000) or with anti-Myc monoclonal antibody (Thermo Fisher Scientific, dilution 1:1000) in TBS-Tween-3% milk. Finally, the membrane was incubated for 1 h at room temperature with goat anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibodies (Abcam, diluted 1:5000) in TBS-Tween-3% milk. Alternatively, the biotinylated proteins were directly probed overnight with HRP-conjugated streptavidin (BD Pharmingen, dilution 1:1000). The blots were developed using the Amersham ECL Prime Western-Blotting Detection Reagent (GE Healthcare) and chemiluminescence was detected using the Amersham Imager 600 (GE Healthcare).

| Mass spectrometry proteomic analysis
Raw data collected during nanoLC-MS/MS analyses were processed and converted into a *.mgf peak list format with Proteome Discoverer 1.4 (Thermo Fisher Scientific). MS/ MS data were analyzed using search engine Mascot (version 2.4.0, Matrix Science) installed on a local server. Searches were performed with a tolerance on mass measurement of 10 ppm for precursor and 0.02 Da for fragment ions, against a composite target-decoy database built with a M. smegmatis mc 2 155 UniProt database (strain ATCC 700084/mc(2)155, taxo 246196, July 20172021, 12666 entries) fused with recombinant trypsin and a list of classical contaminants (118 entries). Cysteine carbamidomethylation, methionine oxidation, protein N-terminal acetylation, and cysteine propionamidation were searched as variable modifications. Up to one missed trypsin cleavage was allowed. The identification results were imported into Proline 2.0 software installed on a local server (http://proli ne.profi prote omics.fr) for validation. 26 Peptide spectrum matches taller than nine residues and ion scores >10 were retained. The false discovery rate was then optimized to be below 2% at the protein level using the Mascot Modified Mudpit score. Spectral counting analyses were performed with Proline.
For BioID analyses, we introduced a corrective factor in order to compare the results, since the total number of spectra may vary between independent biological replicates. For each condition, the number of spectra for one protein was normalized against Replicate 1 and was calculated as follows: Normalized number of spectra (Protein X) = number of spectra (Protein X) in replicate 1 + [number of spectra (Protein X) in replicate 2 × total number of spectra in replicate 1/total number of spectra in replicate 2] + [number of spectra (Protein X) in replicate 3 × total number of spectra in replicate 1/total number of spectra in replicate 3]. Furthermore, as the expression of birA alone may result in nonspecific biotinylation of mycobacterial proteins, 16 the ratio was calculated as follows: Ratio (protein X) = normalized number of spectra with HsaC, HsaD, or MSMEG_1634 in fusion with BirA/normalized number of spectra with BirA alone. For calculation purpose, when the Normalized number of spectra with BirA alone for Protein X was zero, it was arbitrarily set to 0.5 (between 0 and 1 spectrum) in order to calculate a ratio. Only proteins with a normalized number of spectra ≥25 and a ratio >10 were considered as meaningful, implying that Protein X is enriched in that specific condition in comparison to the control (BirA alone). As the heterologous expression of hsaC and hsaD alone could also potentially lead to nonspecific biotinylation of proteins, the ratio of HsaC or HsaD fused to BirA/BirA should be at least two times superior to the ratio HsaC or HsaD alone/BirA.
For pulldown experiments, the ratio was determined as the number of spectra for Protein X in the strain producing recombinant 6His-HsaC divided by the number of spectra for Protein X in the strain containing empty pMV361. When the number of spectra for Protein X was zero in case of M. smegmatis containing empty pMV361, it was arbitrarily set to 0.5 (between 0 and 1 spectrum) in order to calculate a ratio.

| Production of HsaD fusion proteins in M. smegmatis in LB medium
HsaD (MSMEG_6037) from M. smegmatis mc 2 155 was fused to the N-terminal or the C-terminal end of BirA(R118G) in order to catalyze protein biotinylation in a proximity-dependent fashion. 25 The corresponding coding sequences were inserted into the mycobacterial vector pMV361 under the control of the strong hsp60 promoter. All constructions, as well as pMV361_birA and pMV361_hsaD, were introduced into electro-competent cells of M. smegmatis mc 2 155. Western blot analysis using anti-Myc (when BirA was fused to the N-terminal end of HsaD) or anti-HA (for unfused BirA or when BirA was fused to the C-terminal end of HsaD) antibodies confirmed that the fusion proteins were produced in M. smegmatis when grown in LB medium ( Figure 1A).
For BioID analyses, each M. smegmatis culture was grown overnight (16 h) in LB medium, and 200 μM of biotin was added to the culture medium to avoid any biotin depletion during exponential growth. The cultures were then grown for additional 24 h. At 16 and 40 h, the growth of M. smegmatis was not affected by the production of recombinant HsaD or the fusion proteins with BirA, as the OD was similar for all strains of M. smegmatis ( Figure 1B).
The cultures were then subjected to bacterial lysis and the biotinylated proteins were purified using magnetic beads coated with streptavidine. After releasing from the beads, 10 μL of each sample was analyzed by western blotting, and biotinylated proteins were detected by using HRP-conjugated streptavidin ( Figure 1C). Monomers and dimers of bovine serum albumin, which was used to saturate streptavidine free sites on beads, were detected as 66-and 132-kDa proteins, respectively, in all the samples. Additional biotinylated proteins (shown by the asterisks) were detected in M. smegmatis producing the HsaD-fused proteins, but absent with BirA alone, suggesting specific biotinylation of proteins closely associated with HsaD.

| Identification of HsaD interactants by BioID in LB medium
To identify proteins that interact with HsaD, each sample of biotinylated proteins from three independent biological replicates was analyzed by mass spectrometry (Table S3). From our previous study 16 and as confirmed by western blot analyses ( Figure 1C), overexpression of birA alone may result in nonspecific biotinylation of mycobacterial proteins. Therefore, candidate proteins, as quantified using spectral counting, were selected by comparing their enrichment in M. smegmatis producing the bait of interest, BirA-HsaD or HsaD-BirA, over M. smegmatis producing BirA alone. Only a normalized number of spectra equal or above 25 was considered as representative, and only proteins that displayed a ratio superior to 10 of enrichment in the presence of BirA-HsaD or HsaD-BirA compared to BirA alone were selected. Moreover, as recombinant expression of hsaD alone could lead to a nonspecific increase in the amounts of biotinylated proteins, we only considered proteins for which the BirA-HsaD/BirA or HsaD-BirA/BirA ratios were at least two times superior to the HsaD alone/BirA ratio.
These stringent criteria led to the identification of proteins closely associated with HsaD (Table 1). When BirA was fused to the C-terminal end of HsaD, very few interactants were isolated, inferring that a free C-terminal end is important to maintain PPI. However, the HsaD-BirA fusion was still able to interact with HsaD itself (99% identical to BphD), inferring in vivo multimerization of the protein or auto-biotinylation of the HsaD-BirA fusion. When BirA was fused to the N-terminal end of HsaD (BirA-HsaD), two interactants were identified, HsaD itself and HsaC, confirming that HsaC and HsaD interact with each other in vivo and demonstrating that the BioID approach is a robust tool to study PPI.

| Production of HsaC and HsaD fusion proteins in M. smegmatis in 7H9 medium
In order to compare different growth conditions, we also searched for HsaD interactants when M. smegmatis was grown in the chemically defined 7H9 medium, supplemented with glycerol (and without OADC). As HsaC was detected as a HsaD interactant, it was added to the study in order to compare its proxisome with that of HsaD. Similar to HsaD, HsaC (MSMEG_6036) was fused to the N-terminal or the C-terminal end of BirA(R118G) and the corresponding coding sequences were inserted into pMV361. All constructs were introduced into electrocompetent cells of M. smegmatis mc 2 155. Western blot analysis using anti-Myc (when BirA is fused to the Nterminal end of HsaC or HsaD) and anti-HA (when BirA is fused to the C-terminal end of HsaC or HsaD) antibodies confirmed that the fusion proteins were produced in M. smegmatis when grown in 7H9 medium supplemented with glycerol ( Figure 2A).
Each M. smegmatis culture was grown overnight (16 h) to reach the beginning of exponential phase, and 200 μM of biotin was added to the culture medium to avoid any biotin depletion during exponential growth. The cultures were then grown for an additional 24 h, until the beginning of stationary phase. The OD values were similar at 16 and 40 h for the M. smegmatis strains that produced the recombinant HsaC, HsaD, or the fusion proteins ( Figure 2B), indicating that the overexpression of the recombinant proteins did not affect mycobacterial growth.
The cultures were then subjected to bacterial lysis and the biotinylated proteins were purified using magnetic beads coated with streptavidine. After releasing from the beads, 10 μL of each sample was analyzed by western blotting, and biotinylated proteins were detected by using HRP-conjugated streptavidin ( Figure 2C). The profiles of recombinant M. smegmatis containing pMV361_hsaC and pMV361_hsaD were similar to each other and few biotinylated proteins were detected. However, like shown above for LB, further biotinylated proteins (shown by the asterisks) were detected in M. smegmatis producing the HsaC-or HsaD-fusion proteins, but were absent with BirA alone, suggesting specific biotinylation of proteins closely associated with HsaC or HsaD.

| Identification of HsaC and HsaD partners by BioID in 7H9 medium
To identify the proteins that interact with HsaC and HsaD, each sample of biotinylated proteins from three independent biological replicates was analyzed by mass spectrometry (Tables S4 and S5). Protein abundance was evaluated using spectral counting and candidate proteins were selected by comparing their enrichment in M. smegmatis producing the proteins of interest (BirA-HsaC, BirA-HsaD, HsaC-BirA, or HsaD-BirA) over M. smegmatis overexpressing birA alone. When BirA was fused to the C-terminal end of HsaC or HsaD, very few interactants were identified (Tables S4 and S5), suggesting that a free C-terminal end is required to maintain PPI for both proteins. The HsaC-BirA and HsaD-BirA fusions were able to interact with themselves, reflecting their probable ability to multimerize in vivo or to auto-biotinylate.
Interestingly, four proteins (MSMEG_4712, MSMEG_4711, MSMEG_4710, and MSMEG_1684) were common to both the HsaC and HsaD proxisomes, highlighting a potential interconnection between these proteins. MSMEG_4712, MSMEG_4711, and MSMEG_4710 are encoded by bkdA, bkdB, and bkdC, respectively, which form an operon that is conserved in most mycobacteria, including M. smegmatis and Mtb. 27 The corresponding enzymes, BkdA, BkdB, and BkdC, belong to the branched-chain keto acid dehydrogenase (BCKADH) complex, required for Mtb virulence and involved in the catabolism of BCAA. 28 The additional protein, MSMEG_1684, is annotated as probable forkhead-associated protein of unknown function.

| BioID approach to characterize proteins of unknown function such as MSMEG_1634
The use of BioID may be helpful to investigate proteins with unknown function by characterizing their close protein environment. For this purpose, BirA-MSMEG_1634 and MSMEG_1634-BirA were produced in M. smegmatis ( Figure 3A). Their production only slightly affected M. smegmatis growth at 16 and 40 h ( Figure 3B). After purification using streptavidin beads, 10 μL of biotinylated protein solution was analyzed by western blotting and detected using HRP-conjugated streptavidin ( Figure 3C). Several biotinylated proteins (shown by the asterisks) were detected in M. smegmatis producing the MSMEG_1634-fusion proteins, and were absent with BirA alone, suggesting specific biotinylation of proteins closely associated with MSMEG_1634.
Protein abundance was then evaluated using spectral counting, and candidate proteins were selected by comparing their enrichment in M. smegmatis producing the fusion proteins of interest (BirA-MSMEG_1634 and MSMEG_1634-BirA) over M. smegmatis overexpressing birA alone (Table S6). Only two interactants, MSMEG_1634 and MSMEG_6518, were identified when MSMEG_1634 was fused to the N-terminal part of BirA (Table 4), highlighting again the importance of a free C-terminal end to maintain PPI. Interestingly, MSMEG_6518 displays 56% of identity with MSMEG_1634 and both proteins display a very similar structure as predicted by Alphafold 29,30 ( Figure S1). Thus, BioID is able to detect homologous proteins, which may reflect the ability of MSMEG_1634 to form in vivo heteromultimers with MSMEG_6518. Moreover, MSMEG_6518 was also identified as an interactant of BirA-HsaC (Table 2).
For BirA-MSMEG_1634, eight interactants were identified, MSMEG_1634 itself, BkdA/MSMEG_4712, BkdB/MSMEG_4711, MSMEG_6518, MSMEI_1212 (corresponding to MSMEG_1247), HrpA/MSMEG_6587, MSMEG_3726, and BkdC/MSMEG_4710 (Table 4). Interestingly, MSMEG_3726, a putative alcohol dehydrogenase, was previously described as being upregulated in the presence of cholesterol. 11 As shown above, five out of the eight BirA-MSMEG_1634 interactants (BkdA, BkdB, BkdC, MSMEG_1634, and MSMEG_6518) were common to the proxisome of either HsaC or HsaD. T A B L E 1 List of Mycobacterium smegmatis proteins interacting or closely associated with HsaD-BirA or BirA-HsaD and displaying a ratio >10, when bacteria were grown in LB medium. Surprisingly, all the interactants identified with MSMEG_1634-BirA or BirA-MSMEG_1634 were also enriched (ratio ≥7.8) when MSMEG_1634 was produced alone (Table S6). These proteins were not overrepresented when HsaC or HsaD were produced alone (Table S4 and S5), nor in our previous study with HbhA. 16 Moreover, BkdA, BkdB, BkdC, HrpA, and MSMEG_3726 were not overrepresented in the M. smegmatis strain overproducing MSMEG_1634-BirA (Table S6). These results discard any nonspecific binding to the streptavidin beads or any natural biotinylation of these proteins. Hence, the overrepresentation of these interactants appears to be driven by MSMEG_1634 overexpression alone.

M. smegmatis producing 6His-HsaC
The BioID approach allows the detection of PPI in 10-to 20-nm range, 17 without necessarily a direct interaction. Nevertheless, we developed an in vivo pulldown assay, which requires a direct interaction between protein partners, in order to confirm the PPI isolated with the BioID approach. For this purpose, we generated a M. smegmatis strain containing empty pMV361 or producing HsaC with a N-terminal 6His-tag, as only HsaC with a Nterminal BirA was prone to interact with other proteins ( Table 2). After purification on nickel beads, we analyzed the protein content by mass spectrometry, and their abundance was compared to that present in M. smegmatis carrying the empty vector (Table S7). Endogenous HsaC or MSMEG_1634 were not able to bind nickel beads (with 1 and 0 spectrum, respectively) in M. smegmatis containing the empty vector. As expected, using two independent biological replicates, the recombinant 6His-HsaC was the most enriched protein with a ratio of 360 in M. smegmatis producing 6His-HsaC over M. smegmatis containing the empty vector. In the same strain, endogenous MSMEG_1634 was isolated among the top 15 most enriched proteins with a ratio of 52, inferring that HsaC and MSMEG_1634 may be in direct contact with each other in M. smegmatis grown in 7H9 and confirming our results obtained with the BioID approach.

| DISCUSSION
The BioID technology is a new and powerful tool helpful for the characterization of the proxisome of proteins of interest. In this study, we focused on the two enzymes, HsaC and HsaD, which are involved in two consecutive steps of ring cholesterol catabolism in M. smegmatis 11 and Mtb. [8][9][10] Cholesterol catabolism in mycobacteria is a subject of growing interest, as it is a key process for mycobacterial pathogenesis, 3 thereby representing a promising therapeutic target. Using the mycobacterial surrogate strain M. smegmatis mc 2 155, we show here that HsaC and HsaD are able to interact in vivo with each other when the bacilli are grown in LB medium, demonstrating the validity of the BioID approach. This interaction between HsaC and HsaD probably reflects the metabolic channeling of the cholesterol ring degradation. However, no interaction between HsaC and HsaD was detected when a medium specific for mycobacterial growth, such as 7H9 supplemented with glycerol (but no OADC), was used. These results highlight the importance of the medium when carrying out experiments to characterize the proxisome of mycobacterial proteins and suggest that the use of the chemically defined 7H9 medium with glycerol as the sole source of carbon may favor the activation of specific metabolic pathways. This is consistent with other findings showing that the growth of wild-type M. smegmatis and a MSMEG_3811 deletion mutant were reversed when grown in a complete (LB) compared to a defined medium with a single carbon source. 31 Here, the use of complete medium containing multiple carbon sources, such as LB, is likely to activate many catabolic pathways, as shown by the differential proxisomes detected in the two growth conditions. It would have been of interest to use a cholesterol-containing medium as sole carbon source for further characterization of the HsaC and HsaD proxisomes. Thus, the lack of a direct comparison of BioID analyses performed in 7H9 supplemented with glycerol with those performed in 7H9 supplemented with cholesterol, due to the severe growth defect in the latter medium (data not shown), constitutes a limitation of our study. However, this subject may be part of future work. When M. smegmatis was grown in 7H9 supplemented with glycerol, HsaC and HsaD were both strongly interconnected with BkdA, BkdB, and BkdC, which are involved in the catabolism of BCAA. This interconnection was observed only in 7H9 medium and not in LB medium. Since 7H9 contains large amounts of L-glutamate (0.5 g/L), which is used by IlvE to catalyze the final step of leucine, isoleucine, and valine biosynthesis, 32 degradation of excessive BCAA may be a way for the bacilli to feed its central carbon metabolism by increasing the presence of BkdA, BkdB, and BkdC, when grown in 7H9. As cholesterol and BCAA degradation both lead to the generation of propionyl-CoA, which is highly toxic for mycobacteria, this interconnection might help to sequester propionyl-CoA into specialized compartments. Moreover, the interplay between pathways, such as the methylcitrate cycle (MCC), the methylmalonyl pathway (MMP), and the incorporation into methyl-branched cell wall lipids, involved in propionyl-CoA detoxification, has been suggested previously. 12 Hence, such a compartmentalization may avoid dissemination of toxic metabolites within the cytosol of mycobacteria and would allow the metabolic channeling of propionyl-CoA directly from upstream pathways, such as cholesterol and BCAA degradation, to downstream pathways, such as MCC, MMP, and the incorporation into the methyl-branched cell wall lipids (Figure 4).
There is additional evidence indicating that mycobacteria display compartmentalization of their metabolic pathways. For example, the lipid biosynthetic enzymes involved in phosphatidylinositol mannoside biosynthesis were suggested as being compartmentalized at the mycobacterial plasma membrane in M. smegmatis. 33 More recently, the compartmentalization of peptidoglycan precursors was also demonstrated in M. smegmatis, and this organization may be applicable to rod-shaped bacteria that are phylogenetically distant. 34 This concept has now been extended to the biosynthesis of several cell wall components of Mtb. 35 Metabolic compartmentalization is not restricted to lipid biosynthesis, as it was also suggested for the co-catabolism of carbon substrates, and the formation of multi-protein complexes in mycobacteria may occur in T A B L E 4 List of Mycobacterium smegmatis proteins interacting or closely associated with MSMEG_1634-BirA or BirA-MSMEG_1634 and displaying a ratio >10, when bacteria were grown in 7H9 medium supplemented with glycerol. response to the nutritional environment and the physiological needs. 36 Hence, in order to identify these multiprotein complexes, the use of BioID may be a particularly interesting approach. In our previous BioID study, we identified an interaction between HsaC and HbhA, when HbhA-BirA was overproduced in M. smegmatis. 16 However, we were not able to detect the reverse interaction using HsaC fused to BirA. This may be due to the low abundance of endogenous HbhA in M. smegmatis. 37,38 In agreement with that hypothesis, the orthologous HbhA from M. smegmatis (MSMEG_0919) was not detected by mass spectrometry in any of the samples. Moreover, the growth conditions (7H9 or LB) may not be optimal for production of endogenous HbhA and for the detection of the HsaC-HbhA interaction. Alternatively, as BirA covalently binds biotin to lysine residues of the proximal protein, 39 and as HbhA from M. smegmatis is methylated on most lysine residues, 37 these posttranslational modifications may have decreased the efficacy of biotinylation by BirA.

Construct
Finally, the BioID approach may help into characterize bacterial proteins of unknown function. We show here that MSMEG_1634 is interconnected with cholesterol and BCAA degradation, and is closely associated with the homologous MSMEG_6518 protein. The function of MSMEG_1634 is not assigned and blast analyses do not provide any further information. However, the protein contains a probable forkhead-associated domain, which is a signaling motif that mediates PPI and modulates a wide variety of biological processes in many organisms, including mycobacteria. 40 Thus, we hypothesize that MSMEG_1634 and MSMEG_6518 act as regulators of cholesterol and/ or BCAA catabolism, possibly by playing a structural role at the interface of the complexes formed by HsaC, HsaD, BkdA, BkdB, and BkdC ( Figure 4). A similar situation has been described for the mycobacterial forkhead-associated domain-containing protein GarA (Rv1827), which was shown to interact and to regulate several enzymes of the tricarboxylic acid cycle. 41 Since propionyl-CoA and its metabolites are highly toxic for mycobacteria, destabilizing the formation of Hsa/Bkd complexes would constitute a potential antimycobacterial target, as it would lead to their dissociation and drive diffusion of propionyl-CoA into the cytosol, thereby causing self-poisoning of the mycobacteria. Moreover, identifying interconnected pathways may help to design synergizing drugs. Two mycobacterial enzymes have been shown to be lipoylated in Mtb, DlaT, and BkdC/MSMEG_4710. 28 Lipoic acid and biotin display very similar structures and properties (for review). 42 Therefore, lipoylation may increase nonspecific binding of the M. smegmatis DlaT analog (also known as SucB/MSMEG_4283) and BkdC/MSMEG_4710 to streptavidin beads. When MSMEG_1634 alone was overproduced, it interacted with BkdA, BkdB, and BkdC, leading to enhanced binding to the streptavidin beads and potentially creating a snowball effect in which multimers of MSMEG_1634 play a bridging role between the different BCKADH complexes. Moreover, the mycobacterial enzyme Lpd interacts with both DlaT and the BCKADH complex in Mtb, 28 which could amplify this phenomenon in M. smegmatis. Consistent with this hypothesis, all the components of the αketoglutarate dehydrogenase complex (DlaT/SucB/MSMEG_4283, Kgd/SucA/MSMEG_5049, and LpdA/MSMEG_0903) were slightly overrepresented (ratios between 2.8 and 4.4) in M. smegmatis overexpressing MSMEG_1634 or birA_MSMEG_1634 (Table S6), but not in the strain overexpressing MSMEG_1634_birA (ratios between 1.0 and 1.1). As a free C-terminal end of MSMEG_1634 appears to be required to maintain PPI, the MSMEG_1634-BirA construct would no longer be able to play a bridging role, thereby precluding the snowball effect.
The bioinformatics tool STRING (https://strin g-db.org/) allows to deciphering the functional F I G U R E 4 Proposed global scheme of the complex formed by HsaC, HsaD, BkdA, BkdB, BkdC, MSMEG_1634, and MSMEG_6518 in Mycobacterium smegmatis. HsaC and HsaD, involved in the cholesterol ring catabolism, are able to multimerize and interact with each other in vivo. They are closely associated with BkdA, BkdB, and BkdC, which are involved in the catabolism of BCAA (leucine, isoleucine, and valine). We propose that MSMEG_1634 and MSMEG_6518 act as regulators or pivotal links between cholesterol and BCAA degradation complexes. The compartmentalization of these pathways would avoid diffusion of toxic propionyl-CoA into the mycobacterial cytosol and may allow metabolic channeling of propionyl-CoA directly from upstream pathways (cholesterol and BCAA catabolism) to downstream pathways (MCC, MMP, and incorporation into the methylbranched (MB) cell wall lipids).