A periplasmic protein complex mediates arabinofuranosyltransferase activity and intrinsic drug resistance in Mycobacterium tuberculosis

Abstract

The intrinsic drug resistance of Mycobacterium tuberculosis (Mtb) is a major barrier to effective tuberculosis (TB) treatment and is largely due to its complex, impermeable cell envelope. We identified a periplasmic protein complex comprising FecB and Rv3035 that is essential for maintaining envelope integrity and mediating intrinsic multidrug resistance in Mtb. FecB interacts with Rv3035, forming a stable heterodimer that associates with the cell envelope biosynthesis protein AftB. We report the structures of Rv3035 alone and in complex with FecB and identify critical residues for complex formation and function. Coessentiality and genetic interaction analyses support a functional link between FecB, Rv3035, and AftB, an arabinofuranosyltransferase that synthesizes arabinogalactan and lipoarabinomannan. Loss of FecB or Rv3035 disrupted AftB-mediated arabinan synthesis, suggesting that these proteins support AftB’s enzymatic activity. FecB is required for Mtb virulence in mice, underscoring its physiological relevance. These findings highlight FecB, Rv3035, and AftB as promising therapeutic targets.

Periplasmic mediators of cell envelope integrity cause intrinsic drug resistance in Mycobacterium tuberculosis.

INTRODUCTION

Tuberculosis (TB) remains the world’s leading cause of death from a single infectious agent and is responsible for over 1 million deaths annually (1). Unlike most bacterial infections, successful TB treatment requires multiple antibiotics taken for several months, even in drug-susceptible cases. This prolonged therapy is in part due to the intrinsic drug resistance of Mycobacterium tuberculosis (Mtb), the causative agent of TB (2). A key contributor to this resistance is Mtb’s complex, multilayered cell envelope, which acts as a protective barrier against antibiotics and host immune defenses (3). Despite its central role in Mtb’s pathogenicity and drug resistance, the molecular mechanisms governing the structure, assembly, and function of the mycobacterial cell envelope remain incompletely understood (4).

To identify novel targets that could potentiate existing TB therapies and shorten treatment duration, we previously performed a genome-wide chemical-genetic screen using transposon mutagenesis and sequencing (5). This identified fecB (rv3044) as a gene whose disruption sensitized Mtb to multiple antibiotics with various mechanisms of action. These included rifampicin and ethambutol, which are part of the standard treatment regimen for drug-susceptible TB, and pretomanid, which is part of the BPaL (bedaquiline, pretomanid, and linezolid) regimen used to treat drug-resistant TB (5). Functional studies of a Mtb fecB knockout strain (ΔfecB) revealed that fecB is essential for maintaining the impermeability of the Mtb cell envelope (5).

In Mtb, FecB is annotated as a periplasmic substrate-binding protein homologous to FecB from Escherichia coli, which mediates ferric citrate transport (6). However, Mtb lacks several canonical components of the ferric citrate transport system. Instead, Mtb FecB binds heme and the siderophore carboxymycobactin and interacts with components of the mycobacterial siderophore transport machinery (7). Despite these biochemical insights, the physiological role of FecB in Mtb is not fully defined. Here, we investigated how FecB contributes to Mtb’s intrinsic antibiotic resistance and uncovered a critical functional interaction with Rv3035, a previously uncharacterized periplasmic protein. Rv3035 interacts not only with FecB but also with the arabinofuranosyltransferase AftB (Rv3805c), bridging these two proteins to form a larger complex that is important for cell envelope biogenesis.

Given the urgent need for shorter and more effective TB treatments, elucidating mechanisms that underlie Mtb’s intrinsic drug resistance is of high therapeutic relevance (2, 3). By characterizing the function of FecB and Rv3035 in Mtb, this study advances our understanding of periplasmic proteins in mycobacteria and highlights these proteins as promising targets for combination therapies aimed at enhancing the efficacy of existing anti-TB drugs.

RESULTS

FecB is noncritical for mycobactin-mediated iron acquisition in Mtb

To gain insight into the function of FecB, we performed transcriptomic analysis by RNA sequencing (RNA-seq) of wild-type (WT) Mtb, ∆fecB, and a complemented strain (∆fecB::fecB) under growth conditions with and without detergent (0.05% Tween 80) because Tween 80 strips the mycobacterial capsule and some surface exposed lipids (8). Compared to WT and the complemented strain, ∆fecB significantly up-regulated the iniBAC operon, which is known to be induced in response to cell wall stress (9), both in the presence and absence of detergent (fig. S1A and data S1). Furthermore, under detergent-free conditions, ∆fecB exhibited an iron deprivation signature, with 61% of the up-regulated (22 of 36) and 13% of down-regulated (3 of 23) differentially expressed genes (DEGs) related to iron utilization (fig. S1, A and B). Mass spectrometry (MS) confirmed siderophore accumulation in ∆fecB in detergent-free medium (fig. S1C), suggesting impaired siderophore export through the periplasm, outer membrane, and/or capsule. Growth analysis under iron-limiting conditions revealed only minor defects in ∆fecB compared to controls, in both the presence (fig. S1D) and absence of detergent (fig. S1E), contrasting with the severe growth defect of the siderophore mutant ∆mbtK.

FecB can bind ferric-carboxymycobactin and apo-carboxymycobactin with high affinity in vitro (7). To test whether siderophore binding is required for FecB’s function, we expressed FecB variants with reduced ferric-carboxymycobactin binding affinity [glutamic acid (E)339 → serine (S) (E339S) plus arginine (R)240 → S (R240S) and E339S plus tyrosine (Y)242 → S (Y242S)] in ∆fecB. These variants fully reversed the iron deprivation transcriptional signature (fig. S2A) and the subtle growth defect in low iron (fig. S2B) despite undetectable expression by Western blot (fig. S2C). However, these mutants bind apo-carboxymycobactin with similar or slightly higher affinity than WT FecB (fig. S2D). Thus, we cannot rule out that apo-siderophore binding is a function of FecB. Together, these results indicate that FecB plays a condition-dependent, noncritical role in iron acquisition that is likely compensated by other proteins.

FecB’s role in intrinsic multidrug resistance is distinct from its activity in iron homeostasis

FecB is critical for Mtb’s cell envelope impermeability and resistance to multiple antibiotics (5). To test whether FecB’s activity in iron homeostasis relates to its function in mediating cell envelope integrity and multidrug resistance in Mtb, we tested whether the FecB variants with reduced binding to ferric-carboxymycobactin could rescue the hypersusceptibility of ∆fecB to vancomycin and rifampicin (5). Complementation of ∆fecB with the FecB double mutants (E339S_R240S and E339S_Y242S) fully restored ∆fecB’s antibiotic hypersusceptibility (fig. S3A), demonstrating that ferric-carboxymycobactin binding is not required for FecB to mediate antibiotic resistance.

To explore whether siderophore accumulation in ∆fecB (fig. S1C) leads to antibiotic hypersusceptibility, we used CRISPR interference (CRISPRi) to silence expression of the mycobactin biosynthesis gene mbtI in ∆fecB (fig. S3B). This did not alter vancomycin or rifampicin minimum inhibitory concentrations (MICs) (fig. S3C), indicating that siderophore accumulation is likely not the cause of antibiotic hypersusceptibility in ∆fecB. In addition, MIC assays in low iron medium—where siderophore synthesis is induced (fig. S3D)—revealed reduced rather than enhanced differences between ∆fecB and controls (fig. S3E), contrary to expectations if siderophores promoted antibiotic hypersusceptibility. Similar results were observed under detergent-free conditions (fig. S3F), although slowed growth and increased clumping may have affected accuracy of the optical density measurements. In summary, these experiments dissociate FecB’s role in iron homeostasis from its activity in mediating antibiotic resistance.

FecB and Rv3035 interact in Mtb and likely function in the same cellular pathway

FecB is classified as a periplasmic substrate-binding protein that interacts with siderophore-mediated iron transport proteins (7). However, our data suggest an alternative, previously uncharacterized role for FecB in mediating intrinsic drug resistance in Mtb. To identify additional binding partners that might shed light on FecB’s functions, we performed immunoprecipitation (IP) coupled with MS analysis of FLAG-tagged FecB, in comparison to IP of a WT lysate to control for nonspecific interactions.

Rv3035 emerged as the most highly enriched FecB interactor (Fig. 1A). Rv3035 is annotated as an essential hypothetical protein containing WD40 repeats that are predicted to form an eight-bladed β-propeller scaffold—a structural motif that mediates protein-protein interactions and multiprotein complex assembly in prokaryotes and eukaryotes (10). We confirmed binding of FecB to Rv3035 by coimmunoprecipitation of the tagged proteins (fig. S4) and identified Rv3035’s physical interactors using Rv3035-3xFLAG as the bait. FecB was the most enriched protein in the Rv3035 immunoprecipitate compared to the WT lysate control (Fig. 1B). In addition, AftB and Rv0227c, the other two most enriched interacting partners of FecB, were also among the most enriched Rv3035-binding partners (Fig. 1, A and B, and data S2), suggesting a shared molecular complex or functional pathway.

Fig. 1. FecB and Rv3035 interact in Mtb and share essentiality profiles.

Protein interaction profiling of (A) FecB and (B) Rv3035. FecB-FLAG and Rv3035-3xFLAG were immunoprecipitated from live Mtb and interacting proteins were identified by MS. The WT lysate was used a control for unspecific interactions. (C) Heatmap of coessentiality GLS coefficients for selected gene pairs. The genes shown were selected on the basis of three “query genes”: fecB, rv3035, and aftB. We included all gene pairs involving at least one of these queries that met a significance threshold of FDR-adjusted P value < 0.001. Each square’s color represents the GLS coefficient (i.e., the strength of coessentiality), whereas size corresponds to statistical significance (−log₁₀ of the FDR-adjusted P value). Diagonal elements represent self-comparisons (each gene’s log₂ fold change profile against itself); although biologically uninformative, they are displayed at full size as a visual benchmark for perfect correlation. Some gene-gene combinations did not pass the significance threshold and are shown with smaller squares. (D) Venn diagram showing genes that are coessential with fecB (red circle), rv3035 (yellow circle), and aftB (blue circle) and their overlap. N independent experiments = 1 (A and B), each performed with triplicate cultures.

To further inform our proteomics findings, we investigated the genetic interactions of fecB via coessentiality/cofitness analysis. This approach allows the detection of gene pairs that share similar fitness profiles across forward genetic screens and thus can help identify genes that encode proteins with overlapping or related function (11). Using a publicly available CRISPRi chemical-genetic interaction dataset that profiled gene essentiality across 80 drug treatment and CRISPRi predepletion conditions (12), we performed coessentiality analysis using generalized least squares (GLS) regression of fecB, rv3035, and the genes encoding their most enriched common protein interactors, aftB and rv0227c. Whereas rv0227c did not share an essentiality profile with the others, fecB, rv3035, and aftB were all hits in each other’s coessential gene clusters (Fig. 1, C and D). This analysis identified 13 genes coessential with fecB, 19 with rv3035 and 22 with aftB, among which four were shared by all three (Fig. 1D and data S3).

FecB forms a complex with Rv3035 in vitro

To confirm the direct binding of Rv3035 and FecB in vitro, we cloned His-tagged Rv3035 from Mycobacterium thermoresistibile (Mth). Mth Rv3035 shares 67% sequence identity with Mtb Rv3035 (fig. S5A) and was used because full-length or N-terminal truncation constructs of His-tagged Mtb Rv3035 failed to yield soluble protein. Mth Rv3035 is predicted to have an N-terminal signal peptide by SignalP 6.0 (13) and to be a lipoprotein, with the N-terminal cysteine residue forming a thioether linkage to diacylglycerol (14, 15). Here, we have designated this N-terminal cysteine as the first amino acid of the mature form of Rv3035 (fig. S5B). Our Mth Rv3035 protein construct begins at residue 2 of the predicted mature protein and was expressed with a cleavable, N-terminal His-tag. Mth Rv3035-His was purified to homogeneity and the His-tag cleaved, with size exclusion chromatography (SEC) of the resulting protein, suggesting that Mth Rv3035 is monomeric (fig. S6A).

Coexpression of Mth Rv3035-His with untagged Mtb FecB resulted in coelution during Ni-affinity chromatography (fig. S6B). The complex was stable during SEC and displayed a molecular weight of 79.1 kDa, consistent with a heterodimer of Rv3035 and FecB (fig. S6A). Isothermal titration calorimetry (ITC) quantified binding affinity, revealing tight 1:1 interaction with a dissociation constant (K_d) of 27.5 ± 4.6 nM (Fig. 2A).

Fig. 2. Binding affinity of FecB and Rv3035 and x-ray crystal structures of Mth Rv3035 and the FecB-Rv3035 complex.

(A) A 19-point ITC experiment was carried out by titrating 200 μM Mth Rv3035 into 20 μM Mtb FecB. The resulting thermogram is shown (left graph), resulting in the following calculated parameters: K_d = 27.5 ± 4.6 nM; N(sites) = 1.03 ± 0.004 (right graph). (B) Cartoon depiction of the structure of Mth Rv3035. The β-propeller fold is colored in light purple for β strands and white for loops, each blade was assigned a letter, and the additional secondary structure elements are colored in pink. The disulfide bonds are shown in stick with yellow sulfur atoms. (C) Cartoon depiction of the structure of the FecB-Rv3035 complex. Rv3035 is colored as in (B). FecB secondary structure elements are colored in light green and loops in wheat. (D) Zoom in of the interface of the FecB-Rv3035 complex with interacting residues shown in stick; hydrogen bond and salt bridge interactions are represented in black dashed lines.

Structure determination of Rv3035 and the FecB-Rv3035 complex

The structure of Mth Rv3035 was solved by x-ray crystallography to 1.9 Å (table S1). Mth Rv3035 crystallized with two subunits in the asymmetric unit. In both subunits, there is clear electron density for all but the N-terminal seven residues of Rv3035. The two subunits are virtually identical with a root mean square deviation (RMSD) of 0.46 Å over 392 residues.

Rv3035 adopts a β-propeller fold, composed of eight four-stranded β sheets (blades) (Fig. 2B and fig. S5A). At the N terminus of the protein, preceding the β-propeller fold, are ~40 residues that form an ordered loop region (Fig. 2B, light green), which occupies the bottom half of the central cavity formed by the β-propeller. A portion of this N-terminal region (residues 27 to 32) forms β strand β0, which pairs with the interior β strand β26 of blade H. The upper half of the central cavity is capped by a second ordered loop region (residues 171 to 197; cyan) that contains a small α helix (α-cap) (Fig. 2B, salmon), which likely serves as a molecular gate for the central pore.

Mth Rv3035 contains two disulfide bonds (Fig. 2B and figs. S5A and S7A): one between cysteine (C)71 and C87, linking the exterior β strand β5 of blade B to the loop region on the α-cap side, and a second between C187 and C194, tethering the α-cap to the downstream loop region and presumably stabilizing the cap structure.

A structural homology search of Mth Rv3035 using the DALI server (16) (table S2) revealed that the closest hits share extremely high structural homology with Rv3035 and most are accessory proteins that assist in remodeling protein assemblies; for example, two of the top hits are BamB from Gram-negative bacteria.

To determine the protein-protein interacting residues of the FecB-Rv3035 complex, we solved the x-ray crystal structure of the FecB-Rv3035 complex to a resolution of 3.35 Å (table S1). The overall architectures of FecB and Rv3035 remain largely unchanged upon complex formation (fig. S7), with one notable exception: A disulfide bond between C56 and C104 appears in the complex structure but was absent in the previous FecB crystal structure.

The FecB-Rv3035 complex interface is small and involves 729 Ų (5.6%) of the FecB surface and 692 Ų (4.3%) of Rv3035 (Fig. 2C). Although the interface is small, there are multiple interactions between the two proteins including 19 hydrogen bonds and salt bridges that are ≤3.5 Å and six interactions that are <3.0 Å (Fig. 2D and table S3). The interface involves residues from Rv3035 blade B (β4-β5) and residues from the loop that precedes blade B (between β1 and β2) (Fig. 2D and fig. S5A). The interface also involves the backbone nitrogen atom from the disulfide bond, forming C71 (Fig. 2D and fig. S7B). Notably, only nine Rv3035 residues are responsible for the polar interactions at the interface, with six residues forming multiple contacts with FecB: arginine (R)43, lysine (K)46, glutamic acid (E)77, R83, glutamine (Q)84, and R85, where R83 forms four polar contacts at the interface. FecB residues participating in the polar contacts at the interface are localized to the second lobe of FecB and are in α7 [aspartic acid (D)223 and histidine (H)226], the loop between α8 and α9 (D258), the loop between α10B and β11 [alanine (A)291 and D292], and α13 [serine (S)320, R323, and D324] (7). Similar to Rv3035, several FecB residues are responsible for multiple contacts at the interface: H226, D292, S320, R323, and D324, where D292 forms five and D324 forms four polar contacts at the interface. The FecB-Rv3035 interface leaves the ligand binding site of FecB accessible to solvent and thus still able to accommodate a ligand (Fig. 2C).

Key FecB residues for binding Rv3035

To test the functional importance of FecB residues involved in binding Rv3035, we mutated FecB D292 to alanine alone or H226, D292, and D324 to alanine together, and tested binding of these proteins to Rv3035 by protein interaction pull-down assays and by ITC. After determining that the mutants were properly folded by circular dichroism (fig. S8A), we used FecB-His and untagged Rv3035 in a nickel-affinity pull-down assay (fig. S8B). Whereas WT FecB-His copurified with Rv3035, FecB [D292 → A (D292A)] alone and the FecB triple mutant [H226 → A plus D292 → A plus D324 → A (H226A_D292A_D324A)] failed to pull down Rv3035. In addition, binding to Rv3035 was not observed for the single and triple FecB mutants by ITC analysis (fig. S8C). Thus, the nanomolar affinity (K_d of ~30 nM) interaction between FecB and Rv3035 can be completely disrupted by mutating one FecB residue, D292.

Rv3035 mediates intrinsic multidrug resistance in Mtb

Having demonstrated that FecB and Rv3035 interact both in vivo and in vitro, we next investigated the cellular function of Rv3035. In TnSeq screens, rv3035 mutants typically display a “growth defect” (17) and are often absent from TnSeq screen input libraries. This likely explains why rv3035 was not a hit in our TnSeq chemical-genetic interaction screen that identified fecB as important for intrinsic antibiotic resistance in Mtb (5). Because fecB and rv3035 display similar growth defects by CRISPRi (18), we reasoned that it would be possible to generate an rv3035 knockout strain (∆rv3035) to directly test whether Rv3035 mediates intrinsic resistance to different antibiotics. Although ∆rv3035 colonies took longer to grow on agar plates than WT Mtb, the mutant strain exhibited similar growth kinetics as WT and ∆fecB in liquid media (fig. S9A). We also generated a ∆fecB ∆rv3035 double knockout and corresponding complemented mutants.

We performed MIC assays using these strains for several antibiotics having different mechanisms of action to which ∆fecB was hypersusceptible (Fig. 3A). Isoniazid was included as a control because its MIC is unaffected by the absence of fecB (5). The single ∆rv3035 and double ∆fecB∆rv3035 mutants phenocopied ∆fecB, exhibiting antibiotic hypersusceptibility compared to WT. Because the antibiotic hypersusceptibility of ΔfecB correlates with increased cell envelope permeability (5), we also tested whether ∆rv3035 and ∆fecB∆rv3035 are more permeable to fluorescently labeled vancomycin and the reporter dye calcein-AM (19). Both mutants phenocopied ∆fecB with increased uptake of BODIPY-vancomycin (Fig. 3B), and both were hyperpermeable to calcein-AM similarly to ∆fecB (Fig. 3C).

Fig. 3. FecB and Rv3035 function in the same pathway that mediates cell envelope impermeability and antibiotic resistance in Mtb.

(A) MIC assays for vancomycin, bedaquiline, rifampicin, pretomanid, SQ109, and isoniazid; percent growth calculated from no drug control wells (day 10). (B) BODIPY-vancomycin uptake after 30 min of incubation; fluorescence intensity normalized by OD_580nm. Statistical significance was determined by one-way analysis of variance (ANOVA) and Tukey post hoc test. (C) Calcein-AM uptake; fluorescence intensity normalized by OD_580nm at 60 min. (D) Spot assay. (E) MIC assay for vancomycin and rifampicin; percent growth calculated from no drug control wells. ∆fecB complemented with FecB triple mutant that does not bind Rv3035 (∆fecB::fecB-H226A_D292A_D324A) and control strains (day 14), #1 and #2 indicate individual colonies (biological replicates). Error bars: ±SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. N independent experiments = 2 (A to D) and 1 (E), each performed with triplicate cultures.

Notably, ∆fecB displayed altered colony morphology and impaired growth as single colonies on agar plates, appearing smooth on high density spots and often failing to grow in lower spot dilutions (Fig. 3D). This makes experiments that rely on the quantification of ∆fecB colony-forming unit (CFU) titers unreliable. Although the basis of this phenotype remains to be investigated, it could be related to cell envelope alterations. ∆rv3035 and ∆fecB∆rv3035 displayed the same defect (Fig. 3D). Together, these results demonstrate that Rv3035 is important for intrinsic multidrug resistance, mediates cell envelope impermeability, and is required for regular growth on solid media. The absence of synergistic phenotypes in the double mutant (Fig. 3, A to D) indicates that FecB and Rv3035 function in a common pathway.

To test whether binding between FecB and Rv3035 is required for maintaining cell envelope integrity, we complemented ∆fecB with an integrated plasmid expressing the FecB triple mutant H226A_D292A_D324A, which fails to bind Rv3035 in vitro (fig. S8C). After confirming stable expression of the triple mutant in vivo (fig. S9B), we tested whether ΔfecB::fecB-H226A_D292A_D324A could restore vancomycin and rifampicin susceptibility. MICs of both drugs remained unchanged compared to ΔfecB (Fig. 3E), and the fecB triple mutant failed to rescue the solid-media growth defect (fig. S9C). These results establish that FecB-Rv3035 binding is critical for mediating intrinsic drug resistance in Mtb.

FecB and Rv3035 interact with cell envelope biosynthesis proteins

Our IP data revealed that FecB and Rv3035 also interact with the cell envelope biosynthesis proteins AftB and Rv0227c (Fig. 1, A and B), which was confirmed by coimmunoprecipitation (figs. S10 and S11). Moreover, fecB and rv3035 are coessential with aftB (Fig. 1C), suggesting that they function in a shared pathway and may constitute part of a protein complex (11). AftB catalyzes the last step in the biosynthesis of the arabinan domain of arabinogalactan (AG) and lipoarabinomannan (LAM), adding the terminal β-arabinofuranose (β-Araf) residue to those molecules (20). In AG, the terminal β-Araf and the penultimate α2-linked Araf residue provide mycolic acid (MA) attachment sites, which are critical for the formation of the mycolyl-arabinogalactan-peptidoglycan (mAGP) complex (2, 21). Rv0227c (named polar growth factor A, or PgfA, in Mycobacterium smegmatis) is an inner membrane protein that interacts with MmpL3, the transporter for the cell envelope component trehalose monomycolate (TMM), and with a TMM analog, suggesting that it participates in TMM transport across the plasma membrane (22).

In the periplasm, TMMs may donate a mycoloyl residue to AG to form the mAGP complex or to another TMM molecule to form trehalose dimycolate (TDM or cord factor) (23). Whereas MAs attached to AG (mAG) are the major components of the inner leaflet of the mycobacterial outer membrane (MOM), TMMs, TDMs, and free MAs are thought to populate both the outer and inner leaflets of the MOM (3). These lipids are critical determinants of Mtb’s cell envelope impermeability (3, 24). Because AftB and Rv0227c are involved in the biosynthesis of mAG and the MOM, the interaction of FecB and Rv3035 with these proteins suggests that they participate in the assembly or maintenance of mAG and/or the MOM.

We used AlphaFold 3 (25) to predict potential protein complexes formed by FecB, Rv3035, AftB, and/or Rv0227c, modeling different combinations of these proteins (figs. S12 and S13). The FecB and Rv3035 complex showed a very high per-residue model confidence score (plDDT; fig. S12), with structural alignment to the experimentally determined FecB-Rv3035 structure showing an RMSD of 0.91 Å over 696 residues (Fig. 2C). The complexes containing AftB (with or without Rv0227c) yielded higher confidence scores when modeled with Rv3035 than with FecB alone (fig. S13, B and C). Critically, AftB-containing models exceeded the template modeling (ipTM) score threshold (0.8) for high-quality predictions when modeled with FecB and Rv3035, with Rv3035 and Rv0227c, or when all four proteins were modeled together (fig. S13). These results suggest that Rv3035 serves as a structural bridge, mediating interactions between FecB and AftB and likely between FecB and Rv0227c.

AftB is more vulnerable to knockdown in ∆fecB than in WT

Because aftB shares an essentiality profile with fecB and rv3035 (Fig. 1C), we examined the genetic interaction between fecB and aftB. We depleted aftB with a weak CRISPRi single guide RNA (sgRNA) in both WT and ∆fecB. In WT, the weak guide caused minimal growth impairment upon CRISPRi induction, but the same weak guide completely abrogated growth in ΔfecB (Fig. 4A). No differences were observed for strains harboring a nontargeting sgRNA used as a control for anhydrotetracycline (ATc)–specific effects (fig. S14A). This indicates that aftB is more vulnerable to knockdown in ∆fecB than in WT.

Fig. 4. FecB and Rv3035 support AftB’s function.

(A) Growth curves of CRISPRi knockdown (KD) of aftB in WT/∆fecB in the absence or presence of ATc inducer; strains were grown in a 384-well plate under standing conditions. (B and C) Lipidomic analysis by LC-MS of total lipids identifying (B) free MA species and (C) TMM species. (D) TDM spot intensity quantified from fig. S15I by densitometry analysis using ImageJ software. OM refers to the outer leaflet of the outer membrane obtained after AOT treatment. (E to G) Flow cytometry analysis of DMN-Tre–labeled cells using a 405-nm laser and 525/50 channel. Representative histograms (left) and quantification (right) of mean fluorescence intensity of the DMN-Tre probe. Unstained (Unst) WT was used as a control for autofluorescence. (H and I) Relative percentage of 2-arabinofuranosyl (2-Araf)–linked residues in (H) mAGP and (I) lipoglycans (LAM and LM mixture). See fig. S16 for full details of the glycosyl linkage analysis of mAGP and LAM/LM. Statistical significance was determined by two-way (B and C) or one-way (D to I) ANOVA and Tukey post hoc test. Error bars: ±SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. N independent experiments = 2 (A and E to G) and 1 (B to D, H, and I), each performed with triplicate cultures.

We further validated this genetic interaction phenotypically: aftB knockdown in WT conferred hypersusceptibility to vancomycin and rifampicin comparable to ΔfecB (fig. S14B). These data support the coessentiality analysis (Fig. 1C) and establish that AftB is essential for intrinsic multidrug resistance in Mtb.

FecB, Rv3035 and AftB are required to maintain normal TMM levels in Mtb

To test whether FecB is required for proper MOM assembly, we performed lipidomics analysis of total extractable lipids from WT, ∆fecB, and ∆fecB::fecB strains (Fig. 4, B and C, and data S4). Contrary to expectations that mycolate levels would be reduced in ∆fecB due to their role in mediating envelope impermeability, the mutant displayed a higher abundance of TMMs and free MAs compared to WT and ∆fecB::fecB.

Because TMMs accumulate in the plasma membrane of TMM transport mutants (22, 26–28), we next assessed whether lipid trafficking was disrupted in ∆fecB. We treated bacterial pellets with the anionic surfactant dioctyl sulfosuccinate sodium (AOT) dissolved in heptane to strip noncovalently bound lipids off the MOM (29). In Corynebacterium glutamicum, AOT in heptane forms reverse micelles, which solubilize and capture OM surface lipids but does not penetrate other layers of the cell envelope (29). We validated this approach by analyzing known lipids extracted from the supernatants and pellets of AOT-treated samples, corresponding to the outer leaflet of the MOM (labeled OM) and all other extractable lipids (labeled inner membrane or IM), respectively (fig. S15, A to D, and data S4). ∆fecB displayed a higher abundance of TMMs and free MAs in both IM and OM fractions compared to WT and ∆fecB::fecB (fig. S15, E to H, and data S4). We used the OM fraction to visualize and quantify TDMs via thin-layer chromatography (Fig. 4D and fig. S15I). Although TDM levels are inversely correlated with TMMs in mutants with defective TMM transport (22, 26–28), ∆fecB also displayed increased TDM levels compared to controls.

To validate the lipidomics data with an independent method, we used the solvatochromic trehalose probe DMN-Tre to track MA incorporation into TMMs via flow cytometry (30). Upon entering the cell, DMN-Tre is mycolylated to form DMN-Tre monomycolates, which are subsequently inserted into the MOM, where the hydrophobic environment activates the probe’s fluorescence (30). Corroborating the lipid analysis, ∆fecB incorporated more DMN-Tre in its cell envelope than WT, shown as increased mean fluorescence intensity compared to WT and ∆fecB::fecB (Fig. 4E). The same phenotype was observed in aftB knockdown strains upon CRISPRi activation (Fig. 4F) and in ∆rv3035 compared to controls (Fig. 4G). Collectively, these results indicate that fecB, rv3035, and aftB mutants accumulate TMMs in the MOM and suggest that FecB, Rv3035, and AftB are required for proper assembly and/or maintenance of mycolates in the MOM.

FecB and Rv3035 are important for arabinofuranosyltransferase function in Mtb

To investigate whether the FecB-Rv3035 complex influences AftB’s arabinofuranosyltransferase activity, we quantified per-O-methylated alditol acetates derived from the mAGP complex and lipoglycans (LAM and LM mixture) from WT, ∆fecB, ∆rv3035, and complemented strains. Two-linked arabinofuranosyl (Araf) residues in the mAGP and LAM were reduced in ∆fecB and ∆rv3035 compared to WT (Fig. 4, H and I). This defect was fully reversed in the mAGP and partially reversed in the LAM of the complemented strains (Fig. 4, H and I). These findings are consistent with reduced AftB activity in ∆fecB and ∆rv3035 (20, 31). The greater decrease in 2-Araf linkage levels in LAM (1.6-fold) compared to AG (1.2 and 1.3-fold) of the two mutants, and the more efficient restoration of these glycosyl linkages in AG upon genetic complementation, indicate that the available AftB activity in each strain prioritizes the completion of the arabinan domain of AG (which is critical to the formation of the MOM) over that of LAM.

Noticeably, the decrease in 2-Araf linkages in the mAGP of ∆fecB and ∆rv3035 (fig. S16, A to H) was accompanied by a modest decrease in 5-Araf linkages and a reduction in the overall Araf/Galf ratio in ∆rv3035, suggestive of a smaller size arabinan domain compared to WT and ∆rv3035::rv3035 strains (fig. S16, B and H). With regard to LAM (fig. S16, I to O), the decrease in 2-Araf linkages in ∆fecB and ∆rv3035 was accompanied by a decrease in branched 3,5-Araf linkages and t-Araf (fig. S16, I and K) and minor changes in the mannan domain of the lipoglycans (fig. S16, L to O). We observed a marked increase in 6-mannopyranosyl (6-Manp) residues in ∆fecB compared to WT and in 2,6-Manp and 2-Manp residues in ∆rv3035 compared to WT and ∆rv3035::rv3035, typifying a more branched mannan domain elaborated with more mannoside side chains (32) in LM and/or LAM (fig. S16, M to O). Collectively, these structural alterations in the AG and lipoglycans of ∆fecB and ∆rv3035 point to greater alterations in the biosynthetic machinery of these two glycans than a sole reduction in AftB activity and suggest that other arabinosyltransferases and mannosyltransferases may be negatively affected by the loss of FecB or Rv3035.

Because AftB directly controls the addition of mycolylation sites on AG, we determined whether the accumulation of free mycolates observed in ∆fecB and ∆rv3035 (Fig. 4, B to E and G) reflects a redirection of the mycolate pool that does not attach to the cell wall, due to reduced AftB activity, and instead is redirected toward the MOM. We extracted and analyzed mycolic acid methyl esters (MAMEs) from the mAGP complex of the mutants and controls strains (fig. S16P). The two mutants displayed WT levels of cell wall–bound mycolates, indicating that the accumulation of free mycolates—including free MAs, TMMs, and TDMs in ∆fecB—cannot directly be explained by a reduction in the steady-state levels of AG-bound mycolates. However, it is possible that the less efficient attachment of mycolates to AG, when AftB activity is reduced, increases their overall turnover rate, resulting in their accumulation in the MOM.

FecB is essential for virulence of Mtb in mice

The mycobacterial cell envelope is a formidable barrier that protects Mtb not only from antibiotics but also from host immunity (3). Thus, we hypothesized that fecB is also required for establishing infection in mice. The unreliability of ∆fecB to grow as single colonies on agar plates (Fig. 3D), however, impairs the quantification of the bacterial burden in infected mice. To circumvent this problem, we implemented a tetracycline-inducible genetic switch (TetON) to control fecB expression (33), permitting CFU quantification from mice. The conditional expression of fecB was validated via MIC assays with vancomycin and rifampicin, which showed that addition of ATc rescued the antibiotic hypersusceptibility of FecB-TetON (Fig. 5A).

Fig. 5. FecB is required for Mtb virulence.

(A) MIC assays for vancomycin and rifampicin, in the absence or presence of ATc to induce fecB expression in the FecB-TetON strain (day 14). Percent growth calculated from no drug control wells. (B and C) Growth and persistence of FecB-TetON in the (B) lungs and (C) spleens of mice receiving doxy chow, doxy chow until day 42 and then regular chow, or regular chow. Data are average CFU counts from four mice per time point. Black dashed lines represent the lower limit of detection. Statistical significance was determined by unpaired t tests for each time point between doxy and regular chow groups; P value is shown above each pair analyzed. Error bars: ±SD. N independent experiments = 1 performed with triplicate cultures (A) and 1 (B and C).

Mice were infected via the aerosol route with FecB-TetON and received either regular or doxycycline (doxy) supplemented chow (Fig. 5, B and C) to induce fecB expression. Lung and spleen homogenates were cultured on agar containing ATc to activate fecB expression ex vivo. FecB-TetON grew and established a high titer chronic infection in the lungs of mice fed with doxy-containing chow (Fig. 5B). In contrast, in mice receiving regular chow, in which expression of fecB is silenced, FecB-TetON grew during the first 14 days postinfection (p.i.), after which CFU counts remained relatively stable for another 8 weeks p.i. Dissemination to the spleen was delayed, and similarly to the lung, spleen bacterial burden was substantially lower in mice receiving regular versus doxy chow (Fig. 5C).

To probe FecB’s role in Mtb persistence during chronic infection, a group of mice were given doxy chow until day 42 p.i. and then switched to regular chow (Fig. 5, B and C). There were no changes in lung CFUs at day 70 p.i. or day 140 p.i. compared to the group receiving doxy chow continuously (Fig. 5B). At day 140 p.i., spleen CFUs were ~20x lower in mice switched to regular chow compared to the doxy group (Fig. 5C). These results demonstrate that FecB is essential for virulence during acute infection and for bacterial survival in the spleen but not for persistence in the lung.

DISCUSSION

The mycobacterial cell wall, comprised peptidoglycan, AG, and MAs, is a key target of several anti-TB drugs (34). Two of the four first-line TB drugs, isoniazid and ethambutol, target MA and AG biosynthesis and, despite being developed more than 60 years ago, they remain part of the standard TB chemotherapy due to their efficacy in treating drug-susceptible Mtb infections, underscoring the value of identifying new cell wall–targeting inhibitors (9). Many aspects of the Mtb cell wall and cell envelope, including the factors governing their assembly and maintenance, remain poorly understood despite their central roles in pathogenesis and intrinsic drug resistance. In the current study, we describe two periplasmic proteins, FecB and Rv3035, that work together to support arabinofuranosyltransferase activity, mediate cell envelope impermeability, and impart intrinsic multidrug resistance in Mtb.

Our transcriptomic data indicate that Mtb lacking fecB up-regulates the iniBAC operon (isoniazid-induced genes B, A, and C), which is induced by isoniazid, ethambutol, and other inhibitors of cell wall biosynthesis (9). Up-regulation of the iniBAC genes in ∆fecB occurred independently of the presence of detergent in the media and most likely reflects a constitutive defect in cell wall integrity, which results in ∆fecB’s increased membrane permeability (5).

The iron deprivation transcriptional signature and accumulation of siderophores in ∆fecB grown in detergent-free media indicate a role for fecB in iron homeostasis. However, whether this role is direct or indirect remains unclear. FecB can bind ferric-carboxymycobactin with high affinity in vitro (7), but experiments probing the requirement of fecB for iron acquisition in vivo revealed only minor growth defects. FecB variants with reduced affinity for ferric-carboxymycobactin (FecB E339S_R240S and E339S_Y242S) rescued the transcriptional iron deprivation signature and reduced growth of ∆fecB under iron-limiting conditions. Because these variants have slightly higher affinity for apo-carboxymycobactin, we could not directly test the requirement for FecB binding to iron-free siderophores in vivo. Collectively, these findings, together with genetic and microbiological data, indicate that FecB’s role in iron homeostasis is likely separable from its role in mediating antibiotic resistance in Mtb.

Evidence for FecB’s function in cell envelope integrity emerged from its most enriched physical interactors: Rv3035, AftB, and Rv0227c. In Mycobacterium marinum, FecB also interacts with AftB and the Rv3035 homolog MMAR_1667 (35). Although AftB’s function has been well described in the literature (20, 21, 31), the roles of Rv0227c (22, 36) and Rv3035 (35) homologs in mycobacteria are less defined. Here, we report the structures of Rv3035 and the FecB-Rv3035 complex, which identified key FecB residues involved in binding Rv3035 that we validated experimentally. FecB’s binding to Rv3035 is required for its function in mediating antibiotic resistance in Mtb. Mtb∆rv3035 and the double knockout ∆fecB∆rv3035 phenocopied ∆fecB with respect to cell envelope defects, indicating that loss of either protein equivalently disrupts complex function.

Structural homology analysis of Rv3035 revealed two classes of homologs (table S2). One class represents assembly factors that contribute to scaffolding in larger complexes, facilitating the coordinated movement of multiple protein components. This includes BamB of the prokaryotic β-barrel assembly machinery (BAM) complex and the eukaryotic ODA16, an assembly factor involved in outer dynein arm intraflagellar transport (37). These proteins lack both the N-terminal propeller cavity plug and the α-cap present in Rv3035. The second class consists of enzymes—Acidithiobacillus ferrooxidans tetrathionate hydrolase (38), Methylacidiphilum fumariolicum SolV methanol dehydrogenase (39), and the Devosia albogilva quinone-dependent dehydrogenase (40)—that share secondary structure elements with the Rv3035 α-cap and bear an N-terminal propeller central cavity plug. However, in these enzymatic structural homologs, the propeller fold is more extensively decorated than in Rv3035. Thus, Rv3035 more closely resembles its enzymatic structural homologs than assembly factor homologs, suggesting that, despite acting as a scaffold in the FecB complex, it may also harbor an unidentified enzymatic activity.

What is the function of FecB and Rv3035, and how do these proteins contribute to intrinsic multidrug resistance in Mtb? Our findings suggest that FecB and Rv3035 interact with AftB, a known arabinofuranosyltransferase, and together form a functional complex that supports cell envelope biosynthesis. Proteomic data, CRISPRi essentiality profiles, and high-confidence AlphaFold 3 predictions collectively support the interaction of FecB-Rv3035 with AftB. In the predicted model, Rv3035 bridges FecB and AftB, positioning the periplasmic proteins near AftB’s active site cavity (21). This orientation suggests that the FecB-Rv3035 complex may serve as scaffold to promote AftB engagement with its substrate AG/LAM or regulate AftB’s enzymatic activity.

Lipidomics data further implicate FecB in AftB function as ∆fecB exhibited increased levels of TMMs, TDMs, and free MAs compared to controls. This phenotype is also observed in M. smegmatis treated with ethambutol (41) (inhibitor of AG/LAM biosynthesis) and in deletion mutants of aftB (20) and aftC (42) (another arabinofuranosyltransferase) in C. glutamicum and M. smegmatis, respectively, all of which accumulate trehalose mycolates. Notably, fecB, rv3035, and aftB mutants all displayed enhanced incorporation of TMMs into the MOM, a phenotype exacerbated by selective depletion of aftB in the ∆fecB background. This depletion completely inhibited growth in ∆fecB while only mildly affecting WT, indicating that Mtb becomes reliant on full aftB expression in the absence of FecB.

Glycosyl linkage analysis provides additional support for a functional linkage between FecB-Rv3035 and AftB. Mtb mutants lacking fecB or rv3035 showed reduced levels of two-linked Araf residues in AG and LAM, consistent with impaired terminal Araf addition by AftB (20, 31). A decreased ratio of arabinose to galactose in AG was also observed in an M. marinum fecB mutant (35), supporting our findings of additional alterations in length and branching of AG and LAM.

Despite reduced AftB activity, ∆fecB and ∆rv3035 did not show reduced levels of MAs attached to AG. Approximately 20 to 30% of mycolylation sites are unoccupied in in vitro–grown Mtb (43, 44), and terminal sites added exclusively by AftB are not the only mycolylation sites as MAs may also be attached to the penultimate site added by AftC (45). These unoccupied terminal sites and penultimate sites can therefore become mycolylated and buffer modest reductions in terminal Araf abundance, such that only a marked reduction of AftB activity would decrease the overall AG-bound MA levels. Loss of fecB and rv3035 led to a partial reduction of AftB activity, which is insufficient to disrupt this essential cell envelope layer. However, reduced efficiency or altered distribution of mycolates between penultimate and terminal Araf residues may underlie the imbalance in free mycolates detected in the MOM.

Besides the impact on mycolate levels and their localization, structural alterations in AG and LAM can compromise the overall cell envelope integrity and influence host-pathogen interactions (31, 35, 46). In infection experiments, FecB was dispensable for Mtb persistence in mouse lungs but essential for virulence, likely through its role in maintaining cell envelope structure. Differences in survival requirements observed between the lung and the spleen may be explained by the distinct microenvironment and immune cell composition of the two organs. In addition, our data suggest that the disrupted cell envelopes of fecB, rv3035, and aftB mutants underlie their shared multidrug hypersusceptibility, likely through multifactorial mechanisms, involving AG, LAM, and free mycolate alterations. To determine the potential of targeting FecB, Rv3035, and/or AftB to accelerate TB chemotherapy, future studies will evaluate whether conditional mutants of these proteins are killed faster in mice receiving different TB antibiotics or drug regimens.

Together, these findings support a model in which the FecB-Rv3035 complex is essential for optimal AftB activity in vivo (fig. S17). This complex contributes to the biosynthesis and structural integrity of the mycobacterial cell envelope and thereby contributes to Mtb’s intrinsic resistance to antibiotics. Understanding the molecular mechanisms that govern cell envelope impermeability and intrinsic antibiotic resistance in Mtb provides valuable insights for TB drug discovery and the development of more effective regimens that may shorten TB treatment.

MATERIALS AND METHODS

Bacterial strains and culture conditions

Mtb strains used in this study are listed in data S5. For most experiments, the strains were cultured in Middlebrook 7H9 medium (BD, #271310) supplemented with 0.2% glycerol, 0.5% bovine serum albumin (BSA) fraction V, 0.2% dextrose, 0.085% NaCl, and 0.05% Tween 80 (Sigma-Aldrich, #P8074); this is referred to as “regular” medium. For detergent-free conditions, Tween 80 was omitted from the media recipe. For growth under iron-limiting conditions, self-made 7H9 medium without ammonium ferric citrate (AFC) was prepared and chelated for 2 days with the Bio-Rad Chelex 100 resin (20 g/liter). After filter sterilization, metals other than iron were added according to their concentrations in Middlebrook 7H9. To prepare low/high iron media, 1 or 160 μM AFC was added to iron-chelated medium, respectively. For reference, Middlebrook 7H9 medium contains 150 μM AFC. Where required, antibiotics or small molecules were used at the following concentrations: hygromycin (Hyg; 50 μg ml⁻¹), kanamycin (Kan; 25 μg ml⁻¹), zeocin (Zeo; 25 μg ml⁻¹), streptomycin (25 μg ml⁻¹), apramycin (60 μg ml⁻¹), and ATc (200 ng ml⁻¹ for CRISPRi strains and 1000 ng ml⁻¹ for FecB-TetON). The strains were grown at 37°C and 5% CO₂ in a standing incubator, unless otherwise indicated.

Plasmid construction

Plasmids used in Mtb strains are listed under the strain “full name” after the double colon in data S5. Plasmids starting with pGM were constructed using the Invitrogen Gateway Cloning system. Constructs harboring the gene of interest were synthesized by GenScript. CRISPRi constructs were generated as previously described (18).

IP tandem mass spectrometry

Sample preparation

Whole-cell lysates were collected from log-phase cultures grown in the shaking incubator, incubated with 1% n-dodecyl-d-maltoside for 2 hours on a rotator at 4°C, followed by incubation with anti-DYKDDDDK (FLAG) magnetic agarose (Pierce, #A36797). Captured proteins were dissociated from magnetic agarose by addition of 3x DYKDDDDK (3xFLAG) peptide (Pierce, #A36805), according to the manufacturer’s instructions. SDS was added to the peptide eluate at a final concentration of 2%, and samples were boiled at 95°C for 10 min. Boiled samples were transferred to protein LoBind tubes. FLAG-tagged (or 3xFLAG-tagged) bait proteins were detected in eluates on an SDS–polyacrylamide gel electrophoresis gel before MS.

Twenty-three micrograms of proteins per sample was flash frozen on dry ice and subsequently dried using a SpeedVac. Next, 23 μl of lysis buffer [5% SDS and 50 mM triethylammonium bicarbonate (TEAB)] was added to initiate the S-Trap Micro Spin digestion protocol (PROTIFI). The proteins underwent reduction by 4 mM dithiothreitol (DTT) and alkylation by 20 mM iodoacetamide. Following an overnight digestion with trypsin at 37°C, peptides were eluted first by addition of 40 μl of 50 mM TEAB and spun at 4000 rpm for 1 min. This was followed by a 1-min spin with 40 μl of 0.2% formic acid (FA) and, last, 40 μl of 50% acetonitrile (ACN). The three elution were then pooled together, dried using a SpeedVac, and reconstituted in 20 μl of 0.1% FA in 5% ACN. After reconstitution, the samples were centrifuged at 16,000 rpm for 16 min and an 18-μl aliquot of the supernatant was transferred to a nonbinding high-performance liquid chromatography (HPLC) vials.

LC-MS/MS and data analysis workflows

Data from the first experiment (pull-down of FecB-FLAG) were acquired on a TimsTOF Pro2 (Bruker) mass spectrometer, which was coupled to a nanoElute LC system from Bruker. Five microliters of peptides was loaded and separated on an in-house fused silica analytical column [75-μm inside diameter (ID)] packed with 25-cm ReproSil-Pur C18-AQ (Dr. Maisch, GmbH, 120 Å, 3 μm) particles to a gravity-pulled tip, using an aqueous mobile phase (A) of water and 0.1% FA and an organic mobile phase (B) of ACN and 0.1% FA. A 30-min gradient was used for data-dependent acquisition parallel accumulation-serial fragmentation (DDA-PASEF) with a flow rate set at 500 nl/min. The captive nanoelectrospray voltage was maintained at 1600 V, using one column configuration (no trap). The DDA-PASEF method consists of 10 tandem mass spectrometry (MS/MS) PASEF scans per topN acquisition cycle, with ramp and accumulation times of 100 ms, covering a mass/charge ratio (m/z) range from 100 to 1700 and an ion mobility range (1/K₀) from 0.70 to 1.30 V·s/cm². Collision energy settings followed a linear function of ion mobility, ranging from 20 eV at 0.6 V·s/cm² to 59 eV at 1.6 V·s/cm², using default parameters. Calibration of the instrument was performed using three ions from the ESI-L Tuning Mix (Agilent) (m/z 622, 922, and 1222).

The custom workflow “LFQ-MBR” from FragPipe version 21.1 was used, including the database search with MSFragger (version 4.0) and deep learning prediction rescoring with MSBooster, Percolator, and ProteinProphet (Philosopher version 5.1.0) for peptide-spectrum match (PSM) validation and protein inference. The raw files from TimsTOF Pro2 were searched against the database “Mycobacterium_tuberculosis_H37Rv_v4modified” appended with common contaminant proteins. Decoy reversed sequences were appended to the search database. The default MSFragger search parameters were used, except precursor and fragment mass tolerances were set to 20 parts per million (ppm); enzyme: trypsin (not cutting before P); peptide length: 7 to 30. Variable modifications were set to oxidation of Met and acetylation of protein N-t and fixed modifications to carbamidomethylation of Cys. MSBooster, Percolator, and ProteinProphet default options were used, and results were filtered by 1% false discovery rate (FDR) at the protein level. The searched results were then imported to Scaffold 5 Q+S (Proteome Software Inc., v5.3) for data visualization.

Data from the second experiment (pull-down of Rv3035-3xFLAG) were acquired using a NanoAcquity UPLC (Waters Corporation, Milford, MA) coupled to an Orbitrap Fusion Lumos Tribrid (Thermo Fisher Scientific, Waltham, MA) mass spectrometer. Peptides were trapped and separated using an in-house packed precolumn packed with 2-cm ReproSil-Pur C18-AQ (Dr. Maisch, GmbH, 120 Å, 5 μm) particles plus and an in-house fused silica analytical column (75-μm ID) (as previously described) and the same solvent composition. A 3.8-μl sample was injected, and the peptides were trapped at a flow rate of 4 μl/min with 5% mobile phase B for 4 min, followed by a gradient elution with 5 to 35% B at a flow rate of 300 nl/min over 120 min (total run time: 145 min). Mass spectra were acquired over m/z 375 to 1500 Da with a resolution of 120,000 (m/z 200). Tandem mass spectra were acquired using data-dependent acquisition with an isolation width of 1.6 Da, a higher energy collision dissociation (HCD) collision energy of 30%, a resolution of 15,000 (m/z 200), a maximum injection time of 50 ms, and an automatic gain control (AGC) target of 50,000.

Raw data files from the Orbitrap Fusion Lumos Tribrid were peak processed with Proteome Discoverer (version 2.5, Thermo Fisher Scientific) followed by identification using the Mascot Server (Matrix Science, v2.6.2) against the same database used in the previous experiment. Search parameters were as follows: tryptic digestion with up to two missed cleavages; peptide N-terminal acetylation, methionine oxidation, and N-terminal glutamine to pyroglutamate conversion were specified as variable modifications. Carbamidomethylation of cysteines was set as static modification. Assignments were made using a 10-ppm mass tolerance for the precursor and 0.6-Da mass tolerance for the fragment ions. All nonfiltered search results were processed by Scaffold 5 Q+S at 1.0% FDR for peptides and 99% threshold for proteins (two peptides minimum).

For each experiment, the total spectrum count (TSC) was exported from Scaffold 5 Q+S into a Microsoft Excel spreadsheet. Zeros were replaced with ones in WT control samples to allow fold change (FC) calculation of every protein detected in the samples containing the bait protein. For normalization, we summed the TSC values for all proteins in each replicate, calculated the average of the sum of the three replicates, and divided the average by the sum value of each replicate (this is the normalization factor). Then, we multiplied each protein TSC by the normalization factor of that replicate sample and calculated the median, log₂FC (LFC) of the tagged group over WT control, and P value by performing t tests of the normalized values in each row. P values were transformed into −log₁₀(P value) for visualization of the data as a volcano plot.

Coessentiality analysis

We performed coessentiality analysis on CRISPRi data (12), which profiled Mtb fitness under 80 experimental conditions. These conditions span nine antibiotics [bedaquiline (BDQ), clarithromycin (CLR), ethambutol (EMB), isoniazid (INH), levofloxacin (LVX), linezolid (LZD), rifampicin (RIF), streptomycin (STR), and vancomycin (VAN)], each tested under nine screening conditions defined by a combination of predepletion time (1, 5, or 10 days of CRISPRi induction before drug exposure) and drug concentration (“low,” “medium,” or “high” partially inhibitory doses). We used only the LFC values from negative selection screens, which report the fitness effects of gene knockdown.

Each gene was thus represented by an 80-dimensional LFC vector capturing its essentiality profile across the drug/time/concentration matrix. To account for covariance across screens—such as batch effects—we computed an 80 × 80 empirical covariance matrix (Σ) from the LFC data. We then applied GLS, following a previously reported approach (47) to compute pairwise coessentiality relationships among 3979 genes. This yielded a 3979 × 3979 matrix of GLS coefficients and corresponding P values.

We corrected for multiple hypothesis testing using the Benjamini-Hochberg procedure (fdrcorrection in the Python statsmodels package). Gene pairs with a FDR-adjusted P value of <0.001 were considered significantly coessential, resulting in ~16,000 high-confidence gene pairs.

Design of expression constructs for Mth Rv3035

Initial attempts to express Mtb Rv3035 in E. coli using N-terminal or C-terminal 6xHis-tag (His) constructs with full-length and various N-terminal truncations failed to produce soluble Rv3035. To improve solubility, we turned to a thermostable homolog of Rv3035, Mth Rv3035 (accession no. WP_040546537), and constructed two E. coli codon-optimized constructs cloned into pET28a with a tobacco etch virus (TEV)–cleavable N-terminal His-tag (GenScript, Piscataway, NJ). The Mth Rv3035 construct that encodes for the mature form of Rv3035 without its predicted signal peptide [SignalP 6.0 (13)], starting at residue Gly² (fig. S5B), produced soluble Rv3035 only when expressed in E. coli T7SHuffle cells (New England Biolabs, Ipswich, MA). Notably, T7SHuffle cells promote disulfide bond formation (48), and Rv3035 is predicted to harbor two disulfide bonds.

Generation of FecB mutants for in vitro studies

A pET28a FecB construct encoding Mtb FecB (without the signal peptide, mature FecB starting at residue 39) with a thrombin-cleavable N-terminal 6xHis-tag (7) was used as the template for in vitro site-directed mutagenesis using the Pfu Ultra Fusion HS DNA polymerase (Agilent) with the primers listed in table S5 and then confirmed by DNA sequencing (GeneWiz from Azenta Life Sciences). The FecB construct encoding no His-tag was generated using the pET28a FecB plasmid and excising the encoded fecB gene with Nde I and Bam HI and ligating it into a pET22b vector.

Expression and purification of Mth Rv3035 and the Mth Rv3035-Mtb FecB complex

Mth Rv3035-His was expressed in E. coli T7SHuffle cells in Terrific Broth (TB) media containing 0.4% glycerol and Kan (50 μg/ml). To express the Mth Rv3035-Mtb FecB complex, which we refer to as the FecB-Rv3035 complex throughout the manuscript, both pET28a Mth Rv3035-His and pET22b Mtb FecB plasmids were cotransformed into T7SHuffle cells with selection on LB and Kan (50 μg/ml) and ampicillin (Amp; 100 μg/ml) agar plates. A fresh transformant was used to inoculate TB containing 0.4% glycerol, Kan (50 μg/ml), and Amp (100 μg/ml). Cells were grown at 37°C to an OD_600nm (optical density at 600 nm) of ~0.6, before protein expression was induced by the addition of 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG). The cells grown overnight at 20°C and harvested by centrifugation.

Both Mth Rv3035-His and the FecB-Rv3035-His complex were purified using the same protocol. Cell pellets were resuspended in 10x vol(ml)/w(g) cells in Buffer A [150 mM NaCl and 50 mM Tris (pH 7.8)] with 0.25 mM phenylmethylsulfonyl fluoride (PMSF) with stirring at 4°C for 30 min. Cells were lysed by sonication, and the cell debris was pelleted by centrifugation at 14,000 rpm for 1 hour at 4°C. The cell lysate was passed through a 1-μm syringe filter to remove the insoluble debris and applied to a His-trap column (Cytiva, Marlborough, MA) preequilibrated with Buffer A using an AKTA Start FPLC instrument (Cytiva, Marlborough, MA). Following a 10x column volume (CV) wash step with Buffer A, a gradient of Buffer A from 0 to 500 mM imidazole over 20 CVs elute His-tagged proteins. Fractions containing Mth Rv3035-His or the FecB-Rv3035-His complex were concentrated using an Amicon concentrator (MilliporeSigma, 30-kDa cutoff) with buffer exchange into Buffer A.

Expression and purification of Mtb FecB and mutants

Purification of WT FecB and FecB mutant proteins were carried out in a similar manner as previously described (7). FecB WT or mutants were expressed in BL21-(DE3) cells, and cells were grown in TB and 0.4% glycerol and Kan (50 μg/ml) at 37°C to an OD_600nm of ~0.6, before protein expression was induced with IPTG (0.5 mM). Cell growth was continued overnight for 20 hours at 20°C, before cells were harvested by centrifugation. Cell pellets were resuspended in 10x vol(ml)/w(g) Buffer B [350 mM NaCl and 50 mM Tris (pH 7.8), and 10% glycerol] with 0.25 mM PMSF with stirring at 4°C for 30 min. Cells were lysed by sonication, and the cell debris was pelleted by centrifugation at 14,000 rpm for 1 hour at 4°C. The cell lysate was filtered and applied to a Ni-NTA resin preequilibrated with Buffer B using gravity flow. The column was washed with Buffer B and eluted with a stepwise elution gradient of imidazole going up to 600 mM imidazole in Buffer B. Fractions found to contain FecB by SDS-PAGE analysis (expected MW ~37 kDa) were pooled and dialyzed against Buffer B overnight at 4°C.

Removal of N-terminal His-tag from Rv3035-His and FecB-His

For cleavage of His-tags from Rv3035-His and from the FecB-Rv3035-His complex, TEV was added to ~30 to 100 mg of protein at a final concentration of ~0.1 mg/ml TEV and incubated overnight at 4°C. The cleavage reaction was passed over a Ni-NTA column preequilibrated with Buffer A and 10 mM imidazole and chased with Buffer A and 10 mM imidazole. Flow-through fractions containing Rv3035 or the FecB-Rv3035 complex were concentrated and buffer exchanged into Buffer A.

For the isolation of tag-free WT FecB, which was used for some ITC experiments, a similar procedure was used but with thrombin at ~30 U thrombin:1 mg of FecB-His. The cleavage reaction was passed over a Ni-NTA column preequilibrated with Buffer B and 10 mM imidazole and chased with Buffer B and 10 mM imidazole. Flow-through fractions containing WT FecB were concentrated and buffer exchanged into Buffer B.

ITC experiments

ITC experiments were carried out using a Malvern MicroCal PEAQ-ITC instrument (Malvern, UK). Before the experiments, proteins were dialyzed into Buffer B and the same dialysis buffer was used for protein dilutions. Nineteen-point titrations were carried out using 200 μM Mth R3035 in the syringe and 20 μM FecB in the cell, whereas 13-point titrations were carried out using 300 μM Rv3035 and 30 μM FecB-His mutant protein or 200 μM Rv3035-His and 20 μM FecB-His. Notably, the His-tag on FecB did not appreciably change the affinity of FecB for Rv3035. Curve fitting was carried out using the Malvern instrument analysis software.

Structure determination of Mth Rv3035 and the FecB-Rv3035 complex

Initial screening of the Mth Rv3035 (15 mg/ml) and FecB-Rv3035 complex (38 mg/ml) were carried out in 96-well trays using a Mosquito liquid handler (SPT Labtech, Cambridge, MA) and commercially available crystallization screens by the hanging drop vapor diffusion method.

Mth Rv3035 alone

Diffraction-quality Mth Rv3035 crystals were grown with a crystallization reservoir of 0.1 M sodium chloride, 0.1 M Hepes (pH 7.5), 1.6 M ammonium sulfate with Rv3035 at 15 mg/ml, and a reservoir-to-protein ratio of 1:1 at room temperature. The crystals were cryocooled in mother liquor supplemented with 25% glycerol. Data were collected at SSRL 12-2 at a wavelength of 1 Å. Data were processed in XDS (49) to a resolution of 1.9 Å (table S1). A model of Mth Rv3035 was generated on the AlphaFold server (25) and was used for molecular replacement (MR) in phaser (50) in phenix. The MR solution placed two subunits in the asymmetric unit (ASU) and was improved by autobuild in phenix (51), and refined in phenix.refine (52). Following autobuild, the N-terminal residues 10 to 37 were built de novo in coot (53). The N-terminal nine residues of Rv3035 could not be modeled for both subunits, and for subunit B, the last C-terminal residue also could not be modeled due to unresolved electron density. The model was further refined in phenix.refine and coot to a final R_work/R_free of 16.1/20.25 (table S1).

FecB-Rv3035 complex

Diffraction-quality crystals of the FecB-Rv3035 complex were obtained with a crystallization reservoir of 3 M sodium chloride and 0.1 M Bis-Tris (pH 7.0) with a reservoir–to–protein complex ratio of 1:1. To further improve the diffraction quality of the FecB-Rv3035 complex, an additive screen was performed (Hampton Research, Aliso Viejo, CA), with the final conditions used being a crystallization reservoir of 3 M sodium chloride, 0.1 M Bis-Tris (pH 7.0), and 3% trehalose, with the Rv3035-FecB complex at 38 mg/ml and a reservoir-to-protein ratio of 1:1. Crystals were cryoprotected in reservoir solution with 25% glycerol and flash frozen with liquid nitrogen. Data were collected at SSRL 12-2 at a wavelength of 1 Å.

Data were processed in XDS (49) to a resolution of 3.3 Å (table S1). The phases were solved by MR in phaser (50) in phenix, placing a single subunit of the deposited FecB structure [Protein Data Bank (PDB) ID 7UQ0 (7)] in the ASU in complex with a single subunit of the Mth Rv3035 structure solved above. The MR solution was improved in autobuild and refined as described above in phenix.refine (52) and coot (53) to a final R_work/R_free of 25.2/26.3 (table S1).

Knockout construction

∆rv3035 and ∆fecB∆rv3035 mutants were generated using a suicide plasmid approach as previously described (54) with a few modifications. Constructs containing 500–base pair (bp) fragments corresponding to regions upstream and downstream of rv3035 were ordered from GenScript and cloned into a temperature-sensitive deletion vector pDE43-XSTS to flank the Zeo resistance cassette and generate pKO-XSTS-Zeo-rv3035 via Gateway Cloning. Mtb WT and ∆fecB were transformed with both the pKO-XSTS-Zeo-rv3035 and an rv3035 complementation plasmid (pGMCS-T02-Phsp60-SD1-rv3035_A-HA) to generate a merodiploid strain because initial attempts of generating the knockout with pKO-XSTS-Zeo-rv3035 alone had failed. Transformants were plated on Middlebrook 7H10 agar (BD, #262710) containing 0.5% glycerol, 10% Oleic Albumin Dextrose Catalase (OADC) supplement (BD, #212351), and Zeo (25 μg/ml; for WT) or Hyg (50 μg/ml) and Zeo (25 μg/ml; for ∆fecB) and incubated at 37°C until colonies were ~1 mm in diameter. Colonies were tested for reporter gene XylE activity directly on plates by addition of a drop of catechol solution (1 g in 20 ml of water). After 30 min, candidate colonies that turned yellow were selected and grown in regular 7H9 with appropriate antibiotics until they reached an OD_580nm of 1 to 1.5. Candidates were then grown on 7H10 agar with 10% sucrose at 37°C until colonies were ~1 mm in diameter. Colonies were test for reporter gene activity as above, but this time, white colonies (that had successfully lost the suicide plasmid) were selected. Genomic DNA (gDNA) was extracted from a few mutant candidates and allelic exchange was confirmed by polymerase chain reaction (PCR) using primers listed in table S6. To obtain the final knockout mutants, we replaced the rv3035 complementation plasmid integrated into attL5 with an empty plasmid containing an apramycin resistance cassette (pGMCA-OXOX). Transformants were plated on 7H10 containing Zeo (or Zeo and Hyg) and apramycin (60 μg/ml). ∆rv3035 and ∆fecB∆rv3035 mutants took about 6 to 8 weeks to form visible colonies. Candidate mutants were selected and tested for streptomycin sensitivity in liquid culture as an indication that the copy of rv3035 (carrying streptomycin resistance) was successfully replaced in attL5. Last, gDNA was extracted for whole-genome sequencing to confirm deletion of rv3035 and insertion of Zeo and apramycin cassettes.

MIC assays

Mtb strains were inoculated from frozen glycerol stocks and grown with appropriate selection antibiotics in 7H9 medium until the mid-log phase at 37°C. Bacteria were diluted to an OD_580nm of 0.01 in 7H9, and 50 μl of bacteria was added to a 384-well plate in triplicate per antibiotic concentration or to triplicate control wells without antibiotics. The HP D300e digital dispenser was used to deliver antibiotics to plate wells. Bacteria were grown in standing cultures at 37°C, and OD_580nm was measured in a plate reader between days 10 and 21 following treatment (time point is indicated in each figure legend). For MICs under low iron or detergent-free conditions, bacterial cultures were first washed once or twice in phosphate-buffered saline (PBS) before dilution into the respective media of the MIC assay. For CRISPRi strains, cultures were split into two flasks and ATc (200 ng/ml) was added to one of the flasks 2 to 3 days before setting up the MIC assay to predeplete the gene of interest. Fresh ATc (200 ng/ml) was again added to the media of the assay (+ATc conditions) before distributing bacteria into plates.

Cell envelope permeability assays

BODIPY fluorescent (FL) vancomycin (Invitrogen, #V34850) uptake was performed as previously reported (5) with a few modifications. Briefly, mid-log phase Mtb cultures were washed once and adjusted to an OD_580nm of 2 in PBS and incubated with BODIPY FL vancomycin (2 μg/ml). Sample aliquots (350 μl) were taken at the 30-min incubation time point, washed twice in PBS with 0.05% Tween 80, and resuspended in 350 μl of PBS. Negative control samples were treated the same way but did not receive BODIPY FL vancomycin. Each sample (100 μl) was distributed in triplicate wells in a black-sided 96-well plate. Fluorescence was measured (excitation wavelength of 485 nm and emission wavelength of 538 nm) and normalized to the OD_580nm of the final suspension to account for cell loss during washing steps.

For the calcein-AM (Invitrogen, #C3099) uptake assay, mid-log phase cultures were adjusted to an OD_580nm of 0.8 in regular 7H9 and 100-μl aliquots were added to the wells of a black-sided 96-well plate in triplicate. One hundred microliters of calcein-AM (1 μg/ml in 7H9 with 0.4% glucose) was added to the wells, resulting in a final calcein-AM concentration of 0.5 μg/ml. Fluorescence was measured (excitation wavelength of 495 nm and emission wavelength of 520 nm) at 2-min intervals over a course of 60 min. Fluorescence was normalized to OD_580nm measured at 60 min.

Spot assay

Mig-log phase Mtb cultures were adjusted to an OD_580nm of 0.3, and 10-fold serial dilutions in PBS with 0.05% Tween 80 were spotted on 7H10 agar with 5 μl per spot. Plates were incubated at 37°C and imaged after 2 to 3 weeks. To confirm colony morphology alteration and growth defect of ∆fecB, ∆rv3035, and ∆fecB ∆rv3035 mutants, plates were incubated for 4 to 5 additional weeks, after which the phenotypes were unchanged.

Growth curves

Mtb strains were inoculated from frozen glycerol stocks and grown with appropriate selection antibiotics in 7H9 medium until the mid-log phase at 37°C.

For CRISPRi strains, cultures were split into two flasks and ATc (200 ng/ml) added to one of the flasks 2 to 3 days before setting up the growth curve to predeplete the gene of interest. Fresh ATc (200 ng/ml) was again added to the media (+ATc conditions) before distributing bacteria into 96/384-well plates in triplicate.

For predepletion of iron internal stores, bacteria were washed in PBS or iron-chelated 7H9 and diluted to an OD_580nm of 0.05 in iron-chelated 7H9. After four to five generations, bacteria were spun down at 800 rpm for 8 min to generate a single-cell suspension. Bacteria were diluted to an OD_580nm of 0.025 in the indicated media and added to flasks that were incubated standing at 37°C. For serial passaging, when cultures in high iron media reached the stationary phase, low iron cultures were spun down at 800 rpm for 8 min to generate a single-cell suspension, and bacteria were diluted to an OD_580nm of 0.025 in both high/low iron media. OD measurements from cultures in flasks were performed in a spectrophotometer. For growth curves in low iron media without detergent, iron predepleted cultures were washed in PBS or detergent-free iron-chelated 7H9 and diluted to an OD_580nm of 0.05 and 100 μl of bacteria was added to a 96-well plate in triplicate. OD measurements from plates were performed in a plate reader. Cultures grown in 96-well plates were resuspended before OD_580nm readings.

Lipid extraction

Mtb cultures were grown in detergent-free 7H9 containing fatty acid–free BSA until the mid-log phase. Bacterial pellets were transferred to sterile filters (55), which were added directly to 16 mm–by–100 mm glass tubes containing 2:1 chloroform:methanol or to 2-ml screw cap tubes containing 1 ml of 10 mM AOT dissolved in heptane or PBS as a control. For AOT treatment, pellets in AOT or PBS were agitated in a bead-beater at low speed (2800 rpm) for 10 s to knock the cells off the filters followed by removal of filters and incubation in a cooling rack for 1 hour, with mixing by inversion of the tubes every 5 min. Five hundred microliters of the AOT supernatant was added to 16 mm–by–100 mm glass tubes containing 2:1 chloroform:methanol. AOT and PBS pellets were resuspended in 600 μl of 2:1 chloroform:methanol and added to 16 mm–by–100 mm glass tubes containing 2:1 chloroform:methanol.

Liquid chromatography–mass spectrometry analysis

Lipid extracts were centrifuged at 4000 rpm for 15 min at 4°C, and the supernatant decanted into new 16 mm–by–100 mm borosilicate tubes. The tubes were dried using a GeneVac evaporator under vacuum with gentle heating to 40°C. The residual residue was transferred to preweighed 4-ml amber vials using 2 ml of 1:1 chloroform:methanol. The solution was again dried using the GeneVac evaporator. The vials were reweighed to obtain the dry weight of the lipid extract, and this mass was used to prepare a solution (1 mg/ml) in 1:1 chloroform:methanol. Lipids were stored at −20°C as a stock solution. A 150-μl aliquot of stock solution was transferred to a glass insert tube (250 μl, Agilent Technologies) nested within a microfuge tube. The aliquot was evaporated to dryness using an argon air stream then resuspended in 150 μl of a solution of 70% hexanes and 30% isopropyl alcohol. After briefly vortexing to resuspend, the samples were centrifuged at 4000 rpm for 15 min at 4°C. A volume of 100 μl of the supernatant was transferred to a 2-ml amber vial (Agilent) equipped with 250-μl glass insert tubes and capped with 9-mm septum-embedded screw caps (Agilent). A 10-μl injection of this sample was analyzed using liquid chromatography–mass spectrometry (LC-MS) on an Agilent 6520 ESI-QTOF. Separation of lipids using a 3-μm diol column (2.1 mm by 150 mm) and gradient elution from solvent A (70% hexane, 30% isopropanol, 0.02% FA, and 0.01% ammonium hydroxide) to solvent B (70% isopropanol, 30% methanol, 0.02% FA, and 0.01% ammonium hydroxide) as previously described (56). Following chromatography, the sample underwent electrospray ionization at 325°C with a drying gas flow of 5 liters/min at 5500 V and a 30-psig nebulizer pressure and ions were detected by time-of-flight in both positive and negative modes (m/z range: 100 to 3000). Internal calibrants (922.009798 positive mode and 983.0325 negative mode) were infused continuously.

Lipids were identified using unique mass–retention time identifiers for masses exhibiting the expected isotopic distributions as previously reported (56). Integrations of ion abundance were determined using Agilent Profinder 8.0 and Qualitative Analysis 6.0.

DMN-Tre labeling and flow cytometry

Mid-log phase Mtb cultures were adjusted to OD_580nm of 0.5 in regular 7H9. For CRISPRi strains, cultures were split into two flasks and ATc (200 ng/ml) was added to one of the flasks 2 to 3 days before labeling with DMN-Tre to predeplete the gene of interest. The DMN-Tre probe (30) was added to 500 μl of culture at a final concentration of 100 μM. Samples were incubated at 37°C overnight for ~20 hours. On the next day, cultures were spun down at 10,000 rpm for 3 min, and the supernatant containing DMN-Tre was discarded. Samples were fixed with 4% paraformaldehyde (PFA; BioLegend, #420801) for 4 hours at room temperature. After two washes in PBS with 0.05% Tween 80, bacteria were resuspended in PBS. Data collection was performed on a FACSymphony A5 Cell Analyzer (BD Biosciences) at the Weill Cornell Medicine Flow Cytometry Core Facility using the 405-nm violet laser and the 405 [425/50 nm] detector. Fluorescence data were obtained for 200,000 events per sample and analyzed in FlowJo software (BD) by first gating on the bacterial population in the side scatter versus forward scatter plot followed by calculation of the mean fluorescence intensity.

Preparation and analysis of lipoglycans and AG

Lipoglycan and mAGP complex extraction from delipidated cells followed earlier procedures (57, 58). Determination of the glycosyl linkage patterns of lipoglycans and mAGP followed earlier procedures (57). Per-O-methylated alditol acetates were analyzed by gas chromatography–mass spectrometry (GC-MS) on a Thermo Fisher Scientific TRACE 1310 Gas Chromatograph paired with a Thermo Fisher Scientific TSQ 8000 Evo Triple Quadrupole GC-MS/MS. Samples were run on a Zebron ZB-5HT Inferno 30 m–by–0.25 mm–by–0.25 μm capillary column (Phenomenex) at an initial temperature of 100°C. The temperature was increased to 150°C at a ramp rate of 20°C min⁻¹ and then to 240°C at a ramp rate of 5°C min⁻¹ and was held at this temperature for 3 min to be lastly increased to 300°C at a rate of 30°C min⁻¹ and held at the final temperature for 5 min. Data handling was carried out using the Thermo Fisher Scientific Chromeleon Chromatography Data System software.

Mouse infection

Sixty-four female, 8-week-old Mus musculus C57BL/6 mice (the Jackson Laboratory) were infected with ~100 CFUs per mouse using an Inhalation Exposure System (Glas-Col). Mice in the doxy group started to receive doxy chow 2 days before aerosol infection. The FecB-TetON strain was grown to the mid-log phase in the presence of ATc (1 μg/ml), and a single-cell suspension was prepared in PBS with 0.05% Tween 80 and then resuspended in PBS for a final OD_580nm of 0.2. Lungs and spleen were homogenized in PBS and plated on 7H10 agar containing ATc (1 μg/ml) to determine CFUs per organ at the indicated time points. This experiment was performed in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, with approval from the Institutional Animal Care and Use Committee (IACUC) of Weill Cornell Medicine under IACUC protocol 060-441A.

Quantification and statistical analysis

Data analysis and generation of graphs were performed in Prism version 10.5 (GraphPad).

Supplementary Materials

The PDF file includes:

Supplementary Materials and Methods

Figs. S1 to S17

Tables S1 to S6

Legends for data S1 to S5

References

Other Supplementary Material for this manuscript includes the following:

Data S1 to S5