Photoactivatable Glycolipid Probes for Identifying Mycolate– Protein Interactions in Live Mycobacteria

Abstract

Mycobacteria have a distinctive glycolipid-rich outer membrane, the mycomembrane, which is a critical target for tuberculosis drug development. However, proteins that associate with the mycomembrane, or that are involved in its metabolism and host interactions, are not well-characterized. To facilitate the study of mycomembrane-related proteins, we developed photoactivatable trehalose monomycolate analogues that metabolically incorporate into the mycomembrane in live mycobacteria, enabling in vivo photo-cross-linking and click-chemistry-mediated analysis of mycolate-interacting proteins. When deployed in Mycobacterium smegmatis with quantitative proteomics, this strategy enriched over 100 proteins, including the mycomembrane porin (MspA), several proteins with known mycomembrane synthesis or remodeling functions (CmrA, MmpL3, Ag85, Tdmh), and numerous candidate mycolate-interacting proteins. Our approach is highly versatile, as it (i) enlists click chemistry for flexible protein functionalization; (ii) in principle can be applied to any mycobacterial species to identify endogenous bacterial proteins or host proteins that interact with mycolates; and (iii) can potentially be expanded to investigate protein interactions with other mycobacterial lipids. This tool is expected to help elucidate fundamental physiological and pathological processes related to the mycomembrane and may reveal novel diagnostic and therapeutic targets.

Mycobacteria are of enormous medical and biotechnological importance. The most prominent example is tuberculosis-causing Mycobacterium tuberculosis (Mtb), which kills 1.5 million people annually and exists in drug-resistant forms that are extremely challenging to treat.^1–4 Underlying the success of Mtb and related pathogens is a complex cell envelope containing a plasma membrane, peptidoglycan, arabinogalactan, and an outer membrane called the mycomembrane (Figure 1).^5–8 The mycomembrane consists of long, branched mycolic acids, which predominantly exist as mycolate esters linked to carbohydrates.^7–11 The mycomembrane is essential for survival due to its roles in cellular integrity and defense, nutrient acquisition, and cellular communication, including host–pathogen interactions.^9,10 Multiple drugs used to treat tuberculosis act on mycomembrane biosynthesis, highlighting why this membrane is a major focal point for mycobacteria research.¹²

Figure 1.

(A) Metabolism and host interactions of mycolate glycolipids. (B) Strategy for in vivo capture and analysis of mycolate-interacting proteins using photoactivatable probes (see Scheme S1 and Supporting Information Discussion).

Significant progress toward elucidating mycomembrane composition, biosynthesis, and function has been made, although much remains to be learned. Its major mycolate glycolipids, including trehalose monomycolate (TMM), trehalose dimycolate (TDM), and arabinogalactan mycolate (AGM), are synthesized as shown in Figure 1A. TMM is synthesized from trehalose in the cytoplasm via Pks13/CmrA¹³ and then exported by MmpL3^14,15 and processed by Ag85 mycoloyltransferases^16–18 to generate TDM and AGM. However, the identities of many proteins involved in mycomembrane lipid transport, remodeling, turnover, and host interactions have remained elusive. Furthermore, the proteomic composition of the mycomembrane is notoriously poorly defined.¹⁹ Despite computational predictions that the Mtb genome may encode over 100 mycomembrane-associated proteins,^20–22 only a few have been identified and characterized across the Mycobacterium genus.^19,23–27 Most of these proteins exhibit channel activity and/or are important for nutrient influx, including the Mycobacterium smegmatis (Msmeg) porin (MspA),^23,28,29 Mtb CpnT,²⁴ and newly discovered Mtb PPE51.^30,31 The many as-yet unidentified mycomembrane proteins likely have other critical functions as well, including secretion/efflux processes, cell envelope biosynthesis and remodeling, and host–pathogen interactions.¹⁹

New tools are needed to accelerate the identification and functional characterization of mycomembrane-related proteins. Significant efforts have been made to enrich and identify mycomembrane-resident proteins,^27,32–36 mainly through subcellular fractionation and detergent extraction, but the resolution of cell envelope layers remains extremely challenging due to the massive peptidoglycan–arabinogalactan– mycolate covalent complex. Moreover, the lysis conditions, detergents, and centrifugation steps in these methods do not retain all of the protein–lipid interactions that occur in vivo, particularly weaker, transient interactions, which are frequently lost.³⁷ Such methods are also not designed to capture proteins that are not directly associated with the mycomembrane and thus miss an important subset of proteins involved in mycomembrane metabolism or host interactions. Finally, traditional methods are laborious and often incompatible with complex experimental contexts, e.g., biofilm cultures or macrophage/animal infections. Recently, lipid-mimicking probes bearing photoactivatable and clickable groups have emerged as valuable tools for profiling in vivo lipid–protein interactions.^37–41 Here, we merged this photolabeling concept with our mycomembrane-targeting probes to develop the first tool for global analysis of in vivo mycolate–protein interactions, providing a powerful new approach to investigating mycomembrane-related proteins in their native state.

We reported that TMM analogues bearing functionalized mycolate-mimicking chains can metabolically incorporate into mycomembrane components via conserved, substrate-promiscuous Ag85 mycoloyltransferases.^42–44 By altering the linker, we controlled the incorporation mechanism and labeling target, with amide-linked N-AlkTMM-C7 exclusively labeling TDM and ester-linked O-AlkTMM-C7 labeling AGM and TDM (Scheme S1, Supporting Information (SI)).⁴² Capitalizing on the TMM scaffold, we designed the two photoactivatable analogues N- and O-x-AlkTMM-C15 to enable mycomembrane proteomics (Figure 1B, SI Discussion). Both analogues possess the mycomembrane-targeting TMM moiety containing a lipophilic chain, which has a photoactivatable diazirine and a clickable alkyne. We envisioned that N- or O-x-AlkTMM-C15 would metabolically embed into glycolipids in live cells, placing the lipophilic chain in proximity to mycomembrane-related proteins. Upon UV photoactivation, the diazirine would photo-cross-link proteins, enabling click-mediated affinity enrichment from cell lysates and subsequent identification. In principle, this strategy enables capture and analysis of proteins that associate directly with the mycomembrane or that are involved in mycolate synthesis, transport, remodeling, turnover, or host interactions (SI Discussion).

The syntheses of both probes employed bifunctional fatty acid 1,³⁹ which we conjugated to trehalose derivatives 2 and 3^45–47 to produce N- and O-x-AlkTMM-C15 in two steps (Figure 2A). Using bovine serum albumin (BSA) as a model protein,⁴⁸ we confirmed that both probes possessed the requisite functionalities of (i) photo-cross-linking proteins when UV-irradiated and (ii) labeling and detecting the resulting cross-linked products via Cu-catalyzed azide–alkyne cycloaddition (CuAAC) (Figure 2B).

Figure 2.

(A) Syntheses of N- and O-x-AlkTMM-C15. (B) UV-dependent photo-cross-linking of BSA with probes followed by CuAAC-mediated product detection.

We tested whether N- and O-x-AlkTMM-C15 metabolically incorporated into the mycomembrane of live bacteria, focusing on the model organism Msmeg. Both TMM probes labeled Msmeg in a concentration- and time-dependent manner (Figure S1), whereas 1, which lacks the trehalose targeting moiety, did not label the Msmeg surface (Figure S2). Partial growth inhibition for the probes was observed at 250 μM (Figure S3), indicating an optimal concentration of 25–100 μM. The probes were specific, as they efficiently labeled mycomembrane-containing Msmeg and Corynebacterium glutamicum, but not mycomembrane-deficient Bacillus subtilis or Escherichia coli (Figures 3A and S4). Consistent with the hypothesized incorporation routes (Scheme S1), N-x-AlkTMM-C15 labeling was entirely localized to the TDM-containing extractable lipids fraction and a new fluorescent lipid consistent with labeled TDM was observed, whereas O-x-AlkTMM-C15 labeling was detected in both the TDM- and AGM-containing fractions (Figures 3B and S5). The signal from both probes, and the peptidoglycan probe RADA⁴⁹ (positive control), was depleted upon spheroplast formation, which sheds the peptidoglycan–arabinogalactan–mycomembrane complex, leaving a spherical cell with the plasma membrane intact (Figure S6).^50,51 This result indicated that neither probe was detected in the plasma membrane. Incorporation of N-x-AlkTMM-C15 was reduced when Msmeg was co-incubated with an unlabeled TMM competitor or the Ag85 inhibitor ebselen⁵² (Figures S7 and S8). Furthermore, an Ag85 partial knockout mutant⁵³ exhibited reduced labeling by N-x-AlkTMM-C15 compared to a control peptidoglycan probe⁵⁴ (Figure S9). Collectively, these data demonstrate that photoactivatable TMM analogues incorporate into the native mycomembrane as anticipated.

Figure 3.

Mycomembrane labeling with N- and O-x-AlkTMM-C15. (A) Bacteria were cultured in probe (25 μM), reacted with azido-488 by CuAAC, and analyzed by microscopy (Figure S4, flow cytometry). (B) Probe-treated Msmeg was reacted with azido-488 by CuAAC and fractionated into PG-AGM- and TDM-containing fractions, and fluorescence was measured. Error bars denote the standard deviation of three replicates. MFI, mean fluorescence intensity in arbitrary units.

We next performed protein photo-cross-linking experiments in Msmeg using the TDM-targeting N-x-AlkTMM-C15 probe, which we prioritized primarily due to the complexities associated with O-x-AlkTMM-C15 labeling AGM (SI Discussion). N-x-AlkTMM-C15-treated live Msmeg was UV-irradiated, then lysates were collected, subjected to CuAAC with azido-488, and analyzed by SDS-PAGE. These experiments showed that proteins were labeled in a probe-, concentration-, and UV-dependent manner (Figure S10). Msmeg growth and metabolic activity were unaffected by UV irradiation (Figure S11), suggesting that UV-induced crosslinking in live cells occurred with minimal perturbation. To test whether N-x-AlkTMM-C15 photo-cross-linked our validation proteins, Ag85 and MspA, we enriched proteins and performed Western blot analysis. Probe-treated Msmeg was UV-irradiated, and then lysates were obtained and reacted with azido-TAMRA-PEG-biotin (AzTB) by CuAAC, delivering fluorescent and biotin tags to proteins for detection or enrichment. AzTB-treated lysates were analyzed by SDS-PAGE and Western blot prior to (input) and after (output) affinity capture on and elution from avidin beads. Proteins were effectively enriched only in the probe-treated, UV-irradiated (+probe+UV) samples (Figure 4A). Importantly, Ag85 and MspA were detected in all input samples, while both were clearly enriched in the outputs of the +probe+UV samples (Figure 4B and SI Discussion). The plasma membrane-associated mannosyltransferase MptA (negative control) was not detected in the outputs (Figure S12). These results show that N-x-AlkTMM-C15 enables photo-cross-linking, affinity enrichment, and detection of mycolate-interacting proteins.

Figure 4.

N-x-AlkTMM-C15-mediated affinity enrichment of mycolate-interacting proteins. Msmeg was cultured in N-x-AlkTMM-C15 (100 μM), UV-irradiated, and lysed. Lysates were reacted with AzTB by CuAAC and then analyzed using the indicated method before (input) and after (output) incubation with avidin beads to evaluate enrichment of (A) proteins in general and (B) MspA and Ag85. Data are representative of three independent experiments.

Finally, we used N-x-AlkTMM-C15 and label-free quantitative proteomics to identify mycolate–protein interactions in Msmeg, which was grown either for a shorter period to lower density (log phase) or an extended period to higher density (early stationary phase). In each of the two studies, LC−MS/MS analysis identified ~110 proteins that were significantly enriched by ≥4-fold in the +probe+UV group versus the probe−UV control, of which ~75 proteins were identified exclusively in the +probe+UV group (Figure 5 and Tables S1–S4). These identifications included multiple Ag85 isoforms and MspA, consistent with Western blot analysis (Figure 4B) and confirming that N-x-AlkTMM-C15 photo-cross-links mycolate-interacting proteins. Additional proteins with known mycolate-related functions were identified, including CmrA, which is involved in TMM synthesis,^55,56 and MmpL3, which is the TMM flippase.^14,15 We identified multiple relevant hydrolases, including TDM hydrolase (Tdmh), which is involved in stress-induced mycomembrane remodeling,^57–59 and two related proteins, MSMEG_1528 and MSMEG_0194 (55% and 41% sequence identity to Tdmh), which potentially represent novel mycomembrane-remodeling enzymes. Other notable hits include EccA1, whose absence in Mycobacterium marinum reduced mycolate synthesis by 40%,⁶⁰ and the extracellular proteins MTB12, MPT64, and HBHA, all of which have Mtb orthologs involved in host–pathogen interactions that are attractive diagnostic markers and/or vaccine candidates.^61–63 Indeed, most identified Msmeg proteins have Mtb orthologues, ~15–20% of which are essential for growth⁶⁴ and whose major predicted functions include cell wall/cell processes and uncharacterized hypothetical proteins (Tables S1 and S2; Figure S13). The differential protein profiles between our two studies, in terms of both protein identity and predicted functions (Figures 5C and S13), have interesting biological and experimental implications. The observed changes likely reflect a combination of growth-phase-dependent dynamic changes in mycolate–protein interactions and of improved detection of low-abundance interactions in higher-density cultures (SI Discussion). Finally, the successful identification of nearly all known trehalose mycolate-interacting proteins in Msmeg (see Figure 1A) provides high confidence in probe specificity and thus in the biological relevance of the proteins identified through our strategy.

Figure 5.

Volcano plots showing proteins in red that were significantly enriched in N-x-AlkTMM-C15-treated, UV-exposed (+probe+UV) versus nonirradiated (+probe−UV) Msmeg grown to OD₆₀₀ (A) ~1.2 or (B) ~4 using click-mediated protein affinity enrichment, tryptic digestion, and LC−MS/MS analysis. Selected proteins of interest are indicated. (C) Venn diagram of proteins enriched in (A) and (B).

Given the importance of the mycomembrane to mycobacterial physiology and tuberculosis drug development, it is perplexing that such wide gaps in knowledge still exist with respect to its proteomic composition and the identities of proteins involved in its metabolism and host interactions. To date, the extraordinary complexity of the mycobacterial cell envelope, and the lack of suitable tools to experimentally dissect it, have impeded progress toward elucidating the structures and functions of mycomembrane-related proteins. As a new approach to solving this problem, we reported the first probes for capturing lipid–protein interactions in live mycobacteria and demonstrated their ability to identify mycolate-interacting proteins with known functions spanning mycomembrane synthesis, transport, and remodeling. We generated and analyzed protein lists containing numerous candidate mycolate interactors, many with unknown function, which, along with the probes themselves, are valuable for future research. Beyond expanding applications of N-x-AlkTMM-C15, we are further investigating O-x-AlkTMM-C15 and exploring a two-step approach using 6-TreAz⁴⁶ with our photoactivatable cyclooctynes.⁴⁸ Our tools’ in vivo compatibility invites experimentation in diverse contexts (e.g., spatiotemporal proteomics, biofilms, infection models), while their generality motivates application to other mycobacteria, most importantly Mtb, which is labeled by N- and O-x-AlkTMM-C15 (Figure S14). Our approach can also be extended to study endogenous or host protein interactions with other types of mycobacterial lipids, which are widely appreciated for their distinctive structures and biological importance. Ultimately, the ability to elucidate native-state lipid–protein interactions in mycobacteria will advance our understanding of mycobacterial physiology and pathogenesis, and may reveal new targets for the development of urgently needed tuberculosis vaccines, diagnostics, and drugs.