Decision letter: Immunopeptidomics reveals determinants of Mycobacterium tuberculosis antigen presentation on MHC class I

Abstract

Full text Figures and data Side by side Abstract Editor's evaluation Introduction Results Discussion Methods Data availability References Decision letter Author response Article and author information Metrics Abstract CD8+ T cell recognition of Mycobacterium tuberculosis (Mtb)-specific peptides presented on major histocompatibility complex class I (MHC-I) contributes to immunity to tuberculosis (TB), but the principles that govern presentation of Mtb antigens on MHC-I are incompletely understood. In this study, mass spectrometry (MS) analysis of the MHC-I repertoire of Mtb-infected primary human macrophages reveals that substrates of Mtb’s type VII secretion systems (T7SS) are overrepresented among Mtb-derived peptides presented on MHC-I. Quantitative, targeted MS shows that ESX-1 activity is required for presentation of Mtb peptides derived from both ESX-1 substrates and ESX-5 substrates on MHC-I, consistent with a model in which proteins secreted by multiple T7SSs access a cytosolic antigen processing pathway via ESX-1-mediated phagosome permeabilization. Chemical inhibition of proteasome activity, lysosomal acidification, or cysteine cathepsin activity did not block presentation of Mtb antigens on MHC-I, suggesting involvement of other proteolytic pathways or redundancy among multiple pathways. Our study identifies Mtb antigens presented on MHC-I that could serve as targets for TB vaccines, and reveals how the activity of multiple T7SSs interacts to contribute to presentation of Mtb antigens on MHC-I. Editor's evaluation This landmark study uses compelling approaches such as quantitative and screening mass spectrometry to identify peptides from tuberculosis bacteria that are presented by macrophages infected with this pathogen. The authors provide convincing evidence that the presentation of these antigens depends on a specialist bacterial secretion system. The study will be of interest to infectious disease specialists and of particular value for future vaccine development. https://doi.org/10.7554/eLife.84070.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Tuberculosis (TB), caused by Mycobacterium tuberculosis (Mtb), is a leading cause of infectious disease mortality worldwide, causing approximately 10 million new cases of active TB disease and 1.5 million deaths per year (World Health Organization 2022, 2021). Currently, the only clinically licensed vaccine to prevent TB is Bacille Calmette-Guerin (BCG), which protects children against disseminated Mtb infection (Roy et al., 2014), but provides limited and highly variable protection against pulmonary TB in adults (Fine, 1995). More effective vaccines against TB are therefore needed, but identifying Mtb antigens capable of eliciting protective immunity remains challenging. Multiple convergent lines of evidence from experiments in mouse and non-human primate models of TB show that CD8+ T cells can contribute to immune control of Mtb infection (Chen et al., 2009; Flynn et al., 1992; Woodworth et al., 2008b), but the antigenic targets of protective CD8+ T cell immunity to Mtb infection have not been conclusively defined. In murine models, CD8+ T cells specific for some immunodominant Mtb antigens poorly recognize Mtb-infected macrophages (Yang et al., 2018), implying that infected macrophages may not present all Mtb antigens that elicit cytokine-producing CD8+ T cell responses (Kamath et al., 2004; Lewinsohn et al., 2017). These results suggest a need to directly identify which Mtb antigens are presented on MHC-I by infected phagocytes (Flynn et al., 2011). It is currently unknown which Mtb proteins are able to enter MHC-I antigen processing pathways in macrophages infected with virulent Mtb. Whereas some bacterial species are lysed following phagocytosis and expose their internal cell contents to antigen processing pathways (Shen et al., 1998), a high proportion of virulent Mtb remains intact and viable in macrophages (Armstrong and Hart, 1975; Lee et al., 2008; Lewis et al., 2003), leading us to hypothesize that only a subset of Mtb proteins may be accessible for processing and presentation on MHC-I. Here, we use MS-based identification of peptides bound to MHC-I (immunopeptidomics) to directly identify Mtb-derived peptides presented on MHC-I in primary human macrophages infected with virulent Mtb H37Rv, revealing potential targets for CD8+ T cell-mediated immunity. Additionally, we use targeted MS to quantify changes in the presentation of Mtb peptides resulting from genetic perturbations to Mtb and chemical perturbations to the host cell, allowing us to probe host and bacterial determinants of antigen presentation on MHC-I in Mtb infection. Results To identify Mtb antigens presented on MHC-I, we infected primary human monocyte-derived macrophages with Mtb H37Rv, isolated MHC-I by immunoprecipitation 72 hours post-infection, purified the associated peptides, and identified peptides by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). We selected the 72 hour time point over shorter time points to allow sufficient time for antigen processing and presentation on MHC-I. Past experience suggested that choosing a longer time point would have resulted in a high rate of cell death. For an initial set of three analyses, we adapted a previously described protocol (Stopfer et al., 2020; Protocol 1 – see Methods) for use in a biosafety level 3 (BL3) setting. We developed a further optimized protocol (Protocol 2 – see Methods) that we used for 3 additional analyses (Figure 1A). The class I human leukocyte antigen (HLA) genotypes of the 6 donors analyzed were largely distinct from one another (Table 1), enabling us to identify peptides associated with a variety of HLA alleles. Figure 1 with 8 supplements see all Download asset Open asset The MHC-I immunopeptidome of Mtb-infected human macrophages is enriched for T7SS substrates. (A) Schematic representation of two immunopeptidomics workflows used to profile the MHC-I repertoire of Mtb-infected primary human macrophages. IP: immunoprecipitation. MWCO: molecular weight cutoff. LC-MS/MS: liquid chromatography coupled to tandem mass spectrometry. (B) Absolute and relative number of Mtb-derived MHC-I peptides identified for each donor. Macrophages from donor B were pre-treated for 24 hr with 10 ng/mL IFN-γ, and macrophages from donor C were treated with 0.5 µg/mL cycloheximide (CHX) for the final 6 hr of infection. (C) Sequences, source proteins, associated HLA alleles, and donors for each validated Mtb-derived MHC-I peptide. (D) Enrichment analysis of Mtb peptides presented on MHC-I and their source proteins, categorized by protein secretion pathway using SignalP 6.0 (Teufel et al., 2022) and a curated set of known or strongly suspected T7SS substrates. p-Values for enrichment analyses of proteins and peptides were determined using the binomial test and the hypergeometric test respectively (see Methods). Figure 1—source data 1 Source data for enrichment analysis. https://cdn.elifesciences.org/articles/84070/elife-84070-fig1-data1-v2.zip Download elife-84070-fig1-data1-v2.zip Table 1 Class I HLA alleles expressed by primary monocyte donors. DonorHLA-A allele 1HLA-A allele 2HLA-B allele 1HLA-B allele 2HLA-C allele 1HLA-C allele 2ANDNDNDNDNDNDB02:0124:0207:0227:0707:0215:02C01:0102:0144:0357:0106:0216:01D23:0126:0135:0150:0104:0106:02E11:0124:0239:0651:0107:0215:02F01:0126:0135:0138:0104:0112:03 We identified thousands of peptides for each primary cell donor, with a length distribution typical of MHC-I peptides (Figure 1—figure supplement 1A) and retention times that correlated well with hydrophobicity (Figure 1—figure supplement 1B). Unsupervised clustering of identified peptides using GibbsCluster 2.0 (Andreatta et al., 2017) revealed groups corresponding to the known peptide sequence binding motifs of class I HLA alleles expressed by each donor (Figure 1—figure supplement 2), confirming the specificity of our pulldowns. Mtb-derived MHC-I peptides detected in each pulldown were predicted to bind at least one class I HLA allele expressed by the donor (Figure 1—figure supplement 3). Mtb peptides made up less than 0.1% of MHC-I peptides identified, and this proportion was not discernibly increased by pre-treating macrophages with IFN-γ or treating with cycloheximide to inhibit host protein synthesis (Figure 1B). Putative Mtb peptides that passed manual inspection of MS/MS spectra and extracted ion chromatograms (see Methods) were further validated using internal standard parallel reaction monitoring (IS-PRM, also known as SureQuant) (Gallien et al., 2015; Stopfer et al., 2021a; Figure 1—figure supplement 4). Putative Mtb peptides that had MS/MS spectra that closely matched that of a synthetic stable isotope-labeled (SIL) standard, co-eluted with the SIL standard, and were not detected in mock-infected control samples by SureQuant were considered correctly identified, authentic Mtb peptides (Figure 1—figure supplement 5). 77.85% of MASCOT identifications of putative Mtb peptides were rejected after manual data inspection, and of the remaining candidates a further 51.51% were rejected following analysis by SureQuant (Table 2), highlighting the need for rigorous validation using best practices (Fritsche et al., 2021) when identifying pathogen-derived peptides among an MHC repertoire dominated by host-derived peptides. Table 2 Putative Mtb-derived peptides that passed manual inspection of DDA MS data but failed SureQuant validation. Source proteinPeptide sequenceRv0383cAAPGRPVAPGpyrDGDRLALISVRv2303cKHPNVYLELRv0839YTHGYHESkgdAERAAAAAPRv1375EAAQSRITAGabD1AKVGASAAYPE1AAGNLRAAIHlfXIPYDRGDLVRv2807AKWILEGIKRv3818IAPELVRTRv1065YTRIHGDEELRv0293cDELIAGLAYRv3779VAIAVGPALTPPE55TVAPINLNPRv2263QEIEEGILRv0333GEDPGIAR Of the 16 Mtb-derived MHC-I peptides we identified, 13 (81.25%) derived from proteins secreted via type VII secretion systems (T7SS) (Figure 1C). The Mtb genome encodes five of these protein export machines (designated ESX-1, 2, 3, 4, and 5; Abdallah et al., 2007), and we identified MHC-I peptides derived from proteins known to be secreted by three of these systems [ESX-1 (Guinn et al., 2004; Millington et al., 2011; Stanley et al., 2003), ESX-3 (Siegrist et al., 2009; Tufariello et al., 2016), and ESX-5 (Abdallah et al., 2006; Daleke et al., 2012; Ekiert and Cox, 2014; Shah et al., 2015)]. For several T7SS substrates, we identified multiple peptides from the same protein, and/or identified the same peptide across multiple donors. These antigens included EsxA, PPE20, EspC, and PE35, as well as sequences conserved among the four nearly identical members of the EsxJ family of proteins (EsxJ, EsxK, EsxP, and EsxW – referred to here as EsxJKPW). T7SS substrates were significantly overrepresented in the MHC-I repertoire relative to the whole Mtb proteome, both at the peptide level (p<10–10; binomial test with Bonferroni correction) and at the protein level (p<10–7; hypergeometric test with Bonferroni correction; Figure 1D). Proteins without identifiable secretion signals were significantly underrepresented among Mtb-derived MHC-I peptides (p<10–8; binomial test with Bonferroni correction) and source proteins (p<10–6; hypergeometric test with Bonferroni correction). While some of the Mtb antigens we identified are highly abundant compared to the rest of the Mtb proteome (e.g. EsxB), the abundances of other T7SS substrates are near the mean (Figure 1—figure supplement 1C; Schubert et al., 2015), suggesting that the overrepresentation of T7SS substrates is not solely due to greater abundance relative to the rest of the Mtb proteome. These results suggest that T7SS substrates may preferentially gain access to MHC-I antigen processing pathways. To nominate possible mechanisms by which Mtb antigens might be processed and loaded onto MHC-I, we used confocal microscopy to examine the intracellular fate of Mtb in primary human macrophages. We hypothesized that Mtb antigens could either (1) be processed by endolysosomal proteases and loaded onto MHC-I in Mtb-containing compartments, or (2) gain access to cytosolic antigen processing pathways via permeabilization of the phagosome membrane (Grotzke et al., 2010; Grotzke et al., 2009; van der Wel et al., 2007; Figure 2A). In a subset of macrophages at both early and late timepoints, a subset of Mtb co-localized with Galectin-3, a marker of phagosomal membrane permeabilization (Chauhan et al., 2016; Watson et al., 2012; Figure 2B–C; Figure 2—figure supplement 1A). A subset of Mtb also co-localized with P62, an autophagy adaptor protein that recognizes cytosol-exposed bacteria (Watson et al., 2012; Zheng et al., 2009; Figure 2B–C; Figure 2—figure supplement 1B). Since bacteria become ubiquitinated upon exposure to the cytosol and recruitment of autophagy adaptor proteins to the Mtb phagosome has previously been associated with damage to the phagosomal membrane (Watson et al., 2012), these results suggest that Mtb gains access to the host cytosol in primary human macrophages, as has previously been shown in murine macrophages (Mittal et al., 2018) and in cell lines (Beckwith et al., 2020). A subset of Mtb-containing phagosomes co-localized with the late endosome and lysosome marker LAMP-1 (Figure 2B–C; Figure 2—figure supplement 1C), suggesting access to endolysosomal proteases, but Mtb-containing phagosomes did not co-localize with MHC-I itself (Figure 2B–C; Figure 2—figure supplement 1D), suggesting that Mtb antigens were unlikely to be loaded onto MHC-I in the Mtb-containing phagosome. These results suggested that access to the host cell cytosol represented a likely route for processing and presentation of Mtb antigens. An Mtb strain deficient in the activity of the ESX-1 secretion system (eccCa1:Tn) exhibited reduced co-localization with Galectin-3 and P62 (Figure 2D–F). Our results are consistent with prior studies demonstrating that ESX-1 activity is required for Mtb to damage the phagosome membrane (Augenstreich et al., 2017; Simeone et al., 2012; Watson et al., 2012) and show that this phenomenon occurs in our primary human macrophage infection model. Figure 2 with 1 supplement see all Download asset Open asset Mtb co-localizes with markers of phagosome membrane damage in an ESX-1-dependent manner and does not co-localize with MHC-I. Primary human macrophages were infected with GFP-expressing wild-type Mtb, fixed at 24, 72, or 120 hr post-infection, stained by immunofluorescence (IF), and imaged by spinning-disk confocal microscopy. (A) Schematic showing markers associated with each of two possible pathways of Mtb antigen processing and presentation. (B) Representative images of Mtb-infected macrophages stained for Galectin-3, P62, MHC-I, or LAMP-1. Scale bar indicates 10 µm. White arrows indicate Mtb-containing phagosomes co-localizing with each marker. (C) Automated quantification (see Methods) of the proportion of co-localized Mtb-containing phagosomes for each marker for n=4 donors (* p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001; one-way ANOVA with Tukey’s multiple comparisons test). (D) Representative images of macrophages infected with wild-type (H37Rv) or ESX-1-deficient (eccCa1:Tn) Mtb stained for Galectin-3 or P62. Scale bar indicates 10 µm. White arrows indicate Mtb-containing phagosomes co-localizing with each marker. (E–F) Automated quantification of the relative proportion of GFP +objects co-localized with IF staining for Galectin-3 (E) or P62 (F) as a function of time post-infection for n=3 donors, normalized to wild-type (H37Rv) (* p<0.05, ** p<0.01, *** p<0.001; paired t-test). Error bars indicate standard deviation. Figure 2—source data 1 Single-channel images. https://cdn.elifesciences.org/articles/84070/elife-84070-fig2-data1-v2.zip Download elife-84070-fig2-data1-v2.zip Figure 2—source data 2 Colocalization quantification source data. https://cdn.elifesciences.org/articles/84070/elife-84070-fig2-data2-v2.zip Download elife-84070-fig2-data2-v2.zip Our microscopy results led us to hypothesize that ESX-1-mediated phagosomal membrane damage might be required for Mtb antigens to access MHC-I antigen processing pathways (Figure 3A). If this were the case, ESX-1 activity would be required for presentation not only of peptides derived from ESX-1 substrates (e.g. EsxA28-36 – sequence LLDEGKQSL), but also peptides derived from substrates of other T7SSs (e.g. EsxJKPW24-34 – sequence QTVEDEARRMW – which is derived from ESX-5 substrates that do not require ESX-1 for secretion; Champion et al., 2006). To test this hypothesis, we turned to quantitative targeted MS (SureQuant) to quantify changes in the presentation of EsxA28-36 and EsxJKPW24-34 across multiple experimental conditions (Figure 3B). We used primary macrophages from donors expressing HLA-A*02:01 and HLA-B*57:01 for these experiments to ensure presentation of the target peptides. Because the other Mtb epitopes detected in our untargeted MS experiments are not expected to bind these HLA alleles, we only targeted EsxA28-36 and EsxJKPW24-34. While this targeted approach is limited in the number of epitopes it can detect, it enables reliable and accurate quantification of peptides across experimental conditions with low sample input. Because methionine residues of peptides can oxidize during sample handling, we targeted both the oxidized and non-oxidized form of EsxJKPW24-34 where possible. In addition to the SIL synthetic trigger peptides required for SureQuant, we also spiked pre-formed heavy isotope labeled peptide-MHC complexes (hipMHCs) into the lysates prior to immunoprecipitation to provide an internal standard that can be used to normalize out technical variation, which improves the accuracy of label-free quantification (Stopfer et al., 2020). Figure 3 with 1 supplement see all Download asset Open asset ESX-1 activity is required for presentation of EsxA28-36 and EsxJKPW24-34 on MHC-I, independently of type I interferon signaling. (A) Schematic representation of the localization of EsxA and EsxJKPW in macrophages infected with wild-type Mtb H37Rv or the ESX-1-deficient eccCa1:Tn transposon mutant. (B) Schematic showing our workflow for targeted detection and quantification of Mtb-derived MHC-I peptides by SureQuant, using stable isotope labeled peptide-MHC complexes (hipMHCs) as internal standards. SIL: stable isotope labeled. (C) Relative quantification of EsxA28-36 and EsxJKPW24-34 by SureQuant in macrophages infected with no Mtb (mock), wild-type Mtb H37Rv, or eccCa1:Tn for n=3 donors (all HLA-A*02:01+, HLA-B*57:01+). As oxidation of methionine is common during sample handling, both the oxidized and non-oxidized form of EsxJKPW24-34 were quantified. (D) Luminescence as a function of time measured for macrophages infected with luciferase-expressing Mtb, in a wild-type H37Rv or eccCa1:Tn background, with or without the addition of 10 ng/mL IFN-β in the culture media. Addition of 25 µg/mL rifampicin (RIF) to the culture media was used as a control showing reduced luminescence with bacterial death. Data points and error bars represent the mean and standard deviation of n=3 donors, each of which represents the mean of three technical replicates. (* p<0.05, one-way ANOVA with Dunnett’s multiple comparisons test, relative to H37Rv as the reference condition). (E) CXCL10 concentration in the culture media 72 hr post-infection quantified by ELISA. Data points each represent the mean of three technical replicates for a given donor. (* p<0.05, ** p<0.01, one-way ANOVA with Tukey’s multiple comparisons test on log-transformed concentrations.) (f) Relative quantification of EsxA28-36 and EsxJKPW24-34 by SureQuant in macrophages infected with no Mtb (mock), wild-type Mtb H37Rv, or eccCa1:Tn for n=3 donors (all HLA-A*02:01+, HLA-B*57:01+). (**** p<0.001, one-way ANOVA with Tukey’s multiple comparisons test.). Error bars indicate standard deviation. Figure 3—source data 1 Relative abundances of Mtb-derived MHC-I peptides determined by SureQuant. https://cdn.elifesciences.org/articles/84070/elife-84070-fig3-data1-v2.zip Download elife-84070-fig3-data1-v2.zip Figure 3—source data 2 CXCL10 ELISA raw data. https://cdn.elifesciences.org/articles/84070/elife-84070-fig3-data2-v2.zip Download elife-84070-fig3-data2-v2.zip Targeted MS analysis revealed that macrophages infected with ESX-1-deficient Mtb (eccCa1:Tn) did not present either EsxA28-36 or EsxJKPW24-34 on MHC-I (Figure 3c). We assessed alternative explanations for this difference in presentation of EsxA28-36 and EsxJKPW24-34 and observed no difference in Mtb outgrowth (Figure 3D), the proportion of cells infected (Figure 3—figure supplement 1A–B), or total MHC-I surface expression (Figure 3—figure supplement 1C–D) between macrophages infected with wild-type or ESX-1-deficient Mtb. These results show that ESX-1 activity is required for presentation of EsxA28-36 and EsxJKPW24-34 on MHC-I in macrophages. To determine whether the absence of EsxA28-36 and EsxJKPW24-34 in the MHC-I repertoire of macrophages infected with ESX-1-deficient Mtb could be attributed to a loss of type I interferon signaling (Stanley et al., 2007), we added exogenous IFN-β to macrophages infected with ESX-1-deficient Mtb and again quantified presentation of EsxA28-36 and EsxJKPW24-34 by SureQuant. Addition of exogenous IFN-β restored a type I interferon response as measured by production of CXCL10 (Figure 3E), but did not rescue presentation of EsxA28-36 or EsxJKPW24-34 on MHC-I (Figure 3F). These results show that presentation of EsxA28-36 or EsxJKPW24-34 on MHC-I is dependent on ESX-1 activity but independent of downstream type I interferon signaling, consistent with a model in which ESX-1-mediated phagosomal damage enables Mtb antigens to access MHC-I antigen processing pathways. Many MHC-I peptides are proteolytically processed by the proteasome (Kloetzel, 2001), while others are processed by endosomal or lysosomal proteases (Grotzke et al., 2017; Shen et al., 2004). To determine whether these mechanisms contribute to presentation of Mtb peptides on MHC-I, we treated HLA-A*02:01+, HLA-B*57:01+ macrophages with inhibitors of proteasome activity (MG-132), cysteine cathepsin activity (E64d), and lysosomal acidification (bafilomycin; Figure 4A). We quantified presentation of EsxA28-36 and EsxJKPW24-34 on MHC-I along with five host-derived HLA-A*02:01-binding peptides identified in previous studies (Stopfer et al., 2021b) that could be reliably detected in the MHC-I repertoire of macrophages. We began drug treatment of the macrophages prior to infection with Mtb and limited the duration of infection to 24 hr so that the cells could be treated with drug for the full duration of the infection without excessive cytotoxicity (Figure 4B). Treatment with MG-132 inhibited proteasome activity, as measured by accumulation of proteins modified with K48-linked polyubiquitin (Figure 4—figure supplement 1A–B). Treatment with E64d inhibited cathepsin B activity, as measured by a fluorometric assay (Figure 4—figure supplement 1C). Treatment with bafilomycin inhibited lysosomal acidification, as measured by lysotracker staining (Figure 4—figure supplement 1D–E). All three drugs exhibited minimal cytotoxicity in macrophages at the doses used in our immunopeptidomic experiments (Figure 4—figure supplement 2), did not inhibit phagocytosis of Mtb or bacterial outgrowth (Figure 4—figure supplement 3), and did not have a significant effect on surface MHC-I levels (Figure 4—figure supplement 3). Figure 4 with 4 supplements see all Download asset Open asset Inhibition of conventional antigen processing proteolytic pathways does not impair presentation of EsxA28-36 and EsxJKPW24-34 on MHC-I. (A) Schematic representation of proteolytic pathways inhibited by the proteasome inhibitor MG-132, the V-type ATPase inhibitor bafilomycin, and the cysteine cathepsin inhibitor E64d. (B) Schematic showing the timing of drug treatment and Mtb infection for targeted MS experiments. (C) Heatmap showing relative abundance of self and Mtb-derived MHC-I peptides determined by SureQuant in mock-infected macrophages or Mtb-infected macrophages treated with MG-132, bafilomycin, E64d, or DMSO-only control. Colors represent the mean fold change relative to the DMSO-treated, Mtb-infected condition for n=3 donors (all HLA-A*02:01+, HLA-B*57:01+). (D–E) Relative abundance of EsxA28-36 (D) and EsxJKPW24-34 (E) determined by SureQuant in mock-infected macrophages or Mtb-infected macrophages treated with MG-132, bafilomycin, E64d, or DMSO-only control for n=3 donors (all HLA-A*02:01+, HLA-B*57:01+). Error bars indicate standard deviation. Figure 4—source data 1 Relative abundances of MHC-I peptides determined by SureQuant. https://cdn.elifesciences.org/articles/84070/elife-84070-fig4-data1-v2.zip Download elife-84070-fig4-data1-v2.zip Treatment with MG-132 reduced presentation of MHC-I peptides derived from cytosolic or nuclear host proteins, but not peptides derived from Mtb proteins or secreted or endomembrane compartment localized host proteins (Figure 4C). Treatment with E64d decreased presentation of peptides derived from endosomal or secreted proteins and had a range of effects on peptides derived from cytosolic or nuclear proteins, but did not decrease presentation of Mtb peptides (Figure 4C). Treatment with bafilomycin broadly increased presentation of target MHC-I peptides (Figure 4C). This increase was statistically significant for EsxA28-36 (Figure 4D; p<0.05, one-way ANOVA with Dunnett’s multiple comparisons test), though not for EsxJKPW24-34 (Figure 4E). Cell surface levels of HLA-DR and HLA-DQ were not significantly affected by bafilomycin or E64d treatment (Figure 4—figure supplement 4), suggesting that E64d and bafilomycin did not substantially stall MHC-II antigen presentation and their effects of on MHC-I presentation could not be explained as indirect effects of modulating antigen entry into the MHC-II antigen processing pathway. Our results suggest that processing of Mtb peptides for presentation on MHC-I relies on antigen processing proteases other than the proteasome, cysteine cathepsins, or other acidification-dependent lysosomal proteases, and/or that multiple redundant pathways contribute to processing of Mtb peptides for presentation on MHC-I. Discussion Our analysis of the MHC-I repertoire of primary human macrophages infected with Mtb revealed that T7SS substrates are a prominent source of Mtb peptides presented on MHC-I. Our findings contrast with those of previous studies of the immunopeptidome of macrophage cell lines infected with BCG (Bettencourt et al., 2020) and the avirulent strain Mtb H37Ra (Flyer et al., 2002), in which the mycobacterial peptides identified included twin arginine translocation (Tat) pathway substates, Sec pathway secretion substrates, membrane-associated proteins, and cytosolic proteins, with no apparent enrichment for T7SS substrates. Given that we showed ESX-1 activity contributes to presentation of Mtb antigens in infection with virulent Mtb H37Rv, the fact that BCG and Mtb H37Ra are both deficient in ESX-1 activity (Frigui et al., 2008; Gordon et al., 1999; Guinn et al., 2004; Lewis et al., 2003; Pym et al., 2002) may in part explain this difference in the mycobacterial peptides presented on MHC-I. The absence of ESX-1 in BCG therefore could limit the ability of BCG to prime effective T cell responses against Mtb, not only because of the absence of antigens encoded in the ESX-1 locus but also because of altered or reduced presentation of other antigens (Pym et al., 2003). Our findings contrast with previous work arguing that ESX-1 activity is dispensable for presentation of Mtb antigens on MHC-I (Lewinsohn et al., 2006; Woodworth et al., 2008a). ESX-1 activity is not required for Mtb-infected human monocyte-derived dendritic cells to present an epitope derived from TB8.4 (a Sec pathway substrate) to cognate T cells (Lewinsohn et al., 2006). While ESX-1 activity is required for priming of CD8+ T cell responses specific for EsxB in vivo in mice, it is dispensable for priming of CD8+ T cell responses specific for antigens exported via other secretion systems (such as TB8.4 and EsxH) (Woodworth et al., 2008b). While these prior results suggest that the requirements for presentation of Mtb antigens on MHC-I may vary among Mtb antigens and/or vary among antigen presenting cell types, our results show that ESX-1 activity is essential for the presentation of certain Mtb antigens in infected human macrophages, beyond ESX-1 substrates alone. Whereas in axenic culture, secretion of ESX-5 substrates is independent of ESX-1 function (Champion et al., 2006; Shah and Briken, 2016), the fact that presentation of peptides derived from an ESX-5 substrate on MHC-I requires ESX-1 activity supports the hypothesis that the localization of secreted Mtb proteins within a host cell may depend on the activity of multiple T7SSs. This dependence on multiple T7SSs has previously been shown for other ESX-5 substrates such as CpnT (Izquierdo Lafuente et al., 2021), and a functional interdependence between ESX-1 and ESX-3 (which respectively damage the phagosome membrane and prevent host membrane repair) has previously been proposed (Mittal et al., 2018). Our results suggest that the interdependence of Mtb T7SSs in an intracellular context influences the availability of Mtb antigens for processing and presentation on MHC-I. Prior studies have shown that Mtb antigens must access the cytosol to be presented on MHC-I in human monocyte-derived dendritic cells (Grotzke et al., 2010; Grotzke et al., 2009; Lewinsohn et al., 2006), and our results are consiste