Mining the CD4 antigen repertoire for next generation tuberculosis vaccines

SUMMARY

Tuberculosis (TB) is the leading cause of death from infectious disease, and Bacillus Calmette–Guérin (BCG) remains the only clinically approved vaccine. An enduring challenge in TB vaccine development is systematic antigen selection from a large repertoire of potential candidates. We performed an efficacy screen in mice of antigens that are targets of CD4 T cells in humans. We found striking heterogeneity in protective efficacy, and most of the top protective antigens are currently not in clinical development. We observed immunologic cross-reactivity among phylogenetically clustered antigens, reflecting common CD4 epitopes. We developed a trivalent mRNA vaccine consisting of PPE20 (Rv1387), EsxG (Rv0287), and PE18 (Rv1788), which augmented and exceeded BCG protection in multiple mouse models. Finally, we observed cellular immune responses to these antigens in 84% of humans exposed to M. tuberculosis. These data advance our understanding of TB vaccine immunology and define a vaccine concept for clinical development.

Graphical Abstract

In Brief

Systematic antigen selection is an enduring challenge in tuberculosis (TB) vaccine development. An antigen screen defines protective CD4 T cell antigens and yields a novel trivalent mRNA TB vaccine that protects against TB challenges in mice.

INTRODUCTION

TB was the leading cause of death and disability from infectious disease worldwide prior to the coronavirus disease 2019 (COVID-19) pandemic¹ and has increased again in recent years². The burden of TB is particularly high in low- and middle-income countries, where TB is a primary driver of all-cause morbidity and mortality with little change in recent decades³. This global burden of disease is due in part to the lack of an adequately effective TB vaccine^4–6. Indeed, estimates suggest that BCG vaccine efficacy is high against childhood disease but declines significantly in adolescents and adults^7,8, and BCG revaccination was ineffective in three large randomized trials^9–11. The development of improved TB vaccines is therefore a global health priority. Moreover, given high rates of early childhood BCG vaccination in endemic countries, next generation TB vaccines may be administered in the context of BCG immunity¹².

Systematic antigen selection from a large repertoire of candidates is a major challenge in TB vaccine development^4–6. The Mycobacterium tuberculosis (Mtb) genome contains approximately 4,000 open reading frames (ORFs) of which over 100 have been reported to be recognized by T cells during natural infection in humans^13,14. The 6 kDa early secretory antigenic target (ESAT-6), secreted antigen 85-B (Ag85B), and MTB32A vaccine antigens were identified as immunogenic proteins present in secreted culture filtrates^15,16. Other strategies to define vaccine immunogens identified proteins expressed during chronic infection in mice¹⁷, antigens recognized by CD8 T cells in humans¹⁸, or antigens not shared with BCG¹⁹. One study screened antigens in specific functional and structural groups²⁰ that preceded the identification of CD4 antigen targets in humans²¹.

CD4 T cells are a primary adaptive immune responses mediating Mtb control, with additional potential roles for CD8 and other lymphocyte subsets^4–6. Latent TB (LTB) refers to a clinical state following Mtb exposure with adaptive immune control, as evidenced by a positive interferon gamma release assay (IGRA⁺), and LTB is likely dependent on CD4 T cell responses²². We leveraged a genome-scale dataset of CD4 T cell responses in humans with LTB²¹ and developed a pipeline for systematic in vivo vaccine antigen screening for protective efficacy in mice. We identified a series of protective TB antigens, most of which are not currently in clinical development, and we defined a trivalent mRNA-lipid nanoparticle (mRNA-LNP) vaccine concept for clinical development.

RESULTS

In vivo antigen screen

To develop an in vivo pipeline for systematically ranking initial efficacy of candidate TB antigens, we first utilized a DNA vaccine platform that was previously employed for immunogen discovery^23,24, and we screened antigens that were reported to be targets of CD4 T cells in humans with LTB (Fig. S1A)²¹. We screened the 36 top CD4 T cell antigen targets from LTB individuals. We cloned codon-optimized reference genes for these 36 antigens as well as 6 additional antigens that are currently in clinical vaccine development into DNA expression plasmids^6,25. Our screen thus included a total of 42 TB antigens (Table S1) as well as a severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Spike receptor binding domain (RBD) antigen as a negative control. We selected the CB6F1 mouse strain, a first filial generation (F1) cross between BALB/c and C57BL/6 that has increased major histocompatibility complex (MHC II) diversity¹⁷. Mice were immunized intramuscularly (IM) at weeks 0 and 4 and then received a 100 colony forming unit (CFU) H37Rv aerosol challenge at week 6. Protection was measured as fold reduction in lung bacterial load relative to sham controls at week 10.

We observed a spectrum of protective efficacy for the various tested DNA vaccines ranging from no protection to a median of 4.1-fold reduction of bacterial loads for the single antigens tested (Fig. 1A, Table S2). A DNA vaccine expressing SARS-CoV-2 Spike RBD did not detectably reduce bacterial loads (Fig. 1A). The TB antigens currently in clinical vaccine development also showed a range of protective efficacy but only represented 5 of 20 (25%) antigens with >1.5-fold protection (Fig. S1B–C). Unsupervised K-means clustering defined a group of 8 lead antigens with ≥2.5-fold protection: Rv1387, Rv1886c, Rv0287, Rv3804c, Rv0280, Rv3020c, Rv1788, and Rv0256c (Fig. S1D). A focused repeat DNA vaccine study confirmed protective efficacy for these 8 lead antigens (Fig. S1E). Except for Rv1886c and Rv3804c (also called Ag85B and Ag85A, respectively), none of the top 8 protective antigens have previously been evaluated in clinical vaccine trials⁶. Moreover, the top 8 protective antigens demonstrated greater efficacy in this study than those currently in clinical vaccine trials (P=0.0024; Fig. 1B), suggesting the potential of these novel antigens for next generation TB vaccines.

Fig. 1. Large-scale in vivo screening of LTB CD4 and selected clinical antigens.

(A) Protective efficacy of LTB CD4 and selected clinical antigens. Groups of CB6F1 mice were primed and boosted with 50 μg of DNA followed by 100 CFU H37Rv aerosol challenge, lung harvest, and bacterial load quantification. Dots represent individual mice and histograms represent the median fold reduction in lung CFU per vaccine antigen group relative to the median of an internal naïve control group. LTB CD4 antigens are shown in order of decreasing immunodominance²¹. (B) Aggregate protective efficacy of screened antigens. Dots represent the median fold reduction in lung CFU for individual antigens relative to naïve from (A). P value represents a Mann-Whitney U test. ** represents P<0.01. (C) Splenocyte Th1 responses to 8 protective antigens from (A) after 50 μg DNA prime-boost immunization in CB6F1 mice and ex vivo stimulation with overlapping peptide pools. (D) Splenocyte CD8 IFN-γ responses to 8 protective antigens from (A). (E) Boolean analysis of Th1 responses from (C). (F) Splenocyte Th1 responses to selected minimally protective antigens from (A).

To characterize the T cell responses induced by these antigens in vaccinated mice, we performed splenocyte intracellular cytokine staining (ICS) assays after ex vivo stimulation with overlapping peptide pools. All 8 protective antigens yielded robust CD4 Th1 responses with secretion of interferon gamma (IFN-γ), tumor necrosis factor (TNF), and interleukin 2 (IL-2) (Fig. S1F, Fig. 1C). In addition, we observed CD8 responses with Rv1387, Rv1886c, Rv3804c, Rv0280, and Rv0256c, but not with Rv0287, Rv3020c, or Rv1788 (Fig. 1D). Boolean analysis further demonstrated that the CD4 responses were primarily polyfunctional with secretion of ≥2 cytokines (Fig. 1E). We also observed robust T cell responses among five minimally protective antigens showing ≤1.5-fold protection (Table S2, Fig. 1F), suggesting that antigen immunogenicity is necessary but not sufficient for protective efficacy, thus confirming the importance of using protective efficacy rather than immunogenicity as a readout in our screen.

Phylogenetic and epitope analyses of protective antigens

To further investigate the characteristics of the protective antigens from our screen, we performed phylogenetic analyses as well as structural and functional annotation. The 8 top protective antigens segregated into four phylogenetic clusters with distinct structural and functional properties (Fig. 2A, Table S3). Rv1387, Rv0280, and Rv0256c formed a cluster of structurally related proline-proline-glutamic acid (PPE) family proteins implicated in various aspects of immune evasion and nutrient stress adaptation^26–28. Rv0287 and Rv3020c formed a cluster belonging to a family of type VII secretion (Esx) proteins that are critical virulence factors in mycobacteria²⁹. Rv1788 belongs to the structurally related proline-glutamic acid (PE) family proteins that share overlapping functions with the PPE family²⁶. Rv1788 also clustered with Rv1791, another PE family protein than showed 1.9-fold protection in our screen, slightly below our 2.5-fold cut-off (Fig. 1A, Table S2). Finally, Rv1886c and Rv3804c formed a cluster of related Ag85 family proteins, a group of well characterized mycolyl transferases with essential roles in Mtb cell wall homeostasis and intracellular survival³⁰.

Fig. 2. Phylogenetic and CD4 epitope mapping analysis of protective antigens.

(A) Phylogenetic analysis by Tamura-Nei genetic distance model of screened antigens showing clustering of protective outliers. (B-E) Splenocyte CD4 Th1 (IFN-γ, TNF, and IL-2) responses after 50 μg DNA prime-boost immunization in CB6F1 mice among four protective vaccine antigen phylogenetic clusters following ex vivo stimulation with overlapping peptide pool from autologous and heterologous antigen from each cluster in (A). Graph titles represent the DNA vaccine antigen and x-axes represent the stimulation peptide pools. (F-I) Selected sequence alignments of the two most immunodominant and conserved regions from the four phylogenetic clusters in (A). Black boxes represent amino acid positions with 100% sequence homology, grey boxes represent positons with partial homology, and white boxes represent positions with 0% homology. Colored bars below sequence alignments represent the position of the most immunodominant CD4 16mer peptide for each antigen region.

We hypothesized that the protective efficacy of TB vaccine antigens within each phylogenetic cluster was associated with conserved CD4 T cell epitopes. To explore this hypothesis, we immunized mice with DNA vaccines expressing individual antigens and assessed whether the resulting CD4 T cell responses could be stimulated by heterologous peptide pools from other antigens within each phylogenetic cluster. For the PPE antigens, vaccination with Rv1387 stimulated the greatest response to the autologous Rv1387 peptide pool, but we also observed cross-reactive responses to the Rv0280 and Rv0256c pools (Fig. 2B). Similar experiments with the Rv0287-Rv3020c, Rv1788-Rv1791, and Rv1886c-Rv3804c clusters showed that stimulation with either the homologous or cross-reactive peptide pools yielded responses that were cross-reactive (Fig. 2C–E). These studies demonstrated substantial cross-reactivity of CD4 T cell responses within each antigen phylogenetic cluster.

We next performed epitope mapping studies to further characterize the CD4 T cell responses to each of these antigens in mice. Given the larger size of the PPE family antigens (536–556 amino acids), we immunized mice using DNA vaccines and stimulated splenocytes ex vivo with subpools of 10 overlapping peptides (Table S4). We observed responses in 4 to 8 subpools for each antigen (Fig. S2A–C). We selected the two subpools that demonstrated both immunodominance and shared responses among the three antigens for deconvolution, and we observed overlapping regions of CD4 T cell recognition among both subpools (Fig. S2D–I, Fig. 2F). Given the small size of the Esx and PE family antigens (97–100 amino acids), we performed single-round epitope mapping experiments (Fig. S2J–M). For both antigen families, we observed two regions of overlapping CD4 reactivity (Fig. 2G–H). Finally, for the intermediate sized (325–338 amino acids) Ag85 protein family, we performed subpool analysis and observed responses in 6 to 7 subpools (Fig. S2N–O). We again selected two subpools that demonstrated immunodominance and shared responses among the two antigens for deconvolution, and we observed overlapping regions of CD4 T cell recognition among both subpools (Fig. S2P–Q, Fig. 2I). These data demonstrate significant immunologic cross-reactivity of protective antigens in each phylogenetic cluster associated with overlapping CD4 T cell recognition.

Trivalent mRNA-LNP vaccine

The mRNA-LNP vaccine platform has been shown to stimulate high magnitude CD4 T cell responses and provides potency and scalability^31–33. We therefore evaluated the immunogenicity and protective efficacy of our top protective antigens delivered as mRNA-LNP vaccines. Given the observed cross-reactivity of CD4 epitopes within each cluster (Fig. 2), we selected the most protective antigen from each of the 4 clusters (i.e. Rv1387, Rv0287, Rv1788, and Rv1886c) for further evaluation. These genes were cloned into mRNA constructs containing a 5’ cap, 5’ untranslated region (UTR), 3’ UTR, and a polyadenylation (polyA) signal, subjected to in vitro transcription (IVT) using N1-methylpseudouridine, and encapsulated in lipid nanoparticles (LNPs). Following immunization with mRNA-LNPs expressing single antigens and ex vivo splenocyte stimulation with overlapping peptide pools, we observed robust CD4 T cell responses secreting the Th1 cytokines IFN-γ, TNF, and IL-2 (Fig. 3A, Fig. S3A) and a substantial subset with secretion of ≥2 cytokines (Fig. 3B). Similar to our DNA vaccine studies, we observed CD8 T cell responses with Rv1387 and Rv1886c in these mice (Fig. 3C).

Fig. 3. T cell phenotypes and protective efficacy of CD4 antigens mRNA-LNP vaccines.

(A) Splenocyte CD4 Th1 responses after 5 μg mRNA-LNP prime-boost immunization in CB6F1 mice and ex vivo splenocyte overlapping peptide stimulation for the most protective member of each antigen family in Fig. 2. (B) Boolean analysis of Th1 responses from (A). (C) Splenocyte CD8 IFN-γ responses for the same mice from (A). (D) Comparison of the CD4 non-IFN-γ⁺/IFN-γ⁺ ratio after DNA or mRNA-LNP vaccine delivery of CD4 antigens. P values represent Mann-Whitney U tests. ** represents P<0.01. (E) Comparison of the IL-2⁺ fraction after DNA or mRNA-LNP vaccine delivery of CD4 antigens. P values represent Mann-Whitney U tests. ** represents P<0.01. (F) Protective efficacy of monovalent CD4 antigens after 5 μg mRNA-LNP prime-boost immunization in CB6F1 mice, 100 CFU H37Rv aerosol challenge, and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests for comparisons relative to the naïve group. *** represents P<0.001 and **** represents P<0.0001. (G) Comparison of the protective efficacy of eight CD4 antigens delivered with the DNA or mRNA-LNP vaccine platform. P value represents a Wilcoxon matched-pairs rank test. ** represents P<0.01. (H) Splenocyte CD4 IFN-γ⁺ response after prime-boost trivalent or tetravalent mRNA-LNP immunization (aggregate 20 μg dose divided evenly among antigens). P values represent Mann-Whitney U tests for comparisons relative to the matched antigen in the naïve group. ** represents P<0.01. (I) Protective efficacy of CD4 antigens after prime-boost trivalent or tetravalent mRNA-LNP immunization (aggregate 20 μg dose divided evenly among antigens), 100 CFU H37Rv aerosol challenge, and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests for comparisons relative to the naïve group. **** represents P<0.0001. (J) Durable protective efficacy after of 15 μg prime-boost trivalent mRNA-LNP immunization, 12 week interval, 100 CFU H37Rv aerosol challenge, and lung harvest for bacterial load quantification. P value represents Mann-Whitney U test relative to naïve. ** represents P<0.01.

An increased ratio of non-IFN-γ⁺ to IFN-γ⁺ Th1 responses was previously associated with increased TB protection in studies of both natural infection and vaccine protection^19,34. Boolean analysis showed that the ratio of non-IFN-γ⁺ to IFN-γ⁺ Th1 responses was higher for Rv1387, Rv1788, and Rv1886c when delivered as mRNA-LNP vaccines compared with DNA vaccines (Fig. 3D). This was primarily attributable to substantially higher IL-2⁺ cells with the mRNA-LNP vaccines (Fig. 3E) and to a lesser extent TNF⁺ and TNF⁺IL-2⁺ subsets (Fig. S3B–C). These phenotypic differences were associated with enhanced protective efficacy as evidenced by reductions in lung bacterial loads in the 100 CFU aerosol challenge model with mRNA-LNP vaccines expressing the single antigens Rv1387, Rv0287, Rv1788, and Rv1886c (range 4.2- to 7.2-fold reduction in lung CFU; Fig. 3F, Fig. S3D). Rv0280, Rv0256c, Rv3020c, and Rv3804c were also protective as monovalent mRNA-LNP vaccines (range 3.4- to 5.5-fold reduction in lung CFU; Fig. S3E). Overall, the reductions in lung CFU with mRNA-LNP vaccines expressing the top 8 protective antigens were greater than the corresponding reductions with DNA vaccines (P=0.0078; Fig. 3G).

To assess the protective efficacy of a multivalent cocktail of mRNA-LNP vaccines, we combined monovalent mRNA-LNP vaccines into trivalent and tetravalent cocktails while maintaining a fixed total mRNA dose. Multiplexing studies showed that vaccine cocktails maintained CD4 (Fig. S3F) and CD8 (Fig. S3G) T cell responses to all component antigens. Both trivalent (Rv1387, Rv0287, Rv1788) and tetravalent (Rv1387, Rv0287, Rv1788, Rv1886c) cocktails showed robust Th1 responses to all component vaccine antigens (Fig. 3H). Protection studies in the 100 CFU aerosol challenge model showed incremental vaccine protection with antigen cocktails (Fig. S3H), including 8.2- to 10.0-fold reductions of lung bacterial loads with the trivalent and tetravalent mRNA-LNPs (P<0.0001 for both vaccines compared with sham, P=ns for trivalent compared with tetravalent vaccine; Fig. 3I). We also evaluated the trivalent mRNA-LNP vaccine for durability and observed 10.7-fold reductions of lung bacterial loads when mice were challenged 3 months following immunization (P=0.0079, Fig. 3J). Inclusion of additional antigens from our screen did not further increase protection afforded by the trivalent vaccine (Fig. S3I). Taken together, these data show that mRNA-LNP vaccines provided enhanced immunogenicity and protective efficacy compared with DNA vaccines, and that efficacy was increased with trivalent and tetravalent antigen combinations.

Trivalent mRNA-LNP vaccine augments and exceeds BCG protective efficacy

Childhood BCG vaccination is ubiquitous in endemic regions, and thus novel TB subunit vaccines may be administered in the context of BCG immunity¹². To explore whether our trivalent mRNA-LNP vaccine would augment BCG immunity, we first performed experiments in mice primed with BCG and then boosted after 8 weeks with the trivalent mRNA-LNP vaccine (Fig. S4A). BCG alone stimulated modest CD4 T cell responses to Rv0287 and Rv1788, no CD4 T cell responses to Rv1387, and no CD8 T cell responses to these antigens in lung (Fig. 4A–B). Trivalent mRNA-LNP boosting significantly increased CD4 T cells responses to all three antigens and CD8 T cell responses to Rv1387 in lung (Fig. 4A–B). Boolean analysis showed that the lung CD4 T cell responses were primarily polyfunctional (Fig. 4C). In spleen, BCG alone induced low CD4 or CD8 T cell responses, while trivalent mRNA-LNP boosting stimulated robust CD4 T cell responses to all three antigens and CD8 T cell responses to Rv1387 (Fig. 4D–E). Importantly, mRNA boosting improved protection against 100 CFU H37Rv aerosol challenge (Fig. S4B), as evidenced by 63-fold reduction of bacterial loads with trivalent mRNA-LNP boosting compared with 14-fold reduction with BCG alone (P=0.022, Fig. 4F).

Fig. 4. Immunogenicity and protective efficacy of trivalent mRNA-LNP vaccine compared to BCG in the 100 CFU challenge model.

(A)-(B) Antigen-specific lung CD4 and CD8 IFN-γ⁺ response after BCG prime with or without delayed 15 μg trivalent mRNA-LNP prime-boost in CB6F1 mice. P values represent Mann-Whitney U tests for antigen-specific comparisons (except PPD) between the BCG and mRNA-LNP groups. ** represents P<0.01. (C) Boolean analysis of lung CD4 T cell responses from (A) in the trivalent mRNA-LNP prime-boost group. (D)-(E) Antigen-specific spleen CD4 and CD8 IFN-γ⁺ responses in the same mice from (A). P values represent Mann-Whitney U tests for antigen-specific comparisons (except PPD) between the BCG and mRNA-LNP groups. ** represents P<0.01. (F) Protective efficacy of BCG prime with or without 15 μg trivalent mRNA-LNP delayed prime-boost followed by 100 CFU H37Rv aerosol challenge and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests. * represents P<0.05 and **** represents P<0.0001. (G)-(H) Antigen-specific lung CD4 and CD8 IFN-γ⁺ response after BCG prime with or without co-administered 15 μg trivalent mRNA-LNP prime-boost in CB6F1 mice. P values represent Mann-Whitney U tests for antigen-specific comparisons (except PPD) between the BCG and mRNA-LNP groups. ** represents P<0.01. (I) Boolean analysis of lung CD4 T cell responses from (G) in the trivalent mRNA-LNP prime-boost group. (J)-(K) Antigen-specific spleen CD4 and CD8 IFN-γ⁺ responses in the same mice from (G). P values represent Mann-Whitney U tests for antigen-specific comparisons (except PPD) between the BCG and mRNA-LNP groups. ** represents P<0.01. (L) Protective efficacy of BCG prime with or without co-administered 15 μg trivalent mRNA-LNP prime-boost followed by100 CFU H37Rv aerosol challenge and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests. *** represents P<0.001 and **** represents P<0.0001.

We next explored a co-administration strategy for the trivalent mRNA-LNP and BCG vaccines given the potential clinical translatability of this approach (Fig. S4C). BCG alone stimulated modest CD4 T cell IFN-γ responses to Rv0287 and Rv1788 as in the prior experiment (Fig. 4G–H), whereas co-administration of trivalent mRNA-LNP and BCG showed significantly higher CD4 T cell responses to all three antigens and CD8 T cell responses to Rv1387 in lung (Fig. 4G–H), and CD4 responses showed robust polyfunctionality (Fig. 4I). Co-administration of trivalent mRNA-LNP and BCG also showed significantly higher CD4 T cells responses to all three antigens and CD8 T cell responses to Rv1387 in spleen (Fig. 4J–K), and there was no impact of trivalent mRNA-LNP co-administration on the immunodominant BCG antigen Rv0288 (also called TB10.4, Fig. S4D), which was not contained in the trivalent mRNA-LNP vaccine. Similar to delayed boosting, mRNA-LNP co-administration improved protection against 100 CFU H37Rv aerosol challenge (Fig. S4E), as evidenced by 73-fold reduction of bacterial loads with trivalent mRNA-LNP boosting compared with 21-fold reduction with BCG alone, P=0.002, Fig. 4L).

Mtb transmission in humans likely occurs with low bacterial doses³⁵, and our laboratory and others have shown that low dose (1–3 CFU or 1 MID50) infection in mice may better recapitulate these transmission dynamics and yield additional measures of vaccine protection including prevention of detectable infection and bilateral lung dissemination, in addition to lung bacterial load measurements^36–38. We therefore performed low dose challenge studies using BCG with or without co-administered trivalent mRNA-LNP vaccine (Fig. S5A). We also assessed pre-challenge ICS assays in peripheral blood mononuclear cells (PBMC) to assess immune correlates of protection³⁸. For the pre-challenge immunogenicity study, we performed ex vivo stimulation with a combined Rv1387, Rv0287, and Rv1788 overlapping peptide pool given cell limitations in mouse PBMC assays. The combination group exhibited greater antigen-specific CD4 T cell responses compared with BCG alone (Fig. S5B), and CD8 T cell responses were only observed in the group receiving trivalent mRNA-LNP vaccine (Fig. S5C). We also observed low but detectable Th17 responses in PBMC in the combination group (Fig. S5D). In the low dose challenge study, naïve animals showed an infection rate of 18/30 (60%)³⁶, a bilateral lung dissemination rate of 10/18 (56%), and a median infected lung bacterial load of 126,250 CFU (Fig. 5A, D–E, G). BCG trended towards a reduced infection rate of 8/20 (40%) and a bilateral dissemination rate of 1/8 (13%, Fig. 5B, D–E). The addition of the trivalent mRNA-LNP vaccine yielded a reduction in the infection rate to 5/19 (26%, P=0.039, two-sided Fisher’s exact test; Fig. 5C–D) as well as a reduction in the rate of bilateral dissemination to 0/5 (0%, P=0.046, two-sided Fisher’s exact test; Fig. 4C, E). The composite infection outcome incorporating both infection and dissemination rates was 9/40 (23%) for BCG and 5/38 (13%) for BCG with trivalent mRNA-LNP compared with 28/60 (47%) for naïve mice (P=0.020 and P=0.0008 compared to naïve mice, respectively, two-sided Fisher’s exact test; Fig. 5F). Moreover, both vaccine groups reduced median lung bacterial loads by >25,000-fold compared with naïve mice (P<0.0001, negative binomial model, Fig. 5G). Multiple CD4 T cell subset responses following vaccination correlated with bacterial loads following challenge (Fig. S5E). The strongest correlates were total IFN-γ (Spearman ρ=−0.57, P=0.0006) and TNF (ρ=−0.56, P=0.0008) CD4 T cell responses, followed by IFN-γ⁺TNF⁺IL-2⁺ triple-positive and IFN-γ⁺TNF⁺ double-positive CD4 T cell subsets.

Fig. 5. Immunogenicity and protective efficacy of trivalent mRNA-LNP vaccine compared to BCG in the low dose challenge model.

(A)-(C) Left and right lung lobe bacterial loads after BCG prime with or without 15 μg trivalent mRNA-LNP co-administration in CB6F1 mice followed by 1 MID50 H37Rv aerosol challenge and lung lobe harvest for bacterial load quantification. (D) Comparison of infection rates between the vaccine groups in (A)-(C). P value represents a two-sided Fisher’s exact test. * represents P<0.05. (E) Comparison of bilateral lung lobe dissemination rates among infected mice between the vaccine groups in (A)-(C). P value represents a two-sided Fisher’s exact test. * represents P<0.05. (F) Composite outcome of infection and dissemination between vaccine groups in (A)-(C). P values represent two-sided Fisher’s exact tests. * represents P<0.05. (G) Comparison of bacterial loads between the vaccine groups in (A)-(C). P values represent mixed effects negative binomial models for each vaccine group relative to the naïve group. **** represents P<0.0001. (H)-(J) Left and right lung lobe bacterial loads after BCG prime with or without 15 μg trivalent mRNA-LNP co-administration in C3HeB/Fe mice followed by four weekly 0.3 MID50 H37Rv aerosol challenges and lung lobe harvest for bacterial load quantification. (K)-(L) Comparison of infection and dissemination rates between the vaccine groups in (H)-(J). (M) Composite outcome of infection and dissemination between vaccine groups in (H)-(J). P values represent two-sided Fisher’s exact tests. * represents P<0.05. (N) Pre-challenge PBMC CD4 IFN-γ responses for the vaccine groups in (H)-(J) after ex vivo stimulation with an aggregate peptide pool of Rv1387, Rv0287, and Rv1788. P values represents a Mann-Whitney U tests. ** represents P<0.01 and **** represents P<0.0001.

Finally, in a repeated low dose (0.3 MID50) challenge model, we compared the protective efficacy of BCG alone, trivalent mRNA-LNP alone, and co-administration of BCG with trivalent mRNA-LNP (Fig. S5F). BCG alone resulted in 13/20 (65%) infection, 7/13 (54%) dissemination, and 20/40 (50%) composite infection rates (Fig. 5H, K–M). The trivalent mRNA-LNP vaccine alone resulted in lower 8/20 (40%) infection and 1/8 (13%) dissemination rates, and co-administration of BCG and the trivalent mRNA-LNP vaccine yielded 7/20 (35%) infection and 2/7 (29%) dissemination rates (Fig. 5I–L). Both the trivalent mRNA-LNP vaccine alone and co-administration of BCG and the trivalent mRNA-LNP vaccine demonstrated reduced 9/40 (23%) composite infection rates compared with 20/40 (50%) with BCG (P=0.019, two-sided Fisher’s exact test comparing the composite infection outcome for both mRNA-LNP groups to BCG; Fig. 5M). Moreover, both the mRNA-LNP vaccine groups elicited increased antigen-specific CD4 T cell responses compared with BCG alone (Fig. 5N). Taken together, these data show that the trivalent mRNA-LNP vaccine augmented and exceeded BCG protective immunity in standard dose and low dose challenge models, and protection correlated with antigen-specific CD4 T cell responses.

Cellular immune responses to target antigens in humans

To assess cellular immune responses to the antigens in the trivalent mRNA-LNP vaccine in humans, we performed IFN-γ enzyme-linked immunosorbent spot (ELISPOT) assays with PBMCs from participants in the Adolescent Cohort Study (ACS) in South Africa³⁹. We obtained PBMCs from 95 participants with prior Mtb exposure as evidenced by a positive IFN-γ release assay (IGRA), a measure of T cell reactivity to a combined pool of the immunodominant antigens Rv3875 and Rv3874 (also called ESAT6 and CFP10, respectively) and no active TB disease. We observed high rates of cellular immune responses to Rv1387 (73%), Rv0287 (58%), and Rv1788 (64%) in this cohort (Fig. 6A–B). The antigens Rv0125 and Rv1196 (also called MTB32A and MTB39A, respectively), which are in the M72/AS01_E subunit vaccine candidate currently in phase 3 clinical testing, showed cellular immune response rates of 27% and 67%, respectively (Fig. 6A–B). Recognition of any antigen in the trivalent mRNA-LNP vaccine (84%) was higher than any antigen in the M72/AS01_E vaccine (70%, P=0.033, two-sided Fisher’s exact test), and recognition of at least two antigens in the trivalent mRNA-LNP vaccine (68%) was markedly higher than in the M72/AS01_E vaccine (24%, P<0.00001, two-sided Fisher’s exact test, Fig. 6B). These data show high rates of cellular immune recognition to Rv1387, Rv0287, and Rv1788 in humans with exposure to Mtb.

Fig. 6. Trivalent vaccine antigen responses during natural infection in humans.

(A) IFN-γ ELISPOT responses among IGRA⁺ ACS participant PMBCs after overnight stimulation with the trivalent vaccine antigens as well as the M72/AS01_E antigens (MTB32A and MTB39A) or a combined ESAT6 and CFP10 positive control. Bottom dotted line represents assay LOD. Top gray line presents media adjusted cutoff for population reactivity calculations. (B) Calculations of population reactivity to individual antigens or antigen combinations representing the trivalent and M72/AS01_E vaccines.

DISCUSSION

Systematic antigen selection is a major challenge for TB vaccine development^4–6. In this study, we developed an in vivo screening pipeline to rank the protective efficacy of antigens that are most recognized by human CD4 T cell responses. We defined a trivalent mRNA-LNP vaccine concept with a TB antigen combination that augmented and exceeded BCG protective efficacy against infection, dissemination, and bacterial loads in multiple murine challenge models. These data provide insight into TB vaccine immunology and define a next generation TB vaccine strategy for clinical evaluation. This vaccine is currently being manufactured for Phase 1 clinical trials.

We observed substantial differences in vaccine protection among the screened antigens (Fig. 1)⁴⁰. Among the top antigens identified in this screen, we observed phylogenetic clustering and immunologic cross-reactivity among antigens associated with shared CD4 epitopes (Fig. 2). We included in the trivalent mRNA-LNP vaccine the top protective antigens that reflected different phylogenetic clusters: Rv1387 (also called PPE20), Rv0287 (also called EsxG), and Rv1788 (also called PE18). These antigens have not previously been evaluated in clinical vaccine trials, although other antigens from the PPE^19,41 and Esx¹⁹ families have been studied. Rv1387/PPE20 is an outer membrane protein involved in calcium influx²⁷ and was previously implicated as a possible protective BCG antigen in cattle⁴², although we did not observe substantial responses to this antigen after BCG vaccination in mice. Rv0287/EsxG complexes with Rv0288/EsxH to impair phagosome maturation⁴³ and was previously evaluated in mice as a monovalent protein vaccine⁴⁴. Rv1788/PE18 has unknown function and, in contrast to our studies, did not show protection as a protein vaccine with Quil-A adjuvant in mice⁴⁵, suggesting that antigen delivery platforms are critical for TB vaccine protection. Interestingly, antigens that did not show protective efficacy in our screen also were immunogenic, suggesting that immunogenicity is necessary but not sufficient for protective efficacy and that only a subset of antigens that are presented to the immune system will be protective. Certain immunodominant Mtb antigens may also act as decoys by priming T cells that fail to eliminate infected cells^46–48.

Clinical experience with mRNA-LNP vaccines during the COVID-19 pandemic showed robust induction of CD4 T cell responses^31,32,49. The flexibility and scalability of mRNA-LNP vaccines are also well suited for the development of multivalent vaccines, and a preclinical influenza vaccine containing up to twenty strains demonstrated specific responses to all strains included in the vaccine and robust heterologous protection against multiple strains⁵⁰. Moreover, recent studies of mRNA vaccines for TB have shown promising protective efficacy in mice^51–53. The M72⁵⁴, ID93⁵⁵, and H107¹⁹ protein based multivalent vaccines have also shown efficacy in preclinical models and are currently in clinical development. Our data confirm and extend these prior observations by showing that a trivalent mRNA-LNP vaccine expressing a novel antigen combination substantially improves or exceeds BCG protection against bacterial loads in lung in standard dose challenge models and against infection and dissemination rates in low dose challenge models. Taken together, our data provide new insights into the biology of potential TB vaccine antigens and define a novel TB vaccine concept for clinical evaluation.

Limitations of Study

Limitations of our study include the use of a single Mtb challenge strain and mouse strain for antigen screening. Genetic diversity in both challenge stocks and mouse strains may affect vaccine responses⁵⁶, and variations in challenge dose may affect disease outcomes^36–38. The DNA and mRNA vaccine platforms^23,24 may also not be predictive of other vaccine platforms. For example, ESAT-6 protein with a Th17-promoting adjuvant is protective¹⁹, whereas DNA and mRNA⁵² expressing ESAT-6 are minimally protective. Finally, Mtb immunity in mice does not fully represent immunity in humans⁵⁷, and thus further research in humans will be required.

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dan Barouch (dbarouch@bidmc.harvard.edu).

Materials availability

All requests for resources and reagents should be directed to the lead contact. All resources will be made available on request after completion of a Materials Transfer Agreement (MTA). This study did not generate new unique reagents and all reagents are commercial.

Data and code availability

All data are available in the main text or supplementary materials or by contacting the lead author. This manuscript does not include original code.

STAR METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mouse studies

8 week old female CB6F1/J (strain #100007) and C3HeB/FeJ (strain #000658) mice were obtained from Jackson Laboratory and housed in pathogen-free conditions at Beth Israel Deaconess Medical Center and the Harvard School of Public Health. All mouse procedures were performed in accordance with Institutional Animal Care and Use Committee (IACUC) guidance.

Clinical cohort

Human T cell responses to antigens were assessed on cryopreserved PBMCs from a subset of 95 healthy participants of the Adolescent Cohort Study (ACS), an observational epidemiological study among adolescents who attended high schools in Worcester, Western Cape Province, South Africa, a setting endemic for TB³⁹. This subset included adolescents aged 12–18 years of age who tested IGRA⁺ (Quantiferon TB-Gold in-tube test, Qiagen) and who did not develop TB during 24 months of follow-up. The study protocol was approved by the Human Research Ethics Committee of the University of Cape Town. Each adolescent provided written, informed assent and their parents or legal guardians provided written, informed consent.

METHOD DETAILS

Mouse immunizations

DNA vaccine immunizations were performed by diluting 50 μg of vaccine in 100 μL of phosphate buffered saline (PBS) followed by IM injection of 50 μL of vaccine into each of the bilateral quadriceps. Monovalent mRNA-LNP immunizations were performed by diluting 5 μg of mRNA into 100 μL of PBS followed by IM injection of 50 μL into each of the bilateral quadriceps. For mRNA-LNP iterative multiplexing studies, up to four monovalent mRNA-LNP vaccines were mixed with a uniform aggregate dose of 20 μg. The trivalent vaccine used in BCG combination strategies contained a combination of 5 μg of each monovalent vaccine for an aggregate vaccine dose of 15 μg. For BCG immunizations, an aggregate 1×10⁶ CFU of tittered vaccine strain was diluted into 200 μL of PBS followed by subcutaneous (SC) injection of 100 μL into each of the bilateral lower flanks.

Challenge and vaccine strains

H37Rv challenge strain was a gift from the Rubin laboratory (Harvard School of Public Health). BCG Pasteur vaccine strain was a gift from the Urdahl laboratory (University of Washington) and BCG Danish strain was a gift of the Martinot laboratory (Tufts University). BCG Danish was used in the repeated challenge study and BCG Pasteur was used in all other challenge studies. Both challenge and vaccine strains were grown in media consisting of Middlebrook 7H9 (BD Difco) containing 10% Middlebrook OADC (BD BBL), 0.5% glycerol (Sigma Aldrich), and 0.05% tween 80 (Sigma Aldrich). For preparation of conventional H37Rv challenge stock, H37Rv was propagated in growth media to an optical density (OD) of 0.8–1.0, frozen in growth media, and tittered. For low dose challenge studies, H37Rv culture was grown to an OD of 0.8–1.0 followed by passaging through a 5 μm filter to generate a single-cell suspension, resulting in an approximately 2-log10 reduction in titer as measured by agar outgrowth assay. For preparation of BCG vaccine stock, vaccine strain was grown in complete growth medium as above additionally supplemented with 0.05% tyloxapol (Sigma Aldrich). Cells were pelleted twice with resuspension in PBS containing 0.05% tyloxapol and then pelleted a third time with resuspension in PBS containing 0.05% tyloxapol and 15% glycerol. Cells were then passaged though a 40 μm filter followed by a 20 μm filter for clump removal, followed by storage at −80°C and titering by agar outgrowth assay.

Mtb aerosol challenge

For 100 CFU H37Rv challenge, a Glas-Col instrument was used and challenge stocks were titrated to result in a day 1 lung bacterial load of approximately 100 CFU. For single low dose challenge, the same instrument was used and singe-cell suspension challenge stocks were titrated to result in a week 4 infection rate of approximately 50–60% (1–3 CFU or 1 MID50) in accordance with the Poisson distribution as previously described³⁶. For repeated low dose challenge, we further reduced the individual challenge dose to 30% of 1 MID50 (0.3 MID50) in order yield a final infection rate of 50–60% after four weekly challenges. The same instrument and singe-cell suspension challenge stocks were used for single 1 MID50 and repeated weekly 0.3 MID50 studies. CB6F1/J mice were used for all single challenge studies and C3HeB/FeJ were used for the repeated challenge study only.

Lung bacterial load quantifications

Mice were euthanized 4 weeks following both 100 CFU challenge or low dose challenge. For 100 CFU challenge, both lung lobes were dissected en bloc whereas for low dose challenge right and left lung lobes were dissected separately. Tissues were placed into gentleMACS M Tubes (Miltenyi) containing 5 mL of PBS and mechanically dissociated using a gentleMACS Dissociator (Miltenyi) according to manufacturer’s instructions. Lysates were then plated in serial log₁₀ dilutions on 100 mm Middlebrook 7H10 plates (Hardy Diagnostics). In order to achieve an LOD of 5 CFU in low dose challenge studies, we additionally plated 1 mL of lysate onto 150 mm plates containing Middlebrook 7H10 agar (BD Difco), 10% Middlebrook OADC (BD BBL), 0.5% glycerol (Sigma Aldrich), and cycloheximide (Sigma Aldrich) at 100 mg/mL. CFU were counted after a 3 week incubation at 37°C.

DNA vaccine library and screening

Antigens demonstrating >20% CD4 T cell responses in a clinical dataset of LTB²¹ were selected for screening. Among the 39 candidate antigens, Rv1047, Rv3023c, and Rv3115 showed identical amino acid sequences. Similarly, Rv1199c and Rv2512c showed identical amino acid sequences. Therefore, these 5 genes were consolidated into 2 sequences (Table S1), resulting in 36 distinct candidate antigen sequences. We further identified 6 other antigens (Rv3619c, Rv2608, Rv0125, Rv1813c, Rv2660c, Rv3620c) which were in clinical development at the time the screen was performed in 2022⁶. Finally, a SARS-CoV-2 Wuhan Spike RBD sequence was obtained as a negative control, resulting in a total of 43 screened antigens. H37Rv reference sequences were obtained from Mycobrowser²⁵, codon-optimized for expression in Homo sapiens, and cloned into pcDNA3.1(+) expression plasmids (GeneArt, Thermo Fisher Scientific). A subset of ORFs with a valine (V) start codon were replaced with methionine (M) for mammalian expression, and no other modifications were made to the reference primary sequence. To facilitate screening of large numbers of antigens while accounting for potential variations between individual aerosol experiments, the screen was performed in groups of 20 mice distributed equally into subgroups of 4–5 animals including a naïve group in every individual aerosol challenge. To calculate fold reduction in lung CFU for each vaccinated mouse, the lung bacterial load was divided by the median bacterial load of the naïve subgroup within each individual aerosol challenge.

mRNA-LNP design and production

We created a custom mRNA expression vector with non-coding backbone elements that included the regions from the T7 promoter compatible with CleanCap^® AG-3’OMe (TriLink BioTechnologies) to the 5′ and 3′ UTR plus a 110-nucleotide polyA tail interrupted by a linker (A30LA70, 10 nucleotides). mRNA constructs were cloned for Rv1387, Rv0280, Rv0256c, Rv0287, Rv3020c, Rv1788, Rv1791, Rv1886c, and Rv3804c using the same codon-optimized sequences from our DNA vaccine studies. The DNA was linearized, purified, and in-vitro transcribed to make mRNAs. N-1-methylpseudouridine (m1Ψ−5′)-triphosphate (TriLink BioTechnologies) was used to substitute UTP to generate modified nucleoside-containing mRNAs, and capping of in vitro transcribed mRNAs was performed in a one-pot-reaction using CleanCap^® Reagent AG - 3’ OMe (Trilink BioTechnologies). mRNAs were purified by cellulose purification to obtain dsRNA-free mRNA molecules. To synthesize mRNA-LNP vaccines, mRNAs were encapsulated in LNPs using an aqueous solution of mRNA at acidic pH 4.0 mixed with an ethanolic lipid solution at a ratio of 3:1 (aqueous:ethanol) using a NanoAssemblr Ignite⁺ at a flow rate of 12 mL/min (Cytiva Lifesciences). The lipid solution consisted of ALC-0315 (Avanti), 18:1 (Δ9-Cis) PE (DOPE, Avanti), cholesterol (Avanti), and 14:0 PEG2000 PE (Avanti) at a ratio of 50:10:38.5:1.5 mol. After microfluidic mixing, mRNA-LNPs were dialyzed against PBS overnight, concentrated using Amicon ultracentrifugal filters (EMD Millipore), reconstituted using 10% sucrose, and stored at −80°C until further use. All formulations underwent quality control for particle size by dynamic light scattering (DLS) using a Zetasizer Nano (Malvern Instruments). The diameter (z-average) and polydispersity index (PDI) of the mRNA-LNPs was measured, with sizes ranging from 70–100 nm. Quantification of encapsulated mRNAs was performed using Quant-it^™ RiboGreen RNA Assay Kit (ThermoFisher), and all mRNA-LNPs were confirmed to have >95% encapsulation.

Spleen, lung, and PBMC harvesting for immunogenicity studies

For spleen harvest, mice were euthanized and spleens were transferred into R10 medium containing RPMI supplemented with 10% FBS (Gibco) and 1% penicillin-streptomycin (Fisher Scientific). Tissues were mechanically dissociated through a 100 μm filter and red blood cells were lysed with ACK Lysing Buffer (Gibco). Splenocytes were then filtered through a 30 μm filter, pelleted, and resuspended in R10. For lung harvest, mice were euthanized and lungs were thoroughly perfused with 10 mL ice cold PBS via the right ventricle. Lungs were transferred into Gentle MACS C tubes (Miltenyi) containing 4 mL of R10. This was supplemented with 5X digestion buffer consisting of R10 supplemented with 5 mg of type I collagenase (Worthington Biochemicals) and 0.5 mg of DNAse I (Sigma). Lung tissue was then mechanically dissociated using a GentlleMACS instrument according to manufacturer instructions (Miltenyi) following by passaging through a 70 μm filter. The solution then underwent red blood cells lysis with ACK Lysing Buffer (Gibco) and lung cells were then filtered out through a 30 μm filter. For PBMC harvesting, mice were bled via the submandibular route into RPMI (Gibco) containing 5% EDTA (Invitrogen) in accordance with IACUC protocols. The buffy layer containing PBMCs was isolated via Ficoll (GE Healthcare) centrifugation and transferred into R10. Following lymphocyte isolation, viable cell counts were quantified by trypan blue exclusion using a Countess 3 instrument (Thermo Fisher Scientific).

Flow cytometry

Lymphocytes were stimulated with peptide of interest (21st Century Bio) or PPD (Cedarlane) at 400 ng of peptide per test or DMSO control for 1 hour followed by GolgiStop/GolgiPlug (BD Biosciences) overlay for 6 hours at 37°C. Lymphocytes were then stained with live/dead (Aqua or NIR) and cell surface markers in MACS solution (Miltenyi) supplemented with 2% BSA (Miltenyi) and 1.5% penicillin-streptomycin (Fisher Scientific) prior to permeabilization with Cytofix/Cytoperm (BD Biosciences) and staining with intracellular markers in Perm/Wash (BD Biosciences). Cells were then fixed in 2% formaldehyde and stored at 4 °C until flow cytometry on an LSR II flow cytometer (BD Biosciences). PMBC and splenocyte cell surface markers included CD3 (clone 17A2, BD Biosciences), CD19 (clone 6D5, BioLegend), CD4 (clone RM4–5, BioLegend), CD8a (clone 53–6.7, BD Biosciences), CD44 (clone IM7, BD Biosciences), and CD62L (clone MEL-14, BioLegend). PMBC and splenocyte intracellular markers included IFN-γ (cloneXMG1.2, BioLegend), TNF (clone MP6-XT22, BioLegend), IL-2 (clone JES6–5H4, BioLegend), IL17-A (clone TC11–18H10.1, BioLegend), and IL-4 (clone 11B11, BD Biosciences). Lung cell surface markers included CD3 (clone 17A2, BD Biosciences), CD4 (clone GK1.5, BD Biosciences), CD8a (clone 53–6.7, BD Biosciences), CD44 (clone IM7 BD, Biosciences), CD62L (clone MEL-14, BioLegend), TCRg/d (clone GL3, BioLegend), CD45 (clone 30-F11, BioLegend), CD103 (clone 2E.7, BioLegend), PD-1 (clone 29F.1A12, BioLegend), CD11b (clone M1/70, BioLegend), and NK-1.1 (clone PK136, BD Biosciences). Lung intracellular markers included IFN-γ (clone XMG1.2, BioLegend), TNF (clone MP6-XT22, BioLegend), IL-2 (clone JES6–5H4, BioLegend), IL-17A (clone TC11–18H10.1, BioLegend), IL-4 (clone 11B11, BD Biosciences), and CD69 (clone H1.2P3, BD Biosciences).

Human ELISPOT assay

ELISPOT plates were coated with mouse anti-human IFN-γ monoclonal antibody from Mabtech at 1 μg per well and incubated overnight at 4°C. Plates were washed with DPBS and blocked with R10 medium for 2 to 4 hours at 37°C. Mtb peptide pools from Rv1387, Rv0287, Rv1788, Rv0125 (MTB32A), Rv1196 (MTB39A), and combined Rv3875 (ESAT6) and Rv3874 (CFP10, all GenScript) were prepared and plated at a concentration of 2 μg per well, and 100,000 cells per well were added to the plate. The peptides and cells were incubated for 15 to 20 hours at 37°C. All steps after this incubation were performed at room temperature. The plates were washed with ELISPOT wash buffer and incubated for 2 to 4 hours with biotinylated mouse anti-human IFN-γ monoclonal antibody at 1 μg/mL (Mabtech). The plates were washed a second time and incubated for 2 to 3 hours with conjugated goat anti-biotin alkaline phosphatase from Rockland Inc. (1.33 μg/mL). The final wash was followed by the addition of Nitro blue tetrazolium chloride/5-bromo-4-chloro 3 indolyphosphate p-toludine salt (NBT/BCIP, chromagen) substrate solution for 7 minutes. The chromagen was discarded, and the plates were washed with water and dried for 24 hours. Plates were scanned and automatically counted on a Cellular Technologies Limited Immunospot Analyzer.

QUANTIFICATION AND STATISTICAL ANALYSIS

Flow cytometry data were analyzed using FlowJo 10.10.0 software. Statistical comparisons between groups for immunogenicity and 100 CFU challenge studies were performed using GraphPad Prism 9.4.0 software. For determination of fold change cutoff for lead antigens from the screen, we first estimated the optimal number of clusters in the data set using K-means clustering, which locates k centers that minimize within-cluster variance and defines each center as the mean of the data points in that cluster. The cutoffs between clusters were then determined mathematically as midpoints between adjacent cluster centers, and the resulting cutoff for lead antigens was 2.5. For low dose challenge studies, between-group differences in both infection and dissemination rates were calculated by Fisher’s exact test. Composite infection outcomes incorporated both infection and dissemination rates expressed as a percent of total infected lung lobes. Between-group differences in composite infection outcomes were calculated by Fisher’s exact test. Between-group differences in bacterial loads for low dose challenge studies were analyzed using mixed effects negative binomial regression models while controlling for lobe side (right vs. left). Heatmaps were generated using the R package pheatmap. The correlation of CD4 T cell phenotypes with lung CFU was performed using the R package corrplot and Spearman’s method. Statistical evaluation was assessed using a t-test distribution implemented in the R cor.test function. For ELISPOT population coverage calculations, a cutoff of 30.3 spots per 10⁶ PBMC was used based on a standard 50 spot cutoff subtracted by the media condition spot mean of 19.7

Supplementary Material

1

Fig. S2. CD4 subpool and peptide reactivity among protective CD4 antigens, related to Fig. 2. (A)-(C) 10-peptide subpool CD4 T cell reactivity analysis of the high molecular weight Rv1387-Rv0280-Rv0256c protective antigen phylogenetic cluster in CB6F1 mice. (D)-(F) Single peptide deconvolution for subpool 1 from the Rv1387-Rv0280-Rv0256c protective antigen cluster. (G)-(I) Single peptide deconvolution for subpool 3 from the Rv1387-Rv0280-Rv0256c protective antigen cluster. (J)-(K) Single peptide CD4 T cell reactivity analysis of the low molecular weight Rv0287-Rv3020c protective antigen phylogenetic cluster. (L)-(M) Single peptide CD4 T cell reactivity analysis of the low molecular weight Rv1788-Rv1791 protective antigen phylogenetic cluster. (N)-(O) 5-peptide subpool CD4 T cell reactivity analysis of the high molecular weight Rv1886c-Rv3804c protective antigen phylogenetic cluster. (P)-(Q) Single peptide deconvolution for subpools 6 and 13 from the Rv1886c-Rv3804c protective antigen cluster. For all panels, pool stimulation refers to overlapping peptides spanning the entire matching antigen.

2

Fig. S3. Immunogenicity and protective efficacy of CD4 antigens delivered as mRNA-LNP vaccines, related to Fig. 3. (A) Experimental design for splenocyte ICS studies with monovalent mRNA-LNP vaccine studies with selected CD4 antigens. (B) Comparison of the TNF⁺ fraction after prime-boost DNA or mRNA-LNP vaccine delivery of CD4 antigens in CB6F1 mice. P values represent Mann-Whitney U tests. ** represents P<0.01. (C) Comparison of the TNF⁺IL-2⁺ fraction after DNA or mRNA-LNP vaccine delivery of CD4 antigens. P values represent Mann-Whitney U tests. ** represents P<0.01. (D) Experimental design for 100 CFU aerosol challenge studies with monovalent mRNA-LNP vaccine studies with selected CD4 antigens. (E) Protective efficacy of CD4 antigens after 5 μg mRNA-LNP prime-boost immunization in CB6F1 mice, 100 CFU H37Rv aerosol challenge, and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests for comparisons relative to the naïve group. * represents P<0.05 and ** represents P<0.01. (F) Splenocyte CD4 IFN-γ⁺ responses after prime-boost mRNA-LNP immunization (aggregate 20 μg dose divided evenly among antigens for all groups) in CB6F1 mice. P values represent Mann-Whitney U tests for comparisons relative to the matched antigen in the naïve group. ** represents P<0.01. (G) Splenocyte CD8 IFN-γ⁺ response in the same mice from (F). P values represent Mann-Whitney U tests for comparisons relative to the matched antigen in the naïve group. ** represents P<0.01. (H, I) Protective efficacy of CD4 antigens after prime-boost mRNA-LNP immunization (aggregate 20 μg dose divided evenly among antigens for all groups), 100 CFU H37Rv aerosol challenge, and lung harvest for bacterial load quantification. P values represent Mann-Whitney U tests for comparisons relative to the naïve group. ** represents P<0.01 and **** represents P<0.0001.

3

Fig. S4. Immunogenicity and protective efficacy of trivalent mRNA-LNP vaccine compared to BCG in the 100 CFU challenge model, related to Fig. 4. (A) Experimental design for splenocyte ICS studies after BCG prime with or without 15 μg trivalent mRNA-LNP delayed prime-boost immunization. (B) Experimental design for 100 CFU (or 100 MID50) aerosol challenge after BCG prime with or without 15 μg trivalent mRNA-LNP delayed prime-boost immunization. (C) Experimental design for splenocyte ICS studies after BCG prime with or without co-administered 15 μg trivalent mRNA-LNP prime-boost immunization. (D) Splenocyte CD4 T cell responses to the immunodominant BCG antigen Rv0288 after the immunization scheme in (C). (E) Experimental design for 100 MID50 aerosol challenge after BCG prime with or without co-administered 15 μg trivalent mRNA-LNP prime-boost immunization.

4

Fig S5. Immunogenicity and protective efficacy of trivalent mRNA-LNP vaccine compared to BCG in the low dose challenge model, related to Fig. 5. (A) Experimental design for single low dose (1–3 CFU, or 1 MID50) aerosol challenge after BCG prime with or without 15 μg trivalent mRNA-LNP prime-boost immunization and pre-challenge PBMC ICS for correlates of protection in CB6F1 mice. (B) PBMC CD4 IFN-γ responses after the immunization scheme in (A) and ex vivo stimulation with an aggregate peptide pool of Rv1387, Rv0287, and Rv1788. P value represents a Mann-Whitney U test. **** represents P<0.0001. (C) PBMC CD8 IFN-γ responses after the immunization scheme in (A) and ex vivo stimulation with an aggregate peptide pool of Rv1387, Rv0287, and Rv1788. P value represents a Mann-Whitney U test. **** represents P<0.0001. (D) PBMC CD4 IL-17 responses after the immunization scheme in (A) and ex vivo stimulation with an aggregate peptide pool of Rv1387, Rv0287, and Rv1788. P values represent Mann-Whitney U tests. * represents P<0.05 and **** represents P<0.0001. (E) Correlogram showing Spearman correlations between post-immunization, pre-challenge PBMC CD4 T cell subsets and post-challenge bacterial loads among infected animals after single 1 MID50 aerosol. (F) Experimental design for repeated 0.3 MID50 challenge studies after BCG prime with or without 15 μg trivalent mRNA-LNP prime-boost immunization in C3HeB/Fe mice.

5

Fig. S1. In vivo screening pipeline and results, confirmatory DNA vaccine screen, and flow cytometry gating strategy, related to Fig. 1. (A) Schematic of in vivo screening pipeline strategy. 42 codon-optimized ORFs were cloned in pcDNA3.1(+) mammalian expression plasmids followed by 50 μg prime-boost immunization in CB6F1 mice, 100 CFU H37Rv aerosol challenge, lung harvest, and bacterial load quantification by agar outgrowth assay at the indicated timepoints. (B)-(C) Summary fold change data from the DNA vaccine screen in mice. Dots represent the median protection for individual antigens. Red dots denote antigens that have been tested in clinical vaccine studies. (D) Unsupervised K-means clustering highlighting 8 lead antigens with ≥2.5-fold protection. (E) Confirmatory DNA vaccine screen of eight protective outliers with ≥2.5-fold bacterial load reduction in the primary screen. Groups of mice underwent the same immunization, challenge and harvest scheme as in (A). P values represent Mann Whitney U tests. * represents P<0.05, ** represents P<0.01, *** represents P<0.001, and **** represents P<0.0001. (F) Flow cytometry gating strategy used for ex vivo splenocyte ICS analysis throughout the study.

Document S1. Tables S1–S4

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CD3	BD Biosciences	17A2
CD19	BioLegend	6D5
CD4	BioLegend	RM4-5
CD4	BD Biosciences	GK1.5
CD8a	BD Biosciences	53-6.7
CD44	BD Biosciences	IM7
CD62L	BioLegend	MEL-14
IFN-γ	BioLegend	XMG1.2
TNF-α	BioLegend	MP6-XT22
IL-2	BioLegend	JES6-5H4
IL-17A	BioLegend	TC11-18H10.1
IL-4	BD Biosciences	11B11
TCRg/d	BioLegend	GL3
CD45	BioLegend	30-F11
CD103	BioLegend	2E.7
CD11b	BioLegend	M1/70
NK-1.1	BD Biosciences	PK136
CD69	BD Biosciences	H1.2P3
Bacterial and virus strains
H37Rv	Eric Rubin laboratory	N/A
BCG-Pasteur	Kevin Urdahl laboratory	N/A
BCG-SSI	Amanda Martinot laboratory	N/A
Biological samples
Human PBMC	SATVI ACS cohort	N/A
Chemicals, peptides, and recombinant proteins
Fetal bovine serum	Omega Scientific	FB-01
16mers overlapping by 11	21st Century Biochemicals	N/A
Critical commercial assays
(none)
Deposited data
(none)
Experimental models: Cell lines
(none)
Experimental models: Organisms/strains
CB6F1 mouse strain	Jackson Laboratory	100007
C3HeB/Fe mouse strain	Jackson Laboratory	000658
Oligonucleotides
(none)
Recombinant DNA
pcDNA3.1(+) expression plasmid	GeneArt, Thermo Fisher Scientific	N/A
Software and algorithms
FlowJo	FlowJo	10.10.0
Prism	GraphPad	9.4.0
R	R Core Team	4.4.2
Other
(none)

Highlights.

• An antigen screen in mice defines protective CD4 T cell antigens

• CD4 T cell antigens show immunologic cross-reactivity reflecting shared epitopes

• Antigen delivery with mRNA-LNPs results in enhanced protection

• A trivalent mRNA-LNP vaccine augments and exceeds BCG in mouse models