We isolated CD4⁺ T cells from lung tissue of B6 and I/St mice using anti-CD4 magnetic beads, with average purity of 95.2% as assessed by flow cytometry ( Supplementary Figure S2 ; Supplementary Table S1 ). We assessed enriched gene pathways using Gene Set Enrichment Analysis (GSEA) incorporating gene sets from both M5 GO:biological process and M2 Wikipathways as these platforms are based on distinct gene sets. Up-regulated genes in B6 mice exhibited enrichment in the biological processes associated with Type II interferon (IFN-γ) and IL-17A signaling pathways as well as genes associated with humoral response ( Supplementary Figures S1C, D ). Transcriptomic analysis revealed upregulation of several genes associated with a Th2 signature [e.g., Icos, Cxcr4, Ikaros (Ikzf1)] in I/St mice compared to B6 ( Figure 1 ), although a subset of Th2-associated cytokine genes (Il4, Il5, Il6, Il10) was more highly expressed in B6 CD4⁺ cells. We confirmed changes in the expression of Cxcr4, Icos, Ikzf1, Ifnγ, Il5, and Il4 in independent groups of mice by quantitative RT-PCR ( Supplementary Figure S3 ). We also observed lower expression of several genes related with the regulatory T (T_reg) cell transcriptomic program in I/St relative to B6, including Foxp3, Il10, Hmcn1, Ikzf2, and Sema3g ( Figures 1A, B ) (24). Interestingly, expression of IL-16, a CD4 ligand, was increased in I/St CD4⁺ T cells compared to B6 ( Figure 1B ), and this was confirmed by flow cytometry in independent experiment ( Supplementary Figure S4 ). This cytokine has been reported to stimulate innate immune cells for production of proinflammatory cytokines such as tumor necrosis factor (TNF) and IL-6, and to act as a chemoattractant for eosinophils, monocytes, and T cells (25).

Genes linked to Th17 and Th1 cell profiles were heterogeneously expressed between B6 and I/St CD4⁺ T cells. Notably, Ifnγ expression was higher in TB-resistant B6 mice compared to I/St, that coincided with previous measurements of IFN-γ (19) and correlated with prior observations indicating a paramount role of IFN-γ-secreting CD4⁺ T cells in mounting an effective immune response against TB (26).

Interactions between antigen-MHC-II complexes and maturing CD4⁺ T cells during thymic and peripheral selection apparently shape TCR repertoire features at the level of naive T cells (20, 27, 28). This can subsequently alter the adaptive immune response against Mtb antigens in corresponding MHC contexts. We examined differences between lung-infiltrating CD4⁺ T cell repertoires of I/St mice carrying the MHC-II H2-A^j and H2-E^j allelic variants and B6 mice carrying only H2-A^b .

We observed an opposite trend in reciprocal shifting of physicochemical characteristics in CDR3α and CDR3β repertoires between I/St and B6 mice ( Supplementary Figure S5 ). Thus, the average value for the ‘strength’ of the interaction of CDR3β with pMHC was higher in I/St mice than in B6 mice, whereas for CDR3α, this parameter was the opposite ( Supplementary Figure S5С ). Strength is determined based on the frequency of strongly binding amino acid residues in the middle of CDR3. The comparison of the physicochemical characteristics of activated CD4⁺ T cell repertoire relative to the repertoires of B6 H2-A^b and H2-A^j naive CD4⁺ T cells (20) showed different pattern for CDR3α and CDR3β repertoires of I/St. This observation suggests the possibility of an influential role of H2-E^j in shaping the repertoire of lung-infiltrating CD4+ T cells.

The normalized Shannon-Wiener index was significantly lower for TCRα and TCRβ repertoires of I/St mice compared to B6 ( Figure 1C ), and lung-infiltrating CD4⁺ T cells of I/St mice also tended to have lower TCR repertoire diversity compared to B6 mice ( Figure 1D ). The diversity was assessed using Chao1 which takes into account mainly singletons and doubletons while normalized Shannon-Wiener index assesses evenness of clonotype size distribution in repertoire. These results suggest accumulation of expanded T cell clones in TB-susceptible animals. At the same time, convergence was prominently higher in B6 repertoire than in I/St ( Figure 1E ). The combination of these parameters indicated that more clonotypes with different nucleotide sequences had identical sequences at the amino acid level in B6 mice. Notably, the repertoire of lung-infiltrating CD4⁺ T cells exhibited a significantly higher degree of pairwise overlap between B6 mice compared to that observed between I/St mice ( Figures 1E, F ), indicating a convergent antigen-specific response in the former.

To identify TCRs responding to Mtb antigens, we generated Mtb-specific T cell clones in vitro from isolated lymph nodes of B6 and I/St mice that were collected at least 21 days post-immunization with mycobacterial sonicate. After isolation, these T cells were co-cultured with the same mycobacterial sonicate followed by four cycles of expansion in the presence of splenic APC in vitro (see details in SI, Supplementary Figure S6 ). TCRβ repertoires from these antigen-expanded T cells contained 1,504 and 1,114 nucleotide clonotypes for B6 and I/St repertoires, respectively.

In order to identify TCRβ variants potentially selected for recognition to Mtb, we applied the ALICE algorithm, which allowed us to capture convergent TCR clonotypes involved in the immune response to specific antigens (29). We pooled the 950 largest clonotypes of B6 lung-infiltrating CD4⁺ T cells, I/St lung-infiltrating CD4⁺ T cells, antigen-specifically expanded B6 T cells and I/St T cells. The ALICE algorithm assumes that an ongoing immune response is typically driven by groups of convergent T cell clones with highly similar TCR sequences that expand upon antigen recall (30). ALICE further employs a V(D)J-recombination model to account for the presence of non-specific groups of frequently-generated TCR variants (31). ALICE identified 115 TCRβ cluster-associated clonotype hits in B6 mice, 80 hits in I/St mice, 158 hits in antigen-specifically expanded B6 T cells, and 48 hits in antigen-specifically expanded I/St T cells ( Figure 1G ). We next pooled ALICE hits obtained from each of the subsets and built clusters of TCRβ variants with a 1-amino acid (a.a.) mismatch allowed. Strikingly, the convergent clusters that included clonotypes from at least two mice almost exclusively comprised B6 hits. Of the 10 largest clusters, six belonged to B6 mice, and four of those (clusters 1, 5, 6, and 9) included clonotypes from the repertoires of antigen-specifically expanded B6 T cells, confirming their specificity to Mtb antigens. By contrast, only one I/St cluster (cluster 7) from a single mouse was observed in the top 10, and this was not confirmed by data from antigen-expanded I/St T cells. In general, TCRβ clonotypes from antigen-expanded I/St T cells did not overlap with the I/St clusters identified by ALICE in TCR repertoires of lung-infiltrating CD4⁺ T cells. These results clearly show a convergent, antigen-specific, public CD4⁺ T cell response in B6 but not in I/St mice.

As an additional test, we used a TCR neighbor enrichment test (TCRNET) (32) to extract homologous TCR sequences that are abundant in B6 CD4⁺ repertoires compared to I/St and vice versa ( Figure 1H ). We found more B6-enriched clusters than clusters originating from the I/St background, which is in line with our previous observations of high levels of convergence and overlap in B6 repertoires ( Figures 1E, F ; Supplementary Figure S7 ). The intersection of cluster-associated TCRβ clonotypes identified by ALICE and TCRNET showed 22 shared clonotypes in B6 repertoires, which included the members of the two largest B6 clusters (clusters 1 and 2) and only two shared clonotypes from I/St repertoires. Altogether, our cluster analysis suggests that B6 mice repeatedly produce lung infiltrating, Mtb-specific T cell clones of particular specificities that may allow them to control infection more efficiently, while the I/St response is far less focused.

We isolated B cells from the lung tissue of each mouse with anti-CD19 magnetic beads. Flow cytometry analysis showed that the average purity of isolated B cells was 87% ( Supplementary Figures S2C, D ; Supplementary Table S1 ). We then performed bulk transcriptome profiling with extracted RNA from these cells. Using MiXCR (33), we extracted IGH CDR3 (BCR) repertoires from the RNA-seq data and assessed counts and fractions of isotypes in the samples from each mouse group. We found that expression levels of IgG and IgM were higher in transcriptomic data from I/St mice compared to B6 ( Figure 2A ) suggesting high proportion of Ig-producing cells. At the same time, the fraction of CD19⁺ B cells among all lung-infiltrating lymphocytes was lower in I/St mice ( Supplementary Figure S8 ). Notably, IgG was the dominant Ig isotype in I/St repertoires, while IgA prevailed in the extracted repertoires of B6 mice ( Figures 2B, C ). IgG2 was the major IgG subclass in both strains ( Figure 2D ). Although the constant region of the IgG2 heavy chain is encoded by Ighg2c in B6 mice and by Ighg2a in I/St mice, the IgG2c and IgG2a proteins are functionally similar in mice and both are similar to human cytotoxic IgG1 (34, 35).

We observed a higher IgG1 fraction in the IGHG repertoire of I/St mice compared to B6 mice ( Figure 2D ), and this correlated with the measurement of serum Mtb-specific IgG1 concentrations using ELISA ( Supplementary Figure S9 ). Of note, the murine IgG1 isotype is non-cytotoxic and similar to human IgG4 (36).

Consistent with our transcriptomic data, we detected notably higher secretion of antigen-specific IgM in serum from I/St mice relative to B6, which might indicate a delayed or distracted humoral response to Mtb ( Figure 2 ; Supplementary Figure S9 ).

Focusing on the bulk gene expression profile of CD19⁺ B cells, we identified dominant transcriptomic signatures of plasmablasts versus memory B cells in I/St mice, which coincided with the high IGH counts we observed ( Figure 2A ). This suggested more earlier humoral immune response in the infected B6 mice than in I/St. Similar to our observations from CD4⁺ T cells, we found that expression of IFN-γ-inducible genes (e.g., Ly6a, Ly6e) was significantly decreased in CD19⁺ B cells from I/St mice, as was Ifnγ expression ( Supplementary Figure S10 ). We also detected increased expression of Cxcr5 and Cxcr4 in I/St mice compared to B6 ( Supplementary Figure S10 ). These genes encode receptor molecules for Cxcl13 and Cxcl12 chemokines, which are critical for B cell migration and the formation of B cell follicles and granulomata in TB (37, 38). Increased Cxcr4 expression by B cells is indicative of enhanced migration of circulating B cells to pleural tissue (39). Altogether, our analysis of B cells in B6 and I/St mice revealed prominent inter-strain differences in gene expression profiles, reflecting a distinct mode of B cell response to Mtb in susceptible and resistant strains.

We further assessed the number of unique CDR3 nucleotide sequences that encoded identical CDR3 amino acid sequences. In case of immunoglobulin profiling, this metrics reflects activity of somatic hypermutation. Notably, this metrics was significantly higher in IgA but not in IgG2 repertoires of B6 versus I/St mice ( Figures 3A, B ). Сlustering analysis revealed the formation of large, high-density IGH clusters (essentially representing B cell lineages) in B6 repertoires that predominantly incorporated IgA clonotypes ( Figure 3C ). In contrast, the IGH clonotypes of I/St mice formed significantly smaller clusters that consisted mainly of the IgM isotype ( Figures 3D–G ). These data indicate antigen-driven IgA evolution of the BCR repertoire in B6 mice, whereas B cells from I/St mice were most likely unable to generate a focused and protective immune response. Furthermore, the prevalence of IgM clusters might indicate a generally postponed immune response in I/St ( Figure 3D ).

10–12-week-old female mice from the I/StSnEgYCit (I/St) and C57BL/6JCit (B6) strains were used in this study. The animals were bred according to the guidelines from the Russian Ministry of Health #755 and US Office of Laboratory Animal Welfare (OLAW) Assurance #A5502-11. The Animal Facility of the Central Tuberculosis Research Institute (Moscow, Russia) provided water and food ad libitum under conventional breeding conditions. All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC), protocols #1 and 3, on March 1, 2021.

Mice were infected with ~10² CFU of the virulent Mtb strain H37Rv (substrain Pasteur) using the Inhalation Exposure System (Glas-Col, Terre Haute, IN) according to a previously described protocol (56). At eight weeks post-infection, mice were euthanized by injection of thiopental (Biochemie GmbH, Vienna, Austria).

Single-cell suspension samples were acquired separately from each mouse. Blood vessels were washed out by perfusion with 0.02% EDTA-PBS solution introduced through the right ventricle. The lungs were removed and sliced into 1–2-mm³ pieces. These were incubated in supplemented RPMI-1640 containing 200 U/ml collagenase and 50 U/ml DNase-I (Sigma-Aldrich, St Louis, MO) at 37°C for 90 min. The suspensions were washed three times in HBSS (Paneco, Russia) containing 2% fetal calf serum (FBS, General Electric, Boston, MA, USA) and antibiotics (Sigma, St. Louis, MO, USA). Next, the cells (50 x 10⁶ per dish) were incubated in 10 ml of medium (RPMI-1640, 10% FCS, 10 mM HEPES, 2 mM l-glutamine and antibiotics) in 90-mm-diameter cell culture-treated Petri dishes for 1 h at 37°C to achieve cell adhesion. Three repeated rounds of vigorous washing with warm antibiotic-free HBSS containing 2% FCS were performed to collect nonadherent cells. Two additional rounds of washing under the same conditions were then performed to get rid of residual medium. Cells then were resuspended in PBS, calculated, and proceeded for FACS analysis and cell sorting.

Starting with 40 x 10⁶ cells per lung suspension sample, we respectively isolated B and T cells using anti-CD19 (Mojo, Bio-Legend) and anti-CD4 (MACS, Miltenyi Biotec) magnetic bead kits according to the manufacturers’ instructions. Sample purity was evaluated by flow cytometry. After sorting, cells were deposited in 350 ul RLT buffer (Qiagen) for cell lysis, RNA protection, and storage according to manufacture manual.

Single-cell lung suspensions were analyzed by flow cytometry using FACS Cantu II machines (BD Biosciences). Detailed information on the procedure used for staining cells with antibodies can be found in Supplementary Materials .

RNA was isolated with the RNeasy Micro Kit (Qiagen), and the entire sample was used for the synthesis of first-strand cDNA with a mouse TCR profiling kit (MiLaboratories) according to the manufacturer’s protocol (see details in Supplementary Materials ).

RNA was isolated with the RNeasy Micro Kit (Qiagen) from 5 x10⁵ cells. The SMART-Seq v4 Ultra Low Input RNA Kit (Takara) was used to prepare T and B cell cDNA libraries according to the manufacturer’s instructions. We used 5–10 ng of RNA per cDNA synthesis reaction, followed by tagmentation using the Nextera XT DNA library preparation kit (Illumina) and 10 cycles of PCR amplification according to manufacture recommendations. The amplified cDNA libraries were validated using the Agilent 2100 Bioanalyzer, pulled and purified using AMPure XP beads (Beckman Coulter). Final concentration of pulled cDNA library was measured with the QuBit dsDNA HS kit (ThermoFisher Scientific). Samples was sequenced on Illumina MiSeq and NextSeq 550, using 150 + 150 nt paired-end mode.

The sequencing reads were adapter- and quality-trimmed, with a light quality cutoff of 20 Phred. For processed RNA-seq data, principal component analysis (PCA) and other analyses were used to check for batch effects and outliers. For adapter and quality trimming, fastp was used (57). FastQC and MultiQC were used to visualize raw and trimmed fastq files to access the read counts and quality metrics.

Raw RNA-seq data were aligned to the mouse genome using STAR. The FeatureCount tool was used to summarize the resulting bam files into raw read counts for QC (58). The R package DESeq2 was used for raw read count normalization (59). This tool was also used to perform differential gene expression analysis for B6- and I/St-derived T and B cells. Differentially-expressed genes were identified based on a fold-change of > 0.5 or < -0.5 and a false-discovery rate (FDR)-corrected p-value of < 0.1. All parameters were scaled using Z-score normalization. Visualization was performed using the programming language R and the ggplot software package.

Raw data analysis was performed with MiGEC (60), and using MiXCR (33). BCR heavy-chain repertoires were aligned to mouse BCR genes and assembled directly from raw RNA-seq data using MiXCR software (33). R was used to visualize the data and conduct statistical analysis using the Student’s t-test. The Benjamini-Hochberg (BH) procedure was used to correct all p-values for multiple comparisons with false discovery (at the rate ≤ 0.05). Clusterization of IGH chains and calculation of convergence (nucleotide per amino acid CDR3 sequence variants), Shannon–Wiener index, and diversity metrics were performed using in-house python script. Visualization of IGH clusters was performed using Cytoscape software.

All sequencing datasets have been deposited in the Sequence Read Archive (SRA) under BioProject accession number PRJNA1001295.

Neighborhood Enrichment Test (TCR-NET) analysis allowed us to find enriched CDR3 motifs that may encode antigen specificity. This pipeline is described in (61). Briefly, we pooled clonotypes assigned to TCRβ for each separate group of mice. We further calculated the number of equal TCRs that had identical or one amino acid substitution in CDR3 regions (i.e., ‘amino acid neighbors’) in each pooled sample group. We selected clonotypes with increased clonotype numbers based on the following threshold: log10-normalized degree of connectivity fold-change > 0.5 and FDR-adjusted p < 0.005. We then constructed graphs based on these data in which nodes were considered to be connected in cases where the Hamming distance between “identical” amino acid sequences was equal to 1.

A similar search of CDR3 variants potentially responding to Mtb antigens was done with ALICE, which relies on the assumption that the antigen response simultaneously expands different clonotypes with similar CDR3 sequences (29). For each clonotype, ALICE assesses the number of amino acid neighbor clonotypes and compares it with the number expected from a neutral model of V(D)J recombination probabilities to account for the presence of non-specific groups of frequently-generated CDR3 variants (62). Clonotypes with significantly enriched neighborhood are considered as potentially responding to antigens. We applied ALICE as implemented in the package tcrgrapher (https://github.com/KseniaMIPT/tcrgrapher) to the 950 largest lung-infiltrating CD4⁺ T cell clonotypes from three B6 mice and three I/St mice (separately to each mouse), and to the 950 largest clonotypes of antigen-specifically expanded B6 and I/St T cells. To find similarities in response between samples, we pooled ALICE hits (FDR-adjusted p < 0.05) obtained for each of the subsets and built clusters of CDR3 variants with one amino acid mismatch allowed. Cytoscape was used to visualize clusters (63).