BCG vaccination induces antibacterial effector functions among Vδ1/3 T cells that are associated with protection against tuberculosis

Summary

γδ T cells expressing a Vδ1/3+ T cell receptor are enriched at mucosal surfaces, but their role in protection against Mycobacterium tuberculosis (Mtb) is largely unknown. We used multimodal single-cell RNA sequencing, mass cytometry, and flow cytometry to profile γδ T cells from human infants and macaques after protective vaccination with Mycobacterium bovis bacillus Calmette Guerin (BCG). A subset of Vδ1/3 T cells in BCG-vaccinated human infants shows evidence of clonal expansion and differentiation into Mtb-reactive cytotoxic effector cells. In macaques, intravenous BCG induces pro-inflammatory and cytotoxic responses to Mtb among Vδ1/3 T cells that are enriched in the airway compared to the blood. Finally, the frequency of cytokine-expressing Vδ1/3 T cells in the airway is associated with protection against Mtb challenge. Thus, Vδ1/3 T cells are activated by BCG and accumulate in the lung, where they upregulate cytotoxic and pro-inflammatory functions that may contribute to protective immunity against Mtb.

Graphical abstract

Highlights

• •BCG vaccination promotes antibacterial functions in Vδ1/3 T cells
• •Vδ1/3 T cells clonally expand in response to BCG vaccination
• •After BCG vaccination, effector Vδ1/3 T cells accumulate in the airway
• •Vδ1/3 T cells in the airway correlate with protection against tuberculosis

Maerz et al. report that in response to protective BCG vaccination, Vδ1/3 T cells in humans and non-human primates undergo clonal expansion, upregulate pro-inflammatory and cytotoxic effector functions, and accumulate in the airway. Further, cytokine-producing Vδ1/3 T cells in the airway correlate with protection against tuberculosis.

Introduction

The live-attenuated Mycobacterium bovis strain Bacillus Calmette-Guérin (BCG) is the only licensed vaccine for tuberculosis (TB) and was developed over a century ago.¹ BCG shows only partial efficacy against pulmonary disease in adults, which is responsible for most transmission.² However, BCG is approximately 80% effective in preventing severe TB in children and can confer protection when administered intravenously (IV) to rhesus macaques.^3,4,5,6 Thus, BCG vaccination of human infants and macaques provides an opportunity to define cellular mechanisms of protection against TB.⁷

Studies in mice and non-human primates confirm that T cell immunity is essential for mediating the protective effect of BCG.^4,8,9 A dose-ranging study of IV-BCG revealed several immune features correlated with protection, including the frequency of cytokine-producing CD4 and CD8 T cells.¹⁰ These results were supported by a follow-up study demonstrating that treatment with CD4 or CD8ɑ depleting antibody is associated with a loss of IV-BCG-mediated protection.¹¹

γδ T cells express T cell receptor (TCR)-γ and TCR-δ chains and are conserved among most vertebrates.¹² γδ T cells are enriched in barrier tissues, including the lungs, and exhibit features of both innate and adaptive immunity, as they undergo both TCR-dependent and TCR-independent activation.^13,14,15 γδ T cells are composed of two major subsets, Vγ9Vδ2 T cells and Vδ1/3 T cells, that are defined by their TCR gene usage. Vγ9Vδ2 T cells mediate recognition of phosphoantigens through interactions with butyrophilin molecules.¹⁶ The function of Vδ1/3 T cells is less understood, but they are known to target endogenous proteins that are upregulated during bacterial infections and other forms of cellular stress, including CD1b, CD1c, MICA, and MR1.^17,18,19,20

Increasing evidence suggests that γδ T cells are important for TB immunity. In humans and macaques, Vγ9Vδ2 T cells exhibit primary and memory responses after intradermal BCG vaccination or Mtb infection.^21,22 A proportion of Vγ9Vδ2 T cells capable of suppressing intracellular BCG were found to recognize mycobacterial-derived methyl glucose polysaccharides.^23,24 A vaccine specifically targeting the activation and expansion of Vγ9Vδ2 T cells as well as adoptive transfer studies was sufficient to attenuate pulmonary TB in macaques.^25,26 In the absence of cell-specific depleting reagents, these data provide the strongest evidence supporting a causal role for γδ T cells in mediating protection against TB. In human infants, intradermal BCG vaccination is associated with an expansion of interferon gamma (IFN-γ)-expressing γδ T cells in the blood, and natural killer (NK)-like CD8⁺ γδ T cells are expanded in the blood of Mtb-infected South African adolescents.^27,28 . However, very little is known about the role of Vδ1/3 T cells in Mtb control.

Here, we studied the phenotypes, immune functions, and TCR repertoires of Vδ1/3 T cells in two contexts where BCG confers protective immunity against Mtb. Among BCG-vaccinated human infants, we found that a subset of Vδ1/3 T cells express clonally expanded TCRs and exert cytotoxic activity in response to Mtb antigens. Among IV-BCG-vaccinated rhesus macaques, we found concordantly that Vδ1/3 T cells clonally expand and upregulate cytotoxic effector functions in the blood and lungs following vaccination. Additionally, the frequency of cytokine-producing Vδ1/3 T cells in the airway correlates with protection against Mtb challenge. Together, our data reveal that cytotoxic Vδ1/3 T cells respond to protective BCG vaccination and expand the known list of BCG-induced immune correlates of protection.

Results

Single cell analysis of γδ T cells reveals antibacterial transcriptional programs and clonal expansion in BCG-vaccinated human infants

We first used single-cell RNA sequencing (scRNA-seq) and TCR sequencing to characterize unstimulated peripheral γδ T cells from 10-week-old South African infants (n = 6) who received intradermal BCG at birth.²⁹ We have previously shown that the peak CD4 T cell responses to BCG occurs at 10 weeks and that γδ T cells are expanded at this time point compared to age-matched, unvaccinated infants.^30,31 We hypothesized that Vδ1/3 T cells would show evidence of clonal expansion and display an effector memory phenotype.^28,32 After quality control filtering and data integration, a total of 14,302 unstimulated γδ T cells were analyzed with an average of 2,033 cells per infant (Figure S1A). Nine distinct clusters of γδ T cells were defined based on their gene expression (Figure 1A). Forty-three percent of cells expressed a Vγ9Vδ2 TCR (Figure 1B). Paired chain TCRs isolated from 2,329 cells bearing a Vγ9− γδ TCR revealed that 73% expressed a Vδ1+ TCR, 13% expressed a Vδ3+ TCR, and 14% expressed a Vδ2+ TCR (Table S1).

Figure 1.

Analysis of γδ T cells reveals antibacterial transcriptional programs and clonal expansion in BCG-vaccinated human infants

(A) UMAP displaying 14,302 γδ T cells from 10-week-old South African infants who were vaccinated with BCG at birth (n = 6). Cell clusters reflect gene expression, and each cell is color-coded according to its cluster assignment.

(B) Cluster positions of cells expressing Vδ1+, Vδ3+, or Vγ9Vδ2 TCRs.

(C) Cluster positions of cells expressing an expanded TCR (purple) or TCR that was only counted once per subject (gray).

(D) Heatmap displaying the average expression of key genes within each γδ T cell cluster.

(E and F) Background-subtracted frequency of Vδ1/3 T cells expressing cell surface and intracellular proteins (E) after stimulation with Mtb whole-cell lysate or (F) in the absence of stimulation. The frequency in 10-week-old, BCG-vaccinated South African infants (n = 22) is compared to the frequency in unmatched cord blood from South African mothers (n = 21). Boxplots indicate median and interquartile range. Statistical testing was performed using an unpaired Student’s t test. Unadjusted p values are displayed.

As expected, the gene expression patterns of Vδ1+ and Vδ3+ γδ T cells were overlapping and distinct from Vγ9Vδ2+ γδ T cells, which were enriched in clusters 1, 2, and 7 (Figures 1A and 1B). Overall, 8% of Vγ9Vδ2, 1.6% of Vδ1, and 1.2% of Vδ3 clonotypes were expanded as defined by count ≥2 (Table S1). Expanded Vγ9Vδ2 clonotypes were distributed across multiple cell clusters, suggesting that clonal expansion was not associated with a particular transcriptional phenotype (Figure 1C). However, expanded Vδ1/3 clonotypes were enriched in cluster 3 (Figure 1C).

Cell clusters enriched for Vγ9Vδ2+ cells typically expressed low levels of genes associated with a naïve-like phenotype (SELL, CCR7, and TCF7) together with higher levels of cytotoxicity-associated genes (GZMK, GZMA, CST7, PRF1, and GZMM) (Figures 1B and 1D). In contrast, clusters that were enriched for Vδ1+ and Vδ3+ T cells typically expressed higher levels of naïve-associated genes and lower levels of cytotoxicity-associated genes, consistent with published reports (Figures 1B and 1D).³³ However, cluster 3 was uniquely enriched in Vδ1/3 T cells expressing low levels of naïve-like genes, high levels of cytotoxicity genes, high levels of genes associated with T cell activation (HLA-DPA1, HLA-DPB1, HLA-DRB1, and TIGIT), and high expression of TXB21, which encodes a master regulator of pro-inflammatory function in T cells (Figure 1D). In sum, these results suggest that clonally expanded Vδ1/3 T cells express an activated cytotoxic effector phenotype in BCG-vaccinated infants.

As the BCG vaccine is routinely administered to all infants born in South Africa, we were unable to obtain age-matched samples from BCG-unvaccinated South African infants for this analysis. Therefore, to validate the cytotoxic functional programs identified by scRNA-seq, we used flow cytometry to compare γδ T cell responses to Mtb whole-cell lysate in peripheral blood mononuclear cells (PBMCs) from BCG-vaccinated infants (n = 22) and unmatched cord blood controls (n = 21) (Figures 1E and S1B). All samples were collected from the same community residing in a semi-rural region of Cape Town, South Africa, but were not matched by donor.³⁴ γδ T cells were stratified into Vγ9+ or Vγ9− subsets and analyzed for markers of activation (CD137), cytotoxicity (CD107a, granzyme B [GZMB], and GZMK), and cytokine production (IFN-γ and tumor necrosis factor [TNF]). We inferred that Vγ9− γδ T cells are mostly Vδ1/3+ on the basis of paired-chain TCR sequencing, as noted above (Table S1).

We did not observe a significant difference in the median frequency of CD137+ Vδ1/3 T cells after Mtb lysate stimulation (Figure 1E). Consistent with our observations from scRNA-seq, a subset averaging 7% of total Vδ1/3 T cells from BCG-vaccinated infants expressed GZMB in the absence of stimulation, which was half as abundant in cord blood (p = 0.0085) (Figure 1F). Additionally, the frequency of Vδ1/3 T cells expressing GZMB in response to stimulation with Mtb lysate was on average 1% in BCG-vaccinated infants compared to 0% in cord blood (p = 0.0025) (Figure 1E). Further, we observed on average 12-fold higher expression of IFN-γ after stimulation among Vδ1/3 T cells in BCG-vaccinated infants compared to cord blood (p < 0.0001) (Figure 1E). The median frequency of Vδ1/3 T cells expressing TNF in response to stimulation was twice as high in cord blood compared to BCG-vaccinated infants (p = 0.0037). The median frequency of Vδ1/3 T cells expressing CD107a or GZMK was not significantly different between BCG-vaccinated infants and cord blood (Figure 1E). These results reveal that the production of GZMB and IFN-γ by Vδ1/3 T-cells in response to Mtb whole-cell lysate is increased in BCG-vaccinated infants compared to neonates.

We noted that the median frequency of Vγ9+ T cells expressing CD137, CD107a, GZMB, IFN-γ, and TNF in response to stimulation were significantly higher in BCG-vaccinated infants compared to cord blood (Figure S2A). In unstimulated Vγ9+ T cells, GZMB and GZMK were elevated in BCG-vaccinated infants compared to cord blood (Figure S2B). These results are consistent with a prior study showing that Vγ9Vδ2 T cells acquire an activated and cytotoxic functional profile during the first weeks of life.³⁵ Additionally, Vγ9+ γδ T cells show higher responsiveness overall to stimulation with MtbL compared to Vγ9− T cells in cord blood and by 10 weeks after vaccination, as measured by expression of CD137, GZMB, and IFN-γ (Figures S2C and 2D). Further, Vγ9+ γδ T cells express higher levels of GZMK compared to Vγ9− γδ T cells in cord blood and higher levels of TNF by 10 weeks (Figures S2C and 2D).

Single-cell analysis of γδ T cells reveals antibacterial transcriptional programs following IV-BCG vaccination

We next examined γδ T cells in a discovery cohort of IV-BCG-vaccinated Indian origin rhesus macaques (n = 4). We performed multi-modal scRNA-seq on archived PBMCs collected prior to IV-BCG as well as PBMCs and bronchoalveolar lavage (BAL) collected at 4 weeks and 8 weeks post-vaccination (Figure 2A). γδ T cells were sorted immediately after thawing without stimulation. After quality control filtering and data integration, a total of 12,777 γδ T cells were analyzed with an average of 3,200 cells per sample (Figure S3A). Overall, 12 clusters of γδ T cells were defined, which broadly separated as cells derived from PBMCs or BAL (Figures 2B and 2C). γδ T cells bearing a Vδ1+ or Vδ3+ TCR were abundant and populated all 12 clusters. In contrast, Vγ9Vδ2 T cells were enriched in cluster 10 and composed less than 1% of the γδ T cells analyzed (Figure 2D; Table S2). This result was inconsistent with previously published data showing Vγ9+ T cells constituted 25%–75% of γδ T cells in PBMCs and 30%–85% of γδ T cells in BAL when analyzed directly ex vivo by flow cytometry and can represent up to 5% of circulating T cells in animals receiving high-dose IV-BCG.^4,10 We considered several technical explanations for this discrepancy, including age, sex, IV-BCG dose, cryopreservation, and monoclonal antibodies. Age and sex were similarly distributed in the current study and a recently published study .¹⁰ All animals in our study received high-dose (>10⁷ CFU) IV-BCG, which was associated with higher frequencies of Vγ9+ cells in the blood and airway.^4,10 Although previous studies used fresh cells and our study used cryopreserved samples, Vγ9+ frequencies were not reduced by cryopreservation (Figure S4A). Finally, we directly compared the monoclonal Vγ9 and TCR-γδ antibodies used in the current study to those used in previous studies and found that all clones showed similar Vγ9+ frequencies in a rhesus macaque PBMC sample (Figure S4B). Nevertheless, the median frequency of Vγ9− γδ T cells among total T cells in the samples we analyzed was approximately 2% in PBMCs and 0.5% in BAL, which was similar to what was reported previously (Figure S3B).⁴

Figure 2.

Single-cell analysis of sorted γδ T cells in rhesus macaques reveals antibacterial transcriptional programs following IV-BCG vaccination

(A) Study schema depicting experimental modality for each sample.

(B and C) UMAPs displaying 12,777 γδ T cells from IV-BCG-vaccinated rhesus macaques (n = 4). Cell clusters reflect gene expression. Each cell is color-coded according to its (B) cluster assignment and (C) tissue of origin.

(D) Cluster positions of cells expressing Vδ1, Vδ3, or Vγ9Vδ2 TCRs.

(E) Cluster positions of cells derived from pre-vaccination (Pre), week 4 (W4), or week 8 (W8) PBMC samples.

(F) Cluster positions of cells derived from week 4 (W4) or week 8 (W8) BAL samples.

(G) Heatmap displaying the average expression of key genes in each γδ T cell cluster.

(H) Relative proportion of the top five PBMC clusters at each time point.

(I) Relative proportion of PBMC-derived cells populating cluster 0, cluster 3, and cluster 4 at each time point. Each sample donor is shown in a separate line.

(J) Heatmap displaying the average expression of key genes across each tissue and time point.

The transcriptomic profile of γδ T cells changed over time in both PBMC and BAL, suggesting that IV-BCG vaccination directly modulated gene expression (Figures 2E and 2F). The γδ T cell clusters were annotated manually according to their RNA and surface protein expression patterns (Figures 2G and S3C). Cluster 0 expressed low levels of naïve-associated genes and T-effector-like genes and were likely resting cells. Clusters 1 and 9 expressed high levels of CCR7 and low levels of T-effector-like genes, leading to their annotation as naïve-like cells. Cluster 2 expressed TNFRSF9 (encoding CD137), NR4A1 (encoding Nur77), and genes associated with pro-inflammatory activity (XCL1, MX1, and STAT1) together with CD8 protein and were labeled as activated CD8⁺ γδ T cells. Clusters 3, 4, and 7 were cytotoxic T cell subsets differentiated by expression of granzymes (GZMA, GZMK, and GZMM) and perforin (PRF1). Clusters 5 and 8 expressed high levels of TOX, PDCD1, and ITGAX together with low levels of T cell effector genes, suggesting an exhausted-like phenotype. Cluster 6 expressed the tissue residency markers ITGA1 (CD49a), ITGB7, and ITGAE (CD103) but was populated by PBMC-derived cells, consistent with recent tissue egress. Cluster 10 was enriched in cells using a Vγ9Vδ2 TCR and expressed genes related to type 1, type 3, and cytotoxic activity, including IFNG, TNF, RORC, and GZMB. Finally, cluster 11 expressed high levels of ITGB1 (CD29) and NCAM1 (CD56), two genes that have been associated with cytotoxic activity but expressed relatively low levels of granzyme genes.

Among the six most abundant clusters, naïve-like, cytotoxic, and exhausted-like cells were mostly derived from PBMCs; activated CD8⁺ γδ T cells were mostly derived from BAL; and resting cells included both PBMC- and BAL-derived cells (Figures 2B, 2C, and 2G). Among PBMCs, resting cells were relatively enriched pre-vaccination, and cytotoxic cells appeared to be expanded after vaccination (Figures 2H and 2I). Within BAL, resting cells were relatively enriched at week 4, and activated CD8⁺ γδ T cells were expanded at week 8 (Figures S3D and S3E). Overall, multi-modal scRNA-seq revealed cytotoxic transcriptional programs among γδ T cells that partially tracked with time after IV-BCG vaccination and tissue of origin.

Having observed changes in cluster abundance across tissues and time, we next investigated how IV-BCG vaccination and tissue of origin influenced the gene expression profiles of Vδ1/3 T cells independent of cluster assignment. Differential gene expression analysis revealed down-regulation of naïve-like genes (LEF1, ID3, TCF4, and IL7R) and upregulation of cytotoxic effector genes (GZMA, GZMB, GZMK, and PRF1) and genes associated with activated T cells (NR4A2/3, CD38, and MAMU-DRB1) in Vδ1/3 T cells derived from PBMCs at 4 and 8 weeks compared to baseline (Figure 2J). Vδ1/3 T cells in BAL expressed lower levels of cytotoxicity-associated genes (GMZA, GZMB, GZMK, NKG7, and PRF1) compared to PBMC at week 4. However, genes that were upregulated in BAL at week 4 included signatures of response to type I IFN (IFI16, IFIT2, IFIT1) as well as KIR2DL4, SRGN, and XCL1, which are expressed by cytotoxic cells. Other genes that were elevated in BAL compared to PBMCs were associated with tissue-resident cells (ITGA1 and ITGAE), recent activation through the TCR (TNFRSF9 and NR4A1), and chronic stimulation (CTLA4, LAG3, and TIGIT). Notably, one of the most highly upregulated genes at 4 and 8 weeks in BAL compared to PBMCs was TNFSF8 (CD153), which has been associated with Mtb control in both mice and humans.^36,37 These results suggest that Vδ1/3 T cells in the blood and airway acquire antibacterial and cytotoxic effector programs in response to IV-BCG vaccination.

To extend the results of transcriptome analysis, the same samples were analyzed using a 42-parameter mass cytometry (CyTOF) panel measuring the surface and intracellular protein expression of CD45⁺ leukocytes including γδ T cells (Figures 2A, S5, and S6; Table S3).³⁸ Total PBMCs and BAL samples were either left unstimulated or were stimulated for 18 h using Mtb whole-cell lysate. γδ T cells were identified and stratified into Vγ9+ or Vγ9− subsets and analyzed for expression of cytotoxicity (GZMB, GZMK, and perforin), cytokine (IFN-γ and TNF) and activation markers (CD69). Although we did not directly identify Vδ1 T cells, we demonstrated through paired-chain TCR sequencing of 5,738 cells bearing a Vγ9− γδ TCR that 74% expressed a Vδ1+ TCR, 22% expressed a Vδ3+ TCR, and only 4% expressed a Vδ2+ TCR (Table S2).

In unstimulated PBMCs, the median frequency of Vδ1/3 T cells expressing GZMB, GZMK, and perforin was higher after IV-BCG (Figures S6A–S6F). After stimulation with Mtb whole-cell lysate, Vδ1/3 T cells showed increased upregulation of CD69, IFN-γ, and TNF in both PBMCs and BAL compared to pre-vaccination (Figures S6G–S6L). Vδ1/3 T cells producing IFN-γ and TNF in response to Mtb lysate appeared to be enriched in BAL compared to PBMCs at both 4 and 8 weeks post-vaccination (Figures S6K and S6L). These results confirm and extend those obtained using scRNA-seq and suggest that IV-BCG promotes cytotoxic and antimicrobial activity among Mtb-reactive Vδ1/3 T cells.

IV-BCG promotes cytotoxic and pro-inflammatory responses to mycobacterial antigens among Vδ1/3 T cells in the blood and airway

We next analyzed γδ T cell activation (CD69, HLA-DR, and CD137), pro-inflammatory cytokine production (IFN-γ and TNF), and cytotoxicity (GZMB, GZMK, and CD107a) in the presence or absence of stimulation with Mtb whole-cell lysate for 18 h using flow cytometry (Figure S7; Table S4). We analyzed PBMCs from IV-BCG-vaccinated rhesus macaques (n = 16) before and at weeks 4 and 8 after IV-BCG. We also analyzed BAL obtained at week 8 after IV-BCG and compared this to BAL samples from a separate cohort of unvaccinated rhesus macaques (n = 9) (Figure 2A). Consistent with our scRNA-seq analysis, the median frequency of Vγ9+ γδ T cells was lower than previously reported in vaccinated and unvaccinated animals (Figure S8). The expression of CD137, CD107a, or cytokines in unstimulated samples was not significantly increased after IV-BCG in either PBMC or BAL at any time point (Figures S9A–S9F). However, we observed greater than a 2-fold increase in unstimulated Vδ1/3 T cells expressing granzymes from pre-vaccination to week 2 in PBMCs (p = 0.002) but not BAL (Figures S9C and S9D). These results suggest that IV-BCG vaccination promotes cytotoxic potential of circulating Vδ1/3 T cells in vivo.

The median frequency of Vδ1/3 T cells expressing CD137 in response to stimulation with Mtb lysate (Mtb-reactive) increased by up to 6-fold in PBMCs (p < 0.0001) and 10-fold in BAL (p = 0.0024) (Figures 3A and 3B).^39,40 The median frequency of Vδ1/3 T cells expressing granzymes in response to stimulation was not significantly altered after IV-BCG in BAL but increased from 0% at baseline to 8.7% by week 8 in PBMCs (p = 0.043) (Figures 3C and 3D). CD107a median expression was increased up to 8-fold in PBMCs (p = 0.0009) and 50-fold in BAL (p = 0.0057) after IV-BCG, compared to baseline (Figures 3C and 3D). Further, the median frequency of Vδ1/3 T cells producing IFN-γ οr TNF in response to stimulation with Mtb lysate was markedly increased after IV-BCG, rising over 20-fold in PBMCs (p = 0.0011) and over 40-fold in BAL (p = 0.0006) (Figures 3E and 3F). Mtb-reactive Vδ1/3 T cells were on average enriched 13-fold in BAL compared to PBMCs after IV-BCG (p = 0.0002) while there was no significant difference seen before vaccination (Figures 3G and S10A). The median expression of granzymes in response to stimulation did not significantly differ between PBMCs and BAL in vaccinated animals, but in unvaccinated animals it was 16.6% in BAL compared to 0.1% in PBMCs (p = 0.006) (Figures 3G and S10B). Median expression of CD107a in response to stimulation did not differ between PBMCs and BAL in vaccinated or unvaccinated animals (Figures 3G and S10C). Finally, cytokine-producing Vδ1/3 T cells were on average enriched over 10-fold in BAL compared to PBMCs after IV-BCG (p = 0.0001), which was not observed in unvaccinated animals (Figures 3G and S10D). These results suggest that IV-BCG augments Vδ1/3 T cell responses to Mtb antigens in the blood and airway to exert cytotoxic and pro-inflammatory immune functions.

Figure 3.

IV-BCG promotes cytotoxic and pro-inflammatory responses to mycobacterial antigens among Vδ1/3 T cells in the blood and airway

Background-subtracted frequency of Vδ1/3 T cells expressing cell surface and intracellular proteins in response to stimulation with Mtb lysate among IV-BCG-vaccinated (n = 16) and unvaccinated (n = 9) rhesus macaques.

(A) Representative flow cytometry plots (left) and summary line plots over time (right) for CD137 in PBMC.

(B) Representative flow cytometry plots (left) and boxplots (right) comparing CD137 expression in BAL between unvaccinated macaques (Unvax) and week 8 (W8) post-vaccination.

(C) Representative flow cytometry plots (left) and summary line plots over time (right) for CD107a or granzyme (GZMB or GZMK) in PBMCs.

(D) Representative flow cytometry plots (left) and boxplots (right) comparing CD107a or granzymes in BAL between unvaccinated macaques (Unvax) and week 8 (W8) post-vaccination.

(E) Representative flow cytometry plots (left) and summary line plots over time (right) for IFN-γ or TNF in PBMCs.

(F–H) (F) Representative flow cytometry plots (left) and boxplots (right) comparing IFN-γ or TNF in BAL between unvaccinated macaques (Unvax) and week 8 (W8) post-vaccination. Comparison of background-subtracted frequencies of Vδ1/3 T cells expressing CD137, granzyme (GZMB or GZMK), CD107a, and cytokine (IFN-γ or TNF) (G) after stimulation with Mtb lysate or (H) in the absence of stimulation between PBMC and BAL at week 8. In summary line plots, each sample donor is displayed as a gray line, and median frequencies are shown as a black line. Boxplots display median and interquartile range.

In (A, C, E, G, and H), statistical testing was performed using the paired Wilcoxon signed-rank test. In (B, D, and F), statistical testing was performed with an unpaired Wilcoxon signed-rank test. Unadjusted p values are displayed.

Vδ1/3 T cells respond to IV-BCG earlier than conventional T cells

We previously demonstrated that the peak functional response to IV-BCG in conventional T cells occurs at 8 weeks after vaccination in BAL and 4 weeks after vaccination in PBMCs.⁴ Additionally, the peak functional response of Vγ9+ T cells in PBMCs appeared to occur between 2 and 4 weeks after vaccination.¹⁰ However, the time point of peak functional response among Vδ1/3 T cells remains uncharacterized. We found that Mtb-responsive CD8 T cells were significantly more numerous than γδ T cells at nearly all tissues and time points (Figure S11). At most time points in PBMCs and BAL, Vδ1/3 T cells expressing CD137 were significantly more numerous than Vγ9+ T cells expressing CD137 (Figures S11A and 11B). There were no significant differences in the total T cell frequency of Vδ1/3 T cells and Vγ9+ T cells expressing granzymes, except in week 8 BAL (Figures S11C and S11D). Vδ1/3 T cells expressing CD107a were more numerous than Vγ9+ T cells expressing CD107a at late time points in both PBMCs and BAL (Figures S11E and S11F). Finally, at all time points and tissues, Vδ1/3 T cells expressing IFN-γ or TNF were significantly more numerous than Vγ9+ T cells expressing IFN-γ or TNF (Figures S11G and S11H).

Among peripheral blood Vδ1/3 T cells stimulated with Mtb lysate, we found that peak expression of CD137 and CD107a occurred on average at week 2, earlier than in CD8 T cells (Figures 4A and S12A). While no differences were observed in the expression of granzymes, peak expression of CD107a also occurred earlier than in CD8 T cells, at week 2 (Figures 4B, 4C, S12B, and S12C). Peak expression of cytokines in Vδ1/3 T cells occurred at 2 to 4 weeks after vaccination, while peak expression of cytokines in CD8 T cells occurred at 4 to 8 weeks, on average (Figures 4D and S12D). Despite showing a peak functional response earlier than CD8 T cells, Vδ1/3 T cells did not have significantly lower functional responses to Mtb lysate compared to CD8 T cells, except for 3- to 5-fold lower expression of IFN-γ or TNF in week 4 (p = 0.03) and week 8 (p = 0.012) PBMCs (Figures 4A–4D and S12). CD8 T cells expressing CD137 (p = 0.006), CD107a (p < 0.0001), and IFN-γ or TNF (p < 0.0001) were 2- to 3-fold more frequent compared to Vδ1/3 T cells in BAL of vaccinated animals (Figure 4H). Compared to Vδ1/3 T cells, Vγ9+ T cells expressed higher levels of CD137 in PBMCs from unvaccinated animals (p = 0.036), consistent with an enhanced ability to recognize Mtb in an unprimed state (Figures 4A and S12A). Vδ1/3 T cells and Vγ9+ γδ T cells expressed similar levels of granzymes and CD107a in PBMCs, except for higher levels of CD107a in Vδ1/3 T cells at week 2 (p = 0.0095) (Figures 4B, 4C, S12B, and S12C). Vγ9+ T cells expressed higher levels of IFN-γ or TNF at week 2 in PBMCs (p = 0.032) but not at any other time point (Figures 4D and S12D). In BAL, there were no differences in the expression of any functional marker between Vδ1/3 T cells and Vγ9+ T cells (Figures 4E–4H). Thus, Vδ1/3 T cells show peak functional responsiveness to IV-BCG at an earlier time point compared to CD8 T cells, yet the durability of these responses may be comparable. However, CD8 T cells may exert stronger pro-inflammatory activity in the blood and airway at late time points.

Figure 4.

Vδ1/3 T cells respond to IV-BCG earlier than CD8 T cells

Background-subtracted frequency of CD8, Vγ9+, and Vδ1/3+ T cell subsets in PBMCs expressing (A) CD137, (B) granzymes (GZMB or GZMK), (C) CD107a, and (D) cytokine (IFN-γ or TNF) after stimulation with Mtb whole-cell lysate among IV-BCG-vaccinated (n = 16) and unvaccinated (n = 9) rhesus macaques. Median frequencies are plotted pre-vaccination (Pre) and at week 2 (W2), week 4 (W4), and week 8 (W8) post IV-BCG. The frequency of Vγ9+ γδ T cells or CD8 T cells is compared to the frequency of Vδ1/3 T cells at each time point. Background-subtracted frequency of CD8, Vγ9+, and Vδ1/3+ T cell subsets in BAL expressing (E) CD137, (F) granzymes (GZMB or GZMK), (G) CD107a, and (H) cytokine (IFN-γ or TNF) after stimulation with Mtb whole-cell lysate. Boxplots compare median and interquartile range within unvaccinated macaques (Unvax) and at week 8 (W8) post-vaccination. Statistical testing was performed using a paired Wilcoxon signed-rank test. Unadjusted p values are displayed.

Vδ1/3 T cells clonally expand in response to IV-BCG

γδ T cells are known to be activated by TCR-dependent or TCR-independent signals, including NK receptor ligands, TLR ligands, and cytokines.⁴¹ Therefore, we next sought to assess to what extent the Vδ1/3 T cell repertoire is modulated by IV-BCG. We used single-cell TCR sequencing to analyze γδ T cells collected from rhesus macaques (n = 4) pre-vaccination and at 4 and 8 weeks post-vaccination (Figure 2A). In total, 5,517 TCR sequences were resolved in PBMCs, and 2,406 TCR sequences were resolved in BAL.

We found 9 Vδ1/3+ clonotypes and 10 Vδ1/3+ clonotypes that were expanded at weeks 4 and 8, respectively, compared to baseline. In some cases, clonotypes were absent from baseline samples but counted three or more times post-vaccination. In other cases, clonotypes were present at low frequency at baseline but underwent further expansion after IV-BCG (Figures 5A and 5B; Table S2). On average, we identified seven clonotypes within each IV-BCG recipient that were expanded after vaccination, together representing roughly 0.6% of the total Vδ1/3 clonotypes per donor. Additionally, in all four of the animals analyzed, the proportion of the Vδ1/3 TCR repertoire represented by expanded clonotypes appeared to increase at week 4 compared to baseline (Figure 5C). Clonotypes were compartment-specific, with only 0.75% and 0.6% of clonotypes present in both PBMCs and BAL at week 4 or week 8, respectively. Surprisingly, none of the circulating Vδ1/3 clonotypes that were expanded compared to baseline were detectable in BAL at either time point (Table S2). These data suggest that IV-BCG leads to the expansion of mycobacteria-responsive Vδ1/3 T cells in which the TCR repertoire of responding T cells is distinct between the blood and lung.

Figure 5.

Vδ1/3 T cells clonally expand in response to IV-BCG

Analysis of the Vδ1/3 TCR repertoire IV-BCG-vaccinated rhesus macaques (n = 4) and the effect of TCR inhibition on Vδ1/3 effector functions (n = 3–6).

(A) Number of cells per animal expressing each unique Vδ1/3 clonotype in pre-vaccination (Pre) vs. week 4 (W4) and pre-vaccination (Pre) vs. week 8 (W8) PBMCs. The size of each dot is proportional to the number of clonotypes plotted at each point. Clonotypes that are expanded greater than 3-fold compared to pre-vaccination are depicted in black. Count values were natural-log-transformed after adding one.

(B) Proportion of the Vδ1/3 TCR repertoire occupied by the top 10 clonotypes at each time point in PBMCs. Each clonotype is represented by a colored bar. Dynamics are shown in two representative animals in which clonotype colors are not matched.

(C) Frequency of expanded (non-singlet) clonotypes within the Vδ1/3 TCR repertoire at each time point in PBMCs. Each donor is shown as a separate line.

(D) Gene expression UMAP displaying the cluster position of clonotypes that are unexpanded (UnExp), clonotypes that expand more than 3-fold post-vaccination (“Responding,” ExpR), and clonotypes that are expanded pre-vaccination but do not further expand after IV-BCG (“Expanded non-responding,” ExpNR).

(E) Bar plot displaying the cluster distribution of cells in the UnExp, ExpR, and ExpNR categories.

(F) Volcano plot depicting genes that are upregulated (red) or downregulated (blue) in ExpR clonotypes compared to ExpNR clonotypes.

(G) Background-subtracted frequency of Vδ1/3 T cells expressing CD137 (left) or IFN-γ (right) after stimulation with Mtb whole-cell lysate in the presence or absence of cyclosporin A (CsA). Frequencies are shown pre-vaccination (Pre) and at week 2 (W2), week 4 (W4), and week 8 (W8) post IV-BCG. Each sample donor is shown as a separate gray line. Median frequencies are shown in black.

We then assigned each peripheral blood clonotype to one of three categories: “responding” clonotypes, which expanded greater than 3-fold after IV-BCG (ExpR); “expanded non-responding” clonotypes, which were counted at least three times in unvaccinated animals but did not expand more than 3-fold after IV-BCG (ExpNR); and “unexpanded” clonotypes, which included any clonotype that did not fall into the previous two categories (UnExp) (Figure 5D). While UnExp clonotypes were predominantly located in cluster 0 (resting), cluster 1 (naïve-like), and cluster 5 (exhausted-like), ExpR and ExpNR clonotypes were preferentially distributed in cluster 3 (activated granzyme K-high), cluster 4 (activated granzyme A-high), and cluster 7 (cytotoxic MAMU-DRA-high) (Figure 5E). Additionally, ExpR clonotypes were enriched in cluster 4, while ExpNR clonotypes were enriched in cluster 7 (Figure 5E). Genes that were preferentially enriched among ExpR clonotypes compared to ExpNR clonotypes included GZMA, ZBTB16, and FGFBP2, characteristic of cytotoxic lymphocytes (Figure 5F). Taken together, these data suggest that Vδ1/3 T cells that undergo clonal expansion after IV-BCG preferentially express genes associated with cytotoxic function.

Given this evidence of in vivo clonal expansion in response to IV-BCG, we next examined whether Vδ1/3 T cell responses are TCR-mediated. PBMCs were stimulated with Mtb lysate in the presence or absence of cyclosporin A, an inhibitor of TCR signaling.⁴² Consistent with our prior results, cells obtained post-vaccination were more responsive to Mtb lysate than cells obtained pre-vaccination (Figures 3 and 5G). Treatment with cyclosporin A reduced both CD137 and IFN-γ expression in response to stimulation with Mtb lysate by approximately 90% at all post-vaccination time points (Figure 5G). Notably, Vδ1/3 T cells were not more responsive to stimulation with PHA over time, suggesting that the increased responsiveness to Mtb lysate after IV-BCG is dependent on the presence of mycobacterial antigens (Figure S13). In Vδ1/3 T cells stimulated with a cocktail of IL-12, IL-15, and IL-18, treatment with cyclosporin A reduced CD137 expression by approximately 50% but did not reduce IFN-γ expression (Figure S14). Thus, Vδ1/3 T cell responses to Mtb lysate are not likely the result of bystander activation in response to inflammatory cytokines. Rather, our results suggest that IV-BCG leads to clonal expansion of Mtb-reactive, cytotoxic Vδ1/3 T cells in the blood.

Vδ1/3 T cells in the airway correlate with protection against Mtb challenge

Finally, we asked whether Vγ9− γδ T cells correlate with protection against Mtb challenge after IV-BCG. We leveraged a published dataset examining BAL collected from 34 rhesus macaques who were vaccinated with a range of IV-BCG doses, resulting in variable degrees of protection against Mtb challenge.¹⁰ In this study, Vγ9+ T cells were identified as an immune correlate of protection, but Vγ9− T cells were not analyzed.¹⁰ We obtained the original FCS files and gated Vγ9− cells, which we inferred to be 96% Vδ1/3 + T cells (Figures 6A, 6B, and S15; Table S2). In accord with the published study, we defined ‘protected’ as <100 total thoracic Mtb CFU enumerated at necropsy. Protected (n = 18) and unprotected (n = 16) animals showed similar distribution of sex and age at vaccination (Table S4). Animals who were protected after challenge showed 1.5- to 2-fold higher frequency of Vγ9− γδ T cells in BAL at week 2 (p = 0.027) and week 8 (p = 0.046) after vaccination compared to animals who were not protected (Figure 6C, left). At week 4, protected animals showed 2-fold higher frequency of Vγ9− γδ T cells producing IFN-γ, IL-2, or TNF in response to stimulation with Mtb purified protein derivative (PPD) (p = 0.0091) (Figure 6C, right). The frequency of cytokine-producing Vγ9− γδ T cells did not significantly differ between protected and unprotected animals at week 2 or week 8 (Figure 6C, right).

Figure 6.

Vδ1/3 T cells in the airway correlate with protection against Mtb challenge

(A) Representative bivariate plot showing the identification of Vγ9− γδ T cells using FCS files obtained from Darrah et al.¹⁰

(B) Representative bivariate plots showing the expression of cytokines (% positive) among Vδ1/3 T cells in BAL in response to stimulation with Mtb purified protein derivative (PPD).

(C) Frequency of Vδ1/3 T cells among total T cells at week 2 (W2), week 4 (W4), and week 8 (W8) post-vaccination with IV-BCG (left). Background-subtracted frequency of Vδ1/3 T cells expressing IL-2, IFN-γ, or TNF in response to stimulation with Mtb whole-cell lysate (right). Boxplots display the median and interquartile range of each feature in protected (<100 total thoracic CFU, n = 18) or unprotected (≥100 total thoracic CFU, n = 16) animals. Statistical testing was performed with a Welch’s t test. Unadjusted p values are displayed.

(D) Heatmap displaying Spearman correlations between the measured Vγ9− T cell immune features, cohort metadata, and previously published correlates. Significance levels are indicated for each pair of correlates (∗p < 0.05; ∗∗p < 0.01, ∗∗∗p < 0.001).

(E) Volcano plot showing the normalized coefficient of protection status incorporated in the mixed linear model (x axis) versus the adjusted p value of the likelihood ratio test (LRT) for the model fit difference between the two nested models (y axis). See Table S3 for a description of each immune feature.

A prior analysis defined a minimal signature of immune protection in the airway: frequency of IFN-γ+ and TNF+ CD4 T cells, frequency of IL-17+ and TNF+ CD4 T cells, absolute number of NK cells, and IgA titers.¹⁰ In addition, the absolute number of airway Vγ9+ T cells and Vγ9+ T cells producing IFN-γ, IL-2, TNF, or IL-17 were highly correlated with NK cell counts. Considering these results, we sought to determine how the correlations with protection among Vγ9− γδ T cells compared with these six published measurements. Using Spearman correlations, we first investigated the relationship between Vγ9− γδ T cells and total thoracic Mtb CFU counts after challenge, IV-BCG dose, sex, age, and the six published immune measurements (Figure 6D; Table S3).¹⁰ To maintain consistency with the previous dose-ranging study, we calculated the normalized area under the curve (nAUC) to generate summary statistics describing Vγ9− cell counts, % Vγ9− T cells, % IFN-γ+ Vγ9− T cells, % IL-2+ Vγ9− T cells, % TNF+ Vγ9− T cells, and total % IL-2+, TNF+ or IFN-γ+ (Cytokine+) Vγ9− T cells in the airway between 2 and 8 weeks post-vaccination.¹⁰ We observed that Mtb CFU counts were negatively correlated with Vγ9− cell count, % IFN-γ+ Vγ9− T cells, % TNF+ Vγ9− T cells, and % Cytokine+ Vγ9− T cells, consistent with our univariate analysis (Figures 6C and 6D). Age and female sex were positively correlated with cytokine-producing Vγ9− T-cells. Of the tested Vγ9− cell measurements, the strongest and most significant correlate of protection was Vγ9− cell counts, which showed a similar correlation coefficient and significance level to the strongest published correlates. IV-BCG dose also positively correlated with Vγ9− cell count, but not any other measurement among Vγ9− T cells. Further, Vγ9− cell count was strongly positively correlated with all six previously published immune correlates (Figure 6D). Intriguingly, Vγ9− γδ T cells expressing IFN-γ, TNF, or any combination of IFN-γ, TNF, and IL-2 negatively correlated with Mtb CFU but did not correlate with any of the six previously published immune measurements, suggesting that Vγ9− cells may be an independent correlate of protection.

Because Vγ9− γδ T cell counts were strongly correlated with IV-BCG dose as well as protection, we next compared two linear models, a full model including protection and a reduced model without it, to evaluate whether the selected immune measurements were associated with protection while adjusting for IV-BCG dose and vaccination cohort (Figure 6E; Table S3). In agreement with previously published results, we found that NK cell counts, Vγ9+ T cell counts, and IL-17+ TNF+ CD4 T cell counts were significantly correlated with protection after controlling for IV-BCG dose.¹⁰ Additionally, Vγ9− γδ T cell count and Vγ9+ T cells expressing any combination of IFN-γ, IL-2, TNF, or IL-17 remained significant correlates of protection after adjusting for dose (Figure 6E). Vγ9+ T cell counts, cytokine+ Vγ9+ T cells, IL-17+ TNF+ CD4 T cells, and Vγ9− γδ T cell counts showed similar degrees of enrichment in protected animals, and NK cell counts were most strongly associated with protection (Figure 6E). These results suggest that both the overall frequency of Vγ9− γδ T cells and the frequency of PPD-reactive Vγ9− γδ T cells in the airway after IV-BCG are associated with protection against Mtb. Additionally, Vγ9− γδ T cell counts correlate with protection irrespective of IV-BCG dose at a magnitude comparable to the strongest published T cell correlates.

Discussion

Identifying mechanistic immune correlates of protection against Mtb has been a long-standing goal for the field.⁴³ We focused on understanding the role of γδ T cells by studying their phenotypes and functions among infants and macaques after protective BCG vaccination. Our results advance the field in at least three respects. First, we show that BCG promotes a pro-inflammatory and cytotoxic response to mycobacterial antigens in both human infant and macaque Vδ1/3 T-cells. Second, we report that Mtb-reactive Vδ1/3 T cells are clonally expanded in both species and provide evidence that effector functions are mediated through the TCR. Finally, we find that Vδ1/3 T cells correlate with IV-BCG-induced protection against Mtb challenge in a manner that appears to be independent of known correlates of protection.

Although the protective role of cytokine-producing CD4 T cells has been well established, the contribution of cytotoxic immune cells such as CD8 T cells, NK cells, MAIT cells, and γδ T cells is less appreciated.⁴⁴ In a dose-ranging study of IV-BCG, NK cells, Vγ9+ T cells, MAIT cells, and CD8 T cells were among the immune subsets correlated with protection against Mtb challenge after IV-BCG.¹⁰ Additionally, GZMB^high CD4 T cells in the lungs and a blood transcriptional module enriched in NK cell pathways were associated with protection against Mtb after IV-BCG.^45,46 GZMK was found to be enriched in innate-like immune populations compared to conventional T cells and decreased after TCR stimulation, consistent with a role in pre-programmed antimicrobial responses.⁴⁷ By contrast, granzyme A binds directly to Mtb bacilli to promote phagocytosis independent of its enzymatic activity.⁴⁸ Consistent with this, we show that expression of GZMA, GZMB, and GZMK transcripts and expression of GZMB and GZMK proteins were increased in Vδ1/3 T cells following IV-BCG. Additionally, we noted that IV-BCG led to the expansion of two distinct populations of cytotoxic Vδ1/3 T cells, one expressing high levels of GZMK and GZMM and the other expressing high GZMA.

We report evidence of clonally expanded Vδ1/3 T cell clonotypes in BCG-vaccinated infants and non-human primates, which express genes associated with TCR activation and cytotoxic activity. Treatment with cyclosporin A virtually eliminated Vδ1/3 T cell effector function after Mtb lysate stimulation but only partially inhibited responses to cytokines, suggesting that activation through the TCR is required. Clonal expansion of Vδ1/3 T cells has been observed in diverse human disease settings, including solid tumors, CMV infection, and febrile malaria, consistent with an adaptive-like immune role.^33,49,50,51 Notably, the Vδ1/3 TCR ligands identified thus far include several human proteins that are regulated by cellular stress, such as CD1b, CD1c, MR1, MICA, MICB, endothelial protein C receptor, and annexin A2.^{17,19,52,53,54,55,56} While the specific ligands promoting the TCR-dependent Vδ1/3 T cell response to BCG vaccination remain to be defined, we speculate that stress-regulated, endogenous proteins are a likely target. It is also possible that Vδ1/3+ T cells recognize mycobacterial lipid antigens complexed with CD1 molecules.^18,19

Our results contrast with a recent report describing circulating Vδ1 T cells in persistent Mtb infection that are refractory to TCR stimulation yet exert cytotoxic activity through CD16.²⁸ Chowdhury et al. hypothesize that innate-like Vδ1 T cells may be a feature of chronic inflammation, while our study has characterized Vδ1/3 T cells in a setting of acute infection, potentially explaining the difference in TCR-dependent effector functions in these two contexts. Nevertheless, our results are consistent with other studies describing circulating Vδ1/3 T cells as a largely adaptive-like immune subset.⁵⁷

Our findings have direct implications for the design of new vaccines for TB that are largely focused on boosting CD4 T cell responses to mycobacterial protein antigens.⁵⁸ γδ T cells are not directly targeted by these products but may be activated indirectly by adjuvants or signals from other immune cells.⁵⁹ The demonstration that boosting Vγ9Vδ2 T cell responses via a modified Listeria vaccine protects against TB in macaques provides proof of concept for products focused on this T cell subset.²⁵ Vδ1/3 T cells are not activated by the phosphoantigens that stimulate Vγ9Vδ2 T cells, but our data suggest they may also be important to mediating protection. Whole-cell mycobacterial vaccines currently in phase 3 clinical trials, such as VPM001 and MTBVAC, may be able to engage Vδ1/3 T cells through both TCR-dependent and TCR-independent mechanisms.⁶⁰ Finally, induction of cytotoxic effector functions is not typically a criterion for evaluating the immunogenicity of candidate vaccines, but our data add to an emerging body of literature, suggesting that this could be a key feature of a protective immune response to TB.

Limitations of the study

We were unable to analyze BAL samples collected at week 2 after vaccination, potentially precluding us from studying γδ T cells at the time point of peak responsiveness in the airway.⁴ Additionally, our analysis only includes samples collected until 8 weeks after vaccination, so we were unable to study the durability of Vδ1/3 T cell responses to IV-BCG. We have previously shown that circulating Vδ2 clonotypes are expanded up to 1 year after BCG vaccination in humans.⁶¹ Additionally, prior analysis of Vδ1 T cell clonotypes that are expanded after CMV infection have demonstrated that these clonotypes persist in circulation for at least 18 months, but whether similar durability is observed after mycobacterial infection is unknown. Finally, since BCG is routinely delivered at birth to infants in TB endemic areas, our experiments lacked a control group of unvaccinated 10-week-old South African infants. Thus, we are unable to evaluate whether the observed clonal expansion and effector functions are the result of BCG vaccination, T cell development in the early weeks of life, or other factors. Despite these limitations, the concordant findings of clonal expansion, Mtb reactivity, and cytotoxic effector functions in Vδ1/3 T cells across species suggest that at least some of the effects we observed among infants were due to BCG.

Resource availability

Lead contact

Further information and requests for resources and reagent should be directed to and will be fulfilled by the lead contact, Chetan Seshadri (seshadri@uw.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •All the data generated in support of the reported findings can be found at Fairdomhub: https://fairdomhub.org/studies/1421. Raw sequencing data has been submitted to the NCBI Sequence Read Archive under Accession NumbersPRJNA1207031 and PRJNA985785Cytometry time-of-flight (CyTOF), flow cytometry, and processed sequencing data are available on Zenodo https://zenodo.org/records/14583675.
• •Code that was used to produce analysis and figures for this article is available on GitHub.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank Benjamin Bimber for sharing custom γδ TCR primers and Michael Vilme for facilitating sequencing. This study was supported by the Intramural Research Program of NIAID.

Author contributions

M.D.M., conceptualization, methodology, software, formal analysis, investigation, writing—original draft, writing—review and editing, visualization, project administration, and funding acquisition. M.S.M., methodology and investigation. M.S.S., investigation and data curation. A.N.B., investigation and data curation. E.B., methodology, software, and writing—original draft. Z.T., writing—review and editing, visualization, software, and formal analysis. E.D.L., methodology. M.R., resources. A.K.S., resources. R.A.S., resources. T.J.S., resources. C.W., writing—review and editing, visualization, software, and formal analysis. P.A.D., investigation and data curation. C.S., conceptualization, resources, writing—original draft, writing—review and editing, supervision, and funding acquisition.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
116Cd anti-CD45 (clone D058-1283)	BD Biosciences	Cat #552566
106Cd anti-CD68 (clone KP1)	BioLegend	Cat #916104
172Yb anti-CD206 (clone 19.2)	BD Biosciences	Cat #555953
154Sm anti-CD163 (clone GHI/61)	Fluidigm	Cat #3154007B
151Eu anti-CD14 (clone M5E2)	Fluidigm	Cat #3151009B
144ND anti-CD11b (Mac-1) (clone ICRF44)	Fluidigm	Cat #3144001B
209Bi anti-CD16 (clone 3G8)	Fluidigm	Cat #3209002B
146ND anti-CD11c (clone 3.9)	Fluidigm	Cat #3146014B
161Dy anti-CD123 (clone 6H6)	BioLegend	Cat #306002
176Yb anti-HLA-DR (clone LN3)	BioLegend	Cat #327002
169Tm anti-NKG2A (clone Z199)	Fluidigm	Cat #3169013B
147Sm anti-CD20 (clone 2H7)	Fluidigm	Cat #3147001B
153Eu anti-IgD (Polyclonal)	Southern Biotech	Cat #2030-01
170Er anti-CD3 (clone SP34-2)	Fluidigm	Cat #3170007B
APC anti-MR1-5-OP-RU Tetramer	NIH Tetramer Core	N/A
163Dy anti- anti-APC	Fluidigm	Cat #3163001B
171Yb anti-TCRγδ (clone B1)	BioLegend	Cat #331202
112Cd anti-TCR Vγ9 (clone B3)	BioLegend	Cat #331301
160Gd anti-CD161 (clone HP-3G10)	BioLegend	Cat #339902
111Cd anti-CD8α (clone RPA-T8)	BioLegend	Cat #301053
113Cd anti-CD8β (clone 2ST8.5H7)	Novus Biologicals	Cat #NB100-65928
114Cd anti-CD4 (clone L200)	BD Biosciences	Cat #550625
174Yb anti-CD45RA (clone 5H9)	BD Biosciences	Cat #556625
175Lu anti-CD28 (clone CD28.2)	BioLegend	Cat #302902
159Tb anti-CCR7 (clone G043H7)	Fluidigm	Cat #3159003A
156Gd anti-CXCR3 (clone G025H7)	Fluidigm	Cat #3156004B
145ND anti-CD69 (clone FN50)	BioLegend	Cat #310902
173Yb anti-Granzyme B (clone GB11)	Fluidigm	Cat #3173006B
165Ho anti-Granzyme K (clone GM26E7)	BioLegend	Cat #370502
164Dy anti-Perforin (clone Pf-80/164)	Mabtech	Cat #3465-3-250
143ND anti-CD107a (clone H4A3)	BioLegend	Cat #328601
162Dy anti-CD154 (clone 24–31)	BioLegend	Cat #310802
152Sm anti-TNF (clone Mab11)	Fluidigm	Cat #3152002B
168Er anti-IFN-g (clone B27)	Fluidigm	Cat #3168005B
158Gd anti-IL-2 (clone MQ1-17H12)	Fluidigm	Cat #3158007B
148ND anti-IL-17A (clone BL168)	Fluidigm	Cat #3148008B
Cell-ID Intercalator Ir	Fluidigm	Cat #201192A
Cell-ID Cisplatin	Fluidigm	Cat # 201064
FITC anti-TNF (clone Mab11)	BD Biosciences	Cat #554512
BB700 anti-CD8 (clone RPA-T8)	BD Biosciences	Cat #566452
PE anti-Granzyme K (clone GM26E7)	BioLegend	Cat #370512
PE-CF594 anti-Granzyme B (clone GB11)	BD Biosciences	Cat #562462
PE-Cy7 anti-TCRgd (clone B1)	BioLegend	Cat #331222
APC anti-Vg9 (clone B3)	BioLegend	Cat #331310
APC-H7 anti-CD4 (clone L200)	BD Biosciences	Cat #560837
V450 anti-IFN-g (clone B27)	BD Biosciences	Cat #560371
Zombie Yellow Fixable Viability Kit	BioLegend	Cat #423103
BV605 anti-CD137 (clone 4B4-1)	BioLegend	Cat #309822
BV711 anti-CD107a (clone H4A3)	BioLegend	Cat #328639
BV785 anti-CD69 (clone FN50)	BioLegend	Cat #310932
BUV395 anti-CD3 (clone SP34-2)	BD Biosciences	Cat #564117
BUV737 anti-HLA-DR (clone LN3)	ThermoFisher	Cat #367-9956-42
LIVE/DEAD Fixable Green Cell Stain Kit	ThermoFisher	Cat #L23101
BV421 anti-CD3 (clone SP34-2)	BD Biosciences	Cat #562877
PE-Dazzle anti-TCRgd (clone B1)	BioLegend	Cat #331225
Anti-CD8b (clone SIDI8BEE)	eBiosiences	Cat #14-5273-82
ECD anti-CD3 (clone UCHT1)	Beckman Coulter	Cat #NC0720544
PE-Vio770 anti-TCRgd	BioLegend	Cat #130-113-513
TotalSeq-C0006 anti-human CD86 Antibody	BioLegend	Cat #305447
TotalSeq-C0045 anti-human CD4 Antibody	BioLegend	Cat #344651
TotalSeq-C0090 Mouse IgG1 Isotype Control	BioLegend	Cat #400187
TotalSeq-C0091 Mouse IgG2a Isotype Control	BioLegend	Cat #400293
TotalSeq-C0092 Mouse IgG2b Isotype Control	BioLegend	Cat #400381
TotalSeq-C0100 anti-human CD20 Antibody	BioLegend	Cat #302363
TotalSeq-C0358 anti-human CD163 Antibody	BioLegend	Cat #333637
TotalSeq-C0085 anti-human CD25	BioLegend	Cat #302649
TotalSeq-C0140 anti-human CD183 (CXCR3)	BioLegend	Cat #353747
TotalSeq-C0143 anti-human CD196 (CCR6)	BioLegend	Cat #353440
TotalSeq-C0146 anti-human CD69 Antibody	BioLegend	Cat #310951
TotalSeq-C0148 anti-human CD197 (CCR7)	BioLegend	Cat #353251
TotalSeq-C0149 anti-human CD161	BioLegend	Cat #339945
TotalSeq-C0386 anti-human CD28	BioLegend	Cat #302963
TotalSeq-C0390 anti-human CD127 (IL-7Rα)	BioLegend	Cat #351356
TotalSeq-C0251 anti-human Hashtag 1	BioLegend	Cat #394661
TotalSeq-C0252 anti-human Hashtag 2	BioLegend	Cat #394603
TotalSeq-C0256 anti-human Hashtag 6	BioLegend	Cat #394671
TotalSeq-C0257 anti-human Hashtag 7	BioLegend	Cat #394673
TotalSeq-C0258 anti-human Hashtag 8	BioLegend	Cat #394675
TotalSeq-C0007 anti-human CD274 (B7-H1, PD-L1) Antibody	BioLegend	Cat #329751
TotalSeq-C0053 anti-human CD11c Antibody	BioLegend	Cat #371521
TotalSeq-C0080 anti-human CD8a Antibody	BioLegend	Cat #301071
TotalSeq-C0081 anti-human CD14 Antibody	BioLegend	Cat #301859
TotalSeq-C0083 anti-human CD16 Antibody	BioLegend	Cat #302065
TotalSeq-C0156 anti-human CD95 (Fas) Antibody	BioLegend	Cat #305651
TotalSeq-C0159 anti-human HLA-DR Antibody	BioLegend	Cat #307663
TotalSeq-C0161 anti-human CD11b Antibody	BioLegend	Cat #301359
Bacterial and Virus Strains
BCG SSI (Danish strain 1331)	Aeras (IAVI)
Mtb Erdman barcode library	BEI Resources	Cat #NR-50781
Critical Commercial assays
Chromium Next GEM Chip G Single Cell Kit	10x Genomics	Cat #1000127
Chromium Next GEM Chip K Single Cell Kit	10x Genomics	Cat #1000287
Chromium Next GEM Single Cell 5ʹ Library and Gel Bead Kit v1.1	10x Genomics	Cat #1000165
Chromium Next GEM Single Cell 5′ Reagent kit, v2	10x Genomics	Cat #1000263
Deposited data
Custom code used for analysis	This paper	https://github.com/seshadrilab/gammadelta_bcg
Raw sequencing data	This paper	NCBI Sequence Read Archive, Accession #PRJNA1207031 and #PRJNA985785
CyTOF data, Flow cytometry data, and processed sequencing data	This paper	Zenodo, https://doi.org/10.5281/zenodo.14583675
Software and algorithms
FlowJo v10.8	BD Biosciences	RRID:SCR_008520
Cell Ranger	10x Genomics	RRID:SCR_017344
FACS Diva	BD Biosciences	RRID:SCR_001456
R project for Statistical Computing v4.4.1	R Core Team	RRID:SCR_001905
R package Seurat	CRAN	https://cran.r-project.org/web/packages/Seurat/index.html
R package scRepertoire	GitHub	https://github.com/BorchLab/scRepertoire
R package ggplot2	CRAN	https://cran.r-project.org/web/packages/ggplot2/index.html
R package Hmisc	CRAN	https://cran.r-project.org/web/packages/Hmisc/index.html
R package corrplot	CRAN	https://cran.r-project.org/web/packages/corrplot/index.html
Other
Fc Receptor Binding Inhibitor Polyclonal Antibody	eBioscience	Cat #14-9161-73
FACSAria II Cell Sorter	BD Biosciences	RRID:SCR_018934
Qubit 3 Fluorometer	Invitrogen	RRID:SCR_020311
4200 TapeStation System	Agilent	RRID:SCR_018435
NovaSeq 6000 System	Illumina	RRID:SCR_016387
NextSeq 500 System	Illumina	RRID:SCR_014983
Guava easyCyte	MilliporeSigma	RRID:SCR_025377
Mycobacterium tuberculosis whole-cell lysate strain H37Rv	BEI Resources	NR-14822
GolgiStop Protein Transport Inhibitor	BD Biosciences	Cat #554724
Helios	Fluidigm	RRID:SCR_019916
LSR Fortessa	BD Biosciences	RRID:SCR_018655

Experimental model and study participant details

Non-human primate cohorts

Details regarding the non-human primate cohorts have been previously published (Table S4).^10,11 Animal work was approved by the Institutional Care & Use Committees of AAALAC (American Association for Accreditation of Laboratory Animal Care)-accredited institutions (NIH Vaccine Research Center, Bioqual, Inc., and the University of Pittsburgh) and determined to be in accordance with the guidelines outlined by the Animal Welfare Act and Regulation (USDA) and the Guide for the Care & Use of Laboratory Animals, 8th Edition (NIH).

Indian-origin rhesus macaques (Macaca mulatta) aged 3–5 years were vaccinated with BCG Danish Strain 1331 (Statens Serum Institut, Copenhagen) into the saphenous vein. Actual BCG doses were quantified by dilution-plating and are reported in Table S4. Animals were randomized into vaccine groups based on birth colony, gender, and pre-vaccination CD4 T cell responses to Mtb purified protein derivative in the BAL. 6 months after vaccination, macaques were challenged intrabronchially with an average of 12 CFU (range 4–17 CFU) Mtb Erdman strain. All granulomas and other lung pathologies, all thoracic LNs, and peripheral LNs were collected for quantification of Mtb bacterial burden. CFUs were enumerated and summed to calculate the total thoracic bacterial burden for each macaque.

Human cohorts

Details regarding the BCG-vaccinated infant cohort have been previously published (Table S5).⁶² HIV-uninfected infants in the Cape Town region of South Africa received routine intradermal vaccination with BCG (Statens Serum Institut, Copenhagen) at birth. Infants born to HIV-positive mothers, infants known to be HIV positive, infants with suspected or confirmed TB disease, infants with possible TB exposure, and infants with any other illnesses at the time of enrollment were excluded. Human participation was according to the US Department of Health and Human Services and good clinical practice guidelines. This included protocol approval by the University of Cape Town Research Ethics Committee and written informed consent.

Cord blood samples were collected from HIV-uninfected mothers enrolled in a previously published study held in the Cape Town region of South Africa (Table S5).³⁴ Pregnant mothers with TB disease or exposure or who were undergoing treatment for any other disease were excluded from the study. Human participation was approved by the Research Ethics Committee of the University of Cape Town and written informed consent was obtained from pregnant women and parents/legal guardians of infants.

Method details

Sample collection and processing

Non-human primate samples

Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque PLUS gradient separation (GE Healthcare Biosciences) as we have previously described.¹¹ Bronchoalveolar lavage (BAL) wash fluid (3 × 20 mL washes of PBS) was centrifuged and cells were combined before counting. Samples were cryopreserved in fetal bovine serum containing 10% dimethyl sulfoxide (DMSO) in liquid nitrogen until use.

Human samples

Sodium heparinized blood was collected from each infant approximately ten weeks after BCG vaccination. PBMC were isolated from whole blood using Ficoll-Paque PLUS gradient separation (GE Healthcare Biosciences) and cryopreserved in FBS containing 10% DMSO in liquid nitrogen until use. Sodium heparinized cord blood samples were collected immediately after birth by Caesarian section. Samples were treated for 10 min with 2mM EDTA (Sigma Aldrich), washed, and treated again for 10 min with 2mM EDTA. Red blood cells were then lysed using red blood cell lysis buffer before washing and cryopreserving in fetal calf serum containing 10% DMSO in liquid nitrogen until use.

Single-cell RNA sequencing

Sample preparation

Samples were thawed in RPMI medium (Thermo Fisher Scientific) plus 10% fetal bovine serum (Atlas Biologicals) and cell counts and sample viabilities were measured using a hemocytometer. Three million viable cells from each sample were plated in a 96-well plate. The cells were then washed with PBS and stained using a LIVE/DEAD Fixable Green Dead Cell Stain Kit (Invitrogen) for 15 min at room temperature. Next, the cells were blocked using Fc Block (eBioscience) for 15 min at 4°C.

Human samples were stained with an antibody cocktail containing fluorescent-tagged antibodies for cell sorting (Table S6) for 30 min at 4°C. Next, live γδ T cells (Live, CD3⁺, TCR γδ+) were sorted using a FACSAria cell sorter (BD Biosciences) configured with blue (488 nm), red (641 nm), and violet (407 nm) lasers. After sorting, samples were counted using a hemocytometer and up to 33,000 cells were loaded per well in a Chromium Chip G (10x Genomics).

Non-human primate samples stained with an antibody cocktail containing fluorescent-tagged antibodies for cell sorting, hashtags to allow sample multiplexing, and oligonucleotide-tagged antibodies for CITE-seq analysis (Table S7) for 30 min at 4°C. Next, live Vγ9+ γδ T cells (Live, CD3⁺, TCR γδ+, TCR Vγ9+) and live Vγ9-negative γδ T cells (Live, CD3⁺, TCR γδ+, TCR Vγ9-negative) were sorted using a FACSAria cell sorter (BD Biosciences) configured with blue (488 nm), red (641 nm), and violet (407 nm) lasers. After sorting, samples were counted using a hemocytometer and up to 33,000 cells were loaded per well in a Chromium Chip K (10x Genomics).

Library preparation

For human samples, Chromium Next GEM Single Cell V(D)J Reagent kit (v1.1, 10x Genomics) was used to prepare mRNA, TCR, and surface protein libraries. γδ TCR libraries were prepared using custom primers (Table S8). cDNA amplification and target enrichment were quantified using a Qubit 3 Fluorometer (Invitrogen) and assessed for quality using a 4200 TapeStation System (Agilent). Libraries were constructed and sequenced to a depth of approximately 50,000 reads per cell using the NovaSeq 6000 system (Illumina).

For non-human primate samples, the Chromium Next GEM Single Cell 5′ Reagent kit (v2, 10x Genomics) was used to prepare mRNA, TCR, and surface protein libraries. γδ TCR libraries were prepared using custom primers (Table S8). cDNA amplification and target enrichment were quantified using a Qubit 3 Fluorometer (Invitrogen) and assessed for quality using a 4200 TapeStation System (Agilent). Libraries were constructed and sequenced to a depth of approximately 40,000 reads per cell using the NextSeq 500 system (Illumina).

Sample stimulation

Cryopreserved samples were thawed in sterile-filtered RPMI 1640 (Thermo Fisher Scientific) with 10% FBS (Hyclone) and 0.2% Benzonase (MilliporeSigma) and then centrifuged at 500 relative centrifugal force for 5 min at room temperature. Cells were washed using PBS and centrifuged at 500 RCF for an additional 5 min at room temperature and were then rested for two hours at a density of 1–4 million cells/mL. Viable cells were counted using the Guava easyCyte (MilliporeSigma) with guavaSoft 2.6 software. After resting, up to 3 million cells were plated in a 96-well plate. Next, cells were stimulated with 5 μg/mL Phytohemagglutinin (Sigma Aldrich), 100 μg/mL Mycobacterium tuberculosis whole-cell lysate (strain H37Rv, BEI Resources), or a cytokine cocktail including 33.3 ng/uL each of recombinant IL-12 (Peprotech), recombinant IL-15 (Miltenyi Biotec) and recombinant IL-18 (R&D Systems). After 12 h of stimulation, cells were treated with 1.4 μg/mL GolgiStop (BD Biosciences) and 10 μg/mL Brefeldin A (Sigma Aldrich) and stained with anti-CD107a antibody (Table 9) for the remaining 6 h.

TCR inhibition

Cryopreserved samples were thawed in sterile-filtered RPMI 1640 (Thermo Fisher Scientific) with 10% FBS (Hyclone) and 0.2% Benzonase (MilliporeSigma) and then centrifuged at 500 relative centrifugal force for 5 min at room temperature. Cells were washed using PBS and centrifuged at 500 RCF for an additional 5 min at room temperature and were then rested for two hours at a density of 1–4 million cells/mL. Viable cells were counted using the Guava easyCyte (MilliporeSigma) with guavaSoft 2.6 software. After resting, up to 3 million cells were plated in a 96-well plate. Cells were then treated with 1.6 μg/mL Cyclosporin A for one hour prior to stimulation. Next, cells were stimulated with 5 μg/mL Phytohemagglutinin (Sigma Aldrich), Mycobacterium tuberculosis whole-cell lysate (strain H37Rv, BEI Resources), or a cytokine cocktail including 33.3 ng/uL each of recombinant human IL-12 (Peprotech), recombinant IL-15 (Miltenyi Biotec) and recombinant IL-18 (R&D Systems). After 12 h of stimulation, cells were treated with 1.4 μg/mL GolgiStop (BD Biosciences) and 10 μg/mL Brefeldin A (Sigma Aldrich) and stained with anti-CD107a antibody (Table S9) for the remaining 6 h.

Mass cytometry

Metal-labelled antibodies were either obtained from Standard BioTools or were conjugated in-house by labeling purified antibodies using the Maxpar Antibody Labeling kit (Standard BioTools). Immediately following stimulation, cells were washed twice with PBS before viability staining with Cisplatin for 5 min at room temperature. Cells were then washed twice in FACS buffer before staining with MR1-5-OPRU-APC tetramer (NIH Tetramer Core Facility) for 60 min at room temperature. Next, cells were washed three times with cell staining buffer (Standard BioTools) before staining with a surface antibody cocktail for 30 min at room temperature (Table S10). Cells were then treated with Fix 1 buffer (Standard BioTools) for 20 min at room temperature. Next, cells were washed three times with Perm-S buffer (Standard BioTools) before staining with an intracellular antibody cocktail for 30 min at room temperature (Table S10). Stained cells were then washed three times in Perm-S buffer and fixed using a 1% paraformaldehyde solution for 15 min at 4°C. Cells were held in a DNA-intercalator stain at 4°C for up to 7 days until sample acquisition. Immediately before acquisition, cells were washed three times with UltraPure water (ThermoFisher) and were resuspended in UltraPure water with 1:5 EQ calibration beads (Standard BioTools). Cells were then filtered into cell strainer cap tubes (Fisher Scientific). Data were acquired on a Helios (Fluidigm) at up to 500 cells per second. The resulting FCS files were normalized using EQ Four Element Calibration Beads (Fluidigm), as previously described.⁶³ Data were analyzed using FlowJo v10.8 software (BD Biosciences).

Flow cytometry

Immediately following stimulation, cells were washed twice with PBS and then stained with Zombie Yellow fixable viability dye (BioLegend) for 15 min at room temperature. Samples were then washed twice using FACS buffer and incubated in Fc Block (eBiosciences) according to the manufacturer’s instruction. Next, samples were washed once with FACS buffer and stained with a cell surface antibody cocktail for 30 min at 4°C (Table S9). After the incubation period, samples were washed twice with FACS buffer and fixed using CytoFix Fixation buffer (BD Biosciences) according to the manufacturer’s instructions. Next, samples were washed twice with Perm/Wash buffer (BD Biosciences) and stained with an intracellular antibody cocktail for 30 min at 4°C (Table S9). Finally, cells were washed twice with PBS, fixed in a 1% paraformaldehyde solution, and held in 2mM EDTA at 4C for up to 36 h prior to sample acquisition. Cells were acquired on a BD LSRFortessa (BD Biosciences) equipped with a high-throughput sampler and configured with blue (488 nm), green (532 nm), red (628 nm), violet (405 nm), and ultraviolet (355 nm) lasers using standardized good clinical laboratory practice procedures to minimize variability of data generated. For all flow cytometry experiments, study groups were evenly distributed to avoid batch effects and operators were not blinded to study group assignments. Data were analyzed using FlowJo v10.8 software (BD Biosciences).

Quantification and statistical analysis

Single-cell RNA sequencing

Non-human primate samples

The Cell Ranger Multi pipeline (10x Genomics) was used to conduct alignment and feature expression quantification of the single-cell sequencing data. We generated a custom Macaca mulatta reference transcriptome using Ensembl genome assembly Mmul_10 (GCA_003339765.3) in Cell Ranger. We included protein coding, lncRNA, antisense, immunoglobulin, T cell receptor, and mitochondrial genes. Additionally, we created a custom V(D)J reference using the Rhesus macaque reference sequences from IMGT.⁶⁴ Constant regions that were not in IMGT were included from the reference Mmul_10 transcriptome. Hashtag oligo (HTO) and antibody-derived tag (ADT) reads were aligned to a feature reference containing the relevant barcode sequences.

Quality control and analysis of the single-cell mRNA and surface protein expression data were performed using the R package Seurat.⁶⁵ First, we excluded HTOs that did not appear in at least 50 cells as well as cells with fewer than three HTO counts. Next, we excluded cells with fewer than 200 feature counts, greater than 4000 feature counts, or greater than 10% mitochondrial reads. Next, HTO data were normalized using a centered log-ratio transformation and samples were demultiplexed based on HTO enrichment using the MULTIseqDemux function, which demultiplexes samples based on the classification method from MULTI-seq.⁶⁶ We then removed cell barcodes that could not be assigned to a single HTO.

Next, the ADT counts were normalized using a centered log-ratio transformation and visualized using the RidgePlot function. Antibodies with counts that did not exceed the expression level of respective mouse IgG isotype controls were excluded. Next, for each cell barcode, the mRNA count data were log-normalized. For each sample donor, cells were then clustered according to their gene expression. Non-T-cell clusters were then excluded based on low CD3E RNA expression combined with high RNA expression of MS4A1 or CD14 and high ADT expression of CD20 or CD14.

Next, we quality checked our TCR libraries. We first excluded TCR sequences from any cell barcodes that were excluded based on the criteria above. We also excluded any TCR sequences that failed to meet the following criteria: 1. The arrangement is productive, 2. The TCR is amplified from at least three templates corresponding to three unique molecular identifiers, and 3. The TCR includes at least 20 reads. We next used the scRepertoire package to identify the single most highly expressed TCR clonotype for every cell barcode.⁶⁷ Cell barcodes expressing at least one quality-checked γδ TCR chain were annotated as γδ T cells. γδ T cells were further stratified into Vδ1, Vδ3, Vγ9^negVδ2, and Vγ9+Vδ2 subsets according to their TCR expression.

We applied the Seurat Integration procedure to combine the cell barcodes from all four sample donors and used differential gene expression analysis to verify that sample integration did not weaken the major biological axes of variation, such as tissue or origin and sample timepoint.⁶⁵ Next, we clustered the integrated data based on gene expression using the first 12 principal components obtained from a PCA for dimensionality reduction. The Seurat FindAllMarkers function was used to identify differentially expressed genes across cell clusters, sample tissues, and timepoints. Cell clusters were annotated according to differential expression of key T cell related genes and ADTs.

Next, we analyzed our TCR libraries to define γδ T cell clonotypes that were expanded in response to vaccination. Expanded clonotypes were defined as clonotypes that were counted at least three times at week 4 or week 8 and were expanded at least 2-fold relative to pre-vaccination. Additionally, clonotypes that were absent pre-vaccination but were counted at least three times post-vaccination were considered expanded. We also identified clonotypes that were counted at least twice in pre-vaccination samples but did not expand at least 2-fold after vaccination. Finally, unexpanded clonotypes included any clonotype that did not align with either of these criteria. The scRepertoire function compareClonotypes was used to visualize changes in frequency among the most highly expressed TCR-δ clonotypes at each timepoint. Plots were generated using the ggplot2 R package.⁶⁸

Human samples

The Cell Ranger Multi pipeline (10x Genomics) was used to conduct alignment and feature expression quantification of the single-cell sequencing data. We used data analysis services provided by Azenta Life Sciences (Burlington, MA) to align mRNA reads to the GRCh38 human reference genome and V(D)J reads to a custom reference using the human γδ TCR reference sequences from IMGT.⁶⁴

Quality control and analysis of the single-cell mRNA expression data were performed using the R package Seurat.⁶⁵ We first excluded cell barcodes with fewer than 200 feature counts, greater than 3500 feature counts, fewer than 500 RNA counts, greater than 15,000 RNA counts, or greater than 10% mitochondrial reads. Next, for each cell barcode, the mRNA count data were log-normalized. We then clustered all cell barcodes according to their gene expression using the first 10 principal components obtained from a principal component analysis (PCA) for dimensionality reduction. The Seurat FindAllMarkers function was used to identify differentially expressed genes across cell clusters. One cell cluster was excluded due to low expression of CD3E and high expression of MS4A1. Freemuxlet software was used to demultiplex pooled infant PBMC samples according to common single-nucleotide polymorphisms (PMID: 29227470).

We then quality checked our TCR libraries. We first excluded TCR sequences from any cell barcodes that were excluded based on the criteria above. We next used the scRepertoire package to identify the single most highly expressed TCR clonotype for every cell barcode.⁶⁷ γδ T cells were stratified into Vδ1, Vδ3, Vγ9^negVδ2, and Vγ9+Vδ2 subsets according to their TCR expression.

After excluding one cell cluster with low CD3E expression and high MS4A1 expression consistent with B cells, we re-clustered the remaining cells according to their mRNA expression. Cell clusters were defined after principal component analysis reduction using the first 10 principal components. The FindAllMarkers function was used to distinguish genes that were differentially expressed across the resulting cell clusters. Relative expression of key T cell related genes was used to determine cell cluster identities. Plots were generated using the ggplot2 R package.⁶⁸

Correlation analyses

Normalized area under the curve (nAUC) Transformation

Longitudinal immune measurements were summarized using a normalized area under the curve (nAUC) transformation. For each variable, the area under the curve (AUC) was computed using numerical integration over the available measured timepoints. The resulting AUC was then normalized by dividing by the time duration. This approach provided a standardized, single-summarized measurement that effectively captured the overall response across the observation time period.

Spearman correlation analysis

Pairwise Spearman correlation coefficients were computed using the rcorr function from the Hmisc R package.⁶⁹ This method calculates Spearman’s rho correlation, a non-parametric measure assessing monotonic relationships between pairs of variables. Missing data are managed through pairwise deletion rather than complete-case removal, thereby maximizing data utilization. Significance of correlations was approximated using asymptotic p-values derived from the t-distribution. The correlation coefficients with statistical significance were visualized using the pheatmap function from the corrplot R package.⁷⁰

Linear mixed model analysis

To assess the significance of the association between scaled immune feature measurements and protection group (protected vs. non-protected) by controlling potential confounders, we employed a model comparison approach using two linear mixed models: a full model including the protection group term and a reduced model excluding it. Both models included log-transformed intravenous (IV) BCG dose as a fixed covariate and vaccination cohort as a random effect. The full model additionally incorporated the protection group as a fixed effect. The null and full models were specified as follows.

• •Null model: measurement ∼1 + Log IV BCG Dose + (1 | Cohort)
• •Full model: measurement ∼1 + Log IV BCG Dose + Protection? + (1 | Cohort)

Models were fitted using the lmer function from the lme4 package in R,⁷¹ and improvements in model fit were evaluated by likelihood ratio tests (LRT), calculated as:

$$\text{LRT}=-2\times \ln \left(\frac{{\text{MLE}}_{\text{Full}\text{model}}}{{\text{MLE}}_{\text{Null}\text{model}}}\right)\sim {\mathrm{\chi }}^2$$

The significance (p-values) obtained from the LRT and the associated t-values (normalized coefficients) representing protection group effects from the full model were adjusted using Benjamini-Hochberg correction to control for multiple comparisons. Results were visualized using volcano plots created with the ggplot2 package in R.⁶⁸ All statistical analyses were conducted in R (version 4.4.1).

Published: January 12, 2026