Probing Dermal Immunity to Mycobacteria through a Controlled Human Infection Model

Abstract

Cutaneous mycobacterial infections cause substantial morbidity and are challenging to diagnose and treat. An improved understanding of the dermal immune response to mycobacteria may inspire new therapeutic approaches. We conducted a controlled human infection study with 10 participants who received 2 × 106 CFUs of Mycobacterium bovis bacillus Calmette-Guérin (Tice strain) intradermally and were randomized to receive isoniazid or no treatment. Peripheral blood was collected at multiple time points for flow cytometry, bulk RNA sequencing (RNA-seq), and serum Ab assessments. Systemic immune responses were detected as early as 8 d postchallenge in this M. bovis bacillus Calmette-Guérin-naive population. Injection-site skin biopsies were performed at days 3 and 15 postchallenge and underwent immune profiling using mass cytometry and single-cell RNA-seq, as well as quantitative assessments of bacterial viability and burden. Molecular viability testing and standard culture results correlated well, although no differences were observed between treatment arms. Single-cell RNA-seq revealed various immune and nonimmune cell types in the skin, and communication between them was inferred by ligand-receptor gene expression. Day 3 communication was predominantly directed toward monocytes from keratinocyte, muscle, epithelial, and endothelial cells, largely via the migration inhibitory factor pathway and HLA-E-KLRK1 interaction. At day 15, communication was more balanced between cell types. These data reveal the potential role of nonimmune cells in the dermal immune response to mycobacteria and the utility of human challenge studies to augment our understanding of mycobacterial infections.

FIGURE 1.

Study schema and primary microbiology outcomes. (A) Study schema. Ten participants received intradermal BCG, and half were randomized to receive INH therapy. Blood samples were collected at multiple time points, and skin punch biopsies were collected at days 3 and 15 post-BCG injection. (B) Photos of the injection site at days 3 and 15 from a representative participant. Biopsy sites (arrows) are located 2–4 mm above and below the injection site (“X”). Day 15 photo was taken prior to biopsy. (C) CFUs on 7H10 solid media. (D) Maximum pre-rRNA/rDNA ratio (MVT), which indicates the presence of viable bacteria. (C and D) Results for INH-treated (n = 5) and untreated (n = 4) participants. Significance comparing treatment arms was determined by two-sample, two-sided, unpaired Wilcoxon test.

FIGURE 2.

Peripheral immune responses seen as early as 8 d post-BCG. Results from serum binding Ab multiplex assay (BAMA), flow cytometry ICS, and flow cytometry COMPASS analysis. (A) Area under the curve (AUC) of serum IgG binding levels in response to M. tuberculosis WCL (TB WCL), live BCG Tice, or M. tuberculosis LAM as measured by BAMA (n = 10 for days 0, 8, 28, 56, and 84; n = 9 for day 114). Black lines indicate median values across participants. (B) Proportion of CD4⁺ T cells expressing IFN-γ and/or IL-2 as measured by ICS after stimulation with TB WCL or live BCG Tice. (C) COMPASS PFSs in CD4⁺ T cells after stimulation with TB WCL or live BCG. PFS reflects the proportion of cells expressing a range of cytokines, weighted by the number of cytokines expressed per cell. (B and C) Boxplots represent the median and interquartile range. Gray lines indicate the same individual across time points. n = 8, days 0 and 8; n = 9, days 28, 56, 84, and 114. Statistical significance comparing results to day 0 was calculated using paired Wilcoxon signed rank tests. *Unadjusted p < 0.05, **unadjusted p < 0.01.

FIGURE 3.

Genes related to monocytes and neutrophils are downregulated in whole blood at day 56 post-BCG. Bulk RNA-seq of whole blood was performed at multiple time points (n = 10). (A) Heatmap showing log₂FC gene expression for all 1497 differentially expressed genes (p < 0.05, FDR < 0.05, and absolute log₂FC > 0.5), averaged across participants. A subset of monocyte and neutrophil BTM genes are labeled LogFC = log₂FC. (B and C) Average normalized expression of all genes in the (B) neutrophil (I, II) and (C) monocyte (II, IV) BTMs. Black lines indicate median values across participants. Statistical significance comparing results with day 0 was calculated using paired Wilcoxon signed rank tests. *Unadjusted p < 0.05.

FIGURE 4.

Immune cell frequencies at injection site change over time. Skin biopsies were subject to assays including CyTOF, using an immune-focused panel. Immune cell types present at the injection site and changes in their frequencies between days 3 and 15 were identified by CyTOF. (A) CD45⁺ cell subsets were identified using the FlowSOM clustering algorithm and manually annotated (n = 18). “Undefined” represents CD45⁺ cells that could not be confidently annotated. (B) Cell subset frequencies as a fraction of total CD45⁺ cells were calculated at each time point (day 3, n = 10; day 15, n = 6). Samples with low cell counts (day 15 samples for BCG05 and BCG08) were excluded. Comparisons were made using unpaired Wilcoxon rank-sum test with unadjusted p values displayed.

FIGURE 5.

scRNA-seq reveals changing myeloid expression and cross-talk at injection site. (A) All cells that passed quality filtering were automatically annotated on a cell-cell basis using the SingleR package with the “HumanPrimaryCellAtlasData” reference. “Unknown” represents unable to annotate (day 3 = 12,178 cells, day 15 = 4731 cells). (B) Lymphoid and myeloid cells were subclustered and automatically annotated using the Azimuth package with the “pbmcref” reference (day 3 = 6662 cells, day 15 = 2090 cells). (A and B) For visualization purposes, all cells were integrated using Harmony and are shown in UMAP space (n = 15). Notably, no day 3 cells could be annotated as granulocytes and so these do not appear in the integrated data shown here. (C) Select immune cell proportions as a fraction of total parent cell type (lymphoid or myeloid) at days 3 (n = 9) and 15 (n = 6). Statistical testing was performed with unpaired Wilcoxon rank-sum test, and unadjusted p values are displayed. (D) Violin plots of module scores, a measure of relative expression of a gene set, for genes enriched in neutrophil (top) and monocyte (bottom) blood transcription modules from Fig. 3B and 3C at days 3 (n = 9) and 15 (n = 6). Statistical significance was calculated using two-sided Wilcoxon rank-sum tests, and results were adjusted for multiple comparisons with the Bonferroni method. ****Adjusted p < 0.0001. (E and F) Cell-cell interactions were identified by CellChat at (E) day 3 (n = 9) and (F) day 15 (n = 6). Significant interactions were identified using a permutation test and p < 0.05. All interactions shown are weighted by communication strength and scaled to maximum strength for that day. Extracellular matrix pathway interactions are not shown. Notably, no day 3 cells could be annotated as granulocytes.