The clinical stakes are considerable. Sputum culture conversion, defined as two consecutive negative cultures collected at least 30 days apart, is a critical early indicator of treatment efficacy and a predictor of long-term cure. Patients whose cultures remain positive at the two-month mark face a higher risk of treatment failure, relapse, and continued transmission within their communities. The World Health Organization’s End TB Strategy, which aims to dramatically reduce tuberculosis incidence by 2035, depends on improving these unfavourable outcomes, yet the immunological mechanisms governing sputum conversion have remained poorly understood. Host-pathogen interactions in tuberculosis are notoriously heterogeneous, with significant variation between patients in radiographic presentation, symptoms, and disease trajectory, making it difficult to predict who will respond and who will not.

To probe this heterogeneity at unprecedented resolution, the research team enrolled eight patients with confirmed drug-sensitive Mycobacterium tuberculosis infection, dividing them into four treatment responders and four non-responders based on their two-month sputum culture conversion status. The researchers isolated peripheral blood mononuclear cells and subjected them to single-cell RNA sequencing using the 10x Genomics Chromium X platform, sequencing on an Illumina NovaSeq X Plus instrument. After rigorous quality control, 83,999 cells were retained for analysis, with an average of 1,723 genes detected per cell. Unsupervised clustering resolved the data into 17 transcriptional clusters, which the team annotated into six major immune cell types: T cells, NK cells, B cells, myeloid cells, mast cells, and platelets.

The most striking finding emerged from the myeloid compartment. Non-classical monocytes, a subset of monocytes characterized by their patrolling behaviour in the vasculature, were relatively enriched in the non-responder group and carried elevated scores on a previously established tuberculosis progression risk signature. The total monocyte fraction showed a large standardized between-group difference, with a Cohen’s d of 1.04, although the authors caution that with only four patients per group this effect size is descriptive rather than definitive. Trajectory analyses using PAGA and Monocle, combined with CytoTRACE differentiation scoring, indicated that these non-classical monocytes were terminally differentiated, and SCENIC regulatory network analysis highlighted activation of transcription factors including ZNF177, FOXP3, and HES5 within the subset.

The transcriptional programs of these cells diverged sharply between groups. Non-classical monocytes from non-responders showed upregulation of TLR6, TLR1, IFI44, and PARP9, and were enriched for pathways involving Toll-like receptor 2 signalling, type II interferon-mediated signalling, and negative regulation of cytokine signalling through I-kappa-B kinase and NF-kappa-B pathways. In contrast, the corresponding cells from responders expressed higher levels of LAPTM5, CXCL8, and PTGS2 and were associated with cell adhesion and positive regulation of cytokine-mediated signalling. Monocytes from non-responders also displayed elevated expression of GBP2, STAT1, and TAP1, genes previously proposed as candidate biomarkers of tuberculosis infection. Taken together, these signatures suggest that non-classical monocytes in non-responders are locked into a terminally differentiated, senescence-like state biased towards innate inflammatory activation, echoing earlier reports that linked the pro-inflammatory phenotype of this subset to cellular senescence driven by NF-kappa-B and IL-1 alpha signalling.

Perhaps the most conceptually intriguing result came from CellChat-based intercellular communication analysis, which models how cell populations signal to one another through known ligand-receptor pairs. In non-responders, non-classical monocytes received more incoming signals but emitted fewer outgoing signals than their counterparts in responders, effectively behaving as signalling sinks rather than active communicators. The analysis identified the IL16 and TRAIL, encoded by TNFSF10, pathways among the upregulated inferred interactions involving non-responder non-classical monocytes, classical monocyte subsets, and CD4-positive T cells. The authors are careful to frame these findings as transcriptomic ligand-receptor hypotheses rather than demonstrations of protein secretion or causal signalling, but the pattern resembles the behaviour of exhausted or senescent cells that accumulate inhibitory inputs while failing to respond effectively, a phenomenon previously described in severe tuberculosis where CD14-positive CD16-positive monocytes correlate with T-cell inhibition and immune paralysis.

The study also documented profound changes in the lymphocyte compartment. Overall T cell abundance was reduced in non-responders, yet subclustering revealed a significant expansion of CD8-positive cytotoxic T cells alongside CD4-positive regulatory T cells with stronger inhibitory profiles. Exhaustion-associated markers told a consistent story: HAVCR2, which encodes the TIM-3 checkpoint receptor, was broadly elevated in non-responders, particularly in cytotoxic T cells and regulatory T cells, while TIGIT showed more restricted expression in the same subsets. Critically, LGALS9, the ligand for TIM-3, was significantly upregulated in regulatory T cells from non-responders, and CellChat inferred LGALS9-dependent interactions, including LGALS9-P4HB, LGALS9-CD45, and LGALS9-CD44, primarily targeting CD8-positive cytotoxic T cells. At the single-cell level, LGALS9 expression correlated positively with HAVCR2 in non-responders, with a correlation coefficient of 0.471, whereas the corresponding association in responders was weaker and statistically non-significant, and a similar group-specific pattern held for TIGIT.

Innate lymphocytes were not spared. NK cell subclustering identified immature and mature subsets, with NKT cells more abundant in non-responders and a mature NK subcluster showing clear spatial separation between groups. CellChat inferred stronger GALECTIN signalling, including LGALS9-HAVCR2 interactions, from regulatory T cells to this mature NK subset in non-responders, and the affected cells displayed higher HAVCR2 and TIGIT expression alongside reduced CCL5 and CXCR4, with pseudotime analysis placing them towards the terminal end of the inferred differentiation trajectory. To contextualize these cross-sectional findings temporally, the team reanalyzed the public microarray dataset GSE40553 using CIBERSORT deconvolution, which suggested that estimated non-classical monocyte proportions decline after two months of anti-tuberculosis therapy and remain low through twelve months. The authors emphasize that this bulk-transcriptome analysis provides descriptive temporal context only and cannot validate the single-cell-defined functional states or test individual conversion outcomes.

An exploratory in silico layer added further hypothesis-generating depth. Virtual knockout analysis using scTenifoldKnk predicted that perturbing IL16 would disturb 24 genes and TNFSF10 would disturb 34 genes at false discovery rate significance, with representative affected genes including TCF7, IL7R, and CCR7 for IL16, and NKG7, PRF1, and GNLY for TNFSF10, pointing towards consequences for T cell and cytotoxic lymphocyte function. Ninety perturbation-associated genes were shared between the two analyses and were enriched for cellular defence response and response to molecules of bacterial origin. The authors stress that these computational results prioritize candidates for experimental study but do not establish gene function, protein effects, or therapeutic benefit.

The authors are admirably forthright about the limitations of their work. With eight patients in total, four per group, the study lacks the statistical power for definitive conclusions, and no protein measurements of IL16, TNFSF10, or LGALS9 were performed, nor were experimental gene perturbations conducted. Trajectory inference methods are sensitive to sampling density and modelling assumptions, making the terminal differentiation interpretation provisional, and observational cross-sectional data cannot determine whether the observed immune changes are causes or consequences of persistent infection. Nevertheless, the study nominates non-classical monocyte composition and IL16 and TNFSF10-related transcripts as candidate biomarkers worthy of prospective validation, and proposes a testable network model in which dysregulated monocyte signalling coexists with regulatory T cell-mediated galectin signalling and exhaustion features in cytotoxic lymphocytes. If larger cohorts with protein-level measurements and functional assays confirm these associations, the findings could open the door to host-directed therapies or early immunological monitoring that identifies at-risk patients before treatment failure becomes entrenched, a development that would materially advance the global effort to bring tuberculosis under control.

Subject of Research: Single-cell immune profiling of tuberculosis patients to identify immunological determinants of sputum culture conversion and treatment response

Article Title: Single‐Cell Immune Profiling Suggests Non‐Classical Monocyte Is Associated Immunodepression in Tuberculosis Treatment Non‐Responders

Article References: Single‐Cell Immune Profiling Suggests Non‐Classical Monocyte Is Associated Immunodepression in Tuberculosis Treatment Non‐Responders. (n.d.). https://doi.org/10.1111/jcmm.71345

Image Credits: AI Generated

DOI: 10.1111/jcmm.71345

Keywords: tuberculosis, single-cell RNA sequencing, non-classical monocytes, sputum culture conversion, immune suppression, IL16, TNFSF10, LGALS9, regulatory T cells, treatment non-responders, immunodepression, host-directed therapy