Tuberculosis (TB), one of the most prominent communicable diseases in existence, is caused by Mycobacterium tuberculosis (M.tb) infection (1). While some people remain asymptomatically infected, others develop active disease. A major aim of TB research efforts is to characterize host immune components contributing to TB disease susceptibility and resistance (2). Host immune modulators are currently explored as a popular approach to enhance immune responses through novel vaccines and drugs thereby reducing susceptibility to infection, delaying disease progression and accelerating cure (2–4).

Myeloid derived suppressor cells (MDSC) are a heterogenous population of immature myeloid cells, with potent immune suppressive functions (5). MDSC are further grouped into subsets, such as immature MDSC progenitor cells known as early-stage MDSC (e-MDSC), polymorphonuclear MDSC (PMN-MDSC) and monocytic MDSC (M-MDSC) (6–8). The most recently described subset of eosinophilic MDSC (Eo-MDSC) expand our understanding of the suppressive granulocyte lineage (9). A multitude of immunosuppressive mechanisms have been reported for tumor-derived MDSC, presumably reflective of the heterogeneity and functional redundancy of this cell population. While these mechanisms are designed to suppress excess inflammation, they are frequently demonstrated to rather inactivate pivotal immune responses required for protection against tumor cells and pathogenic infections alike (10, 11).

As a major immunotherapeutic target in the field of oncology, studies of tumor-derived MDSC suggest that the factors involved in their differentiation, expansion and activation can be characterized in two classes. Firstly, factors stimulating myelopoiesis and inhibiting differentiation from immature myeloid cells (IMC) into mature myeloid cells (MMC) (5, 12, 13) enhance the accumulation of IMC in lymphoid organs and the tumor microenvironment (cancer). Such factors include interleukin (IL)-6 and vascular endothelial growth factor alpha (VEGFA) which activate the signal transducer and activator of transcription factor 3 (STAT3) signaling pathway (14–17). Secondly, factors such as such as the transcription factors c-myc and C/EBP homologous protein (CHOP), reactive oxygen species (ROS), inducible nitric oxide synthases (iNOS), arginase, IFN-y, TNF-α, IL-13, S100A8/9, transforming growth factor beta (TGF-β), IL-10, and prostaglandin E2 (PGE2) are involved in the pathological expansion and activation of hematopoietic stem cells (HPC) into IMC and eventually MDSC (5, 16). In cancer, these factors are secreted by activated T cells and tumor stromal cells (5, 14–16).

Suppressive myeloid cells have long been known to be expanded in cancer (18, 19), but the identification of their expansion in other inflammatory conditions including infectious diseases such as active TB disease, are more juvenile. Increased MDSC frequencies have been measured in the peripheral blood of patients with active pulmonary TB disease (20–22), however, little is known about the true phenotype and immunosuppressive mechanisms of these cells in TB. Importantly, the frequency and signaling pathways of MDSC at the site of TB disease (the lung) is also unknown with no indication how this compares to those identified in the blood or to those of related myeloid counterparts such as alveolar macrophages, monocytes, dendritic cells, or neutrophils. In recent studies it was shown that MDSC exhibit monocytic and granulocytic morphologies, although M-MDSC appear more relevant in the context of M.tb-infected participation, due to its phagocytic potentials (20, 23, 24).

In the field of immuno-oncology, the minimally accepted set of surface markers for phenotypic classification of human MDSC is HLA-DR, CD33, CD11b and CD14 (M-MDSC) or CD15 (PMN-MDSC) (25). These markers have also been adopted in the human TB MDSC research field, specifically to characterize blood-derived MDSC. Many additional phenotypic and functional markers have, however, recently been described in oncological studies, relating to the function and ontology of MDSC, the most promising being LOX-1 (PMN-MDSC) and S100A9 (M-MDSC) (26–28). Here we set out to evaluate if the expression of these and other functional markers and functional properties associated with MDSC in cancer, may also be relevant to TB-derived MDSC. Additionally, we hypothesize that MDSC from the blood of TB patients could display similar characteristics to lung-derived MDSC but would differ to those of other control myeloid subsets. We make use of myeloid cells from peripheral blood cells and bronchoalveolar lavage cells from of individuals with active TB or with non-TB MDSC-inducing lung diseases, other than TB, to assess expression of genes previously shown to be linked to MDSC in lung malignancies. Specifically, we compare expression between MDSC, alveolar macrophages (AM) and monocytes, to determine how these compare between myeloid subsets.

The phenotype and function of MDSC in the context of TB remains incompletely defined. Most importantly, PB blood has long been used for measurement of soluble mediators and immune cell profile during TB disease, as proxy for the immunological profile at the site-of-disease ( 38). Previous studies have focussed on profiling actively expressed receptors on the surface of MDSC in the periphery during active TB and their soluble mediators, but few have investigated the gene expression profiles of MDSC isolated from different immunological compartments ( 20, 24, 39, 40). The immune phenotype of MDSC in relation to other related immune cell types have also not been compared between the periphery and the lungs during active TB disease. To address these shortfalls, we performed a pilot, observational study in which we assembled a panel of 43 genes from pathways previously shown to contribute to MDSC phenotype, activation, and function in tumor biology. Since MDSC are actively targeted in immune-oncology, owing to the comprehensive knowledge available on the role of these cells in failed host immunity to cancer cells (41), we aimed to identify which MDSC-related pathways also applies to TB. We measured the expression of these genes in MDSC, AM and monocytes from BALF and PB of patients with TB, and those of patients with lung disease other than TB, previously described to show an upregulation in MDSC. This information is important to inform the selection of HDT strategies in TB, specifically those that may be repurposed from the immune-oncology field should the gene expression profiles of MDSC in TB and cancer (OLD) be similar.

In this study, we measured the expression of 43 target genes previously shown to have relevance in MDSC phenotype and function in immune-oncology, to determine their relevance in TB-derived MDSC. We compared these MDSC expression signatures to those observed in control monocyte and AM populations owing to the majority of MDSC in the blood being of the monocyte lineage. Studies have demonstrated that M-MDSC are the predominant subset in the periphery (20, 21, 24), and data from immune-oncology studies have demonstrated that M-MDSC entering the tumor are capable of differentiating into tumor-associated macrophages (TAMs), supporting the use of alveolar macrophages as a control population from the lung ( 42).

In this study we did not observe any significant differences in gene expression in blood-derived MDSC between active TB patients and those with OLD. We also did not measure any significant changes in gene expression of BAL-derived MDSC between active TB patients and those with OLD. The same was also observed for blood- and BAL-derived monocytes between TB patients and OLD. Although there were only two OLD patients in the dataset, we observed a clustering capable of distinguishing between the two groups of patients using PB. This was true for both MDSC and monocytes which suggests that it may be worthwhile investigating the possibility of a PB, cell type-specific RNA diagnostic signature for TB in a larger study. Interestingly, no such discriminating pattern was observed for either MDSC or AM isolated from BALF. These results suggest that there may still be value in investigating PB-specific RNA biomarkers, specifically for the discrimination of OLD from TB disease, and that perhaps previous studies looking at secreted soluble mediators should rather investigate at a gene expression level.

Monocytes and MDSC derived from the PB of TB patients recruited in this study demonstrated 21 differentially expressed genes at both the 90% and 95% Bayesian HDI level. It is interesting to note that of the 21 differentially expressed genes between MDSC and monocytes in PB, many of the genes were associated with the PMN-MDSC subset specifically. These included FUT4, IFNG, ITGAM, MIR223, PTGS2, VEGFA and MAPK14 to name a few. Conventional studies using PB rely on the cryopreservation of PBMC for batched downstream analyses which is known to result in the loss of the majority, if not the entire, PMN-MDSC subset (43). Because this study made use of PBMC immediately stored in an RNA-preserving buffer, we were able to preserve PMN-MDSC-specific RNA for differential expression analyses. It is clear from these results that the PMN-MDSC subset is severely under-reported in the context of active TB disease, and its functional role has subsequently not been accounted for. It would be prudent for future studies to identify realistic PMN-MDSC proportions in the PB of patients with active TB disease and focus on their immunosuppressive role.

Among the differentially expressed genes were also IFNG and IL17A, which were 21.8 times lower and 10.8 times higher in MDSC than in monocytes, respectively. The former finding was not unexpected. Current knowledge suggests that, during active TB disease, ARG1 and NOS2 play a pivotal role in the suppression of T cell proliferation and IFN-γ production, thereby preventing the efficient killing of intracellular pathogens in macrophages and impairing T cell activation (21, 44–46). The 10-fold upregulation of IL17A alludes to the role of IL-17-producing CD4⁺ T cells (Th17 cells) during tumor development, which drive tumorigenesis (47). In contrast, IL-17A is known to be protective against mycobacterial infection in the host, and a lack thereof increases TB susceptibility (48). Given that the main producing cell type of IL-17A remains CD4⁺ and CD8⁺ T cells, evidence suggests that TAMs (derived from M-MDSC) are capable of producing IL-17A and may therefore be present in the PB of active TB patients (49). Other differentially expressed genes of note were FASLG (9 times lower in MDSC than monocytes), and CPT1A (8.8 times higher in MDSC than monocytes). The expression of FASLG on the surface of MDSC is responsible for inducing the apoptosis of T cells by binding with T cell-expressed Fas (50, 51). Data from FASLG deficient mice demonstrate that the lack thereof results in a significant reduction of MDSC in the tumor microenvironment, but more so of the PMN-MDSC subset, suggesting that the lack of FASLG skews the remaining MDSC population to the M-MDSC, and therefore more immunosuppressive, phenotype (52). Considering that this 9 times reduction in MDSC compared to monocytes occurred within the PB suggests that circulating MDSC lack the FasL-Fas immunosuppressive mechanism, which was supported by our data in Table 5 . This demonstrated that the expression of FASLG on MDSC was 20 times lower in PB compared to BALF. CPT1A is a mitochondrial enzyme involved in fatty acid metabolism, the lipid metabolic state preferred by MDSC to which their immunosuppressive potentials can be attributed (53, 54). The upregulation of this gene was not unexpected as it is in line with immune-oncology findings, but it is promising that these findings are similar between cancer and TB.

In contrast to the abundance of differentially expressed genes between MDSC and monocytes from PB, we identified only 6 genes that were differentially expressed between MDSC and AM isolated from the BALF of active TB patients. These included CAV1, CD36, IL10, IL1B, IL6, and MIR146A. Physiologically, the suppressive function of MDSC is likely to feature most notably at the inflammation site, which in cancer is at the tumor site, or in the case of TB, the lung. We have demonstrated that lung-derived MDSC and macrophages from active TB patients do indeed display differential enrichment in selected gene transcripts, which could inform the immunosuppressive mechanisms of lung-derived MDSC during active TB disease. IL-6, IL-10 and IL1-β are well known cytokines involved in the induction/expansion of MDSC, with IL-10 also playing a role in their effector functions at the site of disease (44, 55–58). Of interest is the observed downregulation of CAV1 and CD36, both of which are involved in lipid metabolism and metabolic reprogramming of MDSC, and mycobacterial infections, and are associated with poor clinical outcomes in malignant tumors (59, 60). Previous mycobacterium-specific research demonstrated that vesicular TLR2/Cav-1 signalling, while dispensable for Mycobacterium bovis internalization specifically, is required for T cell suppressive functions within the MDSC milieu (61). Five out of the six genes were less differentially expressed in the MDSC population compared to the AM, suggesting that AM are responsible for the expression of these genes, including IL-10, possibly to induce the expansion of MDSC at the site of disease, while utilising CAV1/CD36-driven lipid metabolism pathways to create a niche for mycobacterial survival (62, 63), thereby exacerbating disease progression.

The fate of circulating MDSC in active TB patients has not been determined, but the theory is that MDSC released from the bone marrow migrate via the blood stream to the site of infection in the lung. How the transcriptional profile of circulating MDSC compares to MDSC entering the disease site, therefore remains to be determined and is of importance to future HDT strategies. Our comparison of gene expression levels in MDSC from PB vs. BALF showed that, at least for the 10 genes listed in Table 5 , the gene expression pattern observed in PB does not reflect levels seen in the lung. This was confirmed by the observation that there was little to no correlation between the expression levels in MDSC from the two sites for all genes ( Supplementary Table 1 ), as well as the high expression factor differences between the two compartments. It is clear from this data that there may be some phenotype and functional profile differences depending on the anatomical compartment from which MDSC were derived. Unfortunately, though, hierarchical cluster analysis demonstrated that, at least during this study, the gene expression profiles of MDSC in PB compared to BALF could not discriminate patients with active TB from those with OLD. Of note from this comparison is the observation of OLR1 being lower by a factor of 180.9 in MDSC in PB compared to BALF. OLR1 codes for the lectin-type oxidized low-density lipoprotein receptor 1 (also known as LOX-1), and is highly expressed in PMN-MDSC (27). With this in mind, previous studies suggesting M-MDSC are the dominant subset of MDSC within the PB compartment are supported, further suggesting the PMN-MDSC subset may dominate the lung compartment, but this needs to be validated.

A significant limitation of this study is that it was unfortunately not possible to obtain paired BALF and PB samples from the same patients which would have enhanced the power of this study. Working with clinical samples is known to be challenging in the TB field; this is especially true for the collection of BALF samples as the bronchoscopy procedure is more invasive and uncomfortable than venepuncture. Often, paired blood specimens cannot be collected in tandem, and additionally, it is likely that the inability to discriminate gene expression profiles of MDSC between other cells or anatomical compartments, is as a result of the small sample size available during this study. A follow-up study is currently being conducted by our research group making use of paired samples from a larger number of participants to validate the findings observed during this study. In addition, our sample size for the comparative OLD group was incredibly small and each individual presented with a distinct diagnosis. This is problematic for representative sampling and should be considered when interpreting the results obtained during this study. Future studies may do well to focus on obtaining patient samples specifically from those with lung cancer to draw stronger parallels between the TB and OLD group. Lastly, this study would benefit from additional validation studies with a particular need being the use of a larger cohort with paired samples.

The results achieved from this study support the hypothesis that anatomical compartments may drive compartment-specific immunological responses, which have considerable effects on the MDSC population and their immunosuppressive functions specifically. The most interesting findings are that the gene expression profiles of MDSC and/or monocytes from PB are able to discriminate between patients with active TB disease and OLD, and that differential gene expression in MDSC vs monocytes is highly variable compared to differential gene expression in MDSC vs AM, with the lung compartment demonstrating a preferential trend to upregulate genes involved in MDSC expansion and lipid metabolism. Considering that the gene expression profiles of MDSC in PB and BALF are not capable of clustering according to disease groups (TB or OLD), it is feasible to speculate that MDSC from TB share similar gene expression profiles with those from cancer but needs to be validated in a larger study. These are important observations for the TB field, possibly lending support to repurpose and validate cancer immunotherapies as HDT for active TB disease. This study provides a backbone upon which future studies into MDSC-related gene expression during active TB disease should be based, as a guide for both improvement and targeted investigations.

The data presented in the study are deposited in the GEO repository, accession number GSE197113 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE197113).

The studies involving human participants were reviewed and approved by Health Research Ethics Committee of Stellenbosch University. The patients/participants provided their written informed consent to participate in this study.

This study was conceived by NP and designed with support from GS and GW. BL, MM, and ME conducted the experiments and analysed the data with support from AP-N, GS, and TS. LK, GS, and NP wrote the manuscript with comments from GW. SA and STM were responsible for participant recruitment, while BA and CK were responsible for the bronchoscopies performed to acquire the BAL fluid. All authors contributed to the article and approved the submitted version.

This work was supported by the National Institute of Health (NIH) International Collaborations in Infectious Disease Research (ICIDR): Biology and Biosignatures of anti-TB Treatment Response (U01 AI115619), European & Developing Countries Clinical Trials Partnership (EDCTP; CDF1546), and South African Research Chair Initiative (SARChI): Biomarkers for TB (SANRF86535).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.