Monocyte pathology in human tuberculosis is due to plasma milieu changes and aberrant STAT signalling

Monocyte‐derived macrophages contribute centrally to immune protection in Mycobacterium tuberculosis infection and changes in monocyte phenotype characterize immunopathology in tuberculosis patients. Recent studies highlighted an important role of the plasma milieu in tuberculosis immunopathology. Here, we investigated monocyte pathology in patients with acute tuberculosis and determined tuberculosis plasma milieu effects on phenotype as well as cytokine signalling of reference monocytes. Patients with tuberculosis (n = 37) and asymptomatic contacts (controls n = 35) were recruited as part of a hospital‐based study in the Ashanti region of Ghana. Multiplex flow cytometry phenotyping of monocyte immunopathology was performed and effects of individual blood plasma samples on reference monocytes prior to and during treatment were characterized. Concomitantly, cell signalling pathways were analysed to elucidate underlying mechanisms of plasma effects on monocytes. Multiplex flow cytometry visualization characterized changes in monocyte subpopulations and detected higher expression of CD40, CD64 and PD‐L1 in monocytes from tuberculosis patients as compared to controls. Aberrant expression normalized during anti‐mycobacterial treatment and also CD33 expression decreased markedly. Notably, higher CD33, CD40 and CD64 expression was induced in reference monocytes when cultured in the presence of plasma samples from tuberculosis patients as compared to controls. STAT signalling pathways were affected by the aberrant plasma milieu and higher levels of STAT3 and STAT5 phosphorylation was found in tuberculosis plasma‐treated reference monocytes. Importantly, high pSTAT3 levels were associated with high CD33 expression and pSTAT5 correlated with CD40 as well as CD64 expression. These results suggested plasma milieu effects with potential implications on monocyte phenotype and function in acute tuberculosis.

aberrant plasma milieu and higher levels of STAT3 and STAT5 phosphorylation was found in tuberculosis plasma-treated reference monocytes. Importantly, high pSTAT3 levels were associated with high CD33 expression and pSTAT5 correlated with CD40 as well as CD64 expression. These results suggested plasma milieu effects with potential implications on monocyte phenotype and function in acute tuberculosis.

INTRODUCTION
Immune surveillance in pulmonary granulomas is central for protection against Mycobacterium tuberculosis and to prevent tuberculosis disease progression after infection [1]. The granuloma is characterized by complex cellular composition including several specialized monocyte-derived macrophage subsets. Both, tissue resident macrophages as well as circulating monocyte-derived macrophages (MDM), contribute to local anti-mycobacterial host response, and MDM require steady replacement of shortlived macrophages from the blood monocyte pool [2]. Changes in blood monocyte phenotype and function would, therefore, likely also affect local MDM subsets with potential implications for immune surveillance.
Immunopathology of peripheral blood immune cells in acute tuberculosis are well described. These include alterations in the distribution of immune cell populations as well as changes in phenotype and function [2][3][4]. Amongst others, higher monocyte proportions (and increased monocyte/T-cell ratios) have been described [4]. In addition, aberrant expression of monocyte subset markers CD14/CD16 was detected, and a subset of small 'inflammatory' monocytes were found in acute tuberculosis [3,[5][6][7][8]. Furthermore, functional relevant receptors (e.g., Programmed cell Death Ligand (PD-L)1, FC-gamma receptor 1 (CD64)) were differentially expressed in monocytes of tuberculosis patients [9,10]. Higher PD-L1 expression was shown to have implications on antimycobacterial effector functions [11,12]. High CD64 expression was detected in some studies and was identified as a potent marker that may contribute to diagnosis of tuberculosis disease [10,[13][14][15]. Own previous studies showed that monocytes and T cells of tuberculosis patients had lower Interleukin (IL)-7 receptor expression and impaired response to IL-7 [5,16]. These findings raised the question about a common cause for changes in different cell subsets and mechanisms that affect immune cells from the peripheral blood without direct contact to the pathogen. For T cells, we could show that cytokine signalling was affected in tuberculosis. High constitutive STAT3 phosphorylation as well as high Suppressor of Cytokine Signalling (SOCS)3 levels were found in acute tuberculosis patients and aberrant pSTAT3 levels were associated with high IL-6/IL-10 plasma concentrations [17]. In addition, we could show recently that aberrant high IL-6 plasma levels were associated with impaired mitogen-induced IFN-γ expression by T cells in patients with tuberculosis [18]. Changes in the plasma milieu and aberrant cytokine concentrations in tuberculosis patients have been described [19][20][21]. These results rendered a role of the plasma milieu in tuberculosis immunopathology likely and suggested similarity to inflammatory syndromes in different diseases (reviewed in Ref. [22]). Recently, we established an in vitro assay based on tuberculosis patients' plasma samples, which were added to the culture medium of reference monocytes and compared to matched control plasma samples [23]. This assay confirmed tuberculosis plasma specific effects on monocyte HLA-DR/IL-7 receptor expression and strengthened the hypothesis that the plasma milieu in tuberculosis is responsible for immunopathology seen for T cells and monocytes [23].
Against this background, we investigated if the aberrant monocyte phenotype in tuberculosis patients is caused by plasma milieu factors in the present study. Several monocyte markers were included for phenotyping of monocytes from tuberculosis patients and controls as well as for analysing the effects of plasma samples on reference monocytes. STAT1, STAT3 and STAT5 phosphorylation was determined to characterize the underlying mechanisms.

Study cohorts and clinical characterization
We recruited tuberculosis patients (n = 40) and asymptomatic contacts of indexed tuberculosis patients (controls, n = 35) from July 2019 to March 2022 at the Agogo Presbyterian Hospital, the St Mathias Catholic Hospital, the Atebubu District Hospital, and the Sene West District Hospital in Ghana. Diagnosis of active tuberculosis was based on patient history, clinical examination, chest x-ray and sputum smear test. GeneXpert (Cepheid) analyses were done for all tuberculosis patients. All patients were included prior to initiation of treatment and blood was taken at baseline (BL), as well as 6 and 16 weeks after treatment onset. EDTA plasma samples were collected at baseline as well as during treatment and were cryopreserved in À80 C freezer until used. Controls were close relatives living in the same household with indexed tuberculosis patients according to self-report and direct observation. Controls had no history of tuberculosis and showed no symptoms at recruitment. None of the controls presented with symptoms of tuberculosis during the study period at one of the participating hospitals. Study group details are summarized in Table 1  Cells were subsequently washed and measured using a four-laser flow cytometer (CytoFlex S; Beckman Coulter). FlowJo software (Version 10, Becton Dickinson) was used to analyse the data. A representative gating strategy is depicted in Figure S1.
Plasma milieu effects on reference monocyte phenotypes and cytokine signalling Using our previously established monocyte in vitro assay [23], we characterized the effect of plasma milieu on monocyte phenotypes and cytokine signalling. A schematic depiction of the assay is shown in Figure 3. Monocytes were enriched (using magnetic cell sorting (EasySept Monocyte negative selection kit; Stemcell Technology) as described before [23]) from peripheral blood mononuclear cells (PBMCs) purified from peripheral blood of healthy individuals by density gradient centrifugation (Histopaque-1077; Sigma-Aldrich) according to manufacturers' guidelines. Enriched monocytes (5 Â 10 4 per well) were cultured in RPMI 1640 medium (Thermo Fisher Scientific) supplemented with L-Glutamine (2 mM, Sigma Aldrich), Hepes (10 mM, Thermo Fisher Scientific) and 10% of heterologous plasma either from a patient with tuberculosis or a control. After overnight incubation (24 h at 37 C, 5% CO 2 ), monocytes were incubated in ice-cold PBS containing 10 mM EDTA and 0.5% BSA for 30 min to detach adherent cells. Cells were then stained on ice for 30 min in the dark using the same antibody panel as for ex vivo phenotyping (see above). Thereafter, cells were washed and measured using LSR-Fortessa flow cytometer (BD Bioscience). Analyses were performed with FlowJo software (Version 10, Becton Dickinson). A representative gating strategy is depicted in Figure S2.

Data visualization and clustering for identification of putative monocytes subsets and candidate marker analysis
Data analysis was performed using FlowJo software (version 10; BD Biosciences, Franklin Lakes). For phenotyping of monocytes from study groups of tuberculosis patients and controls, monocyte changes under antimycobacterial treatment as well as reference monocytes treated with tuberculosis patients' or controls' plasma supplemented media, we performed combined multiplex data analyses. The following steps (shown as part of Figures S1 and S2) were done. Initially, IDs were assigned to each study group and time points during treatment. Next, viable HLA-DR positive cells were down-sampled to a maximum of 7000 cells (for ex vivo phenotyping) or 2000 cells (for reference monocytes) per replicate using the Downsample v3.3 plugin for FlowJo. All samples were then concatenated to one sample for comparison of study groups (Figures 1a,b and 4b,c) and time points (Figure 2b). We then performed fast interpolation-based t-distributed Stochastic Neighbour Embedding (fitSNE) [24] for reduction of data complexity and visualization of cell phenotype pattern. We applied default settings of the fitSNE FlowJo plugin for all parameters but set the number of iterations to 600 (default 1000). FitSNE depicts similarity of cells based on two parameters (fitSNE parameters 1 and 2; Figures 1a and 4b) and the similarity of cells is illustrated by their distances in the respective graphs.
Smoothed density plots (Figures 1a and 4b), colour density plots ( Figure 1a) and contour plots ( Figure 4c) were used to depict different data sets. Algorithm supported unbiased identification of monocyte subsets was done using flow cytometry self-organizing maps (FlowSOM) algorithm (FlowJo plugin downloaded from www.flowjo. com/exchange/). The number of clusters was arbitrarily set to five. All other parameters were left at manufacturer's default settings. Analyses of the three main subsets predicted by FlowSOM was performed by backgating clustered cells based on assigned IDs for individuals, study groups and replicates.

Graphical depiction and statistics
GraphPad Prism v9 software (GraphPad Software) was used for all statistical analyses. Due to the non-normal distribution of data (tested by Kolmogorov-Smirnov and Sha-piroWilk test), nonparametric tests were used throughout. Study group comparisons were performed by the Mann-Whitney U-test while the Wilcoxon matched-pairs signed rank test was used for paired comparisons. Spearman rank correlation was used to assess association between phenotype marker expression and plasma sample induced candidate markers and STAT phosphorylation levels. A p-value below 0.05 was considered statistically significant.

RESULTS
Increased frequency of putative M1 cells and high CD64, CD40 and PD-L1 expression in monocytes from tuberculosis patients   controls. For data complexity reduction and visualization of study group differences, we applied fast interpolationbased t-Stochastic Neighbourhood Embedding (fitSNE) (see Methods section and Figure S1 for details). FitSNE calculates two parameters from all included markers and density plots depict distribution of monocytes with high similarity for tuberculosis patients and controls (Figure 1a). FitSNE plot comparison revealed marked differences between the study groups with distinct regions showing high density for tuberculosis patients (dark red arrow; left graph) and controls (dark blue arrow; right graph) (Figure 1a). Self-organizing map clustering by FlowSOM was then applied to classify potential monocyte subpopulations based on similarities.
Starting with an arbitrarily set number of five, three subsets (i.e., bright green, orange, red) classified by FlowSOM algorithm made up more than 95% of monocytes in tuberculosis patients and controls (Figure 1a; lower graphs). The bright green subpopulation had a phenotype similar to M1 (CD14 high ), was largely distinct from putative M2 (red; CD16 high ) and showed some overlap with putative M1/2 (CD14 medium /CD16 high ; orange) ( Figure 1b). Distribution of monocyte subpopulation showed marked differences. Whereas the putative M1 population was more frequent in tuberculosis patients (71.2%) as compared to controls (54.2%) (p = 0.002), the putative M2 subset was higher in controls (TB: 13.1%; Controls: 27.5%; p = 0.002; Figure 1b). No differences were detected for the putative M1/2 subpopulation (Figure 1b).
Next, we compared protein expression level of candidate markers on monocytes between the study groups. Notably, three markers, CD64, CD40, and PD-L1, were significantly higher in monocytes from tuberculosis patients whereas CD33, CD16, CD70, HLA-DR, CD11b, CD14 and CD11c showed no differences between the study groups (Figure 1c). Phenotype comparisons between putative M1, M1/2 and M2 subpopulations also showed differential expression for identified markers ( Figure S3a). Higher CD40, CD64 and PD-L1 expression was seen for all monocyte subpopulations in tuberculosis patients except putative M2 cells where CD40 differences did not reach significance levels ( Figure S3a). Interestingly, whereas CD16 expression was not generally different between the study groups (Figure 1b), higher CD16 was detected for putative M1 cell from tuberculosis patients but not for the other subpopulations ( Figure S3b). We concluded that tuberculosis patients were characterized by high expression of CD64, CD40 and PD-L1 as well as increased proportions of putative M1 cells with higher CD16 expression.
CD64, CD40, CD33 and PD-L1 expression in monocytes decreased during treatment of tuberculosis patients Next, we compared monocyte phenotypes from tuberculosis patients prior to treatment (baseline, BL) with an L1, p = 0.0003) whereas a significant CD40 decline was only seen between BL and W16 (p = 0.0193). At W16, CD64, PD-L1 and CD40 expression largely normalized and median values were comparable to healthy controls (dotted line; Figure 2a). Interestingly, CD33 (an inhibitory receptor of the Siglec family) although not differentially expressed between tuberculosis patients and controls (Figure 1a), showed decreased expression between BL and W6 (Figure 2a; p = 0.0093). CD11b, CD11c and CD70 were not different between the time points ( Figure S4). Next, we analysed putative monocyte subpopulations for candidate marker expression differences during treatment. Putative M1, M1/2, M2 subpopulations showed significant decrease in CD64, PD-L1 and CD40 expression between BL and W16 (Figure 2b). For CD33, however, significant differences were only detected for putative M2 between BL and W16 (Figure 2b). These results suggested normalization of the aberrant monocyte phenotype during treatment and strengthen the assumption that differential monocyte marker expression is a specific feature of immunopathology in human tuberculosis.
Overnight culture of reference monocytes supplemented with plasma samples from tuberculosis patients induced high CD64, CD40 and CD33 expression Previous studies suggested plasma milieu effects on immune cell phenotypes in tuberculosis [17,23]. In order to directly test plasma effects on monocyte phenotypes in tuberculosis, we performed cell culture of reference monocytes in medium supplemented with plasma samples from either a tuberculosis patient (n = 30) or a control (n = 30) (a schematic depiction is shown as Figure 3). After overnight culture, we analysed the monocyte phenotype by flow cytometry. Notably, three of four candidates (i.e., CD64, CD33 and CD40) showed higher expression in reference monocytes cocultured with plasma samples from tuberculosis patients as compared to healthy controls (CD64: p = 0.0005; CD33: p < 0.0001; CD40: p = 0.0066; Figure 4a). PD-L1 was not different between plasma samples from the study groups (Figure 4a). Higher expression was also detected for HLA-DR and CD14 in the presence of tuberculosis patients' plasma (both p < 0.0001). This confirmed previous results for plasma-induced HLA-DR expression [23], and indicated that plasma effects on reference monocytes may not completely resemble monocyte phenotype changes in acute tuberculosis. Next, we visualized phenotypic changes of treated reference monocytes and identified marked differences between reference monocytes treated with plasma from tuberculosis patients or controls (Figure 4b). Two main populations, one dominant in reference monocytes cultured with plasma from tuberculosis patients and one in reference monocytes cultured with plasma from controls, were identified (Figure 4b, density plots). Notably, FlowSOM confirmed the visual impression and classified these two main subsets as independent clusters, which comprised approximately 90% of all cells (Figure 4c). Notably, differences in HLA-DR, CD33, CD40 and CD64 expression were the main factors for classification and this confirmed concomitant upregulation of these markers on reference monocytes treated with tuberculosis patients' plasma (Figure 4c; histograms). These results suggested causative effects of the plasma milieu in acute tuberculosis for immunopathology seen in monocytes.

STAT3 and STAT5 signalling is higher in reference monocytes cocultured with plasma samples from tuberculosis patients
To characterize underlying mechanisms of plasma milieu effects, we next analysed STAT mediated signalling in reference monocytes cocultured overnight with plasma samples from both study groups. STAT1, STAT3 and STAT5 phosphorylation was determined directly thereafter. Whereas STAT1 showed comparable phosphorylation levels between the study groups, both, STAT3 and STAT5, showed significantly higher phosphorylation in the presence of plasma samples from tuberculosis patients as compared to controls (Figure 5a). Next, we compared plasma samples from tuberculosis patients prior to treatment with samples taken at W6 and W16 after start of therapy. The capacity of plasma samples to induce STAT1 phosphorylation was similar at different time points whereas STAT5 phosphorylation decreased significantly already within 6 weeks of antimycobacterial treatment (Figure 5b). For pSTAT3, a tendency of decreased levels was detected for both W6 and W16 (Figure 5b). These results suggested a role of plasma cytokine signalling via STAT3 and/or STAT5 as underlying mechanisms for the aberrant monocyte phenotype.
To determine a possible association between plasmainduced STAT3/STAT5 signalling and phenotypic changes, we correlated CD64, CD40 and CD33 expression with pSTAT3/pSTAT5 levels in reference monocytes incubated with respective plasma samples. CD40 and CD64 showed a positive correlation with pSTAT5 for both, plasma samples from tuberculosis patients and controls (Figure 5b). For pSTAT3, no correlation was seen for CD64 and only in controls plasma samples for CD40 (Figure 5b). Notably, CD33 expression showed strong positive correlation with pSTAT3 in both cohorts but not with pSTAT5 in tuberculosis patients (Figure 5b). This suggested an association of CD33 expression with induced STAT3 phosphorylation. In summary, different pathways and causative plasma cytokines affect monocyte phenotype in tuberculosis and strengthened the hypothesis that the plasma milieu in acute tuberculosis exerted immunopathology via candidate cytokines, which signal via the STAT3 or STAT5 pathway.

DISCUSSION
The present study provided evidence that the plasma milieu in acute tuberculosis contributes to the immunopathology phenotype of monocytes seen in tuberculosis patients. Higher STAT3 and STAT5 phosphorylation was induced in reference monocytes by plasma samples from  tuberculosis patients as compared to controls and associations were seen with distinct phenotypic markers. This finding suggested that different plasma cytokines are involved in tuberculosis immunopathology. In accordance, a previous study failed to identify individual plasma cytokines from tuberculosis patients associated with induced IL-7Rα expression in monocytes [23]. Distinct association pattern seen for pSTAT3, which correlated positively with CD33 expression, and pSTAT5, which correlated with CD40 and CD64 in the present study, narrowed down potential causative factors.
Evidence for functional relevance of plasma immunopathology in tuberculosis has been provided by previous studies in T cells and monocytes from tuberculosis patients [17,22]. Our group detected constitutive STAT3 phosphorylation in T cells associated with high IL-6 and IL-10 plasma levels [17]. Consequently, increased SOCS3 expression and inhibition of IL-2 induced STAT5 mediated signalling affected M. tuberculosis specific T cells [17]. Constitutive STAT3 phosphorylation of monocytes was detected in a previous study, by Lastrucci et al., that characterized influential factors produced by M. tuberculosis infected monocyte-derived macrophages (MDMs) in reference monocytes [25]. This study detected high pSTAT3 levels together with upregulation of CD16 in monocytes and this resembled monocyte immunopathology features seen in tuberculosis [25]. Functionally, monocytes treated with M. tuberculosis infected MDM supernatants were prone to become anti-inflammatory macrophages in this study and IL-10 was key for these effects [25]. In addition, they detected impaired anti-mycobacterial effector functions of treated monocytes as well as worsened interaction with M. tuberculosis specific T cells [25]. A role of IL-10 was also seen in previous studies on tuberculosis plasma cytokine levels [19] and pathognomonic effects on T cells were found [17]. These findings possibly indicate immunomodulatory mechanisms but since IL-10 is accompanied by aberrant high IL-6 levels in acute tuberculosis [17,19], our findings may alternatively be interpreted as a feature of inflammation [22]. We did not characterize the function of MDM derived from plasma treated monocytes and this was partly due to negative effects of non-inactivated heterologous plasma on monocyte culture. However, others detected functional effects of plasma from tuberculosis patients on monocyte derived dendritic cells [26]. This rendered our assumption likely that aberrant cytokine signalling and accompanied monocyte phenotype changes have functional implications on derived macrophage populations.
Previous studies showed high complexity of peripheral blood monocytes displayed by differences in phenotype, size, and functions [3]. A subpopulation of 'small' monocyte has been identified and characterized by specific effector functions including increased production of pro-inflammatory cytokines [27,28]. This subset of so-called 'inflammatory monocytes' was found to be increased in inflammatory diseases, sepsis and infections [3]. We and others found higher proportions of 'inflammatory monocytes' in tuberculosis [2,5,8,25,29]. Self-organizing map clustering performed in the present study indicated that this subset, which is characterized by concomitant expression of CD14 and CD16, is more similar to M1 (or classical monocytes) and less to M1/2 (or intermediate monocytes characterized by CD14 medium /CD16 high expression). In addition, we showed that M1 proportions are higher in tuberculosis patients as compared to healthy controls. This argued for upregulation of CD16 in enriched putative M1 cells as the mechanisms underlying generation of inflammatory monocytes in tuberculosis. More importantly, the differences in monocyte phenotype from tuberculosis patients were detected for all main monocyte subsets. This suggested subset-independent immunopathology in monocytes and plasma milieu effects on references monocytes strengthened this assumption. In addition, we provided initial evidence for complex mechanisms of plasma and monocyte pathology in acute tuberculosis. Two markers, CD64 and CD40, were strongly induced by tuberculosis plasma samples and also showed higher expression in monocytes from acute tuberculosis patients. Since CD64/ CD40 expression was positively correlated with plasma induced pSTAT5 levels, it is likely that pSTAT5 signalling dependent plasma cytokines are causative for aberrant high levels in acute tuberculosis. Notably, and in contrast to pSTAT3, previous studies did not detect higher constitutive pSTAT5 levels in monocytes (or T cells) from tuberculosis patients [5,17]. This could either be explained by impaired response to causative cytokine(s) or negative regulation of STAT5 phosphorylation. Both has been described for IL-7-mediated T-cell response in tuberculosis [16]. T cells from tuberculosis patients are characterized by low IL-7 receptor expression and impaired response to high IL-7 plasma levels [16]. In addition, the key regulator of STAT signalling SOCS3 was increased in T cells from tuberculosis patients [30] with potential negative effects on pSTAT5 induction [17] and IL-7 receptor expression [31]. Monocytes from tuberculosis patients also have low IL-7 receptor expression and show impaired IL-7-induced pSTAT5 [5]. These findings render IL-7 a promising plasma cytokine candidate involved in the induction of STAT5 phosphorylation.
Another tuberculosis plasma-induced factor in the present study was CD33. Monocytes from tuberculosis patients showed similar CD33 expression as compared to healthy controls but expression decreased in monocytes during treatment. CD33 was strongly induced by plasma samples from tuberculosis patients and this effect was associated with high STAT3 phosphorylation. Although the role of CD33 is not finally defined, its expression on suppressive monocyte subsets [32] may also indicate that alternative monocyte polarization is induced by the plasma milieu in tuberculosis and normalized during antimycobacterial treatment.
In summary, we provided evidence for plasma milieu effects on monocyte signalling and pathology in acute tuberculosis. Further studies are needed to investigate the functional implications in monocytes. The application of identified candidates as markers for tuberculosis diagnosis and treatment is promising since CD64 has already been proven as part of a biomarker signature of tuberculosis [10]. Future studies will investigate the capacity of identified monocyte markers for monitoring treatment efficacy in tuberculosis patients. Furthermore, these studies should also analyse associations between plasma induced signalling pathways and monocyte pathology during treatment and recovery of patients with tuberculosis.