Activin A levels are raised during human tuberculosis and blockade of the activin signaling axis influences murine responses to M. tuberculosis infection

Serum activin A levels were measured in Gambian TB patients at recruitment and 6 months post-treatment, as well as in healthy tuberculin-skin test negative (TST^-) and TST⁺ contacts, the latter being a heterogeneous population which may include people who are infected but healthy (who may progress to disease), as well as those who have cleared infection (29, 30). Full details of the Gambian TB cohort are in Table 1. Serum activin A was significantly higher in patients with active TB than in the healthy TST^- and TST⁺ controls (P < 0.0001) (Fig. 1A). Serum activin A concentrations were diminished in 90% of patients by 6 months post-treatment, with all but three patients showing decreasing serum Activin A after treatment (Fig. 1B). Within 2 years of recruitment, four of the healthy TST⁺ contacts progressed to active TB, diagnosed at days 174, 494, 529, and 671. Only the individual diagnosed with TB at day 174 post-recruitment had raised activin A levels at recruitment (808 pg/mL). Serum activin A levels were higher in patients with a higher X-ray score (XRS) grading of disease severity, both at recruitment (Fig. 1C) and post-treatment (Fig. 1D); therefore, serum activin A levels most likely reflect inflammation in the lungs or severity of disease. The Jonckheere-Terpstra test did not find a significant trend between smear grades and activin A levels (Fig. 1E). A chi-square independence test showed that there was no significant relationship between the smear grades and XRS (P = 0.389). In addition to activin A, 27 other serum cytokines and chemokines were measured using multiplex assays, and tested for their ability to discriminate patients with active TB and healthy TST⁺ contacts. IP-10 and IL-9 levels were raised in those with active TB compared to healthy TST⁺ contacts (P = 0.004; P = 0.011) (Fig. S1A), while IP-10 and IL-1Rα levels were decreased after TB treatment (P = 0.030; P = 0.029) (Fig. S1B). There were no significant differences in the other cytokines measured (data not shown).

We performed receiver operating characteristic (ROC) curve analyses to assess the ability of activin A levels, IP-10 levels, and a combination of activin A and IP-10 levels to discriminate patients with active TB disease from the TST⁻ household contacts and those who were TST⁺ (with possible latent TB infection) (Fig. 1F). Serum activin A levels could discriminate those with active TB from both TST⁻ and TST⁺ contacts (see Table S1). There was a positive correlation between levels of serum activin A and IP-10 (P < 0.0001, R² = 0.395) (Fig. S1C). IP-10 levels were similarly effective at discriminating between patients with active TB from healthy TST⁺ contacts (see Table S1) but not as effective as activin A levels in discriminating TST^- healthy controls from active TB patients. The combination of activin A and IP-10 raised the AUC from 0.85 to 0.93 but was not significantly better in discriminating individuals with active TB from healthy TST⁺ controls than activin A alone (P = 0.100) (Fig. 1C; Table S1). Overall, in Gambian patients, activin A shows promise as a biomarker for active TB.

To validate our findings of increased levels of serum activin A in TB, we examined an independent cohort of TB patients, from Germany. A full description of this cohort is in Table 2. It differed from the Gambian cohort in that not all TB patients were smear positive, though all had TB confirmed by a positive sputum and sputum culture PCR (Xpert). The average age of the German patients and the proportion of males to females was higher (Table S2). Furthermore, while all Gambian patients had drug-sensitive TB, the German patients had an assortment of drug-sensitive, multi-drug resistant (MDR) and extensively drug-resistant (XDR) TB. Despite these differences, serum activin A levels were increased in TB patients compared to controls in the German cohort (Fig. 2A), confirming our finding in the Gambian cohort. However, activin A levels were substantially lower in German patients than in Gambian patients (P < 0.0001), with a median of 425 pg/mL compared to 811 pg/mL in Gambian patients (Table S2). While in the Gambian cohort, the control individuals were sub-divided into TST⁻ and TST⁺, in the German cohort the control individuals were a heterogeneous group of IGRA⁻, IGRA⁺, and untested individuals. Activin A levels tended to be higher in the IGRA⁺ controls than in the IGRA⁻ controls but sample numbers were too low to determine whether this was significant (Fig S2A).

In the German TB patients, we found a positive correlation between serum activin A levels and the general inflammation marker CRP (Fig. 2B). CRP levels were not measured in the Gambian cohort. There was also a positive correlation between X-ray scores (Fig. 2C) and activin A levels. In this study, the Ralph score (31) was used to assess lung pathology. This score is calculated from the percentage of the affected lung, with an extra 40 points if a cavity is present. In this cohort of TB patients, there was a trend toward increased activin A levels at higher smear grade levels (Fig. 2D) as well as a trend toward higher XRS with higher smear grades (Fig. S2B). Overall, the data suggest that activin A levels are a potential biomarker for the severity of TB disease.

As in the Gambian patients, IP-10 levels were elevated in the German TB patients compared to healthy controls (Fig. 2E) and there was a positive correlation between activin A and IP-10 levels (Fig. 2F). ROC curve analysis demonstrated that serum activin A levels were less able to discriminate between TB patients and healthy controls in the German cohort, with an AUC of 0.76 (P < 0.0001) (Fig. 2G). When we stratified the patients according to TB subtype (drug sensitive, MDR or XDR), we found the greatest variance in the drug-sensitive TB group in this cohort (Fig. S2C and D). Activin A levels were significantly higher in the drug-resistant TB groups only (Fig. S2C), while IP-10 was significantly higher in the drug-sensitive TB group (Fig. S2C). For drug-sensitive TB, the combination of activin A and IP-10 could better discriminate TB than activin A alone (P = 0.024) (Fig. S2E; Table S1). For MDR and XDR TB, measuring IP-10 did not add value (MDR P = 0.808, XDR P = 0.889). These findings emphasize the importance of validating biomarker results across different regions and different TB subtypes.

To determine how serum activin A levels in TB compare to those of other lung diseases, we investigated activin A levels in pneumonia and sarcoidosis, infectious and non-infectious diseases, respectively. Pneumonia and sarcoidosis can present with symptoms similar to TB and are relevant for differential diagnosis. No data currently exist in the public domain on serum activin A levels in these diseases; thus, we measured serum activin A levels in German pneumonia and sarcoidosis patients as well as healthy controls. Full details on the pneumonia and sarcoidosis cohorts are in Tables 3 and 4. Overall, serum activin A levels were increased in pneumonia patients compared to healthy controls (P < 0.001; Fig. 3A) but there was high heterogeneity among patients. There was no difference in serum activin A levels within the pneumonia group according to the CRB65 score, a clinical score used to estimate the severity of community-acquired pneumonia (32)(Fig. S3A) or age (Fig. S3B). To determine whether differences were related to pathogen type (viral versus bacterial), we compared levels in patients with proven influenza or pneumococcal pneumonia only (Fig. 3B). In both groups, activin A levels were raised compared to controls (influenza, P = 0.003; pneumococcal, P = 0.011), and there was no significant difference between the groups.

Serum activin A levels in German sarcoidosis patients were also compared with those of healthy controls collected during the TB and pneumonia studies. Sarcoidosis is a systemic inflammatory disorder of unknown etiology, mainly affecting the lungs, and is characterized by tissue infiltrates of lymphocytes and macrophages, and the formation of non-caseating granulomas (33). Sarcoidosis patients showed only a slight increase in serum activin A levels compared to controls (P < 0.01) (Fig. 2C). The individual with an outlying value of 1291 pg/mL activin A had acute renal failure in the month of blood withdrawal, with unknown etiology. There were no differences in serum or bronchoalveolar lavage fluid (BALF) activin A levels between patients with type I and type II X-ray pathology (Fig. S3C). In addition, serum activin A levels did not correlate with BALF activin A levels (Fig. S3D).

Based on the increased levels of serum activin A in active TB disease, we embarked on investigations in mice to characterize possible functional activities of activin signaling pathways in TB. First, it was important to determine whether activin A levels were raised during murine TB, as in human disease. To investigate this, we used stored serum and BALF samples from previous studies (34). Serum activin A levels started to rise as early as day 1, but increased significantly around 2 weeks post-infection and remained elevated up until the end of the experiment (Fig. 4A). BALF activin A levels were higher than those of serum and were also raised by around 2 weeks post-Mtb infection (Fig. 4B). In comparison, BALF activin A levels were increased in Streptococcus pneumoniae (Spn)-infected mice quite quickly compared to naïve controls, peaking at 24 hours p.i., and then beginning to fall (Fig. S3E). Interestingly, the activin A levels mirrored the neutrophil numbers in BAL samples (Nouailles et al, unpublished data). The higher levels of activin A in BALF compared to serum in mice infected with Mtb suggest that activin A may be produced locally in the lungs during TB.

Soluble activin type IIB receptor fused to the Fc portion of human IgG1 (ActRIIB-Fc) has previously been used to ameliorate activin-induced ARDS-like pathology in mice (13). ActRIIB-Fc acts as a ligand trap for several TGF-β/activin family members with overlapping signaling pathways, including activin A, which binds ActRIIB with high affinity (14). To gain a preliminary understanding of the relevance of the activin signaling axis in TB, we investigated the effects of ActRIIB-Fc on acute experimental TB in mice. ActRIIB-Fc was administered i.n. and i.p. prior to infection, and i.p. once a week thereafter, to inhibit activin A systemically and in the lungs (Fig. 5A), which led to decreased lung bacterial loads (Fig. 5B). We analyzed lung cell populations by flow cytometry and found increased numbers of neutrophils compared to PBS-treated mice (Fig. 5C), as well as increased numbers of alveolar macrophages, which can be host cells for Mtb but also play a role in killing the bacilli. The number of CD4 and CD8 T cells (Fig. 5D) and CD4 effector T cells (Fig. 5E) were also increased. Therefore, not only were the activin A levels raised during Mtb infection but the ActRIIB signaling axis influenced cellular responses and bacterial loads.

Because of the numerical increase in T cells after ActRIIB-Fc treatment, we analyzed phenotypes of lung T-cell populations at D14 post-infection using flow cytometry (Fig. S4A through C). The most striking finding was that blockade of ActRIIB signaling pathways in the lungs increased proportions and absolute numbers of CD103⁺ CD4, and CD103⁺ CD8 T cells (Fig. 6A and B). CD103⁺ T cells showed a spectrum of CD69 and CD44 expression (Fig. S4C). Furthermore, absolute numbers of CD4 and CD8 T_RM cells, defined by co-expression of CD103 and CD69, were increased (Fig. 6C), suggesting that ActRIIB-Fc signaling modulates the number of lung T_RM cells. We analyzed CD4 T-cell expression of transcription factors associated with Th1 (T-bet), Th17 (RORγT), and Treg (Foxp3) cells as well as with T-cell proliferation (Ki67), to determine whether sActRIIB-Fc affected T- cell differentiation and proliferation (Fig. 6D; Fig. S5A through C). T-bet expression was significantly decreased in CD4 T cells of sActRIIB-Fc treated mice compared to controls (Fig. 6D), suggesting that the ActRIIB signaling axis may affect Th1 cells, which are considered critical for TB control (35). Interestingly, the downregulation of T-bet is known to promote the expression of CD103 and T_RM development (36, 37). We checked T-bet expression in CD8 T cells and found that it was also decreased in ActRIIB-Fc-treated mice (Fig. 6D), concordant with the numerical increase of T_RM cells. Since TGF-β negatively regulates T-bet expression and directly stimulates CD103 expression via Smad2/3 signaling (36), we measured serum levels of TGFβ1, 2, and 3 to see whether ActRIIB-Fc treatment increased TGF-β levels but found no differences (Fig. S5D through F). By contrast, serum levels of IL-1α, IL-1β, IFN-γ, IL-17, TNF, IL-10, IL-12p40, MIP-1β, and eotaxin were all decreased in ActRIIB-Fc-treated mice (Fig. S6), indicating that ActRIIB signaling pathways affect the cytokine profile during TB, although the lower cytokine levels may also partially reflect the decreased bacterial burdens.

To see whether the effect of ActRIIB-Fc on CD103⁺ T cells was peculiar to Mtb infection or more general, we extended our experiments to examine responses to the vaccine strain M. bovis BCG in the presence of ActRIIB-Fc. T_RM cells are recruited to the lungs following mucosal BCG vaccination (38) and are important in controlling Mtb infection (38, 39). We administered ActRIIB-Fc during i.n. BCG vaccination and checked whether frequencies or numbers of T_RM cells were increased in the lung (Fig. S7A). Compared to PBS treatment, ActRIIB-Fc treatment increased percentages of CD103⁺ cells and T_RM among CD4 T cell populations (Fig. S7B and C), and caused a non-significant increase in the frequency of CD103⁺ cells and T_RM among CD8 T cells (Fig. S7D). Only absolute numbers of CD4 T_RM showed a non-significant increase after ActRIIB-Fc treatment (Fig. S7C and D). We did not pool data from replicate experiments because absolute amounts were different between individual experiments, although the trends were the same. We performed immunohistochemical staining and quantification of CD103⁺ cells in the lungs and found a non-significant increase in CD103⁺ cells in ActRIIB-Fc-treated mice (Fig. S7E and F), consistent with the flow cytometry data. Finally, we performed immunohistochemical staining of lung tissue sections for TGF-beta and did not observe significant differences in TGF-β staining between ActRIIB-Fc-treated and PBS control groups (Fig. S7G and H). The percentage of TGF-β-positive cells and the percentage of TGF-beta-positive pixels were not increased by blocking the ActRIIB signaling axis. In summary, ActRIIB-Fc increased frequencies of CD4⁺ T_RM-like cells after intranasal BCG vaccination, and this appeared to be independent of TGF-β levels.

Gambian pulmonary TB and household contact cohort (Medical Research Council Unit The Gambia - MRCG): The demographic data of the cohort are shown in Table 1. Only HIV-negative patients were included in the cohort. The ethnicity of all the individuals was black African. In all, 30 adults with smear-positive (and subsequently culture confirmed) drug-sensitive TB were recruited and followed up to completion of treatment (standard regimen). Their TB-exposed household contacts were also recruited and analyzed for infection status by tuberculin skin testing (TST). Two independent physicians scored the X-rays according to the guidelines of the National Tuberculosis and Respiratory Disease Association of the United States (62). Blood was centrifuged in BD Vacutainer SST Tubes with Hemogard Closure (Gold) (Becton Dickinson) and serum was collected. Healthy contacts were followed up to 2 years and conversion from TST⁻ to TST⁺ or progression from TST⁺ to TB was noted. All TB cases tested (28/28) were culture negative at 6 months post-treatment (results were not available for two cases). Three individuals among the healthy household contacts were excluded due to pregnancy, which strongly raises activin A levels. One individual in the TST⁺ contact group was excluded due to diagnosis of TB on day 1 post-recruitment. Accordingly, serum activin A levels were measured in 30 TB patients at recruitment and 6 months post-treatment as well as in 29 healthy TST⁻ and 27 healthy TST⁺ contacts. Bioplex was performed on 15 samples per group.

German pulmonary TB cohort: Serum samples from patients with pulmonary TB (n = 47) and healthy controls (n = 27) were provided by the Research Center Borstel. The demographic data of the cohort are shown in Table 2. Since data on patient ethnicity are not usually collected in Germany due to the country’s history of racial profiling during the “Third Reich,” we are unable to provide a complete breakdown of the ethnicity of the patients. However, the majority of patients and all of the controls had Caucasian ethnicity (personal communication from Professor Christoph Lange). Only HIV-negative patients were included in the cohort. All patients had pulmonary tuberculosis and were sputum culture and sputum PCR (Xpert) positive. Of the controls, 4 were interferon-gamma release assay (IGRA) positive, 8 were IGRA negative, and 15 were not IGRA tested. The Ralph score (31) was used to quantify lung disease, calculated as the percentage of the lung being affected plus 40 points, if a cavity was present. Two independent physicians scored the X-rays.

Pneumonia cohort: The CAPNETZ foundation (www.capnetz.de)(63) provided serum samples of adult patients (≥18) with community-acquired pneumonia (CAP) taken within 24 hours after diagnosis. The characteristics of the cohort are shown in Table 3. Patients were identified by clinical signs (cough, purulent sputum) and a positive lung radiograph, and had not had inpatient treatment in the hospital for the previous 28 days. In all, 80 consecutive patients with pneumonia were analyzed; the causative pathogen was not recorded in all patients. Samples were collected before the COVID-19 pandemic. In addition, we analyzed serum from 25 patients considered to have proven streptococcal pneumonia and 25 patients considered to have proven influenza. The CRB65 scores in these patients were 0–2. CRB65 is a clinical score used to estimate the severity of CAP based on confusion, rate of respiration, blood pressure, and age (32). Control samples were obtained from healthy volunteers (n = 20). Clinical and laboratory parameters of the CAPNETZ patients are stored in an electronic database (64).

Sarcoidosis cohort(s): The samples were from the “Orphan Lung Biobank Freiburg” (ethics approval number 3/10). A standardized protocol was used for collecting bronchiolar lavage fluid (BALF) samples (65). Diagnosis of pulmonary sarcoidosis was based on clinical and radiological criteria with histological confirmation in lymph node or lung biopsies, as suggested by the current consensus statement of the American Thoracic Society/European Respiratory Society on sarcoidosis (66). Patients were categorized according to the Scadding radiological types of disease (67). Nine patients had been diagnosed with type I sarcoidosis (bilateral hilar adenopathy), eight with type II (additional involvement of pulmonary parenchyma), and one with type III (involvement of pulmonary parenchyma and fibrosis). The demographic data of the cohort are shown in Table 4.

Mycobacterium tuberculosis (Mtb) H37Rv (American Type Culture Collection; catalog no. 27294) and BCG Danish 1331 (BCG SSI) (American Type Culture Collection; catalog no. 35733) were grown in Middlebrook 7H9 broth (BD) supplemented with albumin-dextrose-catalase enrichment (BD), 0.2% glycerol, and 0.05% Tween 80 or on Middlebrook 7H11 agar (BD) containing 10% (vol/vol) oleic acid-albumin-dextrose-catalase enrichment (BD) and 0.2% glycerol. BCG was grown to the mid-log phase, washed with phosphate-buffered saline (PBS), and stored at −80°C in PBS/10% glycerol. Prior to vaccination, BCG was thawed, washed in PBS, and prepared at a dose of 10⁶ colony forming units (CFU) in 100 µL PBS.

Mice were housed in individually ventilated cages.

Activin A was measured using the Activin A Quantikine ELISA Kit (R&D Systems). Additional cytokines and chemokines were measured using the Bio-Plex Pro TM Human Cytokine 27-Plex, Bio-Plex Pro TM Mouse Cytokine 23-Plex, and Bio-Plex Pro TM Mouse TGF-β3-plex Immunoassays (Bio-Rad).

Spleens and lungs were homogenized in PBS/0.05% Tween80 (PBS-T), prepared as 10-fold serial dilutions, and plated on Middlebrook 7H11 agar containing ampicillin. CFUs were counted 3–4 weeks after incubation of the plates at 37°C.

The left lung was cut into small pieces and digested for 1 hour at 37°C in Iscove’s Modified Dulbecco’s Medium (IMDM) (Gibco) containing 13 µg/mL DNase I (Sigma-Aldrich) and 50 U/mL collagenase IV (Sigma-Aldrich). Single-cell suspensions were prepared by passing digested tissue through 70 µM cell strainers (Corning Falcon). Isolated lung cells were stained with fluorescent antibodies (see Online Supplement) in FACS buffer (PBS/0.1% BSA) containing Fc block (anti-FcγRII/III, clone 24G2, produced in-house) and 1% rat serum and detected by flow cytometry. Intracellular staining of transcription factors was performed using the Transcription Factor Buffer set (BD Pharmingen). Cells were acquired on a Cytoflex (Beckman Coulter) (BSL3 experiments) or LSRII (BD Biosciences)(BSL2 experiments). Data were analyzed using Flowjo software (Tree Star Inc., Ashland, OR).

Paraffin blocks were cut at 1 µm, and sections were mounted and dried on Superfrost Plus slides (Thermo Scientific). After dewaxing and rehydration, sections were treated with the Polyview Plus HRP Kit (Anti-mouse ENZ-KIT160-0150, anti-rabbit ENZ-KIT159-0150 Enzo, Farmingdale, N.Y. USA) following the manufacturer’s manual. CD103 (rabbit) ab224202 Abcam (Netherlands) and TGF beta-1 monoclonal antibody (mouse) MA1-21595, Invitrogen, Rockford, IL, USA, were used as primary antibodies, with an incubation time of 30 minutes at room temperature. Incubation with a blocking buffer instead of primary antibody was used as a negative control in the TGF-β staining. Sections were scanned at 40× magnification with a Zeiss Axioscan Z1 equipped with plan-apochromatic objectives. Quantification was performed with open-source software QuPath v 0.3.2 (69).

GraphPad Prism 8 was used for statistical analysis. For comparing two groups, the Mann-Whitney test or unpaired one-tailed t-test was used, as indicated in figure legends. For comparing multiple groups, the Kruskal-Wallis test with Dunn’s multiple comparison test or the one-way ANOVA with Tukey’s multiple comparisons test was used, as indicated in figure legends. Linear regression was used for correlation calculations. The logistic regression model was used to determine the ability of activin A levels, IP-10 levels or activin A and IP-10 levels in combination with discriminate between groups. For the obtained models, the receiver operating characteristics (ROC) were calculated together with the area under the curve (AUC). The non-parametric DeLongs algorithm was used to assess statistical differences between ROCs (70, 71), using the pROC and ggplot2 package in R 4.2.2. For binary variable dependency, the Jonckheere-Terpstra trend test was used; for other cases, the Spearman rank correlation was calculated, accompanied by linear regression. P < 0.05 was considered significant.

For the clinical samples, all subjects, or their legal representatives, gave written informed consent in accordance with the Declaration of Helsinki, and study protocols were approved by local ethics committees. The tuberculosis study was approved by the MRCG/Gambian government joint ethics committee (reference number SCC1333). The pneumonia study was approved by central and local ethics committees (Ethics Committee of the Hannover Medical School (MHH); registration number: 301–2008). The sarcoidosis samples were taken from the “Orphan Lung Biobank Freiburg” (ethics approval number 3/10). Samples from the Research Center Borstel were collected for a tuberculosis biomarker project (Ethics Committee of the University of Lübeck, AZ 12–233, Germany).

All mouse experiments were ethically approved by the State Office for Health and Social Services, Berlin, Germany (project numbers G0266/11, G0139/14, G0376/13, and G0250/16). Mice were handled in accordance with the European directive 2010/63/EU on the Care, Welfare and Treatment of Animals. All efforts were taken to minimize their discomfort.