Matrix Degradation in Human Immunodeficiency Virus Type 1–Associated Tuberculosis and Tuberculosis Immune Reconstitution Inflammatory Syndrome: A Prospective Observational Study

Summary Matrix metalloproteinase (MMP) activity in tuberculosis (TB) diverges according to HIV serostatus and compartment, releasing the matrix degradation product PIIINP into plasma. Plasma PIIINP and MMP-8 are potential predictive biomarkers of TB immune reconstitution inflammatory syndrome, whereas doxycycline inhibits TB-driven matrix degradation.

In this study, we systematically explored MMP activity and immunopathology in HIV-1-associated TB. We hypothesized that HIV-1-associated TB would be characterized by reduced MMP activity at TB diagnosis compared to HIV-uninfected TB, but that increased MMP activity would associate with inflammatory pathology during TB-IRIS. Our insights inform novel approaches to risk stratify and diagnose TB-IRIS, and also host-directed interventions to prevent pathology.

MATERIALS AND METHODS
Full methods are provided in the Supplementary Data. The study was approved by the University of Cape Town Human Research Ethics Committee (REF 516/2011). Cross-sectional study participants were healthy volunteers, patients with symptoms requiring assessment, or patients recently diagnosed with TB (Supplementary Table 1). Longitudinal study participants were ART-naive HIV-1-infected patients with a CD4 count <200 cells/µL and recently diagnosed TB. Longitudinal study visits occurred at TB diagnosis (TB0), ART initiation (ARV0), and 2 (ARV2) and 4 (ARV4) weeks of ART. Induced sputum and venous blood were collected. TB-IRIS diagnosis was assigned retrospectively on case review, using International Network for the Study of HIVassociated IRIS (INSHI) criteria [3]. Chest radiographic inflammation (0-10) and sputum acid-fast bacilli (0-6) were scored as previously described [12].

Laboratory Analyses
Sputum and plasma samples were analyzed by Bio-Rad Bio-Plex 200 using MMP beads (R&D Systems, Abingdon, United Kingdom). Procollagen III N-terminal propeptide (PIIINP) enzyme-linked immunosorbent assays (Cloud Clone Corp) and urine lipoarabinomannan (LAM) assays (Alere Determine TB LAM assay) were performed as per the manufacturers' instructions.

PBMC Stimulation With H37Rv
Cryopreserved peripheral blood mononuclear cells (PBMCs) from a separate cohort of 22 TB-IRIS patients and 22 non-IRIS controls were stimulated with heat-killed H37Rv Mtb, as previously described [13]. Culture supernatants were harvested at 24 hours and MMP-8 quantified by Luminex.

Extracellular Matrix 3D Modeling
Alginate microspheres incorporating healthy donor PBMCs and collagen were generated by bioelectrospray methodology using a Nisco encapsulator [14]. Stimulated cells were preinfected with ultraviolet-killed Mtb. DQ-labeled fluorescent gelatin or collagen (Invitrogen) was incorporated into the microspheres to quantitate matrix destruction [15]. Doxycycline (10 μg/mL or 20 μg/mL) was added to media after microsphere generation.

Statistical Analysis
Statistical analysis was performed using Prism 6 (GraphPad) and Stata version 14 software. Two-tailed Fisher exact or Mann-Whitney U test was performed for key comparisons. Correlations were assessed by Spearman rank-order correlation coefficients. Unadjusted and adjusted linear regression models were fitted to quantify effects and adjust for age, sex, and smoking status. Repeated-measures 2-way analysis of variance with Tukey posttest comparison compared time-points and conditions in the TB granuloma model.

Cross-sectional Study Participants
In the cross-sectional study, 227 participants were enrolled. Of these, 17 were excluded (unable to obtain samples, n = 8; diagnostic uncertainty, n = 9), leaving 210 for analysis (Supplementary Figure 1). Participant demographic and clinical characteristics are described in (Table 1) However, smoking was more prevalent in TB (HIV − ). TB (HIV − ) and TB (HIV + ) were associated with diverse pulmonary pathologies on chest radiograph. Frequency of cavities and median chest radiograph inflammation score were both reduced in TB (HIV + ) compared with TB (HIV − ). CD4 count and the number of cavities positively correlated (r = 0.357, P = .016), suggesting that destructive pulmonary pathology is reduced in advanced TB (HIV + ). Microbiological confirmation of TB was similar for TB (HIV − ) and TB (HIV + ) (Supplementary Table 2). However, sputum smear positivity was more common in TB (HIV − ).
In TB patients, sputum smear and culture positivity were associated with increased sputum MMP-1 and MMP-3, which were positively correlated with acid-fast bacilli score ( Supplementary Figures 3 and 4). Together, these data support a role for sputum MMPs in pulmonary TB-driven matrix degradation. However, divergent MMP upregulation occurs in TB (HIV − ) and TB (HIV + ), with reduced sputum MMP-1 in TB (HIV + ) associated with lesser pulmonary matrix destruction.

Plasma PIIINP Is Elevated in TB and Further Increased in HIV-1-Associated TB
MMP activity releases matrix degradation products such as PIIINP from type III collagen. Analysis of plasma in a subgroup of 73 patients of mixed HIV serostatus in the cross-sectional study (39 TB, 34 control) showed that PIIINP concentrations were elevated in TB patients compared to control patients ( Figure 2A). Median PIIINP values were 25 278 (IQR, 11 787-45 071) pg/mL and 3888 (IQR, 1278-10 367) pg/mL, respectively (P < .0001). When analyzed according to HIV status, both TB (HIV − ) and TB (HIV + ) patients had higher plasma PIIINP concentrations than corresponding controls. However, we unexpectedly found that plasma PIIINP was further elevated in TB (HIV + ) compared to TB (HIV − ) ( Figure 2B), despite the reduced sputum MMP concentrations. To investigate further, we related plasma PIIINP to clinical features. Plasma PIIINP negatively correlated with peripheral blood CD4 count (r = −0.435, P = .006; Figure 2C) and hemoglobin concentration (r = −0.557, P < .0001), and positively correlated with HIV-1 viral load (r = 0.544, P = .002; Figure 2D). Plasma PIIINP was significantly elevated in TB patients with extrapulmonary TB compared to those without ( Figure 2E). Taken together with the sputum MMP analysis, this suggested that elevated plasma PIIINP in TB (HIV + ) was due to increased extrapulmonary MMP activity.

HIV Coinfection Does Not Suppress Systemic MMP Activity in TB
We therefore measured plasma MMP concentrations in the cross-sectional cohort ( Figure 2F and 2G and Supplementary  Figures 5 and 6). In TB (HIV − ), MMP-1, MMP-7, and MMP-8 were elevated compared with HIV-uninfected controls, while plasma MMP-3, -9, and -10 were similar and MMP-2 was reduced. In TB (HIV + ), the collagenases MMP-1 and MMP-8 were elevated in TB (HIV + ) compared with HIV-infected controls ( Figure 2F and 2G) and, in contrast to the findings in sputum, were not reduced compared to TB (HIV − ).

Paradoxical TB-IRIS Is Associated With Systemic Inflammation at TB Diagnosis
In the longitudinal cohort, 57 ART-naive TB patients with advanced HIV-1 (CD4 count <200 cells/µL) were enrolled (Supplementary Figure 1) Clinical signs that characterized TB-IRIS patients were elevated heart rate ( Figure 3A) and respiratory rate ( Figure 3B) at TB diagnosis, as well as at TB-IRIS presentation, compared to non-IRIS controls. In both TB-IRIS and non-IRIS patients, CD4 count increased in the first 2 weeks of ART ( Figure 3C), and HIV-1 viral load reduced ( Figure 3D), although median HIV-1 viral load was higher in TB-IRIS patients than in non-IRIS patients at TB diagnosis and at ARV2. Markedly increased C-reactive protein occurred at TB-IRIS presentation ( Figure 3E). Median lymphocyte counts were lower in TB-IRIS patients at all timepoints ( Figure 3F), whereas neutrophil counts ( Figure 3G) and monocyte counts ( Figure 3H) were increased at ARV2. Therefore, TB-IRIS was frequent, associated with significant morbidity and characterized by marked features of systemic inflammation at TB diagnosis, which partially resolved with TB treatment but recurred at the time of TB-IRIS.

Immunopathology in Paradoxical TB-IRIS Is Associated With Increased MMP Activity
To investigate the mechanism of immunopathology in TB-IRIS, we measured sputum MMPs and plasma PIIINP longitudinally. We observed no consistent association of increased sputum MMPs with TB-IRIS diagnosis (Supplementary Figure 7 and  Supplementary Tables 8 and 9). In contrast, plasma PIIINP was elevated in TB-IRIS compared to non-IRIS patients both at the  HIV viral load (copies/ml) time of TB diagnosis, and during TB-IRIS (ARV2 and ARV4), but not at ART initiation ( Figure 4A). At TB0, patients who later developed TB-IRIS had a median plasma PIIINP more than double that in non-IRIS controls: 43 600 (IQR, 30 021-63 913) pg/mL vs 21 651 (IQR, 17 757-33 196) pg/mL, respectively (P = .036).
To investigate the hypothesis that the elevated PIIINP resulted from systemic MMP activity in TB-IRIS, we examined plasma MMP concentrations. Plasma MMP-1, -3, and -8 were elevated in TB-IRIS compared to non-IRIS patients ( Figure 4B-D and Supplementary Tables 8 and 10). MMP-8 (neutrophil collagenase) was the most significantly increased, in a similar pattern to PIIINP, suggesting that systemic collagenase activity caused matrix degradation and PIIINP production during TB-IRIS. Supporting this, MMP-8 correlated with plasma PIIINP concentration (r = 0.435, P < .0001) ( Figure 4E). As neutrophils may be a source of MMP-8 and neutrophil counts were elevated in TB-IRIS patients, we assessed the correlation between plasma MMP-8 and neutrophil count and percentage. MMP-8 concentration positively correlated with neutrophil count (r = 0.617, P < .0001; Figure 4F) and percentage (r = 0.664, P < .0001).  Time  TB0  ARV0  ARV2  ARV4   IRIS  -+  -+  -+  -+   Time  TB0  ARV0  ARV2  ARV4   IRIS  -+  -+  -+  -+   Time  TB0  ARV0 ARV2 ARV4 We further hypothesized that increased MMP activity in TB-IRIS patients was secondary to increased mycobacterial antigen load. The frequency of sputum smear positivity, culture positivity, smear score, and time to culture positivity was similar between TB-IRIS and non-IRIS patients. However, these indices represent Mtb antigen in the pulmonary compartment. We therefore measured urinary LAM, indicative of disseminated TB, in patients for whom a urine sample was available. In an adjusted regression analysis, IRIS was associated with increased odds of a positive urine LAM finding (odds ratio, 10.9 [95% confidence interval, 1.02-115.88], P = .048; Supplementary  Table 11). In an analysis of TB-IRIS patients only, those LAM positive had higher plasma MMP-3, MMP-7, and MMP-8 than TB-IRIS patients who were LAM negative ( Figure 5A). We next examined the effect of antigen stimulation on MMP activity in TB-IRIS patients. We studied MMP-8 concentrations in PBMC culture supernatants from a previously published cohort of TB-IRIS and non-IRIS controls sampled at the time of TB-IRIS onset [13]. In TB-IRIS patients, MMP-8 secretion was increased following stimulation with heat-killed H37Rv Mtb compared to non-IRIS controls ( Figure 5B

Doxycycline Suppresses Mtb-Driven Matrix Degradation
Doxycycline is a licensed MMP inhibitor and reduces collagenase activity. We studied the inhibitory effect of doxycycline in a 3D cell culture model of TB that recapitulates key components of human granuloma formation using a functional readout of matrix destruction [15]. Stimulation of PBMCs with ultraviolet-killed Mtb increased degradation of gelatin within microspheres over time compared to uninfected cells ( Figure 5C). Doxycycline in cell culture media around microspheres inhibited this breakdown ( Figure 5D). Similarly, doxycycline suppressed Mtb-driven collagen degradation in a dose-dependent manner ( Figure 5E).

DISCUSSION
Despite some advances, treatment of TB remains a great challenge, due to lengthy regimens, poor side-effect profiles, drug interactions, and drug resistance [16]. These problems are further compounded by HIV-1 coinfection [17]. Hostdirected therapies have been proposed as a novel strategy to improve treatment outcome, but their development requires greater understanding of pathological and protective immune responses [16,18]. In this study, we identified key differences between MMPs that cause immunopathology in TB dependent on HIV-1 status and characterized MMPs in TB-IRIS. In HIVuninfected TB patients, MMP activity was prominent in the pulmonary compartment and MMP-1 was dominant, whereas in HIV-1-infected patients, higher plasma PIIINP may represent MMP-driven tissue destruction at extrapulmonary sites, with MMP-8 as the principal protease. We identified PIIINP as a pathological marker of excessive MMP activity at TB diagnosis and during TB-IRIS, with potential to risk stratify individuals prior to ART and also to diagnose TB-IRIS. In addition, the MMP inhibitor doxycycline has potential as a host-directed therapy to prevent TB-IRIS. Our findings of increased sputum MMPs in TB, and reduced sputum MMPs in HIV-1-associated TB, concur with findings in previous smaller studies [10,12]. We previously reported that multiple MMP genes were upregulated in TB-IRIS PBMCs restimulated with heat-killed H37Rv Mtb [13]. MMP-8 was not among the upregulated genes. However, while most MMPs are regulated at the transcriptional level, MMP-8 is predominantly presynthesized in neutrophils and therefore may not be identified by gene expression analysis. We have previously reported compartmentalized inflammatory responses in HIV-infected patients with tuberculous meningitis (TBM), in cerebrospinal fluid (CSF) and plasma, with elevated CSF MMP-1, -7, and -10 in TBM-IRIS compared with non-IRIS controls, although MMP-8 was not studied [19]. Ravimohan et al found that an increase in plasma MMP-8 at week 4 of ART relative to baseline pre-ART levels was associated with increased TB-IRIS risk and abnormal pulmonary function tests after TB treatment completion, consistent with our finding that MMP-8 is a key collagenase in TB-IRIS [20]. In addition, we have previously demonstrated that neutrophils were an important source of MMP-8 in TB and that neutrophilia was associated with poor outcomes [21,22].
Dysregulated innate immune responses have been implicated in TB-IRIS pathophysiology [6,7,9,23]. Elevated innate proinflammatory cytokines, monocyte activation, cytotoxicity, and inflammasome activation have been associated with TB-IRIS, implying global activation of the innate immune response [6,8,9,24,25], but these studies do not identify the ultimate effectors of tissue destruction. An association between TB-IRIS and mycobacterial antigen load has been demonstrated [9,26]. Our results suggest that either Mtb antigen leads to increased MMP activity or that increased MMP activity in patients who develop TB-IRIS causes increased extracellular matrix destruction, thereby increasing detection of urinary antigen. In HIV-1-infected TB patients on ART, elevated hyaluronic acid, a glycosaminoglycan component of the extracellular matrix, was associated with poor outcomes including death [25]. Therefore, matrix degradation products may have a prognostic role in HIV-1-associated TB and indicate high TB-IRIS risk.
Currently, the only established immunomodulatory strategy for TB or TB-IRIS is corticosteroid therapy, used adjunctively in central nervous system and pericardial TB, and in treatment of paradoxical TB-IRIS [27]. Corticosteroids suppress Mtb-driven MMPs and this may contribute to their beneficial effects [28][29][30]. Our 3D model of TB granuloma formation demonstrates the potential of doxycycline to inhibit Mtb-driven matrix degradation. Extracellular matrix integrity favors host cell survival in Mtb infection, and therefore matrix-protective strategies may improve outcome in TB without exacerbating HIV-1-related immune compromise [31]. Doxycycline, a licensed MMP inhibitor, is cheap, safe, and widely available and could be studied as an immunomodulatory adjuvant in TB treatment, with the added benefit of bacteriostatic antimycobacterial activity [12,32].
We report a large study investigating mechanisms of immunopathology in TB. However, as an observational study, we cannot attribute causality, nor exclude the possibility that unmeasured confounding factors contributed to measured associations. We adjusted for sex and smoking status, factors that have been associated with divergent MMP responses, but cannot exclude alternative confounders [33,34]. The incidence of TB-IRIS (59%) was high, causing significant morbidity. Sputum samples were not available from some severely unwell TB-IRIS patients who were unable to expectorate, which may have resulted in an underestimation of effect in the longitudinal sputum analysis.
In summary, our work supports a central role for MMPs in causing tissue damage in TB and TB-IRIS, generating matrix degradation products. Differential MMP expression and compartmentalization occurs in HIV-1-infected patients. Systemic MMP-8 is the dominant protease in TB-IRIS, in contrast to pulmonary-localized MMP-1 in HIV-uninfected TB patients. Matrix degradation products are promising biomarkers of TB-IRIS risk prior to and during clinical onset. Doxycycline, an MMP inhibitor, may prevent immunopathology in TB-IRIS.

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.