Abstract
Increased access to combination antiretroviral therapy in areas co-endemic for tuberculosis (TB) and HIV-1 infection is associated with an increased incidence of immune reconstitution inflammatory syndrome (TB-IRIS) whose cause is poorly understood.
A case-control analysis of pro- and anti-inflammatory cytokines in TB-IRIS patients sampled at clinical presentation, and similar control patients with HIV-TB prescribed combined antiretroviral therapy who did not develop TB-IRIS. Peripheral blood mononuclear cells were cultured in the presence or absence of heat-killed Mycobacterium tuberculosis for 6 and 24 h.
Stimulation with M. tuberculosis increased the abundance of many cytokine transcripts with interleukin (IL)-1β, IL-5, IL-6, IL-10, IL-13, IL-17A, interferon (IFN)-γ, granulocyte-macrophage colony-stimulating factor (GM-CSF) and tumour necrosis factor (TNF) being greater in stimulated TB-IRIS cultures. Analysis of the corresponding proteins in culture supernatants, revealed increased IL-1β, IL-2, IL-6, IL-8, IL-10, IL-12p40, IFN-γ, GM-CSF and TNF in TB-IRIS cultures. In serum, higher concentrations of TNF, IL-6, and IFN-γ were observed in TB-IRIS patients. Serum IL-6 and TNF decreased during prednisone therapy in TB-IRIS patients.
These data suggest that cytokine release contributes to pathology in TB-IRIS. IL-6 and TNF were consistently elevated and decreased in serum during corticosteroid therapy. Specific blockade of these cytokines may be rational approach to immunomodulation in TB-IRIS.
The dual epidemic of HIV-1 infection and tuberculosis (TB) is one of the most formidable health challenges in sub-Saharan Africa 1. The annual risk of TB amongst HIV-1-infected persons in high-incidence settings can exceed 20% and an estimated 1.37 million new cases of HIV-1-associated TB occurred globally in 2007 2, 3. Combined antiretroviral therapy (cART) reduces the risk of TB in HIV-1-infected persons by 59–80% 2, 4, 5 and integrated (as opposed to sequential) anti-TB and cART therapies are associated with a mortality reduction of >50% 6. Earlier initiation of cART will be liable to further increase observation of the HIV–TB immune reconstitution inflammatory syndrome (TB-IRIS). Initially recognised in the context of non-tuberculous mycobacterial infection 7; the risk factors of TB-IRIS include a high antigen load, low nadir CD4 count and short interval between commencement of anti-TB and cART therapies 8–13. TB-IRIS presents temporally as two forms. Paradoxical TB-IRIS occurs in patients who are diagnosed with active TB prior to cART, are improving on TB treatment and then, during early cART, develop an immune-mediated paradoxical reaction with new or recurrent clinical and/or radiologic manifestations of TB. Less well defined is unmasking TB-IRIS, which occurs in a subset of patients who are diagnosed with active TB while on cART (cART-associated TB), presenting with unusually accelerated or inflammatory features of TB during the first 3 months of cART 14.
Paradoxical TB-IRIS has been reported in 8–43% of patients starting cART while on TB treatment 8, 9, 11–13, 15–18. Common features are recurrence of TB symptoms, fever, lymphadenitis, enlarging serous effusions, new or recurrent infiltrates on chest radiograph and subcutaneous or deep tissue abscesses 19, 20. Multiple organ systems may be involved. Frequently patients experience high fevers, persistent tachycardia and weight loss, indicative of significant systemic involvement 20. The duration of paradoxical TB-IRIS is significant, typically 2–3 months 13, 14, 21, 22, but cases lasting >1 yr have been reported 13, 14, 21. Mortality is difficult to estimate as most series are reported from centres experienced in supportive management; however, it can be as high as 10–13% 12, 23. A consensus clinical case definition for paradoxical TB-IRIS has been published 14, which has been independently validated by two research groups 19, 24.
Given that the incidence of TB-IRIS is likely to increase, its pathogenesis is important to understand in order to inform specific therapies. Studies have tended to be either anecdotal or small, which is an important limitation considering the highly heterogeneous nature of the condition. An early feature associated with TB-IRIS was conversion of a negative tuberculin skin test (TST) to strongly positive after cART 18. Hengel et al. 25 noted expansion of terminally differentiated tuberculin PPD specific CD4 T-cells by flow cytometric analysis in TST-positive patients during cART in the absence of TB-IRIS. Subsequent work has documented highly dynamic Mycobacterium tuberculosis antigen specific T-helper cell (Th) type 1 T-cell expansions are clearly associated with cART-mediated immune restoration in TB co-infected persons 26–29, although their absence from some patients with TB-IRIS and their presence in similar patients without TB-IRIS brings into question whether the association is causal 27, 29. One possibility is that such Th1 expansions are associated with defective restoration of regulatory T-cell function. However, TB-IRIS had no difference in the numbers of CD4+FoxP3+ positive cells (assumed regulatory) when compared to similar patients who did not develop the syndrome 27, observations subsequently replicated by others 30, 31. TB-IRIS has also been associated with killer immunoglobulin receptor (KIR)-negative γδ T-cells and antiphenolic glycolipid antibodies 32, 33. In a small subset analysis, TB-IRIS was also associated with a peak of nonspecific inflammatory cytokines/chemokines (tumour necrosis factor (TNF), interleukin (IL)-6, IL-1β, IL-10, RANTES and monocyte chemotactic protein (MCP)-1) 26. This observation led to the comment that, like H5N1 influenza infection and experimental anti-CD28 therapy 34, 35, TB-IRIS may be associated with a cytokine release syndrome 36. We therefore conducted a case–control study to investigate cytokine gene expression and secretion in vitro and serum cytokine concentrations in vivo to investigate this hypothesis in greater depth.
MATERIALS AND METHODS
Patient enrolment and study design
Participants with suspected TB-IRIS were recruited at GF Jooste Hospital, Cape Town, and at the Ubuntu clinic, site B Khayelitsha (both South Africa) between March 2005 and December 2006. The University of Cape Town Human Research Ethics Committee approved the study (REC 337/2004). TB cases were treated with 6 months of standard first-line therapy consisting of isoniazid, rifampicin, pyrazinamide and ethambutol (HRZE) for 2 months followed by HR for 4 months (2HRZE/4HR) while patients with a previous history of TB had streptomycin added to their treatment regimen. All patients were prescribed first line cART combination of stavudine (d4T), lamivudine (3TC), and either efavirenz (EFZ) or nevirapine (NVP). The 22 paradoxical TB-IRIS patients and 22 non-IRIS controls were recruited subject to the following inclusion and exclusion criteria for this study. We included patients who: 1) were cART naïve; 2) had clinically or microbiologically confirmed TB or conformed to WHO definitions for HIV-associated TB and responded to treatment 37; and 3) in the case of TB-IRIS, had a clear and confirmed diagnosis of paradoxical TB-IRIS based on a validated consensus case definition 14. Patients with confirmed or suspected rifampicin resistance or multidrug-resistant TB were excluded. Cases and controls were prospectively enrolled subject to inclusion and exclusion criteria and the availability of sufficient sample to perform analyses (fig. 1). For case–control analysis, 22 randomly selected (of 32 available) sample sets drawn from patients at the time of diagnosis with paradoxical TB-IRIS were compared with 22 sample sets from non-IRIS control patients who participated in a prospective study of 62 HIV–TB patients commencing cART 27. Observation during this study was 12 weeks (84 days) after starting cART and non-IRIS controls were selected from 28 patients with complete follow-up and sufficient sample who did not develop TB-IRIS (fig. 1). Non-IRIS controls were not individually matched to TB-IRIS cases; instead, the range of nadir CD4, days of anti-TB treatment, duration of TB treatment to cART, disease form, age and sex of TB-IRIS cases were used to define similar non-IRIS controls. Baseline viral load was not available in all cases and was therefore not factored. The median interval between starting antiretroviral therapy and the onset of TB-IRIS was 14 days (table 1) 20. The sample selected from non-IRIS control participants was therefore after 14 days of cART. The basis for the sample size of 22 cases and controls was dual. Effect size and variance and thus our power to detect differences between groups was estimated from prior analysis of interferon (IFN)-γ ELISpot responses 27. The second reason was pragmatic: multiples of 22 accommodate 96-well-based assays including blank and standard wells.
Recruitment of participants to the study. TB: tuberculosis; IRIS: immune reconstitution inflammatory syndrome; cART: combined antiretroviral therapy. #: patients in this limb presented to a 24-h service. If timing meant a blood sample could not be transported to, and processed in, the laboratory within 4 h, blood was not drawn.
Blood processing and cell culture
Peripheral blood mononuclear cells (PBMC) were isolated from 30 mL of blood collected in Na–heparin BD vacutainers using the Ficoll separation method. Following isolation, 2.5×106 cells·mL−1 in RPMI/10% FCS were rested overnight in an incubator at 37°C in 5% CO2. The PBMC were restimulated with heat-killed H37Rv M. tuberculosis for 6 and 24 h by infecting PBMC at a MOI = 1 (H37Rv: PBMC). After re-stimulation, PBMC were harvested and lysed in 350 μL of RLT-lysing buffer (Qiagen, Valencia, CA, USA). Cell lysates from both time points were collected for RNA analysis while only the 24-h tissue culture supernatants were preserved at -80°C until they were required for further analysis.
RNA isolation procedure and quantitative RT-PCR
RNA was extracted from PBMC lysates using the RNeasy Mini Kit Spin Protocol for isolation of total RNA from Animal Cells (Qiagen) as per manufacturer's instructions and stored at -80°C until further use. Primers and probes for RT-PCR were purchased from Applied Biosystems as Predesigned inventoried assay reagents. We used the following TaqMan® Gene Expression Assays: IL-1β, catalogue number Hs00174097_m1; IL-2, Hs00174114_m1; IL-4, Hs00174122_m1; IL-5, Hs00174200_m1; IL-6, Hs00985639_m1; IFN-γ, Hs00174143_m1; TNF, Hs00174128_m1; IL-8, Hs01038788_m1; IL-10, Hs00174086_m1; IL-12p40, Hs01011518_m1; lL-13, Hs00174379_m1; IL-15, Hs00542562_m1; IL-17A, Hs00174383_m1; IL-18, Hs01038788_m1; GM-CSF, Hs00171266_m1; TGF-β1, Hs00171257_m1. RNA concentration was determined by Nanodrop and samples diluted to give an RNA working solution of ∼10 ng·μ−1. RT-PCR was performed using the TaqMan® RNA-Ct 1 Step kit Protocol (Applied Biosystems, Foster City, CA, USA). The reaction mixture was prepared using the following outlined procedure: 1 μL of PDAR Assay, 10 μL of 2× buffer, 0.5 μL of RT-enzyme, and 8.5 μL of diluted mRNA for each reaction. β-actin was employed as an endogenous control throughout. RT-PCR was performed under the following universal thermal cycling conditions: Reverse transcription at 48°C for 15 min, enzyme activation at 95°C for 10 min, denaturation at 95°C for 15 s. (40 cycles), annealing/primer extension at 60°C for 1 min. (40 cycles). To gain an idea of overall transcript abundance the difference in threshold cycle (δCt) values (Ct (β-actin) - Ct (gene of interest)) were compared. Lower δCt values therefore reflect greater transcript abundance. The fold induction of genes was calculated by the δδCt method and values normalised by log10 transformation.
Analysis of tissue culture supernatant and serum samples
Assays for IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, IL-12p40, IL-13, IL-15, IFN-γ, TNF and GM-CSF were performed in 96-well filter plates, on the Bio-Plex platform (Bio-Rad Laboratories, Hercules, CA, USA), using customised MilliplexTM kits (Millipore, St Charles, MO, USA), according to the manufacturer's instructions.
IFN-γ, IL-17A and TGF-β ELISA
IFN-γ in culture supernatants was measured as previously described 38 by ELISA using purified anti-human IFN-γ antibody pairs from BD Pharmingen (Catalogue numbers 551221). IL-17A was determined using the IL-17A ELISA kit (eBioscience, San Diego, CA, USA) as previously described 39. TGF-β1 was measured using a DuoSet ELISA Development System (DY240; R&D Systems, Amersham, UK) according to the manufacturer's instructions. Briefly, 96-well microplates were coated overnight with 2 μg·mL−1 mouse anti-TGF-β1 capture antibody, followed by washing and blocking the wells. Standards and samples were added undiluted (and not acid activated, in order to measure the bioactive TGF-β1) and the plate was incubated overnight at 4°C, followed by washing and addition of 300 ng·mL−1 biotinylated chicken anti-human TGF-β1 detection antibody to all wells for 2 h at RT. The wells were washed again and streptavidin conjugated to horseradish peroxidase was added for 30 min, before a final wash and development of the wells with TMB substrate solution (DY999; R&D Systems). Colour development was stopped using 2N H2SO4 and optical densities read at 450 nm. The sensitivity of the assay was 16–24 pg·mL−1.
IL-13 and IFN-γ ELISpot assays
ELISpotPlus kits (Mabtech AB, Nacka, Sweden) for human IL-13 (product code: 3470-2AW-Plus) and human IFN-γ (product code 3420-2AW) were used for the determination of IL-13 and IFN-γ spot-forming cells respectively as previously described 28.
Statistical analysis
Sample size calculation was based upon existing similar data. For log-transformed gene fold induction, there was a typical standard deviation range of 0.1–1. The sample size therefore allowed us 80% power to detect a ≥2-fold difference between groups when standard deviations were small and a 10-fold difference when they were large. The normality of data was assessed by the D'Agostino and Pearson omnibus normality test. Data are presented as median (IQR) and mean±sd. Paired parametric data were analysed by paired t-tests, or repeated-measures ANOVA. Paired non-parametric variables were analysed by Wilcoxon signed rank test or Friedman test. Unpaired parametric variables were assessed using the unpaired t-test for parametric data while the Mann–Whitney test was used for analysis of unpaired non-parametric data. To factor multiple comparisons, p-values were multiplied by n-1. GraphPad Prism® Version 5 Software (La Jolla, San Diego, CA, USA) was used for all statistical tests.
RESULTS
Participant characteristics
22 TB-IRIS cases sampled on the day of presentation (a median of 14 days after commencing cART) and 22 non-IRIS controls sampled after 14 days of cART were studied. Cases and controls did not differ in baseline CD4, age, sex, TB disease form, days of anti-TB therapy or days of cART. Cases were more likely to have microbiologically confirmed TB (table 1).
Cytokine transcript abundance and fold induction in vitro
PBMC from TB-IRIS and non-IRIS control patients were cultured in the presence or absence of heat-killed M. tuberculosis H37Rv (M. tuberculosis, MOI 1:1) and cultured for 6 and 24 h. At the end of the culture period, supernatant was aspirated, the cells were lysed and RNA extracted and assayed by quantitative RT-PCR. To gain an idea of overall transcript abundance the difference in threshold cycle (δCt) values (Ct (β-Actin) - Ct (gene of interest)) were compared. Lower values indicate high transcript abundance of the gene of interest (tables 2 and 3).
At 6 h, the RNA for several of the 16 genes evaluated (IL-13, IL-15 and IL-17A) in unstimulated cells from non-IRIS patients tended to be slightly but significantly higher than TB-IRIS patients (table 2). Stimulation with M. tuberculosis increased the abundance of all transcripts studied in both groups with the exception of TGF-β. After Bonferroni (n-1, 15) correction of multiple comparisons, the abundance of IL-17A remained significantly greater in stimulated TB-IRIS cultures (pcorr = 0.045, table 2).
At 24 h, the RNA for IL-5 in unstimulated cells from non-IRIS patients was significantly higher than TB-IRIS and conversely the levels of IL-2, IL-15 and TNF higher in TB-IRIS (table 3). Stimulation with M. tuberculosis increased the abundance of all transcripts studied in both groups with the exception of IL-18 and TGF-β (whose level significantly decreased in non-IRIS). The abundance of IL-1β, IL-5, IL-6, IL-10, IL-13, IL-17A, IFN-γ, GM-CSF and TNF were significantly greater in stimulated TB-IRIS cultures (pcorr ≤0.03, table 3).
The fold induction of genes in stimulated relative to unstimulated cultures at the same time point was calculated by the δδCt method and values normalised by log10 transformation (fig. 2). In TB-IRIS patients at 6 h, IL-1β, IL-6 and GM-CSF were >100-fold induced and IL-2, IL-8, IL-12p40, IL-13, IL-17A, IFN-γ and TNF >10-fold. Induction was higher in TB-IRIS than non-IRIS for IL-1β, IL-2, IL-4, IL-6, IL-10, IL-13, IL-15, IL-17A, IFN-γ, GM-CSF and TNF (p≤0.05). After Bonferroni correction of p values the differences in IL-6, IL-12p40, IL-13, IL-17A and IFN-γ (pcorr ≤0.05) remained significant.
Average log-fold induction of cytokine genes by heat-killed Mycobacterium tuberculosis in tuberculosis–immune reconstitution inflammatory syndrome (TB–IRIS) and non-IRIS patients. Peripheral blood mononuclear cells (PBMC) from 22 TB-IRIS and 22 non-IRIS control patients were cultured in the presence or absence of heat-killed M. tuberculosis H37Rv for 6 and 24 h. At the end of the culture period, the cells were lysed, RNA extracted and used in quantitative RT-PCR. The fold induction of genes was calculated by the change in the threshold cycle difference (δδCt) method and values normalised by log10 transformation. At 6 h, induction was significantly higher in TB-IRIS than non-IRIS for interleukin (IL)-1β, IL-2, IL-4, IL-6, IL-10, IL-13, IL-15, IL-17A, interferon (IFN)-γ, granulocyte-macrophage colony-stimulating factor (GM-CSF) and tumour necrosis (TNF) (p≤0.05). At 24 h, significant differences between TB-IRIS and non-IRIS were present for IL-8, IL-10, IL-15 (higher in non-IRIS) and tumour growth factor (TGF)-β1. The p-values are uncorrected.
At 24 h, the fold induction in TB-IRIS patients tended to be similar to the 6-h time point (with the exception of IL-2, IL-15 and IL-18, which showed a reduction). However, fewer differences between TB-IRIS and non-IRIS were observed due to increases that occurred between 6 and 24 h in the latter group. Significant differences between TB-IRIS and non-IRIS persisted however for IL-8, IL-10, IL-15 and TGF-β1. TGF-β1 mRNA levels however tended to be minimally influenced by the presence of M. tuberculosis and all fold values were close to baseline.
Cytokines secreted into cell culture supernatant
The corresponding tissue culture supernatants for 20 IRIS and 19 non-IRIS patients were assayed for cytokine content by multiplex analysis (with the exception of IL-17A and TGF-β which were analysed by ELISA). Beads for IL-18 were unavailable and as transcript levels did not differ between groups this cytokine was not analysed further. IL-5 and IL-15 protein were undetectable except in small quantities in three samples in the case of IL-5 (data not shown). IL-4 and IL-17, whose levels were close to the lower detection limits, did not differ between TB-IRIS and non-IRIS (fig. 3). Constitutive secretion of TGF-β was similar in IRIS and non-IRIS groups, the concentration tended to decrease slightly on restimulation in both groups. Otherwise supernatant cytokine concentrations were consistently and significantly higher in TB-IRIS. After correction of p-values for multiple comparisons the largest and significant fold differences were in IL-12p40, IL-1β, GM-CSF, TNF, IL-10, IL-6, IL-2 and IL-8 (pcorr ≤0.04). Of these cytokines, only IL-2 is exclusively of lymphoid origin, the others being predominantly the products of myeloid cells.
Cytokine content of tissue culture supernatants. Culture supernatants arising from 20 immune reconstitution inflammatory syndrome (IRIS) patients and 19 non-IRIS controls were assayed by Luminex. Interleukin (IL)-4 and IL-17, whose levels were close to lower detection limits, did not differ between tuberculosis (TB)-IRIS and non-IRIS. Otherwise levels were consistently and significantly higher in TB-IRIS. After correction of p-values for multiple comparisons, the largest and significant fold differences were in IL-12p40, IL-1β, granulocyte-macrophage colony-stimulating factor (GM-CSF), tumour necrosis factor (TNF), IL-10, IL-6, IL-2 and IL-8. The p-values are uncorrected in this figure. IFN: interferon.
Internal validity of luminex determination of IFN-γ and IL-13
Luminex analysis is a convenient and powerful technology but has on occasions been reported to correlate poorly with ELISA (arbitrarily defined as gold standard). There was insufficient sample to secondarily test all analytes. However the same supernatant and cells were additionally assayed for IFN-γ secretion by both ELISA and ELISpot using the same stimulus, M. tuberculosis H37Rv. Correlation between ELISA and Luminex values was very strong (Spearman r = 0.82; p<0.0001), although Luminex consistently rendered higher values. IFN-γ ELISpot also correlated positively and significantly with both ELISA (r = 0.49; p = 0.005) and Luminex (r = 0.5; p = 0.003). Furthermore, there was similar association between TB-IRIS and elevation of IFN-γ by ELISA (325, IQR 90–750 pg·mL−1 versus 57, IQR 0–213 pg·mL−1; p = 0.005) and by ELISpot (730, IQR 240–1665 SFC per 106 PBMC versus 20, IQR 0–362 SFC per 106 PBMC; p = 0.001). IL-13 ELISpot (using M. tuberculosis H37Rv) was also performed on PBMC from a subset 16 IRIS and 10 non-IRIS patients. Like the ELISA (fig. 3) the commonest value was 0 SFC/106 PBMC with no significant difference between IRIS and non-IRIS groups (p = 0.38).
Cytokines in serum samples
Based upon the quantitative RT-PCR and supernatant results we next assayed by Luminex the concentration of the 11 most consistently discriminatory cytokines in serum samples taken from the same patients at the same time point. Serum concentrations of IL-1β, IL-2, IL-13 and GM-CSF were consistently close to the lower limit of assay detection and did not significantly differ between TB-IRIS and non-IRIS groups (data not shown). Serum concentrations of IL-6, IL-8, IL-10, IL-12p40, IFN-γ and TNF were significantly higher in the serum of TB-IRIS patients while TGF-β showed a trend towards higher values in non-IRIS (fig. 4). After correction of p-values for multiple comparisons, the largest and significant fold differences were in TNF, IL-6, and IFN-γ (pcorr ≤0.02).
Serum cytokine concentrations. Luminex assays of the serum concentration of 19 immune reconstitution inflammatory syndrome (IRIS) and 20-non-IRIS controls for the most consistently discriminatory cytokines taken from the same patients at the same time. After correction of p-values for multiple comparisons, the largest and significant fold differences were in tumour necrosis factor (TNF), interleukin (IL)-6, and interferon (IFN)-γ. Four tuberculosis–IRIS patients (grey circles) were classified as having localised (usually lymphadenopathic) disease. In these patients, a clear trend towards lower serum cytokine concentrations was seen, with 22 out of 24 (92%; 95% CL 74–97%) cytokine values falling on or below the median. The p-values are uncorrected in this figure. TGF: tumour growth factor.
Stratification by localised or disseminated disease
Most TB-IRIS patients in this study had either TB-IRIS with pronounced systemic signs or disseminated TB (more than two organs involved, table 1). However, four patients were classified as having disease localised to one anatomic site without systemic involvement. In these patients, a clear trend towards lower serum cytokine concentrations was seen with 22 out of 24 (92%, 95% CL 74–97%) cytokine values falling on or below the median (fig. 4). Thus the elevation or otherwise of cytokines appears to relate partially to TB-IRIS extent.
Effect of steroid therapy on cytokines in serum samples
Expert opinion favours the use of adjunctive corticosteroid therapy in some cases of TB-IRIS, an opinion recently provided greater evidence by our randomised placebo-controlled trial of prednisone in TB-IRIS that showed this therapy was associated with more rapid resolution of symptoms 40. To better explore cause and effect we therefore analysed serum concentrations of IFN-γ, IL-6 and TNF in a subset of 10 TB-IRIS trial participants who received corticosteroid therapy for 4 weeks (1.5mg·kg−1 daily for 2 weeks then 0.75 mg·kg−1 for 2 weeks). The concentrations of IL-6 (p = 0.006) and TNF (p = 0.038) significantly declined whereas no effect on IFN-γ concentrations was observed (fig. 5).
Effect of steroid therapy on cytokines in serum samples. Serum concentrations of interferon (IFN)-γ, interleukin (IL)-6 and tumour necrosis factor (TNF) were determined in a subset of 10 tuberculosis–immune reconstitution inflammatory syndrome (TB-IRIS) who received corticosteroid therapy for 4 weeks. The concentrations of IL-6 and TNF significantly declined, whereas no effect on IFN-γ concentrations were observed.
DISCUSSION
We have conducted a case–control analysis of 22 TB-IRIS patients sampled on the day of clinical presentation compared with 22 similar controls in order to determine whether dysregulated cytokine responses might contribute to pathology and symptoms in this condition. Stimulation of PBMC with M. tuberculosis increased the abundance of the majority of transcripts with IL-1β, IL-5, IL-6, IL-10, IL-13, IL-17A, IFN-γ, GM-CSF and TNF being significantly greater in stimulated TB-IRIS cultures at either the 6 or 24 h time points. M. tuberculosis stimulated gene induction in vitro was significantly greater in TB-IRIS patients for IL-6, IL-10, IL-12p40, IL-13, IL-17A and IFN-γ. In tissue culture supernatants, the concentrations of IL-12p40, IL-1β, IL-2, IL-6, IL-8, IL-10, IL-12p40, IFN-γ, GM-CSF and TNF were higher in TB-IRIS patients. In serum, significantly higher concentrations in TB-IRIS patients were observed for TNF, IL-6, and IFN-γ and the serum concentrations of IL-6 and TNF decreased during prednisone therapy of TB-IRIS. Thus many pro- and anti-inflammatory cytokine transcript and protein levels are elevated in TB-IRIS patients strongly suggesting that cytokine release contributes to pathology and symptoms in this condition. IL-6 and TNF were elevated under all conditions and decreased in serum during corticosteroid therapy such that blockade of these cytokines may be a novel and rational approach to immunomodulation in TB-IRIS.
The cytokine-release syndrome (sometimes referred to as a cytokine storm or hypercytokinaemia) occurs in a number of infectious and non-infectious diseases including graft versus host disease (GVHD) 41, acute respiratory distress syndrome (ARDS) 42, sepsis 43, H5N1 influenza 34 and the systemic inflammatory response syndrome (SIRS) 44. The experimental drug TGN1412 also caused serious acute illness likely to be driven by cytokines when given to six participants in a Phase I trial 35. The syndrome does not appear to have a quantitative definition but is characterised as including fever, hypotension, increased endothelial permeability, oedema and vasodilatation. Multiple cytokine mediators of inflammation are described as being elevated. TB-IRIS can be of acute onset especially the unmasking form and fatal ARDS is described 45–47. Patients with paradoxical TB-IRIS frequently have prolonged fever and tachycardia and appear to be at increased risk of venous thromboembolism 20, 48. Whilst the concentrations of serum cytokines that we report are not as high, for example, as those reported in the TGN1412 (anti-CD28) trial or in H5N1, we suggest that the exaggerated cytokine responses we have observed in TB-IRIS patients compared to similar patients who do not experience the syndrome may contribute substantially to pathogenesis.
Antitretroviral therapy effectively suppresses HIV-1 replication and thereby allows recovery of CD4 lymphocyte numbers and function with the most rapid CD4 rise occurring early in therapy 28, 49. It is thus logical to investigate whether dysregulated CD4 responses contribute to TB-IRIS. Very large antigen specific Th1 CD4 expansions accompany cART mediated immune restoration in both TB-IRIS patients and similar co-infected persons who do not develop the syndrome 26, 27. In keeping with these observations we observed an increase in IFN-γ transcript abundance, fold induction and both secreted and serum cytokine in TB-IRIS patients (tables 2 and 3; figs 3 and 4). We also documented increased IL-12p40 fold induction and secreted cytokine and also increased IL-2 in tissue culture supernatants from TB-IRIS patients. However, our observations suggest that lymphocyte subsets other than Th1 may also contribute. Thus transcripts from the “Th2” genes IL-5 and IL-13 were elevated in TB-IRIS patients although that did not translate into elevated protein concentrations. It is also interesting to note that the transcript of IL-17A was, after correction for multiple comparisons, the only significantly elevated in TB-IRIS patients at 6 h (table 2); a difference reflected at 24 h and in fold induction. Again protein concentrations were low and did not differ between groups. IL-17 is markedly pro-inflammatory and has been implicated in the early protective immune response to TB 50. In humans IL-17 secretion in response to TB appears mediated by a distinct T-cell subset with phenotypic characteristics of long-lived central memory cells 39. The lack of protein secretion we observed may have been due to IFN-γ mediated suppression of IL-17 in vitro 51. IL-17 is indirectly chemotactic for neutrophils 52 and the cold abscesses that occur in TB-IRIS are characterised by the presence of neutrophils 53. Further work on the potential for early IL-17 production to contribute to inflammation in TB-IRIS may therefore be of interest.
It is also of interest that elevation of cytokines predominantly of myeloid (e.g. IL-1β, IL-6, IL-8 and IL-12p40) or of dual myeloid/lymphoid (e.g. TNF, GM-CSF and IL-10) origin characterise TB-IRIS. Separating cause and effect in observational studies is difficult but there is a report of unmasking TB-IRIS in which fatal bronchiolitis obliterans was associated with macrophage infiltration 47. cART partially restores all aspects of immune function and others have speculated on the role of the macrophage in this condition 54. It is certainly striking that we have observed TB-IRIS after as few as 3 days cART; well before CD4 restoration can be substantial. It has been previously noted that the numbers of both myeloid and plasmacytoid dendritic cells increase during cART 55. We were also interested to note consistent elevation of immunoregulatory IL-10: the hypothesis that a deficiency of regulation may contribute to TB-IRIS is attractive but again not supported by elevation, rather than suppression, of IL-10.
While ours is one of the largest studies of the pathogenesis of TB-IRIS to date, it remains apparent from the distribution of results we obtained that this is a heterogeneous condition and we caution against causal interpretations of association in observational studies. The purpose of the study was to evaluate differences at presentation of TB-IRIS comparing results to similar participants who did not experience the condition. Prediction of TB-IRIS did not form part of the scope of the study: we are pursuing this in other studies. We do however have the advantage of preliminary data that suggests corticosteroid therapy was associated with a decrease in serum TNF and IL-6 concentrations (fig. 5). It is known that this therapy leads to more rapid symptom resolution in TB-IRIS 40. Not all cases of TB were culture-confirmed although all did conform to WHO guidelines on the diagnosis of HIV-associated TB 37 and all showed response to anti-TB therapy prior to cART. TB-IRIS remains a diagnosis of exclusion: no patients with known drug-resistant TB were included and the case definition we employed for TB-IRIS has been twice independently validated 14, 19, 24. CD4 counts were not re-evaluated at the time of TB-IRIS (and after the corresponding time interval in non-IRIS controls), thus the relationship of cytokine release to the increase in CD4 could not be evaluated. Luminex analysis is a relatively new technology that permits a much greater number of analytes to be assayed than hitherto possible. We based our reagent selection on experience when comparing luminex with ELISA 56, 57 and internal validity at least of IFN-γ estimations by both ELISpot and ELISA was good. Multiple comparisons have also been made but we have restricted our conclusions to results upon which stringent Bonferroni correction was applied and also the biological plausibility of being elevated in all assays that we have performed.
We identified IL-6 and TNF secretion as elevated in all circumstances (fig. 6) and potentially amenable to immune modulation, which is consistent with the finding that polymorphism in these genes may associate with the risk of IRIS 58 and the elevation of IL-6 in other forms of IRIS 59, 60. IL-6 is a key driver of the acute phase response and we have previously documented that the C-reactive protein is invariably elevated at presentation of TB-IRIS 20. TNF has consistently been associated with both protection and pathology in TB 61–63. Recent case reports have documented a beneficial effect of TNF blockade on paradoxically deteriorating TB in HIV-1 uninfected patients 64, 65. However, a clinical study of anti-TNF antibodies in TB-IRIS would face a number of difficult issues because it is recognised this therapy has a prolonged half-life and also leads to the reactivation of TB 62. A phase-II study in patients with known drug-sensitive TB undergoing severe TB-IRIS in whom corticosteroid therapy was either ineffective or contraindicated might be justifiable. But in most environments in which TB-IRIS is encountered, the new knowledge that the condition is contributed to by antigen load and attendant exaggerated cytokine release encourages reinvestigation of the possibility of co-existent TB drug resistance, optimisation of anti-TB therapy, and intensified supportive care together with the judicious use of adjunctive corticosteroid therapy 53.
Summary of results and implication of tumour necrosis factor (TNF) and interleukin (IL)-6 in the pathogenesis of tuberculosis–immune reconstitution inflammatory syndrome (TB-IRIS). While many pro- and anti-inflammatory cytokine transcript and protein levels were elevated in TB-IRIS patients according to experimental circumstances, only IL-6 and TNF were elevated in all circumstances. Thus blockade of IL-6 or TNF may be a rational approach to immunomodulation in this condition. IFN: interferon; GM-CSF: granulocyte-macrophage colony-stimulating factor; M. tuberculosis: Mycobacterium tuberculosis.
Acknowledgments
P. Mouton (University of Cape Town, Cape Town, South Africa) and M. Abrahams (Médecins Sans Frontières, South Africa) are especially thanked for assistance with recruitment and clinical assessment of study participants. A.C. Botella (Institute for Tropical Medicine, Antwerp, Belgium) is thanked for helpful comments on the manuscript.
Footnotes
Support Statement
This work was supported by the Wellcome Trust (references 084323, 081667, 088316, 084670, 083226). Additional support was provided by the Medical Research Councils of the UK and South Africa, by the European Union (Sante/2006/105-061) and by the European and Developing Countries Clinical Trials Partnership (EDCTP 060613).
Statement of Interest
None declared.
- Received June 14, 2010.
- Accepted August 3, 2010.
- ©2011 ERS
REFERENCES
