HIV-tuberculosis-associated immune reconstitution inflammatory syndrome is characterized by Toll-like receptor and inflammasome signalling

Abstract

Patients with HIV-associated tuberculosis (TB) initiating antiretroviral therapy (ART) may develop immune reconstitution inflammatory syndrome (TB-IRIS). No biomarkers for TB-IRIS have been identified and the underlying mechanisms are unclear. Here we perform transcriptomic profiling of the blood samples of patients with HIV-associated TB. We identify differentially abundant transcripts as early as week 0.5 post ART initiation that predict downstream activation of proinflammatory cytokines in patients who progress to TB-IRIS. At the characteristic time of TB-IRIS onset (week 2), the signature is characterized by over-representation of innate immune mediators including TLR signalling and TREM-1 activation of the inflammasome. In keeping with the transcriptional data, concentrations of plasma cytokines and caspase-1/5 are elevated in TB-IRIS. Inhibition of MyD88 adaptor and group 1 caspases reduces secretion of cytokines including IL-1 in TB-IRIS patients. These data provide insight on the pathogenesis of TB-IRIS and may assist the development of specific therapies.

ARTICLE

Received 13 Mar 2015 | Accepted 21 Aug 2015 | Published 24 Sep 2015

Rachel P.J. Lai1,*, Graeme Meintjes2,3,*, Katalin A. Wilkinson1,2, Christine M. Graham1, Suzaan Marais2, Helen Van der Plas2, Armin Deffur2, Charlotte Schutz2, Chloe Bloom1, Indira Munagala4,Esperanza Anguiano4, Rene Goliath2, Gary Maartens2, Jacques Banchereau4, Damien Chaussabel5, Anne OGarra1,6 & Robert J. Wilkinson1,2,3

Patients with HIV-associated tuberculosis (TB) initiating antiretroviral therapy (ART) may develop immune reconstitution inammatory syndrome (TB-IRIS). No biomarkers for TB-IRIS have been identied and the underlying mechanisms are unclear. Here we perform transcriptomic proling of the blood samples of patients with HIV-associated TB. We identify differentially abundant transcripts as early as week 0.5 post ART initiation that predict downstream activation of proinammatory cytokines in patients who progress to TB-IRIS. At the characteristic time of TB-IRIS onset (week 2), the signature is characterized by over-representation of innate immune mediators including TLR signalling and TREM-1 activation of the inammasome. In keeping with the transcriptional data, concentrations of plasma cytokines and caspase-1/5 are elevated in TB-IRIS. Inhibition of MyD88 adaptor and group 1 caspases reduces secretion of cytokines including IL-1 in TB-IRIS patients. These data provide insight on the pathogenesis of TB-IRIS and may assist the development of specic therapies.

1 The Francis Crick Institute Mill Hill Laboratory, London NW7 1AA, UK. 2 Clinical Infectious Diseases Research Initiative, Institute of Infectious Disease and Molecular Medicine and Department of Medicine, University of Cape Town, Anzio Road, Observatory 7925, South Africa. 3 Department of Medicine, Imperial College London, London W2 1PG, UK. 4 Baylor Institute for Immunology Research, Dallas, Texas 75204, USA. 5 Systems Immunology, Benaroya Research Institute, Seattle, Washington 98101, USA. 6 Department of Medicine, National Heart and Lung Institute, Imperial College London, London W2 1PG, UK.* These authors contributed equally to this work. Correspondence and requests for materials should be addressed to G.M. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2015 Macmillan Publishers Limited. All rights reserved.

DOI: 10.1038/ncomms9451 OPEN

HIVtuberculosis-associated immune reconstitution inammatory syndrome is characterized by Toll-like receptor and inammasome signalling

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451

The World Health Organization (WHO) estimates that approximately one-third of the worlds population is infected with Mycobacterium tuberculosis (MTB),

5%10% of whom will develop active disease1 with human immunodeciency virus type 1 (HIV-1) infection being the greatest recognized risk factor for disease2. Seventy-eight per cent of those with HIV-associated tuberculosis (TB) live in Africa1. Increased access to combination antiretroviral therapy (ART) has signicantly improved the clinical outcome of such patients in resource-limited settings3,4. However, up to 54% of patients develop hyperinammatory reactions known as immune reconstitution inammatory syndrome (IRIS) within the rst month of ART initiation5. Paradoxical TB-IRIS is directed towards MTB antigens and is characterized by recurrent, new or worsening symptoms and signs of treated TB.

Three clinical risk factors for HIV-associated paradoxical TB-IRIS are recognized: (i) low baseline CD4 T-cell count (o50100 cells per mm3) before ART68; (ii) a short time interval between commencing TB treatment and ART68; and(iii) dissemination of TB to extrapulmonary sites, possibly reecting higher bacterial load9,10. Nevertheless, the immunopathological basis of the syndrome remains incompletely understood and there is no biomarker to predict which patients will develop IRIS. Studies to date have suggested that both innate and adaptive immunity are involved, leading to hypercytokinemia and severe inammation11.

Although hyperinammation in TB-IRIS is associated with the expansion of TB antigen-specic interferon-g (IFN-g) producing peripheral T-helper 1 cells12, the absence of such expansion in some TB-IRIS cases and the presence in similar patients who do not develop the syndrome suggests that the association might not be causal13. Several studies have suggested that innate immunity also contributes to IRIS. In post-mortem staining of lung tissue sections from a TB-IRIS patient, most of the inammatory cells were identied to be CD68 macrophages14. Increased natural killer cell activation and degranulation activity15,16, and elevated neutrophil counts in the cerebrospinal uid17,18 of tuberculous meningitis-IRIS cases have also been reported. The levels of interleukin (IL)-6 and C-reactive protein are elevated at the time of IRIS in both Mycobacterium avium-associated IRIS19 and paradoxical TB-IRIS5,13,20. TB-IRIS patients were also reported to have higher expression level of pattern recognition receptors (PRRs) such as Toll-like receptor 2 (TLR2)15,2123. Microarray analysis of monocytes isolated from HIV-TB co-infected patients suggested an imbalance between the complement effector C1q and the inhibitor C1-INH in those who developed TB-IRIS24. Notwithstanding these extensive research efforts, the underlying mechanisms mediating TB-IRIS remain to be fully elucidated.

The application of systems immunology has signicantly increased understanding of many diseases. In the last few years, transcriptional proling of blood from TB patients has led to the nding of a type I IFN-inducible neutrophil-driven signature of active TB disease25. In this study, we used a similar unbiased whole-blood transcriptomic proling of HIV-TB co-infected patients commencing ART, to better understand the immune pathways resulting in TB-IRIS. Patients were tracked longitudinally to ascertain development of TB-IRIS. We characterized plasma cytokine proles, which were in keeping with the transcriptomic signature. We further demonstrated the dependence of elevated levels of cytokine on TLR signalling pathways, as well as caspase-1-dependent canonical and caspase-5-dependent non-canonical inammasomes.

ResultsPatient characteristics. Thirty-two patients with HIV-associated TB were included for microarray analysis. Patients in both TB-IRIS

(n 17) and non-IRIS (n 15) groups had similar clinical

variables, including age, sex, CD4 count and HIV viral load, and site of MTB infection (Supplementary Table 1). The majority of patients had both pulmonary and extrapulmonary TB (PTB and EPTB); the latter usually manifest as visceral TB, peripheral adenopathy and pleural effusion, with only a few cases of tuber-culous meningitis (n 7). Patients who developed TB-IRIS had

symptom onset at a median of 13 days (interquartile range (IQR): 721 days) after ART initiation. An additional 31 patients were not included in the microarray analysis, owing to lack of whole-blood samples, but had plasma cytokines analysed. No signicant differences were observed compared with the microarray group with regard to clinical variables (Supplementary Table 1).

Differentially abundant transcripts at 25 days post ART. Longitudinal whole-blood samples were collected from patients with HIV-associated TB commencing ART (Fig. 1a). We investigated the molecular path of immune dysregulation leading to TB-IRIS by inspecting transcriptomic changes in TB-IRIS and non-IRIS patients by using two different normalization approaches. First was a modied method from one previously described25,26, in which signal intensities were normalized to the median of all samples. The second approach normalized each sample to its corresponding baseline (week 0). Differentially expressed genes identied by both approaches are listed in Supplementary Table 2.

At week 0, there was no transcriptomic difference between patients who did or did not develop TB-IRIS; however, within 25 days post ART (week 0.5) we identied 22 differentially abundant transcripts (20 genes) in patients who developed TB-IRIS (Fig. 1b) by using the normalization to median method. Functional analysis of these 22 differentially abundant transcripts indicated over-representation of the JAK/STAT signalling pathways mediated by the IL-6 cytokine family and a role for macrophages and complement (Fig. 1b). Using the alternative normalization to baseline approach, we further identied 56 differentially abundant transcripts (50 genes) in TB-IRIS patients early during ART (week 0.5) (Fig. 1c). Innate immune pathways remained dominant in TB-IRIS, represented by upregulated type I and II IFN signalling, activation of PRRs and the role of macrophages (Fig. 1c), although only 5 of the 50 genes overlapped with the 20 genes identied using the normalization to median approach. Together, the data suggest early activation of a predominantly innate inammatory response at transcriptomic level in TB-IRIS before the onset of clinical symptoms.

Transcriptomic signature at the time of symptom onset. Applying the same analytical approaches described above, we identied a 138-transcript (125 gene) signature associated with TB-IRIS at week 2 by normalizing to the median (Fig. 2a and Supplementary Table 3). Functional analysis of these 125 differentially abundant transcripts indicated upregulation of several innate response pathways, with TLR and TREM1-induced inammasome signalling being most over-represented (Fig. 2a), again suggesting an important role of innate antigen recognition and cytokine signalling in the inammatory response in TB-IRIS. When the transcriptional data were normalized to their corresponding baseline values, we identied 319 transcripts (281 genes) differentially abundant in TB-IRIS patients at week 2 (Fig. 2b and Supplementary Table 4). Functional pathway analysis again indicated innate immune signalling pathways including TLR, IFN and IL-1 signalling to be over-represented in TB-IRIS (Fig. 2b). Of the 281 genes identied, 70 overlapped with the 125 genes identied by normalization to the median. The large number of overlapping genes indicates they are conserved in

2 NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451 ARTICLE

a

Sampling timeline

Antitubercular therapy

ART

Sampling (week)

Week 0.5 (Pre-TB-IRIS) normalized to median

Week 0.5 (Pre-TB-IRIS) normalized to baseline

Top 5 canonical pathways

Top 5 canonical pathways

Interferon signalling

0 0.5 1 2Median time of IRIS onset

b

Colour range

Condition

2.2 0 2.2

(No TBIRIS)

(TBIRIS)

P-value

P-value

Condition

Role of JAK family kinases in IL-6-type cytokine

signalling

Acute phase response signalling

Role of macrophages, fibroblasts and endothelial

cells in rheumatoid arthritis

Role of JAK2 in hormone-like cytokine

signalling

Role of JAK1 and JAK3 in cytokine signalling

1.14E069.96E068.23E05

3.33E041.17E03

2.80E062.79E05

1.48E03

2.39E036.41E03

c

Colour range

Condition

2.9 0 2.9

(No TBIRIS)

(TBIRIS)

Condition

Activation of IRF by cytosolic pattern

recognition receptor

Role of macrophages, fibroblasts and endothelial

cells in rheumatoid arthritis

Role of pattern recognition receptors in

recognition of bacteria and viruses

Role of RIG1-like receptors in antiviral

immunity

Figure 1 | Differentially abundant transcripts associated with progression to TB-IRIS were identied early before onset of clinical symptoms.(a) Whole-blood samples were collected longitudinally and the median onset of TB-IRIS was 13 days. (b) Microarray was performed with 22 whole-blood samples (13 TB-IRIS and 9 non-IRIS). Twenty-two transcripts (20 genes) were differentially abundant at week 0.5 (days 25 post ART) in those patients who eventually developed TB-IRIS, compared with non-IRIS. Transcript intensity values were normalized to the median of all samples. Transcripts are clustered vertically where green represents decreased abundance and red represents increased abundance. Patient arms are clustered horizontally where blue represents non-IRIS and orange represent TB-IRIS. Signicance of pathways over-represented by differentially abundant genes was calculated by Fishers exact test corrected for BenjaminiHochberg false discovery rate (BH-FDR) (P 0.05). (c) Using an alternative analytical approach by normalizing

transcript intensity to its corresponding baseline (week 0) value, 55 transcripts (50 genes) were found to be differentially abundant in TB-IRIS patients, compared with non-IRIS controls. Signicance of pathways over-represented by differentially abundant genes was calculated by Fishers exact test corrected for BH-FDR (P 0.05). One sample was excluded from the analysis, owing to the lack of corresponding baseline.

TB-IRIS regardless of analytical method. Functional analysis of the overlapping transcripts again indicated the innate immune components, with TLR, TREM-1 and IL-1 being the most enriched biological processes. Corticosteroid therapy was prescribed in ten of the patients (four of whom had microarray performed) at the time of ART commencement but it was not a confounding factor in the transcriptomic analysis.

We next performed both regulator and molecule activity prediction27 of the week 2 signature by using the 125 genes identied by the normalization to median approach. Proinammatory cytokines including oncostatin M, IFN-g, tumour necrosis factor (TNF) and IL-1 were predicted to be the upstream regulators of the week 2 signature (Supplementary Table 5). In addition, CSF-3 (granulocyte colony-stimulating factor) and IL-5 were also predicted to be activated upstream at week 2, suggesting activation of adaptive immunity. An array of

proinammatory cytokines and chemokines were predicted to be activated at week 2, including TNF-a, IL-1b, IL-6, IL-8, IL-12,

IL-18, granulocytemacrophage colony-stimulating factor, monocyte chemoattractant protein-1 and macrophage inammatory protein-2 (Fig. 3). Signalling appeared to be mediated by TLR and IL-1/18 receptor via the MyD88 adaptor and TREM1-induced activation of inammasomes and TLR, resulting in the activation of nuclear factor-kB and subsequent production of cytokines and chemokines (Fig. 3).

Plasma cytokine measurement. As the transcriptomic signatures predicted activation of many proinammatory cytokines (Fig. 3), we measured the concentration of these analytes in patient plasma at both week 0 and week 2, to assess whether the transcriptomic prediction was reected in the blood protein.

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451

a

Week 2 (TB-IRIS onset) normalized to median

Colour range

4.4

Condition

0 4.4

(No TBIRIS)

(TBIRIS)

Color range

Condition

Condition

P-value

6.86E072.85E041.02E031.47E031.47E03

b

Week 2 (TB-IRIS onset) normalized to baseline

Top 5 canonical pathways

TLR signalling

TREM1 signalling

Role of macrophages, fibroblasts and endothelial

cells in rheumatoid arthritis

Production of nitric oxide and reactive oxygen

species in macrophages

Role of pattern recognition receptors in

recognition of bacteria and viruses

Top 5 canonical pathways

TLR signalling

IFN signalling

Activation of IRF by cytosolic pattern

recognition receptor

IL-1 signalling

3.9 0 3.9

Condition

(No TBIRIS)

(TBIRIS)

P-value

Role of pattern recognition receptors in

recognition of bacteria and viruses

1.29E06

2.55E052.62E04

4.43E04

8.99E04

Figure 2 | Differentially abundant transcripts in TB-IRIS onset are associated with innate signalling pathways. (a) Microarray was performed with 26 whole-blood samples (13 TB-IRIS and 13 non-IRIS). One hundred and thirty-eight transcripts (125 genes) were differentially abundant in the whole blood of the TB-IRIS patients at week 2. Transcript intensity values were normalized to the median of all samples. Transcripts are clustered vertically where green represents decreased abundance and red represents increased abundance. Patient arms are clustered horizontally where blue represents non-IRIS and orange represent TB-IRIS. Signicance of pathways over-represented by differentially abundant genes was calculated by Fishers exact test corrected for BenjaminiHochberg false discovery rate (BH-FDR) (P 0.05). (b) Using an alternative analytical approach by normalizing transcript intensity to its

corresponding baseline (week 0) value, 319 transcripts (281 genes) were found to be differentially abundant in TB-IRIS patients. Signicance of pathways over-represented by differentially abundant genes was calculated by Fishers exact test corrected for BH-FDR (P 0.05). One sample was excluded from

the analysis, owing to the lack of corresponding baseline.

Figure 3 | Transcriptomic signature of TB-IRIS at week 2 predicts activation of innate signalling pathways and production of proinammatory cytokines and chemokines. The TREM1, TLR and IL-1 signalling pathways are interconnected. Molecules in red were those found to be differentially abundant in the signatures of TB-IRIS patients at week 2 (normalization to median). Molecule activity prediction was performed using Ingenuity Pathway Analysis: molecules in orange were predicted to be upregulated and those in blue predicted to be downregulated.

4 NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451 ARTICLE

Congruent with the transcriptomic signature prediction of week 2, we detected signicantly higher concentrations of IL-12p40, IFN-g, TNF-a and IL-6 in TB-IRIS patients at week 2 (Fig. 4), indicating a circulating inammatory response at the time when clinical symptoms were observed. Our data agree with those previously reported28, further indicating that TB-IRIS is strongly associated with dysregulated cytokine and chemokine production. Although IL-8 and type I IFN were also predicted to be elevated at week 2, we did not detect any differences in plasma proteins between non-IRIS and TB-IRIS patients (Fig. 4).

Validation by technically independent NanoString method. A technically independent method, the NanoString (nCounter) system, was used to further validate the 125 differentially expressed genes identied at the median time of TB-IRIS onset (week 2) by the standard normalization to median microarray analytical approach (Fig. 2a). The 125 genes contained 70 genes that were also identied using the alternative normalization to baseline approach described above (Fig. 2b). Probes to sequences differing from those detected by microarray were designed, with one being excluded from the assay because of nonspecic

complementarity to other genes. Owing to insufcient quantity of RNA, eight samples (ve TB-IRIS and three non-IRIS) were excluded from the NanoString assay. Seventy-six of the 124 probes that were statistically signicantly different in microarray were also found signicant by NanoString (Supplementary Table 6). The remaining 48 genes, although not statistically signicantly different between the TB-IRIS and non-IRIS groups, also showed the same trends in abundance as the microarray. The most over-represented functional pathways of the 76 genes that were signicantly upregulated in the TB-IRIS group were again from the innate immune pathways: TLR, TREM1 and IL-1 signalling.

A 43-transcript signature that characterizes TB-IRIS. To determine a signature that consisted only of the most conserved genes differentially abundant across all technical or analytical approaches, we overlapped the genes identied at week 2 in all three approaches, resulting in a 43-transcript signature. Functional analysis indicated over-representation of several innate immune signalling pathways, including TREM-1 (P 0.0002),

TLR (P 0.0002), IL-1 (P 0.0009) and type I IFN (P 0.0022)

(Fig. 5). A weighted temporal molecular distance algorithm was

103 NS NS

NS

*

**

**

103

102

101

100

0

103

102

101

100

0

IL-12p40 (pg ml1 )

102

101

100

103

102

101

100

IFN-(pg ml1 )

0 Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Week 0

Week 2 Week 0 Week 2

Week 0 Week 2

NS

NS NS

*

TNF-(pg ml1 )

IL-8 (pg ml1 )

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

IFN-2 (pg ml1 )IL-6 (pg ml1 )

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Week 0 Week 2

NS NS

104

102

101

100

103

102

101

100

103

0 Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

0 Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Week 0 Week 2 Week 0 Week 2

Figure 4 | Cytokine production at week 2 of TB-IRIS reects the transcriptomic prediction. Plasma cytokine concentrations in TB-IRIS and non-IRIS patients at week 0 and week 2 were measured. Results were analysed by MannWhitney U-test between non-IRIS and TB-IRIS at each time point. The median value of each group was shown and P-value was designated: *Pr0.05, **Pr0.01.

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451

Weighted temporal molecular

distance

25

Top 5 canonical pathways (43 genes)

TREM1 signalling

TLR signalling

Endothelin-1 signalling

IL-1 signalling

20

15

10

5

0

Non-IRIS

TB-IRIS

Role of pattern recognition receptors in

recognition of bacteria and viruses

P-value

2.01E04

2.01E04

3.84E048.83E048.83E04

**

*

0 0.5 1 2 Week

**

Week 2 (43 consistently overabundant transcripts) ACSL1

ADCY3 AGPAT9

AIG1 ANXA3 APOB48R ASPRV1 B3GNT8

BAMBIBASP1BCL2A1 GPR97

GPR160

RHOBTB1

C19orf35 HECW2 IFIT3 IFITM3 KCNJ15

KLHL2

LOC100128460

MANSC1

MAPK14 OAS1 PFKFB3

PSG9

TUFT1

CARD17

SERPING1

CASP5

SIGLEC9

CDK5RAP2

SIPA1L2

DHRS13

SMARCD3

DKFZp761E198

TLR2

DOK3

TLR5

ECE1

TPST1

GNB4

TRIB1

Figure 5 | Weighted temporal molecular distance to health of 43 consistently overabundant transcripts differentiate TB-IRIS patients from non-IRIS controls. Forty-three genes were consistently overabundant in TB-IRIS by different analytical and technical platforms. These genes are associated with innate signalling pathways, namely TREM1, TLR and IL-1 signalling (Fishers exact test with BenjaminiHochberg false discovery rate (BH-FDR), P 0.05).

The weighted temporal molecular distance measured the total transcriptional perturbation of the 43 genes at each time point relative to its baseline (week0) mean. The mean, s.e.m. and P-values are displayed (unpaired t-test with Welchs correction; *Pr0.05 and **Pr0.01).

further applied to quantify changes in abundance of these 43 transcripts over time (Fig. 5). The weighted temporal molecular distance algorithm measures the transcriptional perturbation of specied genes at a given time point relative to the mean at week0. The non-IRIS patients had no signicant changes in the level of the 43 genes at week 0, 0.5, 1 or 2. By contrast, there was signicant transcriptional perturbation at week 2 in TB-IRIS patients compared with week 0 (P 0.0268). Furthermore, there

was a signicant difference in the transcript levels of these 43 transcripts between the TB-IRIS and the non-IRIS patients as early as 25 days post ART (week 0.5; P 0.008), before IRIS-

associated clinical symptoms were apparent, suggesting that these genes are specically associated with, and represent the earliest pathways involved in triggering clinical progression to, TB-IRIS.

Inhibition of MyD88 reduces proinammatory cytokines. To conrm the involvement of TLRs in the induction of cytokines that associate with TB-IRIS, we performed inhibition of the TLR signalling pathway. We chose to inhibit MyD88 in peripheral blood mononuclear cells (PBMCs) from both TB-IRIS and non-IRIS patients using a specic peptide inhibitor (Pepinh-MYD), compared with using a control peptide and RPMI medium control (mock).

Among the analytes measured, concentrations of IL-12p40, IFN-g and TNF-a in PBMC culture supernatant from TB-IRIS patients signicantly decreased following treatment with MyD88 inhibitor, but not with the control peptide, compared with the RPMI mock (Fig. 6). The same experiments were also performed in the PBMCs from non-IRIS patients, but analytes were below the limits of detection. Similar observations have also been previously reported in the PBMCs from non-IRIS patients28. Although IL-1 was predicted to be activated downstream of the TLR-MyD88 signalling cascade, we failed to detect any difference

in the level of IL-1b following MyD88 inhibition (Fig. 6), suggesting other immunological pathways are also involved in mediating the inammatory response in TB-IRIS.

Inammasomes mediate proinammatory responses in TB-IRIS. We next investigated TREM1 signalling, the other consistently over-represented pathway in TB-IRIS. The TREM1 activation in TB-IRIS consisted of upregulated messenger RNA expression of TLRs, NLR, NOD, CARD and CASP5 (caspase-5), all of which are involved in canonical and non-canonical inammasomes. We detected a signicantly higher concentration of caspase-1 in the supernatant of MTB-stimulated PBMC cultures from TB-IRIS patients at week 2 (Fig. 7a), suggesting the canonical inammasome may be involved in mediating TB-IRIS. For the non-canonical inammasome, PBMC lysates from TB-IRIS were found to have a signicantly higher content of caspase-5, compared with non-IRIS (Fig. 7a), indicating that the non-canonical inammasome is also activated during TB-IRIS. There was no signicant difference in the anti-inamma-tory apoptotic caspase-3 between TB-IRIS and non-IRIS (Fig. 7a). We next quantied the concentration of IL-1b and IL-1a in tissue culture (TC) supernatant as they are cleaved by caspase-1 and caspase-5, respectively. Signicantly higher concentrations of IL-1b and

IL-1a were detected in TB-IRIS, compared with non-IRIS (Fig. 7b). Treatment with an irreversible pan-caspase-1/4/5 inhibitor Z-WEHD-FMK signicantly reduced the production of IL-1b and

IL-1a in both groups, with a greater reduction observed in TB-IRIS. Together, these data suggest that inammasome activation also contributes to the inammatory response in TB-IRIS.

DiscussionWe have identied differentially abundant transcripts that characterize patients who develop TB-IRIS from those that do not as early as week 0.5 (days 25) following ART initiation,

6 NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451 ARTICLE

*** ***

***

***

***

104 104

103

103

102

104

104

**

IL-12p40 (pg ml1 )

103

102

101

IFN- (pg ml1 )

TFN- (pg ml1 )

101

100

100

0 0

RPMI mock

RPMI mock

100

103

102

101

102

0

Control peptide

Control peptide

MyD88-lnh

MyD88-lnh

RPMI mock

Control peptide

MyD88-lnh

RPMI mock

Control peptide

MyD88-lnh

105

IL-1 (pg ml1 )

Figure 6 | Inhibition of the MyD88 adaptor results in reduced cytokine productions. PBMC from TB-IRIS patients (week 2) were treated with MyD88 inhibitor (MYD88-Inh), control peptide or RPMI medium (mock), followed by stimulation with heat-inactivated H37Rv and the supernatant cytokine concentration was determined. An unstimulated background control was included for each treatment per sample. Data are shown as background subtracted (stimulated minus unstimulated) with a value of zero assigned to those negative after background subtraction. The Wilcoxon signed rank test was used for statistical comparisons. The median value of each group was shown and P-value was designated: *Pr0.05, **Pr0.01 and ***Pr0.001.

103

106 105 104 103 102 101 100

0

Vehicle control Pan-caspase inhibitor Vehicle control Pan-caspase inhibitor

a

*

**

Caspase-1 (pg ml1 )

IL-1(pg ml1 )

Caspase-5 fold change

15

10

5

Caspase-3 (pg ml1 )

103

102

101

100

102

101

100

Non-IRIS

Non-IRIS

TB-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

0 Non-IRIS

TB-IRIS

Stimulated Unstimulated Stimulated

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Unstimulated

**** **** ** **

*

b

IL-1(pg ml1 )

103

102

101

100

0

***

Non-IRIS

TB-IRIS

Non-IRIS

TB-IRIS

Figure 7 | Both canonical and non-canonical inammasomes are activated in TB-IRIS. (a) Patient PBMC from TB-IRIS and non-IRIS (week 2) were treated with pan-group 1 caspase inhibitor or dimethyl sulfoxide (DMSO) vehicle control, followed by stimulation with heat-inactivated H37Rv. An unstimulated background control was included for each treatment per sample. Concentrations of caspase-1 (left, on TC supernatant) and caspase-3 (right, with lysate) are shown as background subtracted (stimulated minus unstimulated) with a value of zero assigned to those negative after background subtraction. Caspase-5 (middle, with lysate) is expressed as fold-change of stimulated samples over unstimulated controls, as a standard curve was not available. The Wilcoxon signed rank test was used for statistical comparisons. The median value of each group was shown and P-value was designated: *Pr0.05, **Pr0.01, ***Pr0.001 and ****Pr0.0001. (b) Reecting increased concentrations of caspase-1 and caspase-5, concentrations of IL-1b (left)

and IL-1a (right) produced by patient PBMCs treated above were measured. Concentrations are expressed as that from MTB-stimulated minus the corresponding unstimulated control.

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451

indicating an early transcriptional response before clinical symptoms. These early changes were predominantly related to enriched JAK/STAT signalling pathways mediated by either IL-6 or type I and II IFNs, as well as activation of PRRs and a role of macrophages via IL-1/18R, indicating the restoration of an innate immune response immediately following ART initiation. Although IL-1, IL-6, TNFa and IFN-g have all been shown to be important in protection against MTB2933, type I IFN may adversely affect the host immune response34. The transcriptomic signature of TB patients is characterized by a neutrophil-driven type I IFN-inducible response, which correlates with lung radiographic extent of disease and is downregulated on successful treatment25. Furthermore, in chronic MTB or Mycobacterium leprae infection, type I IFN plays an immunosuppressive role by inducing IL-10 and PDL1, and suppressing the production of IL-1a and IL-1b (refs 30,3538).

This inammatory equilibrium of type I IFN appears to be disrupted in TB-IRIS, in which the negative feedback was not sufcient to control the downstream activation of the proinammatory response, as we observed increased plasma concentration of IL-12p40 and enhanced secretion of IL-1a and

IL-1b in PBMC cultures from TB-IRIS patients.

Of relevance, we also found that early differentially abundant transcripts in TB-IRIS predicted activation of nuclear factor-kB and p38-MAPK signalling pathways, and the production of proinammatory cytokines and chemokines. The molecules of the week 0.5 signatures were also found to be upstream of the week 2 signatures. Indeed, we detected elevated plasma concentrations of some of these proinammatory cytokines in TB-IRIS patients at week 2, compared with non-IRIS controls. These results are in agreement with our previous studies on TB-IRIS18,28,39,40.

We hypothesize that TB-IRIS arises from a combination of a high MTB antigen load at ART initiation17 and differential antigen recognition and immune signalling by innate immune receptors after ART initiation in TB-IRIS patients, which in turn contributes to hypercytokinemia and inammation. Both TLR and IL-1R (Toll/IL-1 receptor (TIR)) share a 200-residue intracellular domain and activate a Rel-type transcription factor on stimulation41. In synergy with TIR, TREM1 modulates the production of cytokines and chemokines, and amplies the proinammatory response induced by TIRs42. Following bacterial sensing, TREM1 signals to activate downstream inammasome activity43. MTB has been shown to activate caspase-1-dependent NLRP3 canonical inammasome during infection that contributes to damaging innate inammatory responses44,45. More recently, a non-canonical inammasome has been described in Gram-negative bacterial infection that activates caspase-11 in mice (orthologue of caspase-4/5 in humans) through a lipopolysaccharide-dependent pathway46,47. Although MTB does not contain lipopolysaccharide, the bacilli have lipid-rich cell walls that contain complex glycosylated lipids, mycolic acids and acylated lipids, among others, and are readily sensed by TLR4 in response to infection48,49. We speculate that MTB may feasibly modulate the activity of non-canonical inammasomes, as signicantly higher levels of caspase-1 and caspase-5, and their prototypic cytokines IL-1b and IL-1a, were detected in PBMCs from TB-IRIS patients stimulated with heat-inactivated MTB, compared with the non-IRIS counterparts. A pan-group 1 caspase inhibitor, Z-WEHD-FMK, signicantly reduced production of IL-1a and IL-1b, with a particularly prominent effect observed in

TB-IRIS (Fig. 7), further indicating that the proinammatory response in TB-IRIS is at least in part mediated by inammasomes. It is established that TLRs and inammasome signals converge to amplify the innate immune response50. More recently, this convergence in signalling has also been shown to amplify CD4 T-cell effector functions in bacterial pathogen

clearance in a MyD88-dependent manner51. With the exception of TLR3, all other TLRs and IL-1R signal through the MyD88 adaptor. Inhibition of the MyD88 in PBMCs from TB-IRIS patients signicantly reduced the production of proinammatory cytokines (Fig. 6), indicating that a TLR-MyD88 signalling pathway contributes to the dysregulated cytokine response in TB-IRIS. Interestingly, PBMCs from the non-IRIS patients produced minimal levels of IFN-a, IFN-g, IP-10 and IL-12, following stimulation, and inhibition of the MyD88 adaptor did not reduce production of other cytokines/chemokines, further suggesting that the dysregulated cytokine/chemokine production via MyD88 signalling contributes to TB-IRIS. Together, our data provide good evidence that inammation and subsequent T-cell expansion in TB-IRIS is driven by innate immune signalling and its amplication, which triggers the activation of transcriptional factors, with the end result being hypercytokinemia contributing to systemic inammation.

Two recent studies on TB-IRIS also investigated the role of innate immunity via microarray analysis of blood monocytes24,52. A few of the differentially expressed genes in TB-IRIS from these studies were also identied in ours, which used whole blood. The authors of these studies also reported PRRs, the complement system and communication between innate and adaptive immunity as among the most signicantly upregulated pathways, further suggesting a prominent contribution of innate immunity to the pathogenesis of TB-IRIS.

Prednisone is used for the treatment of TB-IRIS and has been demonstrated in a clinical trial, to result in more rapid clinical improvement53. However, corticosteroids have potential side effects in patients with advanced HIV-related immuno-suppression such as the development of herpes zoster and Kaposis sarcoma54,55, and therefore therapeutic strategies that offer greater immune specicity should be explored. We acknowledge that our sample size is small and our ndings require further validation; nevertheless, the ability to identify a transcriptomic signature of TB-IRIS within days of commencing ART may be benecial in directing preventive interventions in the absence of a pre-ART predictive biomarker. We have demonstrated that blocking MyD88 and group 1 caspases reduced the production of proinammatory cytokines in TB-IRIS. Based on our ndings, we postulate that immuno-modulators have the potential to interrupt the inammatory pathways that lead to TB-IRIS and their utility as part of the prevention strategies of TB-IRIS could be explored in future studies. Anti-inammatory agents that suppress innate signalling networks are currently being evaluated in clinical trials as treatment options for autoimmune diseases56. An anti-IL-1b monoclonal antibody (canakinumab) is approved for use in autoinammatory syndromes and recombinant IL-1R antagonist (anakinra) for rheumatoid arthritis, suggesting a potential role for these drugs in other inammatory conditions. HIV-TB coinfected patients who display an early transcriptomic signature associated with an increased risk of developing TB-IRIS may potentially benet most from such pre-emptive therapies. In this respect, the recent description of MCC950 (ref. 57), a potent, selective, small-molecule inhibitor of NLRP3, is of particular interest.

Methods

Ethical approval. The University of Cape Towns Faculty of Health Sciences Research Ethics Committee approved this study (HREC No: 049/2009). All participants provided written informed consent.

Study design and participant recruitment. Hospitalized patients with HIV-associated TB were recruited to a prospective, observational cohort study at Brooklyn Chest Hospital (Cape Town, South Africa) between April 2009 and December 2010 (refs 39,58). Eligible patients were adults older than 18 years of age with serologically conrmed HIV-1 infection and a diagnosis of PTB and/or EPTB.

8 NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451 ARTICLE

Patients were ART naive at the time of recruitment and started ART according to WHO guidelines (CD4 count o200 cells per mm3 or WHO stage 4 dening illness before March 2010 and CD4 count o350 cells per mm3 or WHO stage 4 dening illness thereafter). Exclusion criteria included patients with known rifampicin-resistant TB and inability to provide consent. We selected the rst 32 patients(17 TB-IRIS and 15 non-IRIS) enroled, from whom longitudinal Tempus blood samples were available to undertake microarray analysis. We further included another 31 patients (16 TB-IRIS and 15 non-IRIS) to analyse plasma proteins and investigate potential mechanisms mediating TB-IRIS. All but one patient, who was prescribed nevirapine, received efavirenz-based non-nucleoside reverse-transcriptase inhibitor containing ART regimens. Ten patients (four TB-IRIS and six non-IRIS) were on corticosteroid therapy when they commenced ART for an average of 41 days. Corticosteroids were prescribed to most participants who were diagnosed with TB-IRIS.

Case denitions. TB was diagnosed on the basis of smear or culture positivity. Where this was negative or unavailable, diagnosis was according to WHO guidelines for diagnosis of smear negative or EPTB in HIV-1-infected persons59. TB-IRIS was diagnosed according to the International Network for the Study of HIV-associated IRIS case denition60. TB-IRIS patients typically displayed systemic and/or local inammatory features including fever, weight loss, tachycardia, lymphadenitis, enlarging serous effusions and new or recurrent inltrates on chest radiographs60. Non-IRIS patients were dened as those free of IRIS symptoms during the 3-month follow-up period.

Blood sample collection and processing. Whole blood was collected in sodium-heparin and Tempus tubes (Life Technologies, Carlsbad, CA) longitudinally at four different timepoints: week 0 before initiation of ART and weeks 0.5 (25 days post-ART), 1 and 2 for both TB-IRIS and non-IRIS patients post ART. Except for patients who were already receiving corticosteroids at the time of ART initiation, all other patient blood samples before and at week 2 were taken before initiation of corticosteroid therapy. Plasma and PBMCs were separated from whole blood collected in sodium-heparin using Ficoll-Paque (Pharmacia Biotech, Uppsala, Sweden) according to the manufacturers protocol. Plasma was stored at 80 C

and PBMCs were stored in the vapour phase of liquid nitrogen until use. Whole-blood and plasma samples were not available for all time points in some patients because of their discharge or transfer.

RNA extraction and processing for microarray analysis. RNA was extracted from whole blood in Tempus tubes using the MagMAX for Stabilized Blood Tubes RNA Isolation kit (Life Technologies) according to the manufacturers instructions. Total RNA (2.5 mg) was then globin reduced using the GLOBINclear 96-well plate format kit (Life Technologies). The total RNA yield after globin reduction was determined in a NanoDrop 1,000 spectrophotometer (Fisher Scientic, Waltham, MA) and integrity assessed by Agilent 2,100 Bioanalyzer (Agilent Technologies, Santa Clara, CA). Samples with RNA integrity number 46.0 were determined to be of acceptable quality for further analysis. Approximately 250 ng of globin-reduced RNA was used to prepare amplied and biotinylated antisense complementary RNA targets using the Illumina TotalPrep RNA amplication kit (Life Technologies). Seven hundred and fty nanograms of labelled cRNA was hybridized overnight to Illumina human HT-12 v4 BeadChip arrays (Illumina Inc., San Diego, CA), which contained 447,000 probes (transcripts). Some genes were represented by more than one probe (transcript). The arrays were then washed, blocked, stained and scanned on an Illumina BeadStation 500. Signal intensity values from the scans were generated by the Illumina BeadStudio v2 software. All samples were randomized during all stages of processing to avoid any batch effect.

Microarray data analysis. A detail ow diagram on the analytical strategy is shown in Supplementary Fig. 1. Raw microarray data were background subtracted and scaled data were generated using Illumina BeadStudio v2 software. A stringent lter was implemented to select only those transcripts present in 100% of all samples in either the IRIS or the non-IRIS group. The transcripts were further selected to have Z2-fold expression change compared with the non-IRIS samples.

The MannWhitney U-test was applied and corrected with asymptotic Benjamini Hochberg false discovery rate set at P 0.05. The identied transcripts were then

clustered using a hierarchical algorithm with Pearsons uncentred distance metric and average linkage.

Canonical pathway, upstream regulator and molecule activity prediction of differentially abundant transcripts were analysed by Ingenuity Pathway Analysis (Ingenuity Systems, http://www.ingenuity.com

Web End =www.ingenuity.com ). The signicance (P-value) of association between the transcripts and the pathway was calculated using Fishers exact test of probability. The weighted temporal molecular distance was determined for each time point as previously described61, except that the mean baseline value of each group was used as a control. Briey, the algorithm determined whether the raw expression value of each differentially expressed transcript in each sample lay within or outside two s.d. of the baseline mean. To qualify, transcripts had to differ by at least 200 units in raw signal intensity and 2 s.d. from the baseline mean. The weighted distance is the group mean of the sum of the total s.d. for all qualifying transcripts in each sample in each patient.

NanoString. Differentially abundant transcripts identied by microarray analysis were reinvestigated and validated using a customized NanoString nCounter assay (NanoString Technologies, Seattle, WA)62. Housekeeping genes (n 9) and

customized barcoded capture/reporter probe pairs specic for each transcript (n 124) were hybridized overnight at 65 C to 100 ng of total RNA samples

previously used in microarray. Six additional positive and negative control probe pairs were also included. Unhybridized probes were removed and the hybridized probes were puried on an nCounter Prep Station. The barcode on each reporter probe was scanned with an nCounter Digital Analyzer to generate a quantitative measure of the hybridized RNA. Sample signal values were background subtracted and normalized rst to the positive control spikes and then to housekeeping genes. Ninety-nine of the 107 samples from week 0, 0.5, 1 and 2 were included when sufcient RNA was available.

Plasma protein determination. The plasma concentrations of 18 cytokines and chemokines were measured in 33 TB-IRIS and 30 non-IRIS patients in week 0 and week 2 samples. TNF-a, IFN-g, IFN-a2 and IL-1b, IL-6, IL-8 and IL-12p40 were quantied on the Bio-Plex platform (Bio-Rad Laboratories, Hercules, CA), using customized Milliplex kits (HCYTMAG-60 K, Merck Millipore, Germany). The concentration of IL-1a was measured by enzyme-linked immunorsorbent assay (eBioscience, San Diego, CA).

MyD88 inhibition assay. PBMCs from 13 TB-IRIS and non-IRIS patients (week 2, microarrayed) were defrosted and rested overnight at 37 C in RPMI with 10% FCS. Total cell number was counted using a Coulter Counter (Beckman, Brea, CA) and cell viability was determined by Trypan blue staining in a TC-20 automatic cell counter (Bio-Rad, Hercules, CA). Lyophilized MyD88 peptide inhibitor (Pepinh-MYD; InvivoGen, San Diego, CA) and a control peptide (Pepinh-Control; InvivoGen) were dissolved in endotoxin-free water at 1 mM stock concentration. PBMCs were treated with MyD88 peptide inhibitor (50 mM), control peptide(50 mM) or RPMI mock for 5 h at 37 C. Samples were either stimulated or unstimulated with heat-inactivated MTB (H37Rv; multiplicity of infection 1:1) overnight at 37 C. TC supernatant was harvested and puried by centrifugation twice at 2,000 r.p.m. and stored at 80 C until use. Cytokine concentrations in

supernatants were quantied by Luminex analysis. Heat-inactivated H37Rv was used, because patients had undergone intensive antibiotic therapy at the time of sampling and any antigens present would probably be bacterial fragments rather than actively replicating bacilli.

Group 1 caspase inhibition assay. PBMCs from 16 TB-IRIS and 15 non-IRIS patients (week 2, non-microarrayed) were thawed, rested and counted as above. Lyophilized caspase-1/4/5 irreversible peptide inhibitor (Z-WEHD-FMK; R&D Systems, Minneapolis, MN) was dissolved in dimethyl sulfoxide at 1 mM stock concentration. PBMCs were treated with the Z-WEHD-FMK peptide inhibitor (50 mM) or dimethyl sulfoxide vehicle control (50 mM) for 1 h at 37 C. Supernatant was then removed and replenished with a fresh stock of RPMI supplemented with 10% FCS. Samples were either stimulated or unstimulated with heat-inactivated MTB (H37Rv; multiplicity of infection 1:1) overnight at 37 C. TC supernatant was harvested and puried by centrifugations twice at 2,000 r.p.m. The residual PBMC pellets were lysed in lysis buffer on ice for 1 h and recentrifuged to remove all cell debris. Cytokine concentrations in supernatants were quantied by Luminex assay. Concentrations of caspase-1 (R&D Systems) and IL-1a (eBioscience) in super-natant and caspase-3 (eBioscience) and caspase-5 (BioVision, Milpitas, CA) in PBMC lysate were measured by enzyme-linked immunorsorbent assay. The activity of caspase-5 was expressed as the fold change in absorbance (OD405nm) in MTB-

stimulated over unstimulated PBMC lysate, as recombinant protein standards were not supplied in the assay kit.

Statistical analysis. Statistical analyses were performed in GraphPad Prism 6.0 software. Patient baseline characteristics were compared by w2-test for contingency data or KruskalWallis test for continuous data. The weighted temporal molecular distance between IRIS and non-IRIS patients was compared using the unpaired t-test with Welchs correction. Differences in cytokine concentrations between patient or treatment groups were compared by MannWhitney U- test or Wilcoxon signed rank test for paired data. Potential signicance was inferred for results associated with a P-value of r0.05. Microarray data and canonical pathway analysis were corrected for multiple testing using BenjaminiHochberg false discovery rate with q 0.05.

References

1. WHO. World Health Organization: Global Tuberculosis Report 2014. http://apps.who.int/iris/bitstream/10665/137094/1/9789241564809_eng.pdf?ua=1

Web End =http://apps.who.int/iris/bitstream/10665/137094/1/ http://apps.who.int/iris/bitstream/10665/137094/1/9789241564809_eng.pdf?ua=1

Web End =9789241564809_eng.pdf?ua=1 (2014).

2. Selwyn, P. A. et al. A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeciency virus infection. N. Engl.J. Med. 320, 545550 (1989).

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451

3. Suthar, A. B. et al. Antiretroviral therapy for prevention of tuberculosis in adults with HIV: a systematic review and meta-analysis. PLoS Med. 9, e1001270 (2012).

4. Walker, N. F., Meintjes, G. & Wilkinson, R. J. HIV-1 and the immune response to TB. Future Virol. 8, 5780 (2013).

5. Narendran, G. et al. Paradoxical tuberculosis immune reconstitution inammatory syndrome (TB-IRIS) in HIV patients with culture conrmed pulmonary tuberculosis in India and the potential role of IL-6 in prediction. PLoS ONE 8, e63541 (2013).

6. Laureillard, D. et al. Paradoxical tuberculosis-associated immune reconstitution inammatory syndrome after early initiation of antiretroviral therapy in randomized clinical trial. AIDS 27, 25772586 (2013).

7. Luetkemeyer, A. F. et al. Tuberculosis immune reconstitution inammatory syndrome in A5221 STRIDE: timing, severity, and implications for HIV-TB programs. J. Acquir. Immune Dec. Syndr. 65, 423428 (2014).

8. Naidoo, K. et al. The immune reconstitution inammatory syndrome after antiretroviral therapy initiation in patients with tuberculosis: ndings from the SAPiT trial. Ann. Intern. Med. 157, 313324 (2012).

9. Burman, W. et al. Frequency, severity and duration of immune reconstitution events in HIV-related tuberculosis. Int. J. Tuberc. Lung Dis. 11, 12821289 (2007).

10. Manosuthi, W., Kiertiburanakul, S., Phoorisri, T. & Sungkanuparph, S. Immune reconstitution inammatory syndrome of tuberculosis among HIV-infected patients receiving antituberculous and antiretroviral therapy. J. Infect. 53, 357363 (2006).

11. Lai, R. P., Nakiwala, J. K., Meintjes, G. & Wilkinson, R. J. The immunopathogenesis of the HIV tuberculosis immune reconstitution inammatory syndrome. Eur. J. Immunol. 43, 19952002 (2013).

12. Bourgarit, A. et al. Explosion of tuberculin-specic Th1-responses induces immune restoration syndrome in tuberculosis and HIV co-infected patients. AIDS 20, F1F7 (2006).

13. Meintjes, G. et al. Type 1 helper T cells and FoxP3-positive T cells in HIV-tuberculosis-associated immune reconstitution inammatory syndrome. Am. J. Respir. Crit. Care Med. 178, 10831089 (2008).

14. Lawn, S. D., Wainwright, H. & Orrell, C. Fatal unmasking tuberculosis immune reconstitution disease with bronchiolitis obliterans organizing pneumonia: the role of macrophages. AIDS 23, 143145 (2009).

15. Conradie, F. et al. Natural killer cell activation distinguishes Mycobacterium tuberculosis-mediated immune reconstitution syndrome from chronic HIV and HIV/MTB coinfection. J. Acquir. Immune Dec. Syndr. 58, 309318 (2011).

16. Pean, P. et al. Natural killer cell degranulation capacity predicts early onset of the immune reconstitution inammatory syndrome (IRIS) in HIV-infected patients with tuberculosis. Blood 119, 33153320 (2012).

17. Marais, S. et al. Frequency, severity, and prediction of tuberculous meningitis immune reconstitution inammatory syndrome. Clin. Infect. Dis. 56, 450460 (2013).

18. Marais, S. et al. Neutrophil-associated central nervous system inammation in tuberculous meningitis immune reconstitution inammatory syndrome. Clin. Infect. Dis. 59, 16381647 (2014).

19. Barber, D. L., Andrade, B. B., McBerry, C., Sereti, I. & Sher, A. Role of IL-6 in Mycobacterium avium-associated immune reconstitution inammatory syndrome. J. Immunol. 192, 676682 (2014).

20. Andrade, B. B. et al. Mycobacterial antigen driven activation ofCD14 CD16- monocytes is a predictor of tuberculosis-associated immune

reconstitution inammatory syndrome. PLoS Pathog. 10, e1004433 (2014).21. Haddow, L. J., Dibben, O., Moosa, M. Y., Borrow, P. & Easterbrook, P. J. Circulating inammatory biomarkers can predict and characterize tuberculosis-associated immune reconstitution inammatory syndrome. AIDS 25, 11631174 (2011).

22. Porter, B. O. et al. d-Dimer and CRP levels are elevated prior to antiretroviral treatment in patients who develop IRIS. Clin. Immunol. 136, 4250 (2010).

23. Tan, D. B. et al. TLR2-induced cytokine responses may characterize HIV-infected patients experiencing mycobacterial immune restoration disease. AIDS 25, 14551460 (2011).

24. Tran, H. T. et al. Modulation of the complement system in monocytes contributes to tuberculosis-associated immune reconstitution inammatory syndrome. AIDS 27, 17251734 (2013).

25. Berry, M. P. et al. An interferon-inducible neutrophil-driven blood transcriptional signature in human tuberculosis. Nature 466, 973977 (2010).

26. Bloom, C. I. et al. Transcriptional blood signatures distinguish pulmonary tuberculosis, pulmonary sarcoidosis, pneumonias and lung cancers. PLoS ONE 8, e70630 (2013).

27. Kramer, A., Green, J., Pollard, Jr J. & Tugendreich, S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics 30, 523530 (2014).

28. Tadokera, R. et al. Hypercytokinaemia accompanies HIV-tuberculosis immune reconstitution inammatory syndrome. Eur. Respir. J. 37, 12481259 (2011).

29. Ladel, C. H. et al. Lethal tuberculosis in interleukin-6-decient mutant mice. Infect. Immun. 65, 48434849 (1997).

30. Mayer-Barber, K. D. et al. Innate and adaptive interferons suppress IL-1alpha and IL-1beta production by distinct pulmonary myeloid subsets during Mycobacterium tuberculosis infection. Immunity 35, 10231034 (2011).31. Cooper, A. M. et al. Disseminated tuberculosis in interferon gamma gene-disrupted mice. J. Exp. Med. 178, 22432247 (1993).

32. Flynn, J. L. et al. An essential role for interferon gamma in resistance to Mycobacterium tuberculosis infection. J. Exp. Med. 178, 22492254 (1993).

33. Flynn, J. L. et al. Tumor necrosis factor-alpha is required in the protective immune response against Mycobacterium tuberculosis in mice. Immunity 2, 561572 (1995).

34. OGarra, A. et al. The immune response in tuberculosis. Annu. Rev. Immunol. 31, 475527 (2013).

35. Guarda, G. et al. Type I interferon inhibits interleukin-1 production and inammasome activation. Immunity 34, 213223 (2011).

36. McNab, F. W. et al. Type I IFN induces IL-10 production in an IL-27-independent manner and blocks responsiveness to IFN-gamma for production of IL-12 and bacterial killing in Mycobacterium tuberculosis-infected macrophages. J. Immunol. 193, 36003612 (2014).

37. Mayer-Barber, K. D. et al. Host-directed therapy of tuberculosis based on interleukin-1 and type I interferon crosstalk. Nature 511, 99103 (2014).

38. Schreiber, T. et al. Autocrine IL-10 induces hallmarks of alternative activation in macrophages and suppresses antituberculosis effector mechanisms without compromising T cell immunity. J. Immunol. 183, 13011312 (2009).

39. Conesa-Botella, A. et al. Corticosteroid therapy, vitamin D status, and inammatory cytokine prole in the HIV-tuberculosis immune reconstitution inammatory syndrome. Clin. Infect. Dis. 55, 10041011 (2012).

40. Meintjes, G. et al. Corticosteroid-modulated immune activation in the tuberculosis immune reconstitution inammatory syndrome. Am. J. Respir. Crit. Care Med. 186, 369377 (2012).

41. ONeill, L. A. The interleukin-1 receptor/Toll-like receptor superfamily: 10 years of progress. Immunol. Rev. 226, 1018 (2008).

42. Colonna, M. & Facchetti, F. TREM-1 (triggering receptor expressed on myeloid cells): a new player in acute inammatory responses. J. Infect. Dis. 187(suppl 2): S397S401 (2003).

43. Dower, K., Ellis, D. K., Saraf, K., Jelinsky, S. A. & Lin, L. L. Innate immune responses to TREM-1 activation: overlap, divergence, and positive and negative cross-talk with bacterial lipopolysaccharide. J. Immunol. 180, 35203534 (2008).

44. Dorhoi, A. et al. Activation of the NLRP3 inammasome by Mycobacterium tuberculosis is uncoupled from susceptibility to active tuberculosis. Eur. J. Immunol. 42, 374384 (2012).

45. Mishra, B. B. et al. Nitric oxide controls the immunopathology of tuberculosis by inhibiting NLRP3 inammasome-dependent processing of IL-1beta. Nat. Immunol. 14, 5260 (2013).

46. Kayagaki, N. et al. Non-canonical inammasome activation targets caspase-11. Nature 479, 117121 (2011).

47. Kayagaki, N. et al. Noncanonical inammasome activation by intracellular LPS independent of TLR4. Science 341, 12461249 (2013).

48. Reiling, N. et al. Cutting edge: Toll-like receptor (TLR)2- and TLR4-mediated pathogen recognition in resistance to airborne infection with Mycobacterium tuberculosis. J. Immunol. 169, 34803484 (2002).

49. Teixeira-Coelho, M. et al. Differential post-transcriptional regulation of IL-10 by TLR2 and TLR4-activated macrophages. Eur. J. Immunol. 44, 856866 (2014).

50. Schroder, K. & Tschopp, J. The inammasomes. Cell 140, 821832 (2010).51. ODonnell, H. et al. Toll-like receptor and inammasome signals converge to amplify the innate bactericidal capacity of T helper 1 cells. Immunity 40, 213224 (2014).

52. Tran, H. T. et al. The role of monocytes in the development of Tuberculosis-associated Immune Reconstitution Inammatory Syndrome. Immunobiology 219, 3744 (2014).

53. Meintjes, G. et al. Randomized placebo-controlled trial of prednisone for paradoxical tuberculosis-associated immune reconstitution inammatory syndrome. AIDS 24, 23812390 (2010).

54. Elliott, A. M. et al. A randomized, double-blind, placebo-controlled trial of the use of prednisolone as an adjunct to treatment in HIV-1-associated pleural tuberculosis. J. Infect. Dis. 190, 869878 (2004).

55. Volkow, P. F., Cornejo, P., Zinser, J. W., Ormsby, C. E. & Reyes-Teran, G. Life-threatening exacerbation of Kaposis sarcoma after prednisone treatment for immune reconstitution inammatory syndrome. AIDS 22, 663665 (2008).56. Kanzler, H., Barrat, F. J., Hessel, E. M. & Coffman, R. L. Therapeutic targeting of innate immunity with Toll-like receptor agonists and antagonists. Nat. Med. 13, 552559 (2007).

57. Coll, R. C. et al. A small-molecule inhibitor of the NLRP3 inammasome for the treatment of inammatory diseases. Nat. Med. 21, 248255 (2015).58. van der Plas, H. et al. Complications of antiretroviral therapy initiation in hospitalised patients with HIV-associated tuberculosis. PLoS ONE 8, e54145 (2013).

10 NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9451 ARTICLE

59. World Health Organization. TB-HIV: Improving the diagnosis and treatment of smear-negative pulmonary and extrapulmonary tuberculosis among adults and adolescents. Recommendations for HIV-prevalent and resource-constrained settings (Stop TB Department, HIV/AIDS, Tuberculosis and Malaria; World Health Organization, 2006).

60. Meintjes, G. et al. Tuberculosis-associated immune reconstitution inammatory syndrome: case denitions for use in resource-limited settings. Lancet Infect. Dis. 8, 516523 (2008).

61. Pankla, R. et al. Genomic transcriptional proling identies a candidate blood biomarker signature for the diagnosis of septicemic melioidosis. Genome Biol. 10, R127 (2009).

62. Geiss, G. K. et al. Direct multiplexed measurement of gene expression with color-coded probe pairs. Nat. Biotechnol. 26, 317325 (2008).

Acknowledgements

We thank all the participants in this study and the health-care workers and administrators at Brooklyn Chest Hospital. We thank Renaud Prevel for help with sample processing. This work was supported by the MRC (U1175.02.002.00014.01), Wellcome Trust (references 081667, 104803, 098316 and 097254), the European Union (PIRSESGA-2011295214 and FP7 HEALTH-F3-2012-305578) and the National Research Foundation of South Africa (96841).

Author contributions

R.P.J.L., G. Meintjes, A.O.G. and R.J.W. conceived and designed the experiments.G. Meintjes, H.V.d.P., A.D., C.S. and R.G. recruited, sampled and collected data from patients. R.P.J.L., G. Meintjes, K.A.W., C.M.G., I.M. and E.A. performed the experiments. R.P.J.L., G. Meintjes, C.M.G. and C.B. analysed the data. G. Maartens, J.B., D.C.,

A.O.G. and R.J.W. contributed reagents, materials and/or analysis tools. K.A.W., C.M.G., S.M., G. Maartens and C.B. provided intellectual input. R.P.J.L., G. Meintjes, A.O.G. and R.J.W. wrote the paper.

Additional information

Accession codes: The microarray data has been deposited in the NCBIs Gene Expression Omnibus under GEO Series accession number GSE58411.

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Lai, R. P. J. et al. HIVtuberculosis-associated immune reconstitution inammatory syndrome is characterized by Toll-like receptorand inammasome signalling. Nat. Commun. 6:8451 doi: 10.1038/ncomms9451 (2015).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 6:8451 | DOI: 10.1038/ncomms9451 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

& 2015 Macmillan Publishers Limited. All rights reserved.