Human TOLLIP Regulates TLR2 and TLR4 Signaling and Its Polymorphisms Are Associated with Susceptibility to Tuberculosis

Tuberculosis is one of the leading causes of infectious death worldwide, yet its immune pathogenesis remains incompletely understood (1, 2). The innate immune system is critical for the initial host defense against mycobacteria (3). Several lines of evidence suggest that TLR1, TLR2, TLR4, TLR6, TLR8, and TLR9 are important for host defense against Mycobacterium tuberculosis (4-11). For example, in vitro studies in murine macrophages demonstrate that TLR1, TLR2, TLR4, TLR6, TLR9, and MyD88 regulate M. tuberculosis-induced cytokine secretion (5, 12-21). The mycobacterial cell wall contains a number of proinflammatory TLR2 ligands, including lipoproteins, mycolylarabinogalactan–peptidoglycan complex (the cell wall core structure), lipids, and LAM (5, 6, 12–14). Despite the multiple lines of evidence demonstrating a critical role for TLR mediation of M. tuberculosis recognition in vitro, the in vivo significance of individual TLRs has been more difficult to consistently demonstrate. Although some studies demonstrated increased susceptibility to M. tuberculosis in MyD88^−/−, and Tlr2^−/−, and Tlr2/9^−/− mice, these results have not been uniformly observed under all experimental conditions (10, 18, 21, 22-25). Furthermore, the effects appear to be strongest when multiple TLRs are impaired. Together, these in vivo studies suggest an important though not absolute role for the TLR pathway in mediating host protection to in vivo murine M. tuberculosis infection.

Through the TLR pathway, proinflammatory (IL-6, TNF, and IL-12) and anti-inflammatory (IL-10) cytokines are secreted from macrophages stimulated with M. tuberculosis (26, 27). The interplay between proinflammatory and anti-inflammatory cytokines is important in determining the quality and quantity of both the innate and adaptive immune response to pathogens. The signaling mechanisms that specifically modulate IL-10 production after TLR stimulation are unclear, and understanding how IL-10 secretion is modulated by TLR regulators has important implications in M. tuberculosis pathogenesis and vaccine design. IL-10^−/− mice develop severe colitis in response to gut microbiota (28), making extrapolation of the role of IL-10 in humans difficult. Several investigators reported that IL-10^−/− mice have enhanced M. tuberculosis clearance (29), but others have reported no such differences (30, 31). Human studies of the role of IL-10 in tuberculosis (TB) have shown that IL-10 levels are increased in lung tissue of patients with pulmonary TB (32). Furthermore, neutralization of IL-10 enhances IFN-γ and T cell proliferation in vitro (33). Human genetic studies examining whether polymorphisms in the IL-10 gene are associated with susceptibility to TB are inconclusive (34-36). Together, these studies suggest that the role of IL-10 in the pathogenesis of TB remains unclear and further study is required.

Negative regulators have been discovered at nearly every step of the TLR signaling cascade (37–39) and include Toll-interacting protein (TOLLIP, also called IL-1RAcPIP), A20 (40), Single Ig IL-1R (SIGIRR, also called TIR8) (41), and IL-1R–associated kinase 3 (IRAK-3, also called IRAK-M) (42). TOLLIP is a 274-aa protein with highly conserved C2 (amino acids 54-186, similar to that found in phosphoinositol-specific phospholipase C-d1I) and C-terminal ubiquitin-associated domains. It was initially discovered as an IL-1R1–interacting protein that linked IRAK to the IL-1R pathway (43), and later was implicated in suppression of the TLR2 (44) and TLR4 pathways (45). However, in vivo murine knockout models demonstrated that TOLLIP induced proinflammatory pathways, in contrast to in vitro experiments (46). To our knowledge, there are no studies examining the effects of TOLLIP on TLR signaling pathways in humans.

Several lines of evidence suggest that host genetics regulates susceptibility to TB, including twin-based, linkage, candidate gene association, and genome-wide association studies. We, along with other investigators, have identified associations between common polymorphisms in innate immunity genes and susceptibility to TB and clinical phenotypes (22, 47, 48). Much of this work has focused on pattern recognition receptors and their associated adaptor proteins, including TLR 1, 2, 4, 6, and 9. Several candidate gene association studies have shown associations between single nucleotide polymorphisms (SNPs) in TLR2 and TB, including 597CC within a Vietnamese population (8, 49). A polymorphism in TLR1 (T1805G, I602S) regulates lipopeptide-induced cytokine secretion and TLR1 surface expression, and is associated with susceptibility to TB as well as leprosy (48, 50–52). Genetic studies have also found associations of TLR4 polymorphisms with TB susceptibility (10, 24, 25). These studies suggest that genetic control of TLR1/2/4/6 signaling by negative regulators may alter susceptibility to TB. In contrast to TLRs, the functional and clinical significance of genetic variation of TOLLIP and other negative regulators of TLR signaling is poorly studied and only partially understood. In this study, we examined the function of TOLLIP in human monocytes with knockdown studies and examination of common TOLLIP variants. We also examined whether these polymorphisms were associated with susceptibility to TB in Vietnam. These studies suggest that TOLLIP is a critical regulator of the TLR pathway in humans and that this immune regulation has a critical role in the susceptibility to TB.

RPMI 1640 medium, l-glutamine, penicillin, and streptomycin were obtained from Life Technologies (Carlsbad, CA). Ultrapure LPS (TLR4 ligand) isolated from Salmonella minnesota R595 was obtained from List Biological Labs (Campbell, CA). The synthetic lipopeptides PAM2 (PAM2CSKKKK, S-[2,3-bis(palmitoyloxy)-propyl]-(R)-cysteinyl-(lysyl)3-lysine, TLR2/6 ligand) and PAM3 (Pam3CSKKKK, N-palmitoyl-S-[2,3-bis-(palmitoyloxy)-propyl]-(R)-cysteinyl-(lysyl)3-lysine, TLR2/1 ligand) were obtained from EMC Microcollections (Tuebingen, Germany). Whole cell lysate from M. tuberculosis strain H37Rv was obtained as part of National Institutes of Health, National Institute of Allergy and Infectious Diseases Contract No. HHSN266200400091C, entitled Tuberculosis Vaccine Testing and Research Materials (Colorado State University, Fort Collins, CO).

Primary monocytes were isolated from human subjects as described previously (53). Human monocytes were isolated from peripheral blood with Ficoll gradient separation followed by positive selection using human anti-CD14 Ab associated with magnetic beads (Miltenyi Biotec, Auburn, CA). Monocytes were isolated with >95% purity, based on CD14 positivity in flow cytometric analysis and were maintained in RPMI 1640 supplemented with 10% FCS. Cells were incubated with TLR ligands PAM2, PAM3, and LPS, or with M. tuberculosis whole cell lysates for 24 h. Subsequently, supernatants were evaluated for levels of cytokine (IL-6, IL-10, TNF) via ELISA Duoset (R&D Systems, Minneapolis, MN). Each sample was assayed in duplicate or triplicate, and experiments shown were performed at least twice to ensure reproducibility.

Genotyping was performed with Sequenom’s MassARRAY technique as described previously (54). This technique uses allele-specific primer extension reactions to discriminate genotypes. We also confirmed the SNPs of interest using Taqman genotyping technology (Applied Biosystems, Carlsbad, CA). We identified haplotype-tagging SNPs from the CHB (Han Chinese in Beijing) and CEU (Utah residents with European ancestry) populations from the International HapMap Project (http://www.hapmap.org) and other public databases with the Genome Variation Server (http://www.ncbi.nlm.nih.gov/SNP/ and www.innateimmunity.net). We searched a region on chromosome 11p15.5, 10 kb upstream and downstream of TOLLIP for tagged SNPs using an R² cutoff of 0.8 for linkage disequilibrium and a minor allele frequency cut-off of 5%. Stata/Intercooled version 11.0 software program PWLD (StataCorp, College Station, TX) was used to calculate R² and D′ as measurements of linkage disequilibrium between the polymorphisms.

TOLLIP mRNA levels were measured in monocytes from peripheral blood of healthy donors in the Seattle area. RNA was obtained after cell lysis and RNA purification (Qiagen, Valencia, CA). After synthesis of cDNA, real-time PCR was performed. TOLLIP mRNA levels were normalized with a GAPDH control using the following primer-probe sets (5′-J56-FAM/CTCCGGGAT/ZEN/GGT-3′, 5′-CTCAGGCTGTACCACTTG-3′, 5′-GGCGTGGACTCTTTCTAT-3′; Integrated DNA Technologies, Coralville, IA). Quantitative PCR was performed on a Taqman machine.

Knockdown studies were performed using plasmids encoding three TOLLIP-specific short hairpin RNA (shRNA; Santa Cruz Biotechnology, Santa Cruz, CA) that were then packaged into a nonreplicating lentiviral particle to improve transfection efficiency using a standard protocol (UCSF Viracore Facility, San Francisco, CA). These lentiviral particles were then incubated with monocytes at a multiplicity of infection of 2 for 24 h before cells were stimulated and harvested for cytokine analysis. Knockdown was confirmed using quantitative PCR for TOLLIP mRNA transcript.

Approval for human study protocols was obtained from the human subjects review boards at the University of Washington School of Medicine (Seattle, WA), the Hospital for Tropical Diseases (Ho Chi Minh City, Vietnam), Pham Ngoc Thach Hospital for Tuberculosis and Lung Diseases (Ho Chi Minh City,Vietnam), Health Services of Ho Chi Minh City, Hung Vuong Hospital (Ho Chi Minh City, Vietnam), and the Oxford Tropical Research Ethics Committee (Oxford, U.K.).

Genomic DNA was purified from saliva and blood samples. The Seattle study group included 56 healthy volunteers who donated blood and saliva samples. The ethnic composition of these subjects was 70% Caucasian (39/56), 27% Asian (15/56), and 4% African American (2/56). Study subjects with tuberculous meningitis (TBM) were recruited from two centers in Ho Chi Minh City, Vietnam: Pham Ngoc Thach (PNT) Hospital for Tuberculosis and the Hospital for Tropical Diseases (HTD). These 500-bed hospitals serve the local community and act as tertiary referral centers for TB (PNT) and infectious diseases (HTD) for southern Vietnam. Individuals at least 15 y old, admitted to these centers with clinical meningitis (defined as nuchal rigidity and abnormal cerebrospinal fluid parameters), a negative HIV test result, and a positive Ziehl–Neelsen stain for acid-fast bacilli or M. tuberculosis culture, or both, from cerebrospinal fluid (“definite TBM”) were recruited for genetics studies during 2001 to 2008. In addition to definite TBM, the cohort included subjects with “probable TBM,” defined as clinical meningitis plus at least one of the following: chest radiograph consistent with active TB, acid-fast bacilli found in any specimen other than cerebrospinal fluid, and clinical evidence of other extrapulmonary TB (55).

From 2003 through 2004, subjects with pulmonary TB were recruited from a network of district TB control units within Ho Chi Minh City that provide directly observed therapy to TB patients. These subjects were outpatients who were at least 15 y old, had no previous history of treatment for TB, no evidence of miliary or extrapulmonary TB, chest radiograph results consistent with nonmiliary pulmonary TB, negative HIV test results, and sputum smear positive for acid-fast bacilli or M. tuberculosis cultured from sputum. In addition, pulmonary TB subjects enrolled were recruited from PNT hospital from 2006 through 2008.

Control subjects were enrolled at Hung Vuong Hospital in Ho Chi Minh City from blood collected from the umbilical cord of babies after birth. All case and control participants were unrelated, and greater than 95% were of the Vietnamese Kinh ethnicity. Written, informed consent was obtained from patients or their relatives if the patient could not provide consent (i.e., was unconscious). Parents provided consent for cord-blood controls.

Out of the study subjects that self-identified their ancestry as Vietnamese Kinh, we formally tested for levels of background genetic differences between cases and controls by genotyping the cohort for a panel of 24 independent SNP markers (56). The mean χ² statistic for the 24 SNPs comparing allelic frequencies between cases and control was 1.6 (p = 0.20), suggesting that no significant population stratification was present in our study population (57).

All analyzed SNPs were tested for Hardy-Weinberg equilibrium in control subjects using a χ² goodness-of-fit test. In our primary analysis, we examined whether polymorphism genotype frequencies were associated with any type of TB in a case-population study design using Stata MP 11 and the user written package “genassoc” (58). For secondary analyses, SNPs were investigated for associations under additional genetic models (dominant, recessive, and additive) and for association with the clinical subtypes of TB (pulmonary or meningeal). In the recessive model, carriers of allele 1 (00 and 01 genotypes) were compared with homozygous subjects for allele 2 (11 genotype). In the dominant model, carriers of allele 2 (01 and 11 genotypes) were compared with homozygous subjects for allele 1 (00 genotype). Measures of linkage disequilibrium were assessed in controls using “pwld” in Stata.

Nine TOLLIP polymorphisms were genotyped in the Vietnam and Seattle cohorts. Two (rs4744015 and rs5744016) were monogenic and one (rs5743890) was not in Hardy-Weinberg equilibrium and thus not analyzed further. Six are shown in all subsequent analyses. If we correct for multiple comparisons with a Bonferroni test (p value/6), only p values less than 0.008 remain significant. However, as we had two single nucleotide polymorphisms of interest based upon prior functional data, we show uncorrected p values in these tables.

The Pearson-χ² test and Student t test were used to assess categorical and continuous clinical variables, respectively. For analysis of associations between TOLLIP polymorphisms and mRNA levels or cytokines, we used an ANOVA analysis with Mann–Whitney tests for significance.

In order to determine the functional role of TOLLIP in the human innate immune system, we performed shRNA knockdown experiments on peripheral blood monocytes from healthy volunteers (Fig. 1). Using a lentiviral delivery system in primary monocytes, we knocked down TOLLIP mRNA expression by more than 50% compared with monocytes alone or those treated with control lentiviral particles (Fig. 1A; p < 0.001, Student t test). We also tested and confirmed that TLR2 and TLR4 mRNA expression was not altered by TOLLIP shRNA (Fig. 1A; p = 0.78 for TLR2, p = 0.14 for TLR4; Student t test comparing control to TOLLIP shRNA viral infection). After stimulation with PAM2 (TLR2/6 ligand) and PAM3 (TLR2/1 ligand), TOLLIP-deficient monocytes produced elevated levels of IL-6 compared with either untransfected monocytes or control-lentiviral particle-transfected monocytes (p < 0.01; Fig 1B–F). Transfection with control shRNA led to a mild increase in IL-6 that we attributed to the innate inflammatory response to lentivirus infection and cytosolic DNA. In contrast, LPS (TLR4 ligand)-induced IL-6 levels were similar under all conditions (Fig. 1E). TOLLIP knockdown led to increased TNF production after stimulation with PAM2, PAM3, LPS, or M. tuberculosis whole cell lysate (p < 0.01; Fig. 1G–K). In cells stimulated with PAM2, PAM3, or M. tuberculosis whole cell lysate, there were no significant differences in IL-10 production between cells stimulated with nonspecific lentiviral particles or TOLLIP-specific viral particles (Fig 1M, 1N). Intriguingly, TOLLIP knockdown in LPS-stimulated cells led to significant decreases in secreted IL-10 from monocytes (p < 0.01; Fig 1O). In a separate experiment with stimulation with M. tuberculosis H37Rv whole cell lysate, we found that TOLLIP knockdown led to increased IL-6 and TNF secretion, but not IL-10 compared with controls (p < 0.01; Fig. 1F, 1K, 1P).

These data suggest that TOLLIP regulates the innate immune response in humans via two mechanisms—by suppressing proinflammatory cytokines via TLR2 and TLR4 and by inducing IL-10 through a TLR4-specific mechanism.

Next, we examined whether common variants of TOLLIP regulated its function. We genotyped nine haplotype-tagging SNPs in 84 healthy volunteers and examined their association with TOLLIP mRNA expression in monocytes. (Fig. 2, polymorphisms listed in genomic order) For SNP rs3750920, the minor homozygote was significantly associated with increased mRNA expression compared with either heterozygotes or major homozygotes (p < 0.01 by genotypic model; CC = 26, CT = 32, TT = 16; Fig. 2D). This association was also significant in a recessive model comparing TT homozygotes with CT/CC genotypes (p < 0.01). SNP rs5743899 genotype GG was associated with decreased mRNA transcript in a recessive model. (p < 0.01; AA = 49, AG = 31, GG = 4; Fig. 2A). The other 4 TOLLIP SNPs were not significantly associated with mRNA expression (Fig. 2B, 2C, 2E, 2F). Three SNPs (rs3793964, CC = 36, CT = 24, TT = 16; rs3829223, CC = 19, CC = 32, TT = 27; and rs3168046, CC = 15, CT = 22, TT = 14) showed a trend toward an association with mRNA expression (Fig. 2C, 2E, 2F; p = 0.26, 0.22, and 0.14, respectively). These SNPs were in linkage disequilibrium with SNP rs3750920 (Supplemental Fig. 1A; R² = 0.43, 0.55, 0.73 for pairwise comparisons of three SNPs with rs3750920). These results suggest that at least one TOLLIP polymorphism is associated with mRNA expression and is a genetic marker of human TOLLIP deficiency.

We next examined whether polymorphisms rs5743899 and rs3750920 were associated with cytokine responses in monocytes. We stimulated PBMCs with TLR ligands (PAM2, PAM3, and LPS) as well as M. tuberculosis whole cell lysate, and measured secreted IL-6 and IL-10 in the culture supernatants (Fig. 3A, 3B). Minor homozygotes (GG) of polymorphism rs5743899 were associated with significantly higher levels of IL-6 compared with heterozygotes or major homozygotes (Fig 3C, 3D, 3E; AA individuals = 35, AG = 21, GG = 7; p < 0.01, ANOVA with Mann–Whitney U test) after stimulation with PAM2 (TLR2/TLR6) or PAM3 (TLR2/TLR1), but not LPS (TLR4). Furthermore, IL-6 levels were significantly increased after stimulation with M. tuberculosis whole cell lysate (Fig. 3F). In addition, IL-10 levels were significantly decreased in a recessive pattern for GG individuals after LPS stimulation (p = 0.03; Fig 3G).

There were not any significant changes in IL-10 production after PAM2 or PAM3 stimulation. By way of contrast, minor homozygotes from rs3750920 were not associated with IL-6 levels when compared with heterozygotes or major homozygotes (Fig. 4A–D; CC = 18 individuals, CT = 24, TT = 12). Furthermore, no associations were noted in IL-10 production (Fig 4E–H). These data demonstrate that the minor homozygote of rs5743899 is associated with increased levels of IL-6 after PAM2, PAM3, or M. tuberculosis whole cell lysate stimulation, as well as decreased IL-10 after LPS stimulation.

To assess the role of TOLLIP deficiency in human disease, we performed a candidate gene case-population association study to examine associations between TOLLIP polymorphisms and susceptibility to TB. We examined 671 cases (394 pulmonary TB [PTB] and 277 TBM) and 760 cord blood controls in our cohort. We first analyzed polymorphisms rs3750920 and rs5743899 because of their association with functional phenotypes. TOLLIP polymorphisms had similar linkage disequilibrium patterns in Seattle and Vietnam, as well as the Northern European (CEU) and Han Chinese (CHB) HapMap populations (Fig. 5, Supplemental Figs. 1 and 2). Polymorphisms rs3750920 and rs5743899 were associated with susceptibility to TB using a genotypic model (Table I; rs3750920, p = 7.03 × 10⁻¹⁶; and rs5743899, p = 6.97 × 10⁻⁷). Three other SNPs (rs3793964, rs3829223, and rs3168046) were associated with TB, and these SNPs were in a high degree of linkage disequilibrium with SNP rs3750920 in the Vietnamese population (Supplemental Table I). All these SNPs were associated with TB after a conservative Bonferroni correction for multiple comparisons. One SNP (rs5743942) was associated with TB in an unadjusted analysis, but not after a Bonferroni correction.

For the two significant polymorphisms, we examined these associations further under different genetic models and clinical phenotypes. Testing the association with TB under different genetic models for rs3750920, we found that the strongest association was consistent with a dominant model (Table II; odds ratio [OR], 0.453; p = 6.28 × 10⁻¹²). In contrast, SNP rs5743899 was most strongly statistically associated with all forms of TB in a recessive model (OR, 1.641; p = 0.0004). We also examined the associations with different clinical forms of TB (e.g., PTB) and TBM. Both SNPs were associated with PTB and TBM to a similar degree (Table II). Finally, we evaluated whether these two SNPs were associated with death or disability from TB, as well as with CSF and serum cytokine responses at the time of presentation (for subjects with TBM). We did not find any significant associations with these outcomes (data not shown). These data suggest that two TOLLIP polymorphisms are associated with susceptibility to TB as well as TOLLIP expression levels or regulation of TLR-mediated cytokine secretion.

Our study shows that TOLLIP regulates human TLR signaling pathways in monocytes by suppressing proinflammatory cytokines (IL-6 and TNF) and inducing anti-inflammatory cytokine (IL-10) secretion in peripheral blood monocytes. In addition, we found two common genetic variants of TOLLIP that are associated with TOLLIP mRNA expression or TLR-mediated cytokine release, or both. Finally, these genetic markers of TOLLIP deficiency were associated with an increased risk for developing TB in a case-population study in Vietnam.

Our data from knockdown of TOLLIP in peripheral blood monocytes indicates that TOLLIP is a negative regulator of TNF and IL-6 after PAM2, PAM3, and LPS stimulation. Although these data demonstrate that TOLLIP regulates human TLR2 and TLR4 function, the findings appear to be in contrast to murine models, where Tollip^−/− bone marrow-derived macrophages and dendritic cells had decreased LPS-induced IL-6 and TNF in comparison with wild type cells (46). The reasons for this discrepancy are uncertain. It is possible that the unique gene splicing profile of murine TOLLIP led to the proinflammatory profile in the knockout (59) or that cell type or assay conditions were different between the human and murine experiments. Possibly, another set of regulatory molecules may exist apart from TOLLIP and distinguish TLR2 and TLR4 signaling pathways in mice and humans. These species-specific findings underscore the well-recognized and important concept that murine data cannot always be extrapolated and assumed to be similar to humans.

In contrast to the proinflammatory cytokines, TLR2- and TLR4-mediated IL-10 secretion was decreased after TOLLIP knockdown in peripheral blood monocytes. To our knowledge, TOLLIP is the only signal transduction molecule to downregulate IL-6 and TNF while upregulating IL-10. LPS upregulates IL-10, TNF, and IL-6 in monocytes, presumably by activation of similar transcription factors such as NF-κB (60). However, there are several examples of the IL-10 induction pathway diverting from the IL-6 or TNF induction pathway. For example, IL-10 is induced through the TLR2/TLR4 pathway (61) and modified by CARD9 as well as IFN-β. (62) In addition, Card9^−/− mice are unable to control M. tuberculosis infection and do not produce significant IL-10, despite IL-6 and TNF production. (63) Furthermore, differential regulation of proinflammatory and anti-inflammatory signaling has been observed in HIV-infected macrophages that upregulate IL-10 and downregulate IL-12. (64) Several possible mechanisms could explain how TOLLIP could differentially regulate proinflammatory and anti-inflammatory cytokine production. For example, the C-terminal domain of TOLLIP contains a Tom-1 binding domain and likely plays a role in cellular trafficking and ubiquitination of proteins that could affect TLR signal transduction. In addition, TOLLIP can traffic from the cytosol to the cell nucleus via sumoylation, suggesting a role for TOLLIP regulation of transcription factors (65). Although the mechanism by which TOLLIP differentially regulates cytokines is not clear, this finding may help to elucidate the signaling pathway of IL-10.

To our knowledge, this report is the first description of TOLLIP polymorphisms that are associated with a loss of function as they mimic the phenotype elicited by shRNA knockdown. The mechanism that leads to its loss of function is unclear. We found that minor homozygotes of rs5743899 were associated with a trend of less TOLLIP mRNA in comparison with heterozygotes and major homozygotes, suggesting that differences in TOLLIP expression levels regulate cytokine responses. Our analysis of the case-population data supports this theory, because there is an association between rs5743899 and risk for developing TB in a recessive model. The causative polymorphisms are not known, and this complicates interpretation of the data. In addition to regulation of transcription, the possible mechanisms include linkage disequilibrium of the associated SNPs with a nonsynonymous coding region functional SNP that changes protein localization or conformation and leads to abnormal TOLLIP function. The C2 domain of TOLLIP contains a phosphorylation domain that is hypothesized to be important for IRAK-4 binding (66). This may affect TLR signaling transduction and lead to the cytokine differences noted in this study. A mutation in this domain is a potential location for a loss-of-function mutation, and in vitro cell culture models have shown that abolishing the phosphorylation domain reduces the regulatory capacity of TOLLIP (45). Further study is needed to determine the molecular mechanism of action that leads to this phenotype.

In contrast to rs5743899, polymorphism rs3750920 was associated with protection from TB and increased levels of TOLLIP mRNA. The functional change that leads to the association of rs3750920 with TB is a subject of ongoing study. There are several possible mechanisms that could lead to rs3750920 affecting risk for developing TB, but not IL-6 or TNF cytokine responses in monocytes. First, polymorphisms at rs3750920 might affect cytokines that were not examined during this study, such as IL-1β or IL-18, which could affect inflammasome activity. Second, rs3750920 may affect cytokines that are not expressed in monocytes, such as IL-12p70 or IL-23, but that are expressed in much higher quantities in macrophages (67) and dendritic cells. Third, SNP rs3750920 may regulate protein trafficking and functions such as endocytosis and phagocytosis. TOLLIP is known to bind to ubiquitinated proteins and localize with the protein Tom1 to the early endosome, and it may be important in bacterial entry (68–70). Finally, TOLLIP polymorphisms may regulate effector mechanisms that lead to bacterial killing (such as NO production). Studies are ongoing to examine these possibilities and to determine the causative SNP in linkage disequilibrium with rs3750920. The genetic and mRNA data from these two SNPs likely represent two independent effects with distinct mechanisms.

Interestingly, our genetic data suggests that a hypofunctional TOLLIP genotype (rs5743899 GG) is associated with an increased risk of TB, as well as increased levels of proinflammatory cytokines. There is a long history of observations in the TB field of the potentially deleterious effects of too much inflammation. Steroid treatment of patients with TBM and pericarditis is used to dampen the immunopathologic consequences of inflammation (55, 71). In this regard, TNF has been shown to have protective and deleterious effects in different models of TB. For example, although knockout of TNF in murine models is deleterious with increased M. tuberculosis replication, evidence also suggests that a hyperinflammatory state can lead to increased TB replication and worsened disease (72–74). Furthermore, studies of leukotriene A4 hydrolase, an enzyme that regulates eicosanoid and TNF production in zebra fish and humans, suggest that optimal control of M. tuberculosis requires balanced signaling with deleterious outcomes associated with insufficient or excessive TNF secretion (73). TOLLIP regulation of the anti-inflammatory cytokine IL-10 provides an additional mechanism to inhibit inflammatory pathways (61). Card9-deficient mice do not produce any IL-10 and are unable to control M. tuberculosis infection (63). In addition, IL-10–producing T cells promote anergy to PPD in TB patients and may alter the T cell response and overall control of latent tuberculosis infection (33). These data suggest that inhibition of inflammation is partially beneficial for clinical TB outcomes. However, the nuances of how and when to inhibit inflammation during clinical treatment of TB remain poorly understood.

Our study has several potential limitations. The association findings from the case-control study might not be due to polymorphisms within TOLLIP, but this seems unlikely. TOLLIP is fairly isolated on chromosome 11 in humans and is flanked by the mucin 2 precursor and BRSK2, both ∼100 kbp upstream and downstream, respectively. Neither protein has been associated with immune responses. A second limitation of any genetic study relates to multiple comparisons that require adjustment of significance thresholds for p values depending on the number of analyses performed. Even with the most conservative Bonferroni correction of our genetic association data (multiplying the p value by 6), our findings remain highly significant. Furthermore, our experimental data suggests that two polymorphisms are genetic markers of TOLLIP function. A third limitation of this study and case-population studies is the misclassification of controls, as some of the cord-blood subjects may develop TB in the future. Although the use of cord blood samples can lead to a modest loss of power, the misclassification of controls underestimates the genetic risk of polymorphisms. Finally, candidate gene association studies are subject to confounding because of population substructure. However, our study population was the Vietnamese Kinh, a relatively homogenous population in South East Asia. We previously found no evidence of population stratification in this population using control genomic SNPs (57).

These data suggest that TOLLIP has an important role in the pathogenesis of TB by regulating both proinflammatory and anti-inflammatory pathways. Although further study is required, modulating TOLLIP activity and function may lead to important breakthroughs in TB vaccine design and immune drug development.

We thank the clinical staff from the Hospital of Tropical Diseases and Pham Ngoc Thach Hospital who initially diagnosed and studied the patients with TBM and PTB; Dr. Nguyen Thi Hieu from Hung Vuong Obstetric Hospital Vietnam; Dr. Tran Tinh Hien from the Hospital for Tropical Diseases Vietnam; Dr. Guy Thwaites from Imperial College London; all Vietnamese individuals who participated in this study; Dr. Alan Aderem and Dr. Marta Janer from the Institute for Systems Biology for advice and Sarah Li for technical assistance; and Glenna Peterson, Rick Wells, Dr. William Berrington, Dr. Ann Misch, Dr. Dave Horne, and Dr. Chetan Seshadri from the University of Washington for assistance.

The authors have no financial conflicts of interest.