Genetics of human susceptibility to active and latent tuberculosis: present knowledge and future perspectives

Abstract

Tuberculosis is an ancient human disease, estimated to have originated and evolved over thousands of years alongside modern human populations. Despite considerable advances in disease control, tuberculosis remains one of the world’s deadliest communicable diseases with 10 million incident cases and 1·8 million deaths in 2015 alone based on the annual WHO report, due to inadequate health service resources in less-developed regions of the world, and exacerbated by the HIV/AIDS pandemic and emergence of multidrug-resistant strains of Mycobacterium tuberculosis. Recent findings from studies of tuberculosis infection and of patients with Mendelian predisposition to severe tuberculosis have started to reveal human loci influencing tuberculosis outcomes. In this Review, we assess the current understanding of the contribution of host genetics to disease susceptibility and to drug treatment. Despite remarkable progress in technology, only a few associated genetic variants have so far been identified, strongly indicating the need for larger global studies that investigate both common and under-represented rare variants to develop new approaches to combat the disease. Pharmacogenomic discoveries are also likely to lead to more efficient drug design and development, and ultimately safer and more effective therapies for tuberculosis.

Background

Infectious diseases have ravaged mankind throughout the ages; therefore, they represent a major force in human evolution.^1,2 Response to pathogen exposure is highly variable between people, and a better understanding of the biological basis of interindividual differences in infection outcomes could result in new preventive and therapeutic strategies. Exploration of the genetic underpinning of human susceptibility to infections started decades ago, but only increased in size and scope in the past few years due to newly available, more powerful approaches such as genome-wide association studies based on high-throughput genotyping and sequencing technologies.^1,3,4

The potential of host genetic studies has been repeatedly shown for several infections of major public health importance. A prime example is malaria, which has exerted the strongest known selective pressure on the human genome since the start of agriculture, about 109000 years ago.⁵ Multiple genetic discoveries shed light on interactions between Plasmodium spp parasites and erythrocytes, including the protective effects of otherwise deleterious haemoglobin β gene variants (eg, HbS and HbC alleles) against severe disease⁶ or the complete protection against Plasmodium vivax conferred by a single nucleotide change in the Duffy antigen receptor for chemokines (DARC) gene, leading to no expression of this antigen receptor on red blood cells.⁷ Studies of chronic viral diseases also showed the ability of host genetics to uncover pathogenic mechanisms, by highlighting the role played by immune genes in shaping an individual’s response to infection. Differences were observed in immunogenetic susceptibility to HIV and hepatitis C virus: variation in adaptive immunity genes, particularly in the human leucocyte antigen class I gene family, was consistently shown as the most important determinant of HIV control,⁸ whereas innate immune mechanisms appeared to be more important for hepatitis C virus control, for which the strongest host contributor to spontaneous viral clearance was a variant of the interleukin 28B gene, encoding a type III interferon.⁹

Mycobacterium tuberculosis, the causative agent of tuberculosis, infects about a third of the world’s population. Individuals infected with M tuberculosis have a 10% lifetime risk of developing active tuberculosis. Despite remarkable advances in disease control in the past 15 years, tuberculosis ranks alongside HIV as the leading infectious cause of death worldwide: an estimated 1·8 million people died from tuberculosis in 2015, including 0·4 million resulting from tuberculosis disease in people living with HIV.¹⁰ M tuberculosis uses complex strategies for long-term survival within its host. Infection is initiated in the lungs through inhalation of a few bacteria contained within small aerosol droplets, which are engulfed by alveolar macrophages. Detection of mycobacterial pathogen-associated molecular patterns and damage-associated molecular patterns by the host cytosolic sensors within macrophages results in the induction of type I interferon responses, chemokines, and cytokines. These immune responses elicited by the host contribute substantially to bacterial containment and persistence within organised host structures in the lungs called granulomas.¹¹ The progressive development of the granuloma lesion following infection involves a dynamic chemokine-mediated recruitment of myeloid cells—including dendritic cells, polymorphonuclear leukocytes, and myeloid-derived suppressor cells—and T and B lymphocytes, followed by the evolution of multinucleated giant cells and lipid-filled foamy macrophages encapsulated within fibrotic cuffs. Finetuning of pro-mycobacterial and antimycobacterial effects mediated by the cells within the granuloma and by their chemical mediators, while limiting bacterial growth, also suppress immune responses and provide a survival niche from which the bacteria can ultimately disseminate in immuno compromised individuals. Components of the type I interferon and inflammatory cytokine pathways are thus known to mediate host resistance and susceptibility both in animal models and human beings.

Host genetic studies of tuberculosis have been challenging for various reasons, including difficulties in phenotype definition and probable genetic heterogeneity.¹² Nevertheless, recent findings from genome-wide scans and analyses of patients with Mendelian susceptibility to mycobacterial disease have started to reveal human vulnerabilities against tuberculosis. In this Review, we present a perspective on tuberculosis host genomics, based in part on the workshop “Host response to TB-HIV infection: a genomic perspective” organised by the National Institute of Health (Rockville, MD, USA, Jan 13–14, 2015), describing the latest progress and discussing the opportunities and challenges of human genetic approaches in tuberculosis research.

Early observations on the genetic basis for tuberculosis susceptibility

In Lübeck, Germany, from Dec 10, 1929, to April 30, 1930, 251 neonates were accidentally given preparations of the BCG vaccine that had been contaminated with virulent M tuberculosis.¹³ Of the vaccinated children, 72 died of tuberculosis and the surviving 174 children (five children died of non-tuberculosis causes) presented with a wide range of clinical presentations. A careful follow-up investigation of the accident identified variable concentrations of contamination (ie, the infectious dose) as the main source of variable clinical outcomes.¹³ However, even when estimated dose was considered, substantial clinical variability remained, ranging from death to very mild disease. This pronounced variability of outcome in the absence of other known environmental or societal factors might represent differences in the innate, genetically controlled ability of neonates to fight off M tuberculosis. Outbreaks of tuberculosis have been documented in members of the Navy and Marine Corps service, who are at increased risk of exposure because of close living quarters, ventilation systems aboard ships, and their travel to countries where tuberculosis is endemic. In reported instances of outbreak, although about 20% of previously unaffected sailors were found to be newly positive by tuberculin skin test (TST), results from screening of all sailors showed little transmission of M tuberculosis, despite months of potential exposure in a high-risk setting,^14–16 as determined by close sleeping quarters in bunks arranged in stacks with air intake that exhausted directly overboard, and sailors in adjacent compartments connected with the patient’s. Deployments of ships spread over several weeks to months. Maximum theoretical transmission in the absence of genetic-based resistance is difficult to determine, but the closest would be intra-household transmission in susceptible individuals. For example, a simulation study indicates that the prevalence of M tuberculosis infection in household contacts in Pakistan is about 49.4%.¹⁷

Genetic basis of resistance is also substantiated by findings that previously underexposed populations are more susceptible to the disease than overexposed populations, as evidenced by the effects of European colonialism. Tuberculosis brought in by colonising Europeans has played a large part in deaths among Qu’Appelle Indians in the Saskatchewan province, Canada, in 1890, with more than half the families lost in the first three generations of the epidemic before considerable fall in death rates that could be attributed to strong selection against susceptibility genes.¹⁸ Through contact with European explorers, whalers, and missionaries, Inuit populations in Nunavut, Canada, had increasingly succumbed to many new diseases including tuberculosis.¹⁹

Tuberculosis was first introduced to Yanomami Indians in the Amazon rain forests through contact with gold miners of European descent, and since 1980 has become an epidemic even in BCG-vaccinated individuals. A study of this population showed reduced cell-mediated immunity and an increased antibody response relative to populations with extensive previous contact with M tuberculosis, again suggesting that tuberculosis exerts a selective pressure on immune responses and on human evolution.²⁰ The protective efficacy of BCG against pulmonary tuberculosis has varied widely in different parts of the world. BCG vaccinations in Yanomami Indians did not enhance the proportion of low-grade purified protein derivative reactions, commonly observed after BCG immunisation or exposure to environmental mycobacteria in other populations, strongly suggesting an intrinsic un responsiveness of this population on first exposure to mycobacterial antigens.

Earlier studies of twins have shown that the concordance of tuberculosis was higher in monozygotic twins than in dizygotic twins under comparable environmental and social conditions,²¹ and these findings were further confirmed by a reanalysis in the Prophit survey.²² Studies of racially integrated nursing homes in Arkansas, USA, have shown that African Americans are twice as susceptible as people of white descent to infection with M tuberculosis, a finding that could not again be explained by environmental or social factors.²³ These studies have helped to reinforce the concept of genetic susceptibility to tuberculosis.

Insight from inbred animal models of tuberculosis susceptibility

Mice and rabbits have often been used in studies to investigate genetic susceptibility to tuberculosis. Crossbreeding between resistant and susceptible inbred mice strains has shown that resistance to tuberculosis follows a complex, non-Mendelian inheritance.²⁴ In 1981, Gros and colleagues²⁵ found that susceptibility to infection with avirulent BCG derived from Mycobacterium bovis was determined by a host genetic locus Bcg on mouse chromosome 1. Subsequently, the gene encoding the natural resistance-associated macrophage protein 1 (Nramp1) was identified within this locus.²⁶ Targeted disruption of Nramp1 also eliminated resistance to other infectious agents such as Leishmania donovani and Salmonella typhimurium.²⁷ The human orthologue of mouse Nramp1 is designated NRAMP1 or SLC11A1, and is involved in susceptibility to tuberculosis. In the mouse however, Nramp1 does not affect susceptibility to virulent M tuberculosis.^28,29

In a study investigating susceptibility to M tuberculosis in mice, Kramnik and colleagues²⁹ mapped a new locus designated sst1 (susceptibility to tuberculosis 1) on mouse chromosome 1, 10–19 centimorgans distal to Nramp1. The phenotypic expression of sst1 differs from Nramp1 in that sst1 exerted a lung-specific effect. sst1-susceptible C3HeB/FeJ mice developed lung lesions characterised by extensive necrosis and uncontrolled multiplication of M tuberculosis, whereas sst1-resistant mice developed interstitial granulomas with controlled multiplication of the bacilli. Positional cloning identified the Ipr1 (intracellular pathogen resistance 1) isoform of the Ifi75 (interferon-inducible-75) gene within the sst1 locus as imparting increased susceptibility of the C3HeB/FeJ mice to M tuberculosis.³⁰ These genetically defined mouse models provide important tools to study the pathogenesis of tuberculosis.

In the 1940s, Max Lurie^31,32 did detailed studies over a period of 7 years on inbred families of rabbits to investigate natural resistance to tuberculosis. Lurie found that inbred strains exhibited either resistance or susceptibility to infection with virulent M bovis, and that although resistant families developed cavitary tuberculosis, susceptible families developed disseminated disease. Rabbits that were resistant lived about twice as long as susceptible ones. The inheritance pattern suggested that resistance was predominantly genetically determined. Subsequently, recognition of differential susceptibility of inbred laboratory animal strains to tuberculosis infections has enabled linkage studies and targeted gene disruptions to identify host determinants of disease susceptibility. The influence of host genetics on disease response to tuberculosis in rabbit and mouse models is indicated in figure 1.

Figure 1: Influence of host genetics on disease response to tuberculosis in rabbit and mouse models.

(A) Image captured at 40x magnification. A granuloma from a rabbit, showing central necrosis and tissue destruction. (B) Image captured at 100x magnification. A granuloma from a BALB/c mouse. (C) Image captured at 40x magnification. A necrotising granuloma from a C3HeB/FeJ mouse. Differences in pathology and host response in the C3HeB/FeJ mouse were mapped to an allele in the sst locus. (D) Pathological response to Mycobacterium tuberculosis infection in disease models compared with the pattern observed in human disease.³³ (E) Pattern of bacterial burden after infection in the resistant and susceptible inbred rabbits of Max Lurie.³³ (F) Pattern of bacterial burden after infection in BALB/c mice carrying a BCG-susceptible allele mapped to the Bcg locus.³⁴ (G) Pattern of bacterial burden after infection in B6 mice compared with congenic B6 mice engineered with the sst-susceptible locus characterised in C3HeB/FeJ mice.³⁵ ···=human disease hallmark frequently observed in model. ··=human disease hallmark sometimes observed in model. ·=human disease hallmark rarely observed in model. 0=human disease pattern not observed in model.

Infection with M bovis, a member of the M tuberculosis complex, causes tuberculosis in cattle. M bovis has a wide host range, including a number of wild ruminant animals, is infectious in human beings, and causes substantial economic hardship for farmers worldwide. Zoonotic transmission of M bovis occurs primarily through ingestion of unpasteurised milk or close contact with infected cattle. Although the global prevalence of human tuberculosis due to M bovis is low (<2%), increased prevalence has been reported in African countries, Mexico, Turkey, and India.³⁶ In efforts to identify genotypes resistant to M bovis for the generation of resistant animal populations, Bhaladhare and colleagues³⁷ found that single nucleotide polymorphisms (SNPs) in TLR2, a host gene important in immune sensing of pathogen-associated molecular patterns, was associated with susceptibility to M bovis in cattle. Association between TLR mutations and SNPs in TLR2 and TLR4 and increased susceptibility to Mycobacterium avium paratuberculosis has also been previously reported in cattle.³⁸ These studies prompt additional validation in larger groups before screening and breeding of resistant animal populations can be envisaged.

Human genetics of tuberculosis

Genetic control of tuberculosis infection

There is no direct test for infection with M tuberculosis, and the latent tuberculosis infection phenotype is inferred indirectly from quantitative measurements of antimycobacterial immunity.^11,39 TST is the most widely used method to test for latent infection.⁴⁰ Additionally, two in-vitro blood assays measuring either the secretion of interferon γ by lymphocytes or the frequency of interferon-γ-producing blood cells in response to M tuberculosis antigens (interferon-γ release assays [IGRAs]) have been developed.⁴¹ TST and IGRAs assess different aspects of antimycobacterial immunity and are not fully concordant in predicting infection with M tuberculosis.^42,43 Little or no reactivity in these tests in individuals heavily exposed to M tuberculosis is indicative of innate resistance to M tuberculosis infection. In household studies, 30–50% of contacts with heavy short-term exposure do not become infected,⁴¹ revealing substantial heterogeneity in susceptibility to infection.

Several studies have considered TST and IGRA results as quantitative traits and have shown substantial heritability for both tests following exposure to M tuberculosis. In The Gambia, a twin study estimated the heritability of TST response at 71% and IGRA response at 39%.⁴⁴ The heritability of quantitative TST reactivity in children exposed to an active tuberculosis case has been estimated at 92% in Chile.⁴⁵ A complex segregation analysis of TST reactivity in household contacts of tuberculosis cases in a Colombian population provided evidence for a major co-dominant gene accounting for about 65% of TST variability.⁴⁶ The heritability of interferon-γ secretion has been estimated at about 43% following BCG stimulation and 58% following ESAT-6 stimulation in South Africa,⁴⁷ and at between 17% and 48% following stimulation with M tuberculosis antigens, including ESAT-6, in Uganda.^48,49

Few studies have aimed to identify the genetic variants underlying the detected heritability. Candidate gene association studies have focused on TST response as a binary trait, with the most interesting association found with an IL10 promoter haplotype in Ghana⁵⁰ and Brazil.⁵¹ A recent genome-wide association study in a large Icelandic population identified some HLA class II variants as associated with TST positivity with weak effect (odds ratio [OR] 1·14).⁵² In Uganda, a genome-wide linkage analysis suggested that persistent TST negativity was linked to chromosomal regions (2)(q21;q24) and (5)(p13;q22).⁵³ A study of TST reactivity in a sample of multiplex families from South Africa identified two major loci affecting TST responses more than 0 mm (TST1 on chromosome 11p14), and the intensity of TST-reactivity (TST2 on chromosome 5p15).⁵⁴ The TST1 locus was replicated in a household contact study of French families of various ethnic origins.⁵⁵ The TST1 locus might reflect innate resistance to infection with M tuberculosis whereas TST2 might account for T-cell-mediated antimycobacterial immune responses. It was subsequently discovered that a locus affecting the production of tumour necrosis factor (TNF) by blood cells in response to BCG and BCG plus interferon γ, TNF1, was genetically indistinguishable from TST1.⁵⁶ This finding raises the exciting possibility that innate resistance to M tuberculosis infection might involve a TNF-mediated effector mechanism.

Finally, a linkage study identified two loci on chromosomes 8q and 3q that are controlling BCG-triggered and ESAT-6 triggered interferon-γ production, respectively, in populations of various ethnic origins living in different M tuberculosis exposure settings.⁵⁷

Genetic control of severe primary tuberculosis

Many factors are known to influence the risk of progression from latent infection to clinical disease. However, as a general rule of thumb, the lifetime risk of tuberculosis in a child infected with M tuberculosis is approximately 10%.⁵⁸ In a large population survey in the Netherlands, approximately two-thirds of patients who had tuberculosis developed their clinical disease within 2 years after infection,⁵⁹ and it is possible that the majority of tuberculosis cases occur in this period. In highly endemic countries, this rapid onset of primary tuberculosis is particularly common in children, some of whom developing a haematogenous disseminated form (referred to in this Review as severe primary tuberculosis).¹³

The first molecular evidence that childhood tuberculosis might reflect a Mendelian predisposition came from the observation of severe tuberculosis in children with classic primary immunodeficiencies,⁶⁰ such as chronic granulomatous disease⁶¹ or more recently complete TYK2 deficiency.⁶² Further progress came from the study of the syndrome of Mendelian susceptibility to mycobacterial diseases (MSMDs), which is defined by a selective vulnerability to weakly virulent non-tuberculous Mycobacterium species, such as BCG and environmental mycobacteria.^63,64 Since 1996, germline mutations in seven autosomal (IFNGR1, IFNGR2, IL12B, IL12RB1, STAT1, IRF8, and ISG15) and two X-linked (NEMO and CYBB) genes have been discovered in patients with MSMDs.^63,64 These defects are physiologically related, because they all result in an impairment of interferon-γ immunity. Several patients with MSMDs, particularly those with interferon-γR1^65,66 and interleukin-12p40⁶⁷ deficiencies, have been shown to have infections due to both weakly virulent mycobacteria and M tuberculosis, raising the hypothesis that tuberculosis observed in these patients could also be attributed to a monogenic predisposition.⁶⁰

Additional evidence came when several siblings of patients with MSMDs carrying the same genetic defect as the index case were found to display severe tuberculosis as their sole infectious phenotype. This finding was the case in a child with partial interferon-γR1 deficiency,⁶⁸ and in a male participant from a large multiplex X-linked kindred carrying a specific mutation of CYBB impairing the interferon-?-dependent respiratory burst in macrophages.⁶⁹ Several patients with tuberculosis and mutations in MSMD genes were also identified without any familial history of MSMDs.⁶⁰ The most common genetic defect identified in patients with severe tuberculosis to date is complete interleukin-12Rβ1 deficiency,⁷⁰ which was observed in several families.⁶¹ In a more systematic search for IL12RB1 mutations in 50 children with severe tuberculosis, two patients with complete interleukin-12Rβ1 deficiency were identified.⁷¹ Overall, these results provided proof of concept for monogenic predisposition to severe tuberculosis, and raised the possibility that severe tuberculosis could be due in a substantial proportion of children to single-gene inborn errors of immunity. This proportion has been estimated at up to 45% by theoretical calculations,⁷² and can be determined experimentally by next-generation sequencing approaches. These findings have paved the way for new treatments based on physiopathology. The best example is provided by patients with interleukin-12Rβ1 deficiencies, who can be treated by recombinant human interferon-? in addition to antimycobacterial drugs.⁷³

Genetic control of pulmonary tuberculosis

Most patients with latent tuberculosis infections (90–95%) never develop clinical disease. The remaining 5–10% develop clinical tuberculosis later in life, typically pulmonary tuberculosis, due to reactivation of the original infection or, in other cases, to secondary infection.^11,74 The development of pulmonary tuberculosis reflects an impairment of host resistance to M tuberculosis that might be favoured in a minority of patients by acquired immunodeficiency, such as HIV infection or anti-TNF treatment. However, in most people who do not have any overt immunodeficiency, the pathogenesis of reactivation remains elusive. There is strong evidence that the development of pulmonary tuberculosis is influenced by host genetic factors,^12,75 in particular based on familial aggregation studies⁷⁶ such as twin studies.²¹ Most classic genetic association studies investigating pulmonary tuberculosis have focused on candidate genes, and several common risk variants have been reported in immunity-related genes such as those encoding DC-SIGN, TLR1 and TLR2, vitamin D receptor, TNF, interleukin 1β, interferon γ, or some HLA class II molecules.⁷⁷ However, there has been an absence of consistency and replication between most of the reported results of independent studies, partly due to underpowered studies, and strong heterogeneity across phenotype definition and epidemiological settings.^77,78 One of the most convincing findings has been the identification of associated polymorphisms of the NRAMP1 gene, with an effect that is heterogeneous across populations, epidemiological settings, age, and clinical phenotypes.^79–81 The importance of considering age at tuberculosis onset in host genetic analyses was further supported by the positional cloning of a major locus on chromosome 8q conferring predisposition to pulmonary tuberculosis.⁸² The association mapping of the linked region identified variants of the TOX gene as strongly associated with the development of early-onset pulmonary tuberculosis (before 25 years of age) in populations from Morocco and Madagascar.⁸³ TOX encodes a nuclear factor involved in the development of T cells,⁸⁴ particularly the CD4-positive T cells essential for immunity to mycobacteria.

Conversely, genome-wide association studies of pulmonary tuberculosis, considered as a single phenotype, have met with little success to date. Studies from Ghana and The Gambia,^85,86 and another from Russia⁸⁷ have led to the identification of only three signals that reached the genome-wide threshold for statistical significance (59×910^™8). One of the two variants identified in the African genome-wide association study, rs4331426, is located in a gene desert on chromosome 18q11.2⁸⁶ and the other, rs2057178, is near WT1 on chromosome 11p13,⁸⁵ and both had modest effect sizes (OR 1·19 and 0·77, respectively). The chromosome 11 signal was replicated in populations from Indonesia and Russia in the original study,⁸⁵ and in independent investigations in South African⁸⁸ and Moroccan⁸⁹ populations. Attempts to replicate the chromosome 18q11.2 signal provided more conflicting results in Chinese^90,91 and Moroccan⁸⁹ populations. The Russian study identified a cluster of intronic ASAP1 variants in a study population of more than 159000 participants, also with a weak effect size (OR 0·84 for SNP rs4733781), with possible functional involvement in dendritic cell mobility.⁸⁷ The Icelandic genome-wide association study⁵² also reported the association of HLA class II variants with pulmonary tuberculosis, although the precise role of these variants in tuberculosis infection or clinical tuberculosis, or both, remains to be clarified. Finally, a genome-wide association study of 581 patients with HIV (267 with active tuberculosis) from Uganda and Tanzania identified rs4921437 at chromosome 5q33, close to the IL12B gene, with a stronger protective effect size (OR 0·37) than those previously reported in HIV-negative populations.⁹² A striking feature of these studies is the inadequacy of replication of the pulmonary tuberculosis susceptibility factors previously detected in candidate gene analyses.⁴ Overall, the results from the genome-wide association studies suggest that common variants might have a little effect on individual predisposition to adult pulmonary tuberculosis, at least when considered as a single homogeneous phenotype.

Pharmacogenomics

Isoniazid

Human genetic polymorphisms affect antimicrobial pharmacokinetics and pharmacodynamics. Isoniazid is the most extensively studied anti-tuberculosis drug in this regard. It undergoes acetylation by hepatic arylamine N-acetyltransferase 2 (encoded by NAT2),⁹³ and NAT2 loss-of-function alleles are common,^93,94 with slow acetylator allele frequencies of approximately 45% with Asian, 55% with African, and 75% with European ancestry. One copy of such alleles confers an intermediate acetylator phenotype and two copies confer a slow acetylator phenotype, and increased plasma isoniazid exposure.^93,94 These polymorphisms have also been associated with risk for isoniazid hepatotoxicity. In individuals prescribed isoniazid for latent tuberculosis in the USA, raised hepatic trans aminase concentrations to greater than five times the upper limit of normal were reported in approximately one case per 1000 people completing therapy,⁹⁵ although risk increases with concomitant rifampicin and other factors.⁹⁶ Associations between NAT2 polymorphisms and isoniazid hepatotoxicity were shown in meta-analyses involving multiple cohorts with tuberculosis, with ORs ranging from 1·93 to 4·69.^97–99

Evidence for genetic associations with isoniazid hepatotoxicity beyond NAT2 is scarce. Studies have suggested associations with CYP2E1 polymorphisms,^99–102 alone or with NAT2 alleles,^94,97,101 but either associations have not been replicated^103–105 or were seen only in east Asians.⁹⁹ A meta-analysis suggested an association with GSTM1 polymorphisms,⁹⁹ but not in all ancestries.^99,100,106 Associations in other genes have not been replicated.¹⁰⁶ A small study in Japan randomised patients with tuberculosis to standard isoniazid doses (5 mg/kg) versus individualised doses based on NAT2 genotype.¹⁰⁷ Benefit was suggested, with fewer treatment failures in rapid acetylators who received higher doses than standard doses (15% vs 38%) and fewer hepatotoxicity cases in slow acetylators who received lower doses than standard doses (0% vs 78%), although failure and hepatotoxicity percentages seem excessive.

A complex drug–drug interaction involves isoniazid, the antiretroviral drug efavirenz, and several genes. Rifampicin induces hepatic cytochrome P450 (CYP) 2B6, so should decrease plasma efavirenz exposure.¹⁰⁸ However, in some patients who are prescribed rifampicin with isoniazid, plasma efavirenz exposure paradoxically increases, particularly with CYP2B6 and NAT2 loss-of-function genotypes (figure 2).^109–111 High isoniazid concentrations in NAT2 slow acetylators apparently inhibit CYP2A6, a necessary pathway for efavirenz clearance in CYP2B6 slow metabolisers.^110–112 Isoniazid induced peripheral neuropathy might increase with NAT2 slow acetylator alleles.¹¹³ In f^ive Japanese patients who had nerve biopsies for isoniazid-induced neuropathy, all were slow acetylators.¹¹⁴

Figure 2: Complex drug–drug interaction between isoniazid and efavirenz, and involving several pharmacogenes.

(A) Efavirenz is converted to inactive metabolites by cytochrome P450 (CYP) 2B6, with minor metabolism by CYP2A6. Individuals carrying two CYP2B6 loss-of-function alleles (rs3745274 or rs28399499, or both; circle 1) have increased plasma efavirenz exposure, which is further increased with concomitant CYP2A6 loss-of-function polymorphisms (eg, rs28399433; circle 2). When individuals with two CYP2B6 loss-of-function alleles are prescribed isoniazid, inhibition of CYP2A6 by isoniazid can also increase efavirenz exposure (circle 3), particularly with concomitant NAT2 genotypes that increase plasma isoniazid exposure (eg, rs1801280, rs1799930, and rs1799931; circle 4). (B) Depicts approximate plasma efavirenz trough concentrations by CYP2B6, CYP2A, and NAT2 genotype and isoniazid exposure. *Data are absent for CYP2A6 in isoniazid recipients with concomitant CYP2B6 and NAT2 loss-of-function genotypes, but this combination of CYP2B6, NAT2, and CYP2A6 genotypes is expected to confer even higher efavirenz concentrations.

Other anti-tuberculosis drugs

Rifampicin is a substrate for organic anion-transporting polypeptide 1B1 (coded by SLCO1B1),¹¹⁵ and one study associated an SLCO1B1 polymorphism (rs4149032) with rifampicin bioavailability in South Africans.¹¹⁶ It is not known whether immune response genes affect susceptibility to rifampicin hypersensitivity, as with other antimicrobials.¹¹⁷ Parenteral aminoglycosides can cause sensorineural hearing loss,^118–122 and mitochondrial DNA mutations in the gene that encodes 12S ribosomal RNA confer increased risk.^{118,121,123,124} The mitochondrial 1555A→G risk mutation^121,122,124 is present in as many as 5% of southeast Asians and approximately 0·2% of Europeans.

Linezolid is being prescribed for multidrug-resistant tuberculosis. Evidence suggests that neuropathy, myelosuppression, and hyperlactataemia with linezolid might reflect inhibition of mitochondrial protein synthesis. It is conceivable that mitochondrial mutations will affect susceptibility to these adverse events.^125–130

Given the immense global burden of tuberculosis, there is opportunity for pharmacogenomic discoveries to have profound impact. Progress in this area will be facilitated by access to well phenotyped clinical datasets, including datasets that represent newer agents, with linkage to well consented DNA banks. This effect, together with advanced technologies such as high-throughput next generation sequencing, should lead to more efficient drug discovery, design, and development, and ultimately safer and more effective therapies for tuberculosis.

Future direction and perspectives

The identification of the genetic factors involved in tuberculosis susceptibility has proven difficult. As in several other human diseases, the role of common variants in pulmonary tuberculosis seems to be scarce, leading to the concept of missing heritability.¹³¹ Among the several non-mutually exclusive hypotheses that might account for this concept, an important one in tuberculosis is the oversimplification of the studied phenotypes.¹² For example, tests used to assess tuberculosis infection display little concordance, and each assay captures a different aspect of the antimycobacterial response.⁴² In pulmonary tuberculosis, the most commonly used phenotype-defining characteristic is the presence of M tuberculosis in the sputum of patients, regardless of the other clinical, microbiological, and demographical covariables. This approach ignores the dynamic nature of pulmonary tuberculosis, and the likelihood of different stages of this process being under different genetic controls, as shown in mouse models of the BCG infection.¹³² The need for more homogeneous and refined tuberculosis phenotypes for genetic studies is also shown by the strong impact of age at onset of disease on the ability to detect genetic effects in clinical tuberculosis.^81,83,133 Moreover, it is possible that the infectious dose has a pivotal role in tuberculosis susceptibility and that host genetic studies would be more informative if done in sporadic cases that are more likely to have received a low infectious inoculum.¹³

Another major aspect that has been mostly ignored so far is the possible effect of the M tuberculosis strain on latent tuberculosis infection and clinical outcomes.^134,135 Evidence is accumulating that human-adapted M tuberculosis has been coevolving with modern human beings over millennia,^136,137 strongly suggesting the importance of host–pathogen interaction in the different tuberculosis infection and clinical outcomes.^134,138 Although some associations between human genetic variants and specific M tuberculosis lineages have been reported,¹³⁹ large scale studies of interaction between human polymorphisms and pathogen variants that have been done in HIV¹⁴⁰ are missing in tuberculosis. HIV infection has a major influence in tuberculosis infection and clinical outcomes. However, little progress has been made in the genetics of tuberculosis–HIV co-infection, with only few inconclusive association studies.¹³⁹ More comprehensive studies are needed in the context of co-infection that might take advantage of extreme phenotypes such as patients with HIV infection and low CD4-positive T-cell counts who do not develop tuberculosis infection or disease despite sustained exposure. An interesting example was provided by the recent genome-wide association study⁹² done in participants with HIV infection that identified a locus with a strong protective effect that needs to be replicated in other samples. Likewise, the study of the genetic basis of the immune reconstitution inflammatory syndrome occurring in patients with tuberculosis–HIV co-infection after initiation of an antiretroviral treatment,¹⁴¹ which have similarities with reversal reactions observed in leprosy,¹⁴² will be of major interest.

Several genetic explanations might also account for the missing heritability observed in tuberculosis including genetic and allelic heterogeneity, and epistatic (gene × gene interaction) effects. Among these explanations, the contribution of rare variants to tuberculosis pathogenesis deserves attention for two main reasons: conceptually, rare variants bridge the gap between Mendelian and complex inheritance, and could account for the major loci identified through linkage studies;^1,80,143 and experimentally, these variants abound in the genome can now be identified by whole-exome and whole-genome sequencing,^144–146 and analysed by appropriate methods.^147,148 Working on improved specified phenotypes will be pivotal for studies based on next-generation sequencing methods that typically need to be done on extreme phenotypes (eg, existence of familial cases, early age at onset, severity of the disease, or absence of comorbidities).

Transcriptomic studies, including the search for loci involved in the expression of genes (expression quantitative trait loci) in cell-specific or tissue-specific studies,¹⁴⁹ based on RNA-sequencing, in particular,¹⁵⁰ are also of major interest for genetic dissection of tuberculosis. A transcriptomic analysis of peripheral blood cells identified a neutrophil-driven interferon-a-inducible and interferon-β-inducible transcript signature in individuals with active pulmonary tuberculosis.¹⁵¹ This blood signature was validated in independent studies,^152–154 and was shown to be different from that of several other infectious and pulmonary diseases,^151,155 providing new insight into the most relevant pathways and candidate biomarkers for investigation in tuberculosis.^156–158 A recent transcriptomic study provided evidence for a TLR and inflammasome signature in immune reconstitution inflammatory syndrome occurring in patients with tuberculosis–HIV co-infection.¹⁵⁹ Studies done in monocyte-derived dendritic cells have also provided new insights in the identification of expression quantitative trait loci associated with M tuberculosis infection.¹⁶⁰ Infection of dendritic cells with M tuberculosis also resulted in the up-regulation or downregulation of 155 microRNAs (miRNAs), and these changes in miRNA expression were partly under genetic control.¹⁶¹

The transcriptional response is strongly affected by epigenetic modifications of DNA-encoded regulatory elements in the 5'-upstream region of genes and more distal enhancers. Key molecular events are methylation and acetylation of histone proteins that result in opening of chromatin and the binding of transcriptions factors (figure 3). The epigenetic landscape of many cell types has been profiled and is available in public databases such as ENCODE. However, only one study investigated the changes in DNA methylation of human dendritic cells after infection with M tuberculosis.¹⁶²M tuberculosis infection triggered the rapid demethylation of thousands of regulatory sites preferentially corresponding to distal enhancer regions (figure 3). Demethylation affected expression of adjacent genes and was associated with extensive remodelling of the epigenetic landscape of infected cells.¹⁶² A second independent line of investigation has shown a striking association between epigenetic programming of monocyte to macrophage and trained innate immunity.^163–165 Collectively, these studies identify epigenetic changes as a key factor in the understanding of cellular responses to infections.

Figure 3: Interplay between epigenetic remodelling of the host cell by Mycobacterium tuberculosis and the role of germline variants in controlling host gene expression.

Opening of chromatin, a prerequisite for binding of transcription factors, is induced by histone modifications such as monomethylation at lysine 4 in histone H3 (H3K4me1). In the example shown, M tuberculosis triggered H3K4me1 histone modification that mediates chromatin accessibility in a regulatory region that carries a transcription factor binding site (GGGC or GGGA). Only the GGGA allele allows efficient binding of the transcription factor and induction of gene expression, whereas the GGGC allele does not.

Conclusion

Historical observations, heritability estimates, linkage analyses, and genome-wide scans indicate that human resistance to tuberculosis infection has a strong genetic basis, which is likely to be a consequence of evolutionary counter-response to microbial virulence. Mendelian studies have also provided the first evidence that severe tuberculosis in children could be due to single gene inborn errors of immunity. However, much is yet to be learned regarding susceptibility to disease, in particular pulmonary tuberculosis, from studies of human genetic diversity and of interactions between host and pathogen genomes. Despite remarkable progress in genotyping and sequencing technology and bioinformatic analysis, only a few associated genetic variants have been identified, strongly indicating the need for an improved definition of the phenotype, and larger studies investigating not only common variants, but also rare single nucleotide and structural variants (copy number variation, insertions, and deletions) that are under-represented on available genotyping platforms.

Although the Hapmap Project and the 1000 Genomes Project have greatly enhanced our understanding of human genetic variation globally, efforts to study African genetic diversity have been rather scarce. Therefore, synergies with the African Genome Variation Project,¹⁶⁶ an international collaboration examining genetic diversity among 18 ethnolinguistic groups of sub-Saharan Africa where tuberculosis is endemic, become highly imperative. In this regard, a recent transcriptomic analysis of whole blood samples from independent South African and Gambian cohorts prospectively identified people at risk of developing active tuberculosis.¹⁶⁷ In this predictive signature of 16 genes, the gene module that was predominantly represented was interferon responses, suggesting that chronic peripheral immune activation precedes onset of active disease. The tuberculosis risk signature predicted disease progression despite substantial diversity between the study cohorts, raising the possibility for targeted intervention to prevent the disease.

Recent observations in a model population of genetically diverse mice suggest that a single vaccine might not be able to protect all individuals, further emphasising the need to understand immunological variations that determine tuberculosis susceptibility and vaccine efficacy.¹⁶⁸ Finally, whether variants associated with disease are truly causative and of any clinical relevance needs to be determined, emphasising the need to study better specified clinical and tuberculosis-relevant cellular phenotypes.

Search strategy and selection criteria.

We identified data for this Review from MEDLINE and PubMed with the search terms “host genomics AND tuberculosis”, “susceptibility to tuberculosis”, and “innate resistance to tuberculosis”. Because the Review traces the history of tuberculosis through the 20th century including the colonial period, we have included articles published in English from the database search between Jan 1, 1940, and January 31, 2017. We also identified data from our earlier work and from the references cited in these publications.