Tuberculosis (TB) is the most common opportunistic infection associated with human immunodeficiency virus (HIV) infection, and co-treatment of the two diseases is recommended (Harries et al., 2009). However, simultaneous treatment of TB and HIV infections is challenging due to drug interactions and overlapping toxicities (Cohen and Meintjes, 2010). Antiretroviral (ARV) and anti-tuberculosis (anti-TB) Drugs-induced liver injury (DILI) is a common adverse event, which can be fatal if therapy is not interrupted or changed on time (Devarbhavi et al., 2013; Naidoo et al., 2015; Shamanna et al., 2016). TB-HIV co-infected patients on anti-TB and ARV co-therapy are at a higher risk of developing DILI than TB or HIV only infected patients receiving monotherapy (Yimer et al., 2011, 2014; Mugusi et al., 2012). A recent study in TB/HIV patients on anti-TB and antiretroviral therapy (ART) with high levels of immune activation demonstrated impaired isoniazid clearance, implicating the need for exploring immune response and the risk of DILI (Vinnard et al., 2016). Up to 32% of HIV patients on ART discontinue their treatment or switch therapy mainly due to DILI (Bica et al., 2001), and genetic predisposition contributes partly (Lubomirov et al., 2011). Treatment interruption may increase the risk for developing of multidrug-resistant TB (MDR-TB) and HIV/AIDS (Hirpa et al., 2013). Thus, identifying genetic markers for drug-induced liver toxicity is valuable to identify high-risk patients and to introduce appropriate measures.

Both HIV and TB remain a major problem and co-infection is common in most Sub-Saharan African (SSA) countries including Ethiopia, the second most densely populated country in Africa with an estimated population size of 100 million. Ethiopia is listed among the top 20 high-TB burden countries globally, and one of the high MDR-TB burden countries (Biadglegne et al., 2014; World-Health-Organization, 2016). The rate of new HIV infection in Ethiopia is declining with the estimated number of people living with HIV being 769, 600 in 2014 (World-Health-Organization¹). The scale of ART is increasing in the country as part achieving the UNAID/WHO “90-90-90” target: to diagnose 90% of all HIV positive people, provide ART for 90% of those diagnosed and achieve viral suppression for 90% of those treated, by 2020 (UNAIDS, 2014). ART and anti-TB drug-induced liver toxicity is a common problem in Ethiopia causing treatment discontinuation and hence MDR-TB (Hirpa et al., 2013; Yimer et al., 2014).

Genetic variations in HLA gene is implicated with susceptibility to T-cell mediated adverse events to a wide range of pharmaceuticals making it a candidate gene relevant to pharmacogenetic studies (Barbarino et al., 2015). HLA alleles that are reported to be association with increased risk of idiosyncratic DILI include: HLA-DQB1^*02:01 and DQB1^*05 to anti-TB drugs (Sharma et al., 2002; Chen et al., 2015), and HLA-B^*58:01 and DRB1^*01:02 to nevirapine-containing ARV regimens (Phillips et al., 2013). HLA-B^*57:01 and A^*33:03 variant alleles were also reported as genetic markers for idiosyncratic liver injury induced by flucloxacillin (Daly et al., 2009) and ticlopidine (Hirata et al., 2008) respectively. Susceptibility to amoxicillin-clavulanate-induced liver injury is influenced by multiple HLA class I and II alleles (Lucena et al., 2011). A large genome-wide association study found a strong association of amoxicillin-clavulanate induced liver injury with HLA-A^*02:01, HLA-DQB1^*06:02, and DRB1^*15:01 variant alleles (Lucena et al., 2011). Genetic screening for HLA-B^*57:01 and subsequent treatment modifications have been shown to reduce incidence of life-threatening hypersensitivity to abacavir in HIV/AIDS patients carrying the allele (Hughes et al., 2008; Mallal et al., 2008) and HLA-B^*15:02 to carbamazepine in Southeast Asian carriers (Amstutz et al., 2014).

Most previous reports investigating genetic risk factors for anti-TB and ARV drugs-induced liver toxicity focused on drug metabolizing enzymes and transporter proteins (Lee et al., 2010; Yimer et al., 2011, 2012). Previously we reported the association of high efavirenz plasma concentration and CYP2B6^*6 genotype with DILI in TB-HIV patients (Yimer et al., 2011, 2012; Mugusi et al., 2012). However, only a few studies have explored the association of HLA genes with anti-TB or ARV drugs-induced liver toxicity. Therefore, in this study, we aimed to investigate the possible associations between HLA-B alleles, and anti-TB and ARV drugs co-treatment induced liver injury in TB and HIV co-infected patients in Ethiopia.

A total of 92 TB and HIV co-infected patients on anti-TB and ARV drugs co-treatment were involved in this study; 46 treatment induced liver toxicity cases and 46 sex and age matched treatment tolerants. The demographics and clinical characteristics of the study participants are described in Table 1.

In the univariate analysis, there were statistically significant differences in the CD4 count and Karnofsky score between DILI cases and treatment tolerants (P < 0.05). In a multivariate logistic analysis, baseline CD4 count remained as significant predictors of anti-TB and ARV drugs co-treatment induced liver injury. There were no statistically significant differences in the baseline liver enzyme levels and type of ARV regimens used between the DILI cases and treatment tolerants. More than half of the DILI cases developed the cholestatic type of liver toxicity, and 85% of the cases had mild-to-moderate severity of liver toxicity. All of the cases had a minimum score of three (“possible”) in RUCAM scoring system for DILI.

HLA-B genotype result from the low resolution typing for each study participant is presented in Supplementary Table 1. Comparison of HLA-B allele carriers' proportions between patients who developed DILI vs. treatment tolerants is presented in Table 2. A total of 18 HLA-B variant alleles were detected (Table 2). In the univariate analysis, the proportion of HLA-B^*57 and HLA-B^*14 alleles carriers were significantly higher in DILI cases than treatment tolerants. Association of being a carrier of HLA-B^*57 with increased risk for DILI remained significant after correction for multiple testing. On the other hand, HLA-B^*41 was negatively associated with DILI. The multivariate logistic analysis retained HLA-B^*57 and HLA-B^*14 as significant predictors of concomitant anti-TB and ARV drugs induced liver injury. For HLA-B^*57, the association maintained after correcting for multiple comparisons (Pc = 0.036).

Associations between HLA-B^*57, HLA-B^*14, and HLA-B^*41 with the pattern and severity of liver toxicity was evaluated by comparing the proportion of allele carriers between DILI cases and treatment tolerants (Table 3). Compared with treatment tolerants, the proportion of HLA-B^*57 allele carriers were significantly higher in cholestatic and mixed types of DILI. In addition, the proportion of HLA-B^*14 allele carriers were higher in the mixed type of DILI cases than the treatment tolerants. There were no statistically significant differences in the hepatocellular type of DILI. Compared with treatment tolerants, the proportion of patients carrying HLA-B^*57 allele was significantly higher in the mild-to-moderate DILI group, and those carrying HLA-B^*14 in the severe DILI group were over-represented. There was a statistically significant difference in the proportion of HLA-B^*41 allele carriers in the mild-to-moderate DILI group compared with treatment tolerants.

HLA-B variant alleles that showed significant association with DILI from the low-resolution genotyping were further subjected to high resolution (four digit) testing to identify the sub-variant allele. The result from high resolution genotyping was interpreted in conjunction with the results from the prior low-resolution typing. HLA-B^*57 was the only variant allele that exhibited a significant association with DILI after correcting for multiple testing. Thus, high-resolution typing was done for all subjects who were genotyped as carriers of HLA-B^*57 in the low-resolution typing (see Table 4).

Of all HLA-B^*57 alleles identified, HLA-B^*57:03 accounted 58.3% and HLA-B^*57:02 accounted 41.7%. HLA-B^*57:01 was not detected. Of the HLA-B^*57 allele carrier cases, 10 (58.8%) had HLA-B^*57:03 (four homozygous) and the rest seven (41.2%) had HLA-B^*57:02 (two homozygous). There was only one HLA-B^*57 allele carrier (heterozygous for HLA-B^*57:02) out of the 46 treatment-tolerants. The proportion of HLA-B^*57:03 and B^*57:02 allele carriers were significantly higher in the DILI cases than treatment tolerants [P = 0.01, OR = 26.8 (1.5–47.2) and P = 0.03, OR = 8.1 (1.0–68.6)], respectively.

Comparisons of HLA-B^*57, ^*14, and ^*41 (from the low-resolution genotyping) as well as HLA-B^*57:02 and B^*57:03 (from the high-resolution genotyping) allele frequencies between DILI cases and treatment tolerants is presented in Table 5. The overall allele frequency of HLA-B^*57 was higher (13.0%) than that of the other HLA-B alleles (7.6% for B^*14 and 7.1% for B^*41). The allele frequency of HLA-B^*57 was higher in DILI cases (25.0%) compared to treatment tolerants (1.1%). The HLA-B^*57:03 and B^*57:02 allele frequencies in DILI cases (15.2 and 9.8%) were higher than the treatment tolerants (0 and 1.1%), respectively.

In the present study, we investigated the association of HLA-B variant alleles with risk for concomitant anti-TB and ARV drugs induced liver toxicity. The proportion of HLA-B^*57 allele carriers in Ethiopian patients who developed anti-TB and ARV drugs induced liver toxicity (37.0%), particularly in those who developed cholestatic type of liver toxicity (44.8%) was significantly higher compared with those who tolerated the treatment (2.2%). The proportion of HLA-B^*14 allele carriers who developed DILI (21.7%) was also significantly higher compared with the treatment tolerants (6.5%). These indicate that HLA-B^*57 and B^*14 allele carriers might be at a higher risk of developing anti-TB and ARV drugs co-treatment induced liver toxicity. Accordingly, these variant alleles might play important roles in the pathogenesis of immune-mediated liver toxicity during anti-TB and ARV drugs co-treatment. The HLA-B^*57 and B^*14 molecules may function as endogenous antigen presenting molecules for the drugs/metabolites to HLA-restricted cytotoxic T-cell activation (Pichler, 2002). To our knowledge, this is the first report to investigate the association of HLA-B^*57, B^*14, and B^*41 alleles with anti-TB and ARV drugs co-treatment induced liver toxicity.

The HLA-B^*57 variant allele, which was observed in significantly higher proportion among DILI cases than treatment tolerants, had a high specificity (97.8%) and positive predictive value (94.4%). Hence, HLA-B^*57 is likely to be an important predictor for anti-TB and ARV drugs co-treatment induced liver injury. However, there could be additional yet unidentified genetic markers and non-genetic risk factors involved in the pathogenesis of DILI. The matched case-control design used in this study minimizes effects of potential confounders and may increase power to identify genetic associations. Although this limits us from exploring associations of the matching variables such as sex, body mass index, Karnofsky score, CD4 count, and HIV viral load which were independently and significantly associated with the risk of developing DILI (Yimer et al., 2011, 2014). Association of CD4 cell counts and Karnofsky score as risk factors for DILI were also found in this study, although the others were not significant.

Studies suggest that a particular HLA-B allele may exert a protective effect against certain adverse drug reactions as evidenced by lower allele carrier frequencies in cases compared with treatment tolerants. HLA-DQA1^*01:02 was identified as a protective variant for anti-TB drugs induced hepatotoxicity (Sharma et al., 2002). HLA-B^*40:01 and HLA-B^*07:02 were also identified as protective variant alleles for carbamazepine-induced severe cutaneous adverse reactions (Alfirevic et al., 2006; Hung et al., 2006). In our study, statistically significant lower allele carrier rate of HLA-B^*41 was noted in the DILI cases compared with the treatment tolerants (6.5 vs. 21.7%), but this effect did not reach statistical significance after correcting for multiple comparisons. Further, analysis is required to clarify the role of HLA-B^*41 in the prevention of development of anti-TB and ARV drugs co-treatment induced liver injury.

Our result indicated a positive association of HLA-B^*57 allele with mild-to-moderate liver injury and the HLA-B^*14 allele with severe liver injury. On the other hand, HLA-B^*41 allele was negatively associated with mild-to-moderate liver injury. Accordingly, the association of HLA-B alleles with anti-TB and ARV drugs co-treatment induced liver injury may seem to depend on the severity of liver injury. The HLA-B^*57 allele may be critical for the initiation of the immune response to cause DILI and the HLA-B^*14 allele for the progression to severe degree of liver injury. On the other hand, the HLA-B^*41 allele seems to play a role in the prevention of development of mild-to-moderate liver injury due to anti-TB and ARV drugs co-treatment. These findings warrant further investigation in a larger DILI case samples for each severity grade of liver injury.

DILI can be hepatocellular (predominant rise in ALT), cholestatic (predominant rise in ALP), or mixed type liver injury (Hussaini and Farrington, 2014). Recently, we conducted a prospective observational study to evaluate the incidence, type, severity, and predisposing risk factors for DILI in a large well-defined TB and/or HIV patient cohort receiving either anti-TB drugs alone, ARV drugs alone or concomitant anti-TB and ARV therapy (Yimer et al., 2014). We found rates of hepatocellular DILI being highest among patients treated with anti-TB drugs alone than patients treated with ARV drugs alone or co-treated with anti-TB drugs. On the other hand, the rates of cholestatic DILI was highest among patients treated with efavirenz based-ARV drugs alone than patients treated with anti-TB drugs alone or with ARV drugs (Yimer et al., 2014). DILI due to anti-TB drugs in TB mono-infected patients is known to be more of hepatocellular type. In the present study, most of the TB-HIV co-infected patients treated with concomitant anti-TB and efavirenz based-ARV drugs developed cholestatic DILI cases. Apparently, there is a significant contribution from efavirenz based-ARV drugs toward developing cholestatic type of DILI. Indeed, the significant association of HLA-B^*57 variant allele with cholestatic type DILI identified in the present study might reflect for ARV-drugs induced hepatotoxicity. However, the role of HLA variant alleles for predisposition to anti-TB DILI cannot be ruled out (Sharma et al., 2002; Chen et al., 2015). Further, studies are necessary to investigate the association of HLA allele carrier status with anti-TB drugs alone as well as ARV drugs alone-induced liver injury.

Major histocompatibility complex (MHC) class I and class II-mediated immunological reactions are implicated in DILI, particularly in the cholestatic type that involves damage to the biliary system (Andrade et al., 2004; Daly, 2010). In line with this, carrier status of HLA-B^*57 was significantly higher in patients who presented with the cholestatic type of DILI (44.8%) and mixed type (26.7%) compared with those who tolerated the treatment (2.2%). None of the patients who developed hepatocellular DILI were carriers of HLA-B^*57 variant allele.

HLA-B^*57 allele is associated with long-term non-progressive chronic HIV-1 infection by restricting cytotoxic T-lymphocyte response (Goulder et al., 2000). HLA-B^*57:01 and B^*57:03 are the most prevalent HLA-B^*57 subtypes in Caucasian and African populations, respectively (Pelak et al., 2010; Apps et al., 2013). The HLA-B^*57:01 and B^*57:03 alleles are protective against HIV disease progression, and appear to present identical Gag epitopes (Payne et al., 2014). HLA-B^*57:01, B^*57:02, and B^*57:03 share more than 90% sequence homology and as such have peptide-binding repertoires which substantially overlap (Illing et al., 2012; Ogese et al., 2017). Although in ART naïve patients HLA-B^*57 (B^*57:01 in Europe and US, B^*57:03 in black Africans) confers protective effect against HIV-1 disease progression to AIDS (Costello et al., 1999; Migueles et al., 2000; López-Larrea et al., 2005; Frater et al., 2007), it may exert contradictory effect on treatment outcome when the disease course is altered by ARV therapy (Dold et al., 2015). Previous studies reported the association of HLA-B^*57 with increased all causes of mortality (Dold et al., 2015) and reduced virological responses during ARV therapy (Kuniholm et al., 2011). HLA-B^*57 allele is also known to be associated with immune-mediated drug-induced hypersensitivity reactions. Carriers of HLA-B^*57:01 allele are at higher risk of developing abacavir-induced hypersensitivity reactions (Hetherington et al., 2002), whereas HLA-B^*57:03 is associated with spondylarthropathies (López-Larrea et al., 2005). Indeed, genetic screening for HLA-B^*57:01 variant allele has been shown to reduce drug toxicity and subsequently led to a labeled recommendation of routine screening before treatment initiation (Hughes et al., 2008; Mallal et al., 2008).

The frequency and subtypes of HLA-B^*57 variant alleles display wide inter-ethnic variability globally ranging from 0 to 22.5% (http://www.allelefrequencies.net/). The overall frequency of HLA-B^*57 in our study population from Ethiopia is 13% which is relatively high. HLA-B^*57:01 occurs in Asians and Caucasians (up to 5%). The HLA-B^*57:03 and B^*57:02 variant alleles commonly occur in black population reaching up to 3 and 7% allele frequencies, respectively. Interestingly HLA-B^*57:01 was not detected, and it may be rare or absent in Ethiopians similar to other black Africans where the allele frequency is <1%. The overall HLA-B^*57:03 and B^*57:02 allele frequencies in our TB/HIV co-infected study population was 7.6 and 5.4%, respectively, although the frequencies in healthy Ethiopians is yet unknown. Interestingly, HLA-B^*57:03 and B^*57:02 allele frequencies in DILI cases (15.2 and 9.8%) were significantly higher than the allele frequencies in the treatment-tolerants (0 and 1.1%), respectively.

There were some limitations in this study. First, as DILI is a rare event, it was not easy to get large number of cases (four years were required to collect the DILI cases in this study), which subsequently resulted in a small number of samples for sub-group analysis. The second limitation is that as drug combinations are the current treatment protocols for TB and HIV co-infections, and hence we cannot link the risk variant allele to a specific drug or class of drug(s). Since first line anti-TB and HIV treatment regimen consists of combination therapy, it is not possible to study individual drug-induced liver toxicity in TB and HIV co-infected patients for ethical reasons. However, our study represents an important first step in applying HLA-B typing to identify genetic variants for anti-TB and ARV drugs co-treatment induced liver injury.

Identification of genetic risk factors for anti-TB, and ARV drugs co-treatment induced liver injury is essential for patient safety. The HLA risk alleles predisposing to immune-mediated anti-TB and ARV drugs induced liver toxicity in black African population are not well investigated. A common problem encountered in HLA genotyping is inability to determine the variant alleles accurately using a simple genotyping procedure. This is mainly due to the extensive genetic diversity in HLA gene locus. Accurate allele-level HLA typing using the current methods requires high workload, cost and time. Because of extreme genetic variation of the HLA locus, pharmacogenetic testing for routine clinical practice is increasingly challenging in resource limited settings. However, the recent development of second-generation sequencing methods provides the possibility of sequencing a single DNA strand in isolation. Establishing a straightforward and affordable genotyping method for accurate HLA typing to identify patients at risk of developing drug-induced adverse events may lay the ground for the future application of pharmacogenetic testing in clinical practice for globalized personalized medicine.

In conclusion, HLA-B variant alleles may play important roles in determining the risk and severity of concomitant anti-TB and ARV drugs induced liver toxicity. HLA-B^*57 variant alleles (HLA-B^*57:03 and HLA-B^*57:02) are risk factors to develop anti-TB and ARV drugs co-treatment induced liver injury, mainly of cholestatic type and mild DILI cases. The possible risk association of HLA-B^*14 allele with severe DILI and the protective association of HLA-B^*41 require further investigations. Additional studies are necessary to understand the roles of the identified HLA-B variant alleles in the pathogenesis of anti-TB and ARV drugs co-treatment induced liver toxicity.

EA, JK, and EM conceived and designed the study; EA, GY, AH, EM collected the data; EA, JK, and ZP performed the experiment and analyzed the data; EA and ZP wrote this paper. All authors revised/edited the manuscript and approved for submission.

The study was financially supported by grants from European and Developing Countries Clinical Trials Partnership (NL) (CG_TA.05.40204_005) and from Swedish Research Council (Vetenskapsrådet, Grant number: 2015-03295). This work was also supported partly by the NIH/Fogarty International Center Global Infectious Diseases grant D43TW009127.