Patients were recruited at the Medical Research Council (Durban, South Africa), CNHU-PPC (Cotonou, Benin), CNHU Akron (Porto-Novo, Benin), hospital Ignace Deen (Conakry, Guinea), and Ambulatory Treatment Center (Dakar, Senegal) from March 2014 to September 2016. These are national reference centers for the management of TB and TB/HIV co-infected patients. The expertise and infrastructure required for conducting Pharmacokinetic studies in these centers were available and accessible.

TB or TB/HIV co-infected patients who participated in the Pharmacokinetic studies annexed to the OFLOTUB and RAFA clinical trials were included after their informed consent. The inclusion, non-inclusion, and exclusion criteria of both clinical trials are summarized in Table 1.

Whole blood samples were collected from fasting patients in 10 ml EDTA tubes and stored at –80˚C until the genomic DNAs extractions. South Africa patients’ samples were sent to the Mycobacteria Reference Laboratory (MRL) in Cotonou where genomic DNAs were extracted using QIAGEN^® kits QIAamp^® DNA Blood Maxi according to manufacturer’s instructions. On the other hand, the genomic DNAs of samples from Guinea, and Senegal were extracted respectively at the Mycobacteria Reference Laboratory (Hospital Ignace Deen, Conakry, in Guinea), and at the Medical Biology Laboratory (Ouakam Military Hospital, Dakar, in Senegal). All the extracted genomic DNAs were then sent to the MRL in Cotonou for the quality control procedures.

TB: tuberculosis; PK: pharmacokinetic; ARVs: antiretrovirals; MDR-TB: multidrug resistant tuberculosis.

The quality control was performed on all the samples and their DNA concentrations quantified with the NanoDrop 8000 spectrophotometer. All the extracts were then stored at −80˚C until their shipment to the University of Liverpool (United Kingdom) for genotyping analyses.

The double-stranded DNA concentrations of the extracts were quantifid by a fluorimetric method using Quant-iT^TM PicoGreen^TM dsDNA kits following the manufacturer’s instructions. The extracts were then diluted to prepare 96-well microplates, each containing 100 μl of DNA extract at 20 ng/μl.

Genotyping of DNA extracts was performed by an outsourced service based in Oxford using illumina HumanOmniExpress-24 v1.2 chips (illumina, San Diego, CA, USA).

Quality control of genotyping data. It was done according to the method of Coleman et al. for multiple purposes: to search missing data and duplicates, to check gender concordance and heterozygosity in the patients, to check the minor alleles frequency and the adherence to Hardy-Weinberg equilibrium [15] . Genetic polymorphisms with <95% call rate or presenting a deviation from Hardy-Weinberg equilibrium (HWE) were removed [16] .

The genotyping data were visualized on the Manhattan plot (Figure 1) to identify the significant genetic polymorphisms at the threshold of p < 5 × 10⁻⁸ [17] .

The correlation between the genotyping data, plasma clearance, and the AUC was tested by a multiple linear regression model using PLINK2 software [18] [19] .

A total of 408 participants from four countries were included in this study (Table 2). In all the countries, they were mostly men (sex ratio: 1.582). The sex ratio by country varies from 1.093 in Guinea to 5.250 in Senegal, where the number of participants was the lowest. The distributions of participants by country are not related to the national TB prevalences.

OFLOTUB trial participants were approximately twice as numerous as those in RAFA trial (Table 3). They were about 6 years younger than the RAFA trial participants. The RAFA trial conducted in Benin, Guinea, and Senegal did not concern South Africa. Body mass indices (BMI) of the participants randomized in the different arms of both clinical trials were identical.

Three (03) samples with more than 5% missing genotypes, two (02) duplicated samples; 384,586 genetic polymorphisms having failed to the frequency test and to the Hardy-Weinberg equilibrium were excluded. In summary, a total of 378 samples (209 OFLOTUB + 169 RAFA) and 318,751 genetic polymorphisms were included in the genetic analyses.

There was no significant association between the genetic polymorphisms and the pharmacokinetic parameters of Rifampicin. The SLCO1B1 polymorphisms (rs11045819 and rs4149032) were not significantly associated with variations in Rifampicin pharmacokinetic parameters. However, the rs11792875 polymorphism appears to be associated (p = 2.44e-07) with Rifampicin clearance but not with its AUC.

Two NAT2 polymorphisms (rs1801280 and 1495741) were associated with the variation of Isoniazid clearance in OFLOTUB trial patients (Figure 1).

The horizontal red line represents the significance threshold p < 5 × 10⁻⁸ at the genome scale (−log10). The NAT2 polymorphisms associated with isoniazid clearance are in green and located on chromosome 8.

SLCO1B1 polymorphisms frequencies are variable in TB or TB/HIV co-infected patients and from one ethnic group to another. Thus, SLCO1B1 rs11045819 polymorphism frequency is 19% in African TB patients, and 25% in healthy American controls [24] . In previous studies, SLCO1B1 rs11045819 polymorphism was associated with decrease in Rifampicin plasma concentrations [24] . Similarly, SLCO1B1 rs4149032 polymorphism was associated with lower Rifampicin exposures in African populations, suggesting the need for increased therapeutic doses of Rifampicin. This very common polymorphism is present in 70% of South Africans [24] . In children with TB, the SLCO1B1 rs2306283 polymorphism has been associated with lower Rifampicin concentrations [11] .

Unlike Allegra et al., who found an association between SLCO1B1 rs4149056 polymorphism and higher plasma concentrations of Rifampicin, our results showed that there was no association [25] . In addition, our results failed to confirm the effects of SLCO1B1 rs2306283, rs4149032, and rs11045819 polymorphisms identified in previous conflicting studies. On the contrary, SLCO1B1 rs11792875 polymorphism we identified could be a new genetic marker of great interest for further investigations in later studies. Kim et al., in their study on Korean patients with TB, concluded that SLCO1B1 polymorphisms did not influence Rifampicin plasma concentrations and clearances [24] . Therefore, SLCO1B1 polymorphisms are not the only genetic factors responsible for the variability of Rifampicin plasma concentrations in TB and TB/HIV patients. In addition, SLCO1B1 polymorphisms would have no effect on the time to conversion of Acid-Fast Bacilli smears and cultures, as well as on the occurrence of Rifampicin adverse effects [24] .

Polymorphisms of drug metabolizing enzymes can be single nucleotide polymorphisms (SNPs) or haplotypes. In the case of our study, the SNPs associated with the variability of Isoniazid clearance were NAT2 rs1801280 (p = 2.14e-10), rs3008589 (p = 3.346e-08), and rs12641762 (p = 2.565e-08). The SNP rs1495741 was associated with the variability of the AUC (p = 1.14e-09) and clearance (p = 6.18e-14) of Isoniazid. Some previous studies in other populations have also found the existence of a good correlation between NAT2 polymorphisms and Isoniazid pharmacokinetics [1] [26] [27] . This correlation suggests that NAT2 genotyping could allow the optimization of Isoniazid doses in order to increase its efficacy while reducing its toxicity [28] .

NAT2*12A polymorphisms are most frequent in black populations in Sub-Saharan Africa. Unlike the more common SLCO1B1 polymorphisms in South African patients, the frequencies of NAT2 polymorphisms could be similar in all patients from the countries involved in this study. Thus, according to Mthiyane et al., the genotypic frequencies of slow, intermediate, and fast acetylators were 52.5; 35.8, and 11.7% respectively in South African, black and TB/HIV co-infected patients [28] . NAT2*4, NAT2*5B polymorphisms, less frequent in the black ethnicity, were not identified in the patients of this study. Similarly, NAT2*6A and NAT2*7B polymorphisms, common in Caucasian and Asian populations, were not observed in this study.

The microarray-based genotyping used in this study made it possible to identify new genetic polymorphisms that can modify the pharmacokinetic of Rifampicin and Isoniazid. The microarray-based genotyping method would identify a greater number of polymorphisms than other genotyping method. Unlike this genotyping method, some authors carry out a literature review to select the most significant genetic polymorphisms. Participants from Guinea and Senegal were the least numerous because few patients from these two countries had participated in the pharmacokinetic studies.

Many studies have shown that people with low NAT2 activity have a higher risk of developing liver problems than those with high NAT2 activity. However, other authors have observed that rapid acetylation is a risk factor for antituberculosis drug-induced hepatitis [9] .

The NAT2 haplotype rs1041983|rs1801280|rs1799929|rs1799930|rs1208|rs1799931 identified in this study was only associated with the variability of Isoniazid clearance (p = 2.77e-09). This haplotype found in some patients included in this study is associated with Isoniazid hepatotoxicity [29] [30] [31] . Other haplotypes such as NAT2*6A variants have also been associated with hepatotoxicity [32] . In contrast, the wild-type NAT2*4 haplotype is not associated with Isoniazid hepatotoxicity [29] . The lack of concordant evidence between the association of the different NAT2 haplotype variants and the occurrence of hepatotoxicity in patients exposed to Isoniazid requires a study on a larger population. It is also possible that there is an interaction between the SNPs of the haplotype leading to the activation or inactivation of their hepatotoxic effects.

In the patients of this study, we did not detect other NAT2 polymorphisms, probably due to their low frequencies (<1%). The haplotypes, NAT2*4 and NAT2*6A, would be new predictive biomarkers of Isoniazid-induced hepatotoxicity.

This study has certain limitations related to the study arms of OFLOTUB and RAFA trials, but also due to the method used for the genotyping. Indeed, unlike the patients in OFLOTUB trial, those in RAFA trial received, in addition to anti-TB drugs, antiretrovirals initiated earlier at two weeks or later at two months. It is therefore possible that drug interactions between anti-TB drugs and antiretrovirals in RAFA patients affect the pharmacokinetic parameters of Rifampicin and Isoniazid. However, since the participants in OFLOTUB trial are much more numerous, it is possible that these interactions, even if they existed, could not bias the results of this study.

The microarray used in this study is universal, not specific to Africans requiring chips enriched with the genetic polymorphisms frequent in black populations from Sub-Saharan Africa.