N-acetyltransferase 2 genotypes amongst Zulu Speaking South Africans and isoniazid / N-acetyl-isoniazid pharmacokinetics during anti-tuberculosis treatment

Background Distribution of N-acetyltransferase2 (NAT2) polymorphisms varies considerably among different ethnic groups. Information on NAT2 single-nucleotide polymorphisms in South African population is limited. We investigated NAT2 polymorphisms and their effect on isoniazid pharmacokinetics in Zulu black HIV-infected South Africans in Durban, South Africa. Methods: HIV-infected participants with culture-confirmed pulmonary tuberculosis (TB) were enrolled from two unrelated studies. Culture-confirmed participants were genotyped for NAT2 polymorphisms 282C>T, 341T>C, 481C>T, 857G>A, 590G>A and 803A>G using Life Technologies pre-validated Taqman assays (Life Technologies, Paisley, UK). Participants underwent sampling for determination of plasma isoniazid and N-acetylisoniazid concentrations. Results Amongst the 120 patients, 63/120 (52.5%) were slow metabolisers (NAT2*5/*5), 43/120 (35.8%) had intermediate (NAT2*5/12), and 12/120 (11.7%) had rapid genotype (NAT2*4/*11, NAT2*11/12 and NAT2*12/12). NAT2 alleles in this study were *4, *5C, *5D, *5E, *5J, *5K, *5KA, *5T, *11A, *12A/12C and *12M. NAT2*5 was the most frequent allele (70.4%) followed by NAT2*12 (27.9%). 34/40 had both PK results and NAT2 genotyping results. The median area under the concentration-time-curve to infinity (AUC0-∞) interquartile range (IQR) was 7.81 (5.87 – 16.83) μg/ml/hr and maximum concentration (Cmax) 3.14 μg/ml (2.42 – 4.36) μg/mL. Individual polymorphisms were not equally distributed, with some represented in small numbers. Genotype did not correlate with phenotype, rapid genotype showing higher AUC0-∞ than slow but not significant, p=0.43. Conclusion There was high prevalence of slow followed by intermediate then rapid acetylator genotypes. The poor concordance between genotype and phenotype suggests that other factors or genetic loci influence INH metabolism, and warrants further investigation in this population.


Introduction: 54
Tuberculosis (TB) remains a leading cause of global morbidity and mortality, with approximately 55 10 million cases and 1.5 million deaths in 2018 (1). South Africa is a high TB burden country 56 with an estimated 301,000 cases in 2018. The so-called 'short-course' treatment regimen 57 recommended in international guidelines; consisting of 6 months of rifampicin and isoniazid, 58 supplemented by pyrazinamide and ethambutol in the first 2 months, has remained largely 59 unchanged for several decades. Whilst this regimen can achieve high relapse-free cure rates, a 60 range of host and mycobacterial factors can influence treatment outcomes. There is increasing 61 evidence that inter-individual variability in the pharmacokinetics (PK) of drugs within this 62 regimen lead to heterogeneity in clinical outcomes(2, 3). 63 64 Pharmacogenomics describes one cause of PK variability due to polymorphisms in drug 65 metabolising enzymes and transporters. During TB treatment, isoniazid is the paradigmatic 66 case. Isoniazid is acetylated to its major metabolite, N-acetyl-isoniazid (AcINH), by the action of 67 hepatic N-acetyltransferase 2 (NAT2). AcINH is subsequently rapidly hydrolysed to acetyl-68 hydrazine, which is also acetylated, to diacetyl-hydrazine, by the action of NAT2(4). 69 Accumulated acetyl-hydrazine can be oxidised to form other, potentially hepatotoxic 70 metabolites(4-6). Moreover, accumulated isoniazid can be metabolised by an alternative 71 pathway where it is first hydrolysed to hydrazine, which has also been implicated in liver injury, 72 before acetylation to acetyl-hydrazine, again by NAT2(4, 7). Hence, the activity of NAT2 both 73 dictates metabolism of isoniazid, and determines the availability of potentially hepatoxic 74 hydrazine and acetyl-hydrazine metabolites. Within the 870-base pair NAT2 gene, a number of 75 low-activity single nucleotide polymorphisms (SNPs) have been characterised. The NAT2 76 genotype has been shown to determine the rate of acetylation by NAT2 in several 77 populations(8). Individuals homozygous for the wild-type alleles are characterised as 'rapid' 78 acetylators (RAs), those homozygous for low-activity SNPs as 'slow' acetylators (SAs) and 79 heterozygotes as 'intermediate' acetylators (IAs)(9-13). SAs have a higher incidence of side-80 effects, particularly drug-induced hepatitis, during TB therapy, presumably due to higher levels 81 of hepatoxic metabolites (14)(15)(16)(17)(18)(19)(20). Amongst the first-line TB drugs isoniazid has the greatest 82 dose, with samples analysed for INH and AcINH for 60 participants with microbiologically 112 proven pulmonary TB (positive sputum culture or smear) who received a standard first line TB 113 regimen consisting of a FDC as described above .

Haplotype assignment and acetylator genotype inference 150
Haplotype assignment from probe-based SNP data is poorly described in African populations. 151 We elected to employ an unbiased PHASE analysis, which takes the dataset as a whole to assign 152 the most likely haplotype for each individual, alongside a probability for this assignment (53, 153 54). Haplotype for each individual and acetylator genotype for each haplotype were defined as 154 per the NAT gene nomenclature committee (55). Individuals with two rapid alleles were defined 155 as RAs, those with two slow alleles as SAs and those with one fast and one slow allele as IAs. 156 157

Isoniazid and N-acetyl-isoniazid PK and phenotype inference 158
Blood samples were collected and placed on ice immediately, before centrifugation within 60 159 minutes, immediate separation and storage of plasma at -70°C until analysis. Concentrations of 160 isoniazid, AcINH and a 6-aminonicotinic acid internal control were quantified using validated 161 high-performance liquid chromatography and tandem mass spectrometry (HPLC-MS/MS). 162 Sample preparation included a protein precipitation with acetonitrile and subsequent dilution 163 with water. Analytes were chromatographically separated using a Waters Exterra C18, 3.5µm, 164 50mm x 2.1mm column and detected using the AB Sciex 5500 Q-Trap mass spectrometer. All 165 analytes were analysed isocratically with an acetonitrile/water/0.1% formic acid mobile phase. 166 Isoniazid, AcINH and the internal standard were analysed at mass transitions of the precursor 167 ions (m/z) 137.9, 180.1 and 138.7 to the product ions (m/z) 66.0, 78.6 and 50.9, respectively. 168 Chromatographic data acquisition, peak integration and quantification of analytes was 169 performed using Analyst® software version 1.5.2. We constructed time-concentration curves in 170 the PK package in R for windows (version 3.5.1). We characterised the isoniazid and AcINH PK 171 parameters maximum concentration (C max ), time to maximum concentration (T max ), area under 172 the concentration curve from zero to infinity (AUC 0-∞ ) , apparent oral clearance (CL) and 173 elimination half-life and compared C max to published efficacy targets (56). AUC 0-∞ was calculated 174 using the trapezoid rule, apparent oral clearance estimated by dose / AUC 0-∞ and elimination 175 half-life by regression analysis of log 10 concentrations of the terminal exponent of elimination. 176 We analysed the ratio of log 10 AcINH to log 10 isoniazid at two and four hours to assess 177 acetylation phenotype.

Statistical methods 187
All data were entered in Epidata and transferred to either Stata (version 14) or R for windows 188 (version 3.5.1) for statistical analysis. Demographic characteristics were presented as 189 frequencies and percentages for categorical variables, and as means with standard deviations 190 for continuous variables. Descriptive PK data was described as median and inter-quartile 191 ranges. C max and AUC 0-∞ were log-transformed prior to comparison between genotypes. PK 192 parameters were compared, by genotype, using the Wilcoxon rank-sum test or  195

Hepatic adverse events 196
Hepatic adverse events were defined as elevated alanine transaminase (ALT) and

Participant characteristics 203
One hundred and twenty-two individuals living with HIV participating in two PK studies were 204 included in the study. Eighty participants in study 1 were included in the NAT2 genotyping 205 analysis and 60 in the PK analysis (with 58 individuals having both PK and genotype data), while 206 40 participants in study 2 were included in the PK analysis and 40 participants included in NAT2 207 genotyping analysis (with 34 individuals having both PK and genotype data). Key characteristics 208 are outlined in table 1. Participants in study 1 included 60 with pulmonary TB and HIV co-209 infection; 40 with CD4 count >200 cells/mm 3 and, 20 with CD4 count <200 cells/mm 3 as well as 210 20 participants living with HIV and without TB (who contributed only genotype data). All 40 211 participants in study 2 had TB and HIV coinfection, with a CD4 count of 200 cells/mm 3 or below. 212 In the combined studies, 66.7% of participants had CD4 counts <200 cells/mm 3 and 33.3% had 213 CD4 count >200 cells/mm 3 . The Median age was 33.1 years (IQR 18-53). Only 15 (12.5%) of 214 patients had a BMI < 18.86 kg/m 2 . 215 216 NAT2 genotype and deduced phenotype 217 One hundred and twenty participants (80 from study 1 and 40 from study 2) were genotyped. 218 Allele and haplotype frequencies and deduced phenotypes are outlined in tables 2-5. We 219 identified 12 different alleles in the population. The most common allelic group was NAT2*5 220 (70.4%) followed by NAT2*12 (27.9%). From the NAT2*5 group NAT2*5C (21.3%), NAT2*5J 221 (17.5%), NAT2*5D (14.6%) and NAT2*5K (10.4%) were the most common. The NAT2*12 group 222 was predominantly NAT2*12C. The deduced phenotype was 11.7% rapid, 35.8% intermediate 223 and 52.5% slow. 224 225

Isoniazid and N-acetyl-isoniazid PK 226
As above, to assess sample integrity for Study 1 we compared the AUC 0-∞ of the current analysis 227 with that previously reported on the same samples analysed in 2010. The median (IQR) AUC 0-∞ 228 was 5.53 (3.63 -9.12), processed at University of Cape Town (UCT) in 2009 and 5.70 (3.85 -229 7.94), processed at Africa Health Research Institute (AHRI) laboratory in 2014, suggesting that 230 the integrity of the samples was maintained for isoniazid, but cannot be confirmed for AcINH. 231 232 Study 1 showed rapid absorption, with a median (IQR) isoniazid T max of 1 hr(1 -2). Isoniazid 233 exposure was variable amongst individuals with median (IQR) C max 1.47 (1.14 -1.85) µg/ml and 234 AUC 0-∞ 5.53 (3.63 -9.12) µg.h/ml. Median (IQR) elimination half-life was relatively slow at 2.27 235 (1.69 -3.56) h. We compared these isoniazid PK measures to published targets; 98.28% (57/58) 236 failed to attain the minimum 2-hour plasma concentration target of 3 µg/ml (56). PK 237 parameters by genotype are shown in table 8(A), unexpectedly median half-life was slowest, 238 apparent oral clearance lowest and AUC 0-∞ highest amongst genotypically rapid acetylators, 239 with the reverse true for genotypically slow acetylators, although none of these differences was 240 statistically significant. Similarly, there were no statistically significant differences by genotype 241 for AcINH C max , elimination half-life or AUC 0-∞ . Median isoniazid and AcINH time-concentration 242 curves are given in Figure 1(A). 243 244 Absorption was rapid in Study 2, with a median INH T max of 2 hrs. INH exposure was also 245 variable amongst individuals with median (IQR) C max 3.14 µg/ml (2.39 -4.34) and AUC 0-∞ 10.76 246 µg.hr/ml (8.24 -28.96 µg/ml). Median elimination half-life was 2.62hr (2.26 -4.07). Again, we 247 compared these INH PK measures to published PK targets; 47.5% (19/40) failed to attain the 248 minimum 2-hour plasma concentration target of 3 µg/ml. PK parameters by genotype are 249 shown in table 8(B). For both isoniazid and AcINH and across the PK parameters; C max , AUC 0-∞ 250 and elimination half-life, variability (both range and IQR) were increased amongst those 251 genotyped as SAs. Again however, there were no statistically significant differences between 252 these PK parameters by genotype. Median isoniazid and AcINH time-concentration curves are 253 given in Figure 1(B). 254

255
For both studies we calculated the log 10 AcINH: log 10 isoniazid ratio, as a measure of 256 acetylation, at two and four hours post-dose and analysed this ratio by genotype (figure2 & 3). 257 In both studies we saw no statistically significant difference in ratios between genotypes at 258 either two or four hours. In Study 2 we again saw increased variability in this metric amongst 259 those genotyped as SAs. 260 261

Hepatic adverse events 262
There were no grade 3 and 4 hepatic adverse events in Study 1 and only 1 grade 4 hepatic 263 event was reported from the only participant with rapid genotype in Study 2. Although there 264 were more grade 1 hepatic adverse events among the slow genotype participants, as shown in 265 There was a high prevalence of the NAT2*5 allelic group in our population, accounting for the 280 slow acetylator genotype. In well studied Caucasian and Asian populations, four variants; 281 NAT2*4 (wild type, rapid), NAT2*5B, NAT2*6A, and NAT2*7B (all slow), account for most NAT2 282 alleles. In Asian populations there are generally a higher proportion of wild type NAT2*4 alleles 283 and few NAT2*5B alleles, and this difference largely accounts for the much lower prevalence of 284 RAs in non-Asian populations. Consistent with other studies in Sub-Saharan African populations, 285 the wild-type NAT*4 allele was far less prevalent and variant alleles were far more diverse in 286 our Study. In our population, the NAT2*5B allele was relatively rare in comparison to two 287 studies in the black population from Western Cape and North West Province. (45, 60). 288 However, in contrast to these populations, there were a diversity of other NAT2*5 alleles, 289 including a much higher prevalence of the rare NAT2*5J allele (17.5%) and the poorly 290 characterised NAT2*5K allele (10.4%). The NAT2*6A and NAT2*7B alleles, common in 291 Caucasian and Asian populations, were not seen in our cohort. In Caucasian and Asian 292 populations, rapid NAT2*12 alleles are rarely seen, where as in populations in sub-Saharan 293 Africa the NAT2*12A allele is reported at much higher frequencies(35). In our Study the 294 NAT2*12A allele did indeed comprise 5.8% of alleles seen but we saw a much higher frequency 295 of the NAT2*12C allele (21.2%), in contrast to other Southern African cohorts(10, 45, 60, 61). 296 297

Isoniazid C max and AUC 0-∞ demonstrated considerable variability between individuals in both 298
studies and almost all participants in Study 1 and almost half of the participants in Study 2 had a 299 C max below the lower limit of the target range(56). Low isoniazid concentrations during TB 300 treatment are concerning because it is postulated they may lead to poorer treatment 301 outcomes, or the generation of isoniazid resistance, the likely first step in the evolution of 302 multi-drug resistant TB (MDR TB). However, the evidence for either of these concerns is mixed 303 and in this setting the prevalence of INH mono-resistance is relatively low. 304

305
There was a marked difference in PK measures between the two studies analysed, with Study 1 306 having much lower measures than Study 2. There are several reasons that could have 307 contributed to this difference. The difference in isoniazid dosing could explain the lower PK 308 measures, where Study 1 used the FDC dosing as per WHO recommended weight bands, 309 leading to almost half the participants receiving doses <300 mg, as previously reported(49). All 310 participants in Study 2 received 300 mg doses of isoniazid irrespective of weight. Although the 311 samples of Study 1 did not appear to deteriorate during the 5 years between first analysis and 312 subsequent analyses for this study, differences in processing and storage between the studies 313 cannot be excluded. Figure 3  We identified no statistically significant difference by NAT2 genotype in a variety of PK 319 measures, hence in this cohort we found poor correlation between NAT2 genotype and 320 phenotypic acetylation of isoniazid. Previous studies in other populations have shown good 321 correlation between NAT2 genotype and isoniazid PK, suggesting that NAT2 genotyping could 322 be used as a parsimonious way to risk-stratify patients and personalise dosing of isoniazid in an 323 attempt to maximise efficacy whilst minimising toxicity. There are significant practical 324 difficulties to implementing these approaches in this setting, but our data suggest that in this 325 population NAT2 genotyping will not be helpful in guiding TB therapy. A lack of concordance (65). They found that HIV infection was related to an increase in variability of these DMEs. 332 Whilst additional pathways, aside from NAT2 genotype, have been implicated in hepatotoxicity 333 of isoniazid-containing TB treatment regimens, it is not clear that these pathways alter isoniazid 334 PK and thus could account for the lack of genotypic and phenotypic concordance in this study. 335

336
Although there were more hepatic adverse events among the SA, there was no statistical 337 association between genotype and hepatotoxicity in the two studies, with only 1 patient who 338 was a RA having a grade 4 hepatic adverse event and 2 others who were IA having grade 3 339 hepatic adverse events. 340 341 In our study, participants received pyridoxine and cotrimoxazole with the ATT in Study 2, but 342 not in Study 1 as we used the samples collected on day 1 for this analysis when only ATT was 343 given. As both INH and sulfamethoxazole are inhibitors of CYP2C9, this could be one of the 344 reasons for the variations noted. INH also inhibits CYP3A4, which is induced by rifampicin, this 345 interaction has not proven significant except when it relates to hepatotoxicity (66, 67). That the 346 combination of INH and rifampicin leads to an increased risk of hepatotoxicity, has been 347 reported in other studies. In our Study 2, isoniazid was given with Rifabutin which is a less 348 potent hepatic enzyme inducer, which therefore should have less interaction with INH (68). 349 Considering the limited effect on hepatotoxicity, the effect of CYP2E1 was not evident in our 350 study. We cannot confirm or exclude the effect of these CYP450 enzymes on INH metabolism in 351 these participants. 352

353
In our study samples were stored at -80 0 Celsius and loss of compound due to storage would 354 have been minimal (69), although studies have not reported on plasma samples stored longer 355 than 5 weeks, nor sample integrity for the metabolite, AcINH. 356 357

Conclusion 358
Amongst black Zulu TB-HIV coinfected South African patients, most had slow or intermediate 359 NAT2 genotype. There was a diversity of specific NAT2 alleles of a pattern differing from 360 previously studied cohorts in other settings. Despite the rarity of rapid acetylator genotypes, 361 INH PK was variable and a substantial proportion of individuals failed to attain minimum 362 efficacy targets. Importantly NAT2 genotype did not explain PK variability in this cohort or the 363 low C max , which suggests that other factors could be influencing isoniazid bioavailability and 364 metabolism, which require further elucidation.  Tables  382  383  Table 1: Demographic characteristics 384 385           and range (whiskers) for the pharmacokinetic parameters; log 10 maximum concentration 472 (C max ), log 10 area under the time-concentration curve (AUC 0-∞ ), of isoniazid (INH) and N-473 acetyl-INH (AcINH) stratified by acetylator status and logAcINH to logINH ratio at 2 and 4 474 hours stratified by acetylator genotype. 475 476