Effect of NAT2, GSTM1 and CYP2E1 genetic polymorphisms on plasma concentration of isoniazid and its metabolites in patients with tuberculosis, and the assessment of exposure-response relationships

Objectives: Isoniazid is a key drug in the chemotherapy of tuberculosis (TB), however, interindividual variability in pharmacokinetic parameters and drug plasma levels may affect drug responses including drug induced hepatotoxicity. The current study investigated the relationships between isoniazid exposure and isoniazid metabolism-related genetic factors in the context of occurrence of drug induced hepatotoxicity and TB treatment outcomes. Methods: Demographic characteristics and clinical information were collected in a prospective TB cohort study in Latvia (N = 34). Time to sputum culture conversion (tSCC) was used as a treatment response marker. Blood plasma concentrations of isoniazid (INH) and its metabolites acetylisoniazid (AcINH) and isonicotinic acid (INA) were determined at three time points (pre-dose (0 h), 2 h and 6 h after drug intake) using liquid chromatography-tandem mass spectrometry. Genetic variations of three key INH-metabolizing enzymes (NAT2, CYP2E1, and GSTM1) were investigated by application PCR- and Next-generation sequencing-based methods. Depending on variables, group comparisons were performed by Student’s t-test, one-way ANOVA, Mann-Whitney-Wilcoxon, and Kruskal-Wallis tests. Pearson correlation coefficient was calculated for the pairs of normally distributed variables; model with rank transformations were used for non-normally distributed variables. Time-to-event analysis was performed to analyze the tSCC data. The cumulative probability of tSCC was obtained using Kaplan-Meier estimators. Cox proportional hazards models were fitted to estimate hazard rate ratios of successful tSCC. Results: High TB treatment success rate (94.1%) was achieved despite the variability in INH exposure. Clinical and demographic factors were not associated with either tSCC, hepatotoxicity, or INH pharmacokinetics parameters. Correlations between plasma concentrations of INH and its metabolites were NAT2 phenotype-dependent, while GSTM1 genetic variants did not showed any effects. CYP2E1*6 (T > A) allelic variant was associated with INH pharmacokinetic parameters. Decreased level of AcINH was associated with hepatotoxicity, while decreased values of INA/INH and AcINH/INH were associated with month two sputum culture positivity. Conclusion: Our findings suggest that CYP2E1, but not GSTM1, significantly affects the INH pharmacokinetics along with NAT2. AcINH plasma level could serve as a biomarker for INH-related hepatotoxicity, and the inclusion of INH metabolite screening in TB therapeutic drug monitoring could be beneficial in clinical studies for determination of optimal dosing strategies.


Introduction
Tuberculosis (TB) is one of the leading infectious causes of death worldwide.The World Health Organization estimates that 10.6 million people developed TB in 2021 (World Health Organization, 2022).Treatment success rate of 85% for drugsusceptible TB is achieved with 2-month short-course chemotherapy containing isoniazid (INH), rifampicin (RIF), ethambutol (EMB), and pyrazinamide (PZA), followed by INH and RIF for additional 4 months (World Health Organization, 2020).INH is a key drug in the chemotherapy of TB, however, interindividual variability in pharmacokinetic parameters and drug plasma levels may affect drug responses (Ellard and Gammon, 1976;Aarnoutse, 2011).Therapeutic drug monitoring, which utilizes a technique using plasma drug concentration measurements in clinical specimens, was highlighted as a personized treatment approach for the evaluation of the effectiveness of anti-TB treatment and achieving therapeutic goals (Alsultan and Peloquin, 2014;Perumal et al., 2020).However, the current data correlating INH plasma concentrations, adverse drug events and TB treatment outcome are still limited, and the involvement of contributing factors is not sufficiently studied.
Several enzymes are known to metabolize INH, and a detailed model of candidate enzymes in the metabolism of INH in liver cells is described elsewhere (Klein et al., 2016).The main metabolic pathway of INH is acetylation by arylamine N-acetyltransferase 2 (NAT2) to N-acetylisoniazid (AcINH) (Ellard and Gammon, 1976;Daly, 2021).Other INH metabolites, such as hydrazine (Hz), isonicotinic acid (INA), acetylhydrazine (AcHz), and diacetylhydrazine (diAcHz) are generated by hydrolysis, glycine conjugation, hydrazone formation, and further acetylation.None of the INH metabolites are active, while Hz and AcHz are considered to be hepatotoxic (Becker et al., 2007).In turn, INH, AcHz, and Hz are likely oxidized, in part, by cytochrome P450 2E1 (CYP2E1) enzyme into potentially hepatotoxic intermediates, which could be further conjugated by glutathione S-transferase (GST) enzyme family with glutathione and effectively removed by excretion (Klein et al., 2016).Currently, NAT2 is the most extensively studied pharmacogenetic risk factor for the development of INHrelated drug induced hepatotoxicity (DIH) (Daly, 2021).It is well-known that the NAT2 enzyme activity exhibits substantial interindividual variability resulting in different INH plasma halflives (Ellard and Gammon, 1976;Aarnoutse, 2011).Genetic polymorphisms of NAT2 gene are responsible for the variations in the acetylation capacity, and the recent metaanalysis study has shown that NAT2 deficiency was significantly associated with the likelihood of experiencing DIH (Richardson et al., 2019).Also, it was proposed that INH dose adjustment based on the NAT2 genotype before prescribing TB treatment could improve clinical outcome and reduce the risk of INH-induced liver injury, however, the current level of evidence is not sufficient to recommend dose optimization (Azuma et al., 2013;McIlleron and Chirehwa, 2018;Khan and Das, 2022).A number of risk factors for the development of INHrelated liver injury have been identified, including age and some environmental factors, such as alcohol consumption, and genetic variations in other drug metabolizing enzymes, such as CYP2E1 and GSTM1, could also be related to the alterations of the INH metabolism and accumulation of toxic substances, which, subsequently, can lead to adverse drug events and affect treatment outcome (Ellard and Gammon, 1976;Yang et al., 2019).
This study has two objectives: (1) to examine the impact of exposure to INH and its metabolites on TB treatment outcome and DIH; (2) to investigate factors which influence variability in INH pharmacokinetics.Towards these objectives we performed pharmacokinetic profiling of INH and its major metabolites (AcINH, INA) in patients with pulmonary TB, and investigated genetic variations of three key INH-metabolizing enzymes (NAT2, CYP2E1, and GSTM1) along with an array of patient-related factors.The obtained data provide a better insight into INH pharmacogenetics thus contributing to future implementation of personalized approach for enhancing the efficacy, tolerability and limiting the toxicity of existing anti-TB medication.

Study participants and data collection
Study cohort was comprised of 34 patients with diagnosed drug susceptible pulmonary TB admitted to the Riga East University Hospital (REUH), Centre of Tuberculosis and Lung Diseases.In order to minimize the impact of other health conditions on the study results, the following exclusion criteria were applied: age ≤ 18 years, pregnancy or lactation, history of cancer, other infectious diseases (e.g., HVC, HVB, HIV, and AIDS), and renal failure.
The diagnosis of TB was confirmed radiologically, symptomatically, and/or bacteriologically on Löwenstein-Jensen (L-J) solid medium.Xpert/RIF Ultra test was used for detection of TB and resistance to RIF with additional Line probe assay for resistance-conferring mutation/INH resistance detection before phenotypic drug susceptibility testing.Sputum smear/culture findings, laboratory test values and treatment outcomes were recorded.Time to sputum culture conversion (tSCC) was determined by the number of days to the first negative inoculation of Löwenstein-Jensen (L-J) medium (including the last day), indicating the treatment outcome."Treatment success" was defined as the sum of "cured" and "treatment completed" (World Health Organization, 2022).All patients received the WHO-recommended treatment regimen for drug-susceptible TB which consistent of a 2-month chemotherapy containing isoniazid INH (4-6 mg/kg), RIF (8-12 mg/kg), PZA (20-30 mg/kg) and EMB (15-25 mg/kg) followed by INH and RIF for additional 4 months.The anti-TB drugs were all single drug products rather than fixeddose combinations, administered once daily.
Clinical biochemistry tests of liver and renal function were performed at baseline and on day 10-12 of treatment.The reference ranges for transaminases [alanine aminotransferase (ALAT), aspartate aminotransferase (ASAT)], direct and total bilirubin were adopted from specific reference ranges of REUH Laboratory Service laboratory "Gaiļezers" to determine the upper limit of normal range (ULN); the reference range were as follows: alanine aminotransferase (ALAT), 8-41 U/L; aspartate aminotransferase (ASAT), 8-40 U/L; direct (conjugated) bilirubin, < 3.4 µ mol/L; total bilirubin, 5-21 µ mol/L.The DIH case was confirmed based on the increase of ALAT/ASAT values above the ULN on day 10-12 of TB treatment with/without increase of total and conjugated bilirubin levels: mild hyperfermentemia was characterized by elevated serum ALAT/ASAT levels 1.5-5 times above the ULN; moderate hyperfermentemia: 5-10 times above the ULN, and severe hyperfermentemia: >10 times above ULN.
Patient data [age, biological sex, height, weight, sputum microscopy results, time to sputum culture conversion (tSCC), laboratory findings, treatment outcome] were obtained from medical records; smoking status, self-reported alcohol intake were obtained from patients' questionnaires.Patient's nutritional status was reported as body mass index (BMI), which was calculated based on height and body weight.According to the TB treatment guidelines, TB treatment was initiated following the directly observed treatment (DOT) strategy for health facility-based patients, and the video observed directly observed treatment (VDOT) strategy was implemented for outpatient department patients.
Patients were enrolled in the study after providing written informed consent for the use of their phenotypic and genetic data.(approval No. 6-3/1/6 and approval No. 105/28.01.2016.).This study was conducted in accordance with the Declaration of Helsinki.

DNA and blood plasma samples
Blood sample collection was carried on 10-12 days after the treatment onset.Blood samples were collected by medical personnel using vacutainers with EDTA at three time points: pre-dose (0 h), 2 h, and 6 h after drug intake.Blood samples were immediately centrifuged at 4,000 rpm (3,488 × g) for 15 min at 4 °C to separate the plasma; plasma aliquots were immediately frozen and stored at −70 °C until further analysis.Genomic DNA was extracted from the peripheral white blood cells using the standard phenolchloroform method described elsewhere (Sambrook et al., 1989).DNA sample concentrations were measured using a spectrophotometer (ND-1000 UV-VIS Spectrophotometer, NanoDrop Technologies, United States).DNA samples were stored at −20 °C, until it was used as a template for PCR amplification.
For NAT2 gene analysis of heterozygous samples, subsequent cloning of 1211-bp PCR fragment into the pJET1.2/bluntcloning vector was performed using the CloneJET PCR Cloning Kit (Thermo Fisher Scientific, United States) according to the manufacturer's instructions.Plasmids with confirmed NAT2 gene inserts were purified from bacterial cultures using GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific, United States), and the inserts were sequenced.
Sequencing data analysis was performed using CodonCode Aligner software (CodonCode Corporation, United States) with the sequence of human gene for arylamine N-acetyltransferase (EC 2.3.1.5)(GenBank: X14672.1) as the reference.A possible NAT2 phenotype was assigned based on the obtained genotyping data of conventional 7-SNP signature, as the presence/absence of signature mutations were linked to the resultant protein activity (Sim et al., 2014).In addition, a database of human NAT2 alleles was used (http://nat.mbg.duth.gr/Human%20NAT2%20alleles_2013.htm).According to (Stanley and Sim, 2008) individuals were classified as rapid (carrying two rapid NAT2 alleles), intermediate (one rapid and one slow allele) or slow (two slow alleles) acetylators.

GSTM1 genotyping
The GSTM1 gene double deletion genotype (0/0) was detected based on the comparative duplex PCR assay as described previously (Igumnova et al., 2016).Briefly, the PCR master mix contained (per reaction) 1X Taq Buffer with (NH 4 ) 2 SO 4 , 2.5 mM MgCl 2 , 100 mM of each dNTPs, 0.2 mM of each primer Beta1 (5′-GGTTGGCCAATCTACTCCCAG G-3′), Beta2 (5′-GCTCACTCAGTGTGGCAAAG-3′), M1 (5′-CTG CCCTACTTGATTGATGGG-3′) and M2 (5′-CTGGATTGTAGC AGATCATGC-3′), 1.25 U of Taq DNA polymerase and 28 ng of DNA template.PCR assays were performed under the following conditions: an initial denaturation at 95 °C for 10 min; 35 cycles of denaturation at 95 °C for 30 s, primer annealing at 64 °C for 30 s, and elongation at 72 °C for 40 s; and a final elongation step at 72 °C for 2 min.The PCR products were separated and visualized using electrophoresis on a 1.5% agarose gel containing 0.2 mg of ethidium bromide/ml with transillumination under UV light.The homozygous GSTM1 null genotype (0/0) was detected by the absence of the 273-bp.Individuals who were presented with the 273-bp fragment were classified as GSTM1 plus genotype (A/A or A/0) and contained at least one wild-type allele.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS)
Blood plasma concentrations of INH and its main metabolites AcINH and INA were determined at three time points (pre-dose (0 h), 2 h and 6 h after drug intake) using liquid chromatographytandem mass spectrometry (LC-MS/MS) method as previously described (Kivrane et al., 2021).Briefly, plasma samples were analysed using An Acquity UPLC H-Class chromatographic system (Waters, Milford, MA, United States) coupled to a XEVO TQ-S tandem mass spectrometer (Waters, Milford, MA, United States).The chromatographic separation was achieved on the Waters Acquity UPLC BEH C8 column (2.1 mm × 75 mm; 1.7 μm).The multiple reaction monitoring mode (MRM) was used for detection and quantification of the analytes.Data acquisition and analysis were performed using MassLynx software and TargetLynx module (version 4.1.,Waters, Milford, MA, United States).Calibration ranges were 0.16-10.0μg/mL for INH, AcINH and INA, respectively; the lower limits of quantification (LLOQ) were 0.16 μg/mL for INH and both two primary metabolites.The within-and between-run accuracy in all concentration levels ranged from 87.2% to 113.6%, but within-and between-run precision was between 1.6% and 14.9% (at the LLOQ level CV<20%).

Statistical analysis
All statistical analysis were conducted using R statistical software version 4.2.1.(R Core Team (2022), R Foundation for Statistical Computing, Vienna, Austria; https://r-project.org).
Categorical variables were summarized as frequencies and percentages; equality of proportions and association between two-categorical variables were tested using Chi-square tests.When Chi-square test were not applicable, Fisher's exact test was used instead for association test.Shapiro-Wilk test was used to verify the normality assumption, and, where possible, log-transformation was used to achieve the normal distribution in the data.For normally distributed variables, two or more group comparisons were performed by Student's t-test and one-way ANOVA, respectively.Models with rank transformations were used for non-normally distributed variables, two and three or more group comparisons were performed by Mann-Whitney-Wilcoxon and Kruskal-Wallis test, respectively.
Pearson correlation coefficient with the corresponding 95% confidence interval and p-values were calculated for the pairs of normally distributed variables; model with rank transformations were used for non-normally distributed variables (i.e., AcINH/INH and INA/INH ratios).When applicable, during the correlation analysis, the six pharmacokinetic parameters of INH were adjusted for age, sex, and INH dose (mg/kg); AcINH and INA (AUC0-6 h) values were log-transformed and the correlations were also stratified by NAT2 acetylator status and GSTM1 genotypes.
Time-to-event analysis was performed to analyse the tSCC data.Patients whose treatment outcomes were "treatment completed" or "lost to follow-up" were considered as censored.The cumulative probability of tSCC, stratified by GSTM1 genotypes and NAT2 phenotypes, was obtained using Kaplan-Meier estimators.Log-rank tests were performed to assess differences between the groups.Median tSCC were calculated for both GSTM1 and NAT2 groups.Additionally, the mean tSCC, restricted at time point 60 days, was compared between subgroups of GSTM1 and NAT2 (R library survRM2).Confidence intervals were constructed for the group medians and for the differences of restricted mean tSCC.Cox proportional hazards models were fitted to estimate hazard rate (HR) ratios of successful tSCC.In a case of proportional hazards, the HR of successful tSCC can be interpreted as the odds for an individual in the group with higher hazard to reach the endpoint first (Spruance et al., 2004).In the Cox regression models the following predictors were included for the whole study cohort (n = 34): ( 1 3 Results
In total, after 10-12 days of therapy, transaminase levels increased in 4 patients (11.76%): mild and moderate DIH developed in one patient each, and severe DIH-in two patients.In two of these patients ALAT/ASAT levels were increased before therapy onset, and in 1 patient simultaneous changes in total and conjugated bilirubin levels were recorded (Table 2, Supplementary Table S1; Supplementary Figure S1).

INH pharmacokinetic data
Mean plasma values of INH and both metabolites are presented in Table 3. Overall, 64.7% (22/34) of the patients had INH C max 2 h value below the therapeutic range window (<3.0 μg/mL; Supplementary Table S1); the mean INH C max 2 h of the cohort was 2.7 (±1.4) µg/mL.
When selecting tagging SNPs, priority was given to variants found to be associated with impaired drug metabolism, i.e., with potential effects on CYP2E1 enzyme activity and/or potentially leading to adverse drug reactions, including DIH.Based on the MAF and HWE results, as well as previously reported possible effects on CYP2E1 enzyme activity, 7 SNPs were selected for further analysis: rs8192766, rs6413432, rs2515641, rs41299422, rs7081484, rs41299398, and rs915908.

Association of non-genetic and pharmacokinetic factors with TB treatment outcome and tSCC
No significant associations of age, BMI group, self-reported alcohol intake, or INH dose (mg/kg) with treatment outcome or tSCC were found (Table 2).There was a significant association between smoking status and treatment outcome, where smokers in comparison to non-smokers had 13.8 times higher odds of treatment outcome "cured" in comparison to the "treatment completed" and "lost to follow up" outcomes (p = 0.039); however, we did not observe significant associations of smoking status and the overall treatment success (Table 2).Similarly, there was a significant association between sputum microscopy results and treatment outcome: the treatment outcome "cured" could be assigned to 100% of microscopy-positive patients in comparison to the microscopy-negative patients (outcome "cured": 60.0%, "treatment completed": 26.7%, "lost to follow up": 13.3%) (p = 0.02).However, again, sputum microscopy result was not associated with the overall treatment success in our cohort.Further analysis indicated that smoking was significantly associated with positive sputum microscopy (p = 0.025) (Supplementary Table S4).
Within the INH pharmacokinetic parameters, median INA/INH ratio was significantly higher in patients with tSCC < 60 days in comparison to those with tSCC > 60 days (p = 0.026) (

Association of non-genetic and pharmacokinetic factors with DIH
No significant associations of age, BMI group, self-reported alcohol intake, or INH dose (mg/kg) with hepatotoxicity were found (Table 2).Biological sex was slightly associated with DIH, with males having 85% lower odds of hepatotoxicity than women, however, statistical significance was not reached (p = 0.071).Within the INH pharmacokinetic parameters, AcINH AUC0-6 h was significantly lower in patients with DIH (p = 0.019).

Association of factors with INH pharmacokinetic parameters
Further, we analysed the possible impact of various non-genetic and genetic factors on the variability in INH pharmacokinetics.None of the analysed pharmacokinetic parameters differed significantly between patients' groups when stratified by biological sex, smoking status, BMI, and self-reported daily alcohol intake (Table 2).Also, neither age, nor INH dose (mg/kg) correlated with any of INH pharmacokinetics parameters (Supplementary Figures S2, S3).
Correlations between the pharmacokinetic parameters became significantly stronger when the patients were stratified based on the NAT2 acetylator status; the majority of the observed correlation coefficients were statistically significant, and clear separation between NAT2 groups was observed (Figures 2C, D, G, H).On the other hand, stratification of the patients based on the GSTM1 genotype did not improve the correlation values for the pharmacokinetic parameters, and clear group separation was not observed (Supplementary Figure S4).
In the whole cohort, significant correlation was observed between both INH metabolites INA (AUC 0-6h ) and AcINH (AUC 0-6h ), as well as between AcINH/INH and INA/INH ratios (Figures 3A, D).Again, group separation was observed for NAT2 SA and IA acetylators, but not for the GSTM1 genotypes (Figures 3B,  C, E, F).
After adjustment by sex, age and INH dose, the mean and median values of the pharmacokinetics parameters differed significantly between NAT2 SA and IA groups, except for INA AUC 0-6h (Table 4).More specifically, INH AUC 0-6h and INH C max values were significantly higher, while AcINH AUC 0-6h values were significantly lower in NAT2 slow acetylators.Also, median values of both AcINH/INH and INA/INH ratios were significantly lower in the NAT2 SA group.None of pharmacokinetic parameters differed significantly between GSTM1-plus and -null genotypes.
Similarly, pharmacokinetic parameters were compared between patient groups divided according to the presence/absence of the each CYP2E1 SNPs.Overall, only rs6413432 claimed a statistically significant difference for three INH pharmacokinetic parameters: INH AUC 0-6h , values were significantly higher, while median values of both AcINH/INH MR and INA/INH MR were significantly lower in patients with rs6413432 (Table 4).

Association of genetic factors with treatment outcome, tSCC and DIH
As some INH pharmacokinetic parameters were associated with tSCC and DIH, and the impact of genetic factors on INH pharmacokinetics was observed, we further investigated possible association of genetic factors with treatment outcome, tSCC and DIH.
Based on the univariate analysis, there were no significant associations between treatment outcome and any of genetic factors studied (Table 5).DIH occurred only in the patients with NAT2 SA phenotype (N = 4); however, statistically significant difference was not observed between NAT2 SA and IA groups (Table 5).A trend of association between NAT2 phenotypes and tSCC was detected, with SA having 80% lower odds of tSCC within 60 days in comparison to IA (OR = 0.2; 95% CI, 0.03-1.22);however, statistical significance was not reached (p = 0.069) (Table 5).
No significant association between GSTM1 groups and tSCC was detected, and no significant associations were found for any of 7 analysed SNPs of CYP2E1 gene with either treatment outcome, tSCC, or hepatotoxicity (Table 5).

Discussion
This is the prospective study aimed to decipher the impact of pharmacokinetic, pharmacogenetic and patient-related factors on the response to TB treatment, TB treatment outcome and the risk of DIH, and to investigate possible links between these variables.For this task, along with the array of patient-and disease-related factors, we evaluated six pharmacokinetic parameters of the parent drug INH and its two metabolites, AcINH and INA, for more accurate assessments of the data obtained (Akkerman et al., 2014;Sturkenboom et al., 2015).In our study, a high TB treatment success rate was achieved (94.1%).However, we did not observe any associations between patient-related factors with either DIH or treatment outcome.Molla and colleagues have reported that males were significantly less likely to develop hepatotoxicity than females (Molla et al., 2021); in our study similar trend was observed, however, statistical significance was not reached (p = 0.071).Smoking status was considered to be a baseline risk factor associated with negative impact on the TB treatment success, delaying conversion of the sputum culture, and extending the time of TB treatment (Silva et al., 2018;World Health Organization, 2022).Here, smoking was highly associated with positive sputum microscopy results confirming the negative smoking impact.However, both smokers and patients with positive sputum microscopy results were more likely to achieve treatment outcome "cured" in comparison to the "treatment completed" or "lost to follow up" outcomes (p < 0.05).This result could be related to the fact that patients with positive sputum microscopy result were treated in the hospital until the negative sputum microscopy was confirmed, rather than received ambulatory treatment for TB, thus achieving better treatment compliance.These findings could have some implications for TB eradication programmes.NAT2 enzyme plays a key role in biotransformation of xenobiotics including first-line anti-TB drug INH (Mthiyane et al., 2020).The NAT2 gene is highly polymorphic, and more than 65 variants containing one or more SNPs in the 870 bp long coding region have previously been reported (Hein and Doll, 2012); in turn, the state of NAT2 activity in an individual depends on combinations of rapid and slow alleles (Hein and Doll, 2012;Mthiyane et al., 2020).Worldwide, significant differences in the distribution and frequency of NAT2 alleles between ethnic populations exist; the overall proportions of the acetylator phenotypes in our study cohort was 68% for SA and 32% for IA carriers, similar to those reported in Caucasian populations in Europe and the United States (Sabbagh et al., 2011;Bisso-Machado et al., 2016).
Data from our study confirmed previous reports reviewed in (Hong et al., 2020) showing that INH and its metabolites AcINH and INA blood plasma levels were significantly different between the NAT2 slow and intermediate acetylators: INH AUC 0-6h and INH C max were significantly higher, and AUC 0-6h values for both metabolites were significantly lower in the SA group in comparison to the IA group; consequently, the median AcINH/ INH and INA/INH ratio values were also significantly lower in the SA group.Also, significant NAT2 inter-phenotypic differences in INH and metabolites correlation patterns were detected.
According to the proposed INH metabolism pathway, INH, Hz and AcHz are metabolized by NAT2 enzyme and thus are subjects to the same acetylation polymorphism (Balhara et al., 2021).Three metabolites have been proposed to be responsible for INH-induced liver injury, AcHz, Hz and a metabolite resulting from the bioactivation of INH itself (Metushi et al., 2011).In our study, AcINH AUC 0-6h values were significantly lower in patients with DIH suggesting impaired INH acetylation in these cases; in addition, slow acetylation of toxicologically active AcHz to inactive diAcHz could be expected.The frequency of DIH observed in our cohort (11.76%), as well as the association of a slow acetylation profile with DIH, are consistent with the data reported previously (Sistanizad et al., 2011;Perwitasari et al., 2018;Khan et al., 2019).In our small cohort, statistical significance was reached for this INH no other INH pharmacogenetic parameters reached statistical significance, thus these results indicated that AcINH plasma level could serve as a biomarker for DIH.GSTs are important phase II detoxification isozymes that catalyse the conjugation of reduced form of glutathione to toxic metabolites including in the INH metabolic pathway, thereby reducing their reactivity toward cellular macromolecules (Sodhi et al., 1996;Eaton and Bammler, 1999).GST genetic polymorphisms, especially the variants of GSTM1 gene, have been extensively studied, and the absence of GSTM1 activity due to null mutation was associated with accumulation of toxic intermediates leading to DIH in some studies (Wang et al., 2016).The proportion of the GSTM1-null genotype in our Caucasian cohort was 59%; however, we did not observe any effects of GSTM1 genotype on DIH, TB treatment outcome, or any of the six INH pharmacokinetic parameters studied.
CYP2E1 is another enzyme involved in the Phase I biotransformation reactions of INH (Cheng et al., 2013).Previous studies on the pharmacogenetic variability of CYP2E1 have resulted in evidence-based clinical annotations of specific genetic polymorphisms and corresponding allelic variants with consequent different phenotypes in regulatory pathways that effect the pharmacokinetics and treatment response (https://cpicpgx.org/guidelines), but controversial associations of CYP2E1 genotype with risk of INH-related hepatotoxicity were reported (Lei et al., 2021).In our study, statistically significant associations between CYP2E1*6 (T > A) allelic variant (rs6413432) and three INH pharmacokinetic parameters, i.e., increase in INH AUC 0-6h , and decrease in both AcINH/INH and INA/INH ratios, were detected; however, no direct proofs for contribution in either DIH or treatment outcome were observed.None of other detected SNPs, including two allelic variants CYP2E1*4 (rs6413419) and CYP2E1*7A (rs2070673) which were previously associated with potential effects on CYP2E1 enzymatic activity, showed any statistically significant associations.
CYP2E1*6 allelic variant, recognized by Dra I restriction enzyme, is located in intron 6 of the gene (Fang et al., 2017).Previously, it was suggested that this polymorphism is associated with the CYP2E1 gene expression at mRNA level (Uematsu et al., 1994).Our data indicate that CYP2E1*6 was related to the decreased amounts of AcINH and INA; however, the mechanism of interaction between CYP2E1 genetic polymorphisms and INH plasma levels is unclear.Clinical studies have reported the interactions of INH with CY2E1 substrates including observations of interphenotypic difference for INH-related inhibition of CYP2E1 activity, which was more extensive in slow acetylators (O'Shea et al., 1997).However, in vitro studies did not observe any significant inhibition of CYP2E1 by INH or its metabolites (Balhara et al., 2021).Also, treatment of drugsusceptible TB involves several drugs, including RIF, which is a CYP2E1 inducer, thus drug-drug interactions could potentially interfere with INH metabolism or manifestation of INH-related  adverse drug reactions through increased production of reactive metabolites, especially in slow acetylators (Shen et al., 2008;Yamada et al., 2009).Additional studies are required to determine the possible functional implications of the CYP2E1 SNVs by assessing the drug response and clinical manifestations of the identified variants.
Another objective of our study was to determine the possible relationship between INH pharmacokinetic data and TB treatment outcome.First, it should be noted that, in our study, INH C max at 2 h time point was below the therapeutic reference range in the majority of the patients (64.7%).Considering the overall good treatment success rates, our findings support the opinion that the current therapeutic ranges are probably set too high (Perumal et al., 2020).Further, our results showed that INA/INH ratios were significantly higher in patients with SCC time less than 60 days.This assertion was further supported by the findings showing that a one unit increase of the standardized AcINH/INH ratio value gave 58% probability to achieve SCC first.Thus, it could be concluded, that for a patient to have a probability of healing first, the metabolic rate of the standard dose INH should be sufficiently high, while nonefficient INH metabolic rate is not favourable.Also, quite surprisingly, our results from the time-to-event analysis showed that intermediate acetylators recovered more rapidly than slow acetylators; however, the results of univariate analysis did not reach statistically significant difference between NAT2 phenotypes and tSCC.Somewhere controversially, several studies have shown that low INH C max and/or AUC values in rapid acetylators were associated with longer tSCC and a higher likelihood of failure or relapse; however, other studies, and also this study, found no association between the INH concentrations and treatment outcomes (Sekaggya-Wiltshire, von Braun, et al., 2018).INH plasma half-life is approximately 70 min in fast and 2-5 h in slow acetylators (Goodman et al., 2006), thus INH C max measurements at a single 2 h time point may not fully reflect INH exposure for all patients.Thus, in our study, plasma concentrations of AcINH and INA were measured, and INH metabolic ratio values were used in addition to INH C max and AUC values, which could provide more accurate assessments of the data.
On the other hand, inconsistency of the results may be due to the variability of the study population, the involvement of different factors, and the study design.TB treatment outcome should be considered as multifactorial event, considering multiple factors such as bacillary load, M. tuberculosis strain, individual measurement or local population estimate of the infecting strain minimum inhibitory concentration (MIC) values, treatment adherence, insufficient drug absorption, drug-drug interactions, drug concentration at the site of TB localization, age, gender, comorbid conditions, and immune and nutritional status of the patient (Colangeli et al., 2018;Perumal et al., 2020).Importantly, it was reported recently that INH concentrations were 14.6-fold higher in lung/airway epithelial lining fluid than blood plasma, thus accumulation of drugs at the site of disease may explain the success of the first-line regimen (McCallum et al., 2021).
Our study had limitations.First, the number of cohort participants was relatively small, and we were not able to include individuals with fast NAT2 phenotype.Second, the study was conducted in the Latvian population with European ancestry, and the obtained data were not extrapolated to other ethnicities.Also, we were not able to include all possible patient-, disease-, and treatment-related factors, and there is a possibility that the statistical power of the analysis to identify all genetic polymorphisms/SNPs was not sufficient.

Conclusion
In conclusion, the results of our study brought closer the research on anti-TB drug pharmacology and its application in clinical practice.Our findings suggest that CYP2E1, but not GSTM1, significantly affects the INH pharmacokinetics along with NAT2.Also, significant associations between INH pharmacokinetic parameters and both incidence of hepatotoxicity and month two sputum culture positivity were observed.AcINH plasma level could serve as a biomarker for INH-related hepatotoxicity.These findings could have further implications in promotion of the concept of precision medicine in TB management.The inclusion of INH metabolite screening in TB therapeutic drug monitoring could be beneficial in clinical studies for determination of optimal dosing strategies.
1A).Based on the preceding information, the following observations were found for the whole cohort: the odds of an individual with a given standardized MR (AcINH/INH) value to reach SCC first compared to an individual with one unit lower standardized MR value were 1.4 (95% CI, 1.00-2.1),p = 0.052; a one unit increase of the standardized AcINH/INH ratio value gave HR/ (1 + HR) = 58% probability of healing first.There was no significant effect of other predictors on HR of treatment success.Global log-rank p-value of the Cox regression model incorporating age, sex, smear microscopy, and standardized values of INA/INH ratio was p = 0.145 and concordance index was 0.65; the estimated hazard rate (HR) ratios are illustrated (Figure1B).The odds of an individual with a given standardized MR (INA/INH) value to reach SCC first compared to an individual with one unit lower standardized MR value were 1.56 (95% CI, 1.07-2.3),p = 0.02; a one unit increase of the standardized INA/INH ratio value gave HR/(1+HR) = 61% probability of healing first.

FIGURE 1
FIGURE 1 Forest plot demonstrating the hazard rate ratios of successful time to sputum culture conversion obtained from fitted Cox proportional hazards model.Statistically significant p-value is marked with asterisk (*).(A).Interactions between age, sex and sputum microscopy results were used along with the AcINH/INH metabolic ratio.(B) Interactions between age, sex and sputum microscopy results were used along with the INA/INH metabolic ratio (C) Interactions between NAT2 acetylator status, GSTM1 genotype and the presence of CYP2E1*6 (T>A) allelic variant (rs6413432) were used along with the sputum microscopy results.NAT2, N-acetyltransferase 2; GSTM1, glutathione S-transferase mu 1 class; MR, metabolic ratio; INH, isoniazid; AcINH, N-acetylisoniazid; INA, isonicotinic acid.
The study was approved by the Central Medical Ethics Committee of Latvia (approval No. 01-29.1/1and approval No. 01-29.1.2/1736),the Ethics Committee of the Riga East University Hospital (approval No 24-A/15 and approval No. A1/1.1-07.1/20/1809), and the Riga Stradins University Research Ethics Committee Tools/LD?db=core) utilizing genotyping data from the 1000 Genomes Project database for SNP pairwise comparisons through the r 2 coefficient.All the SNPs were filtered using a pairwise tagging algorithm with high LD r 2 measure ≥ 0.8 and minor allele frequency (MAF) > 0.05.All detected SNPs genotypes were tested for agreement with Hardy-Weinberg equilibrium (HWE) by Fisher exact probability test based on the genotype distributions in this study.As a result, seven SNPs were chosen as the representatives to analyse the association with patients' clinical characteristics.p-values ≤ 0.05 were considered statistically significant.

Table 2 )
. Global log-rank p-value of the Cox regression model incorporating age, sex, smear microscopy, and standardized values of AcINH/INH ratio was p = 0.227 and concordance index was 0.6; the estimated hazard rate (HR) ratios are

TABLE 2
Treatment outcome, response and hepatotoxicity in patients with TB (n = 34) depending on various factors.

TABLE 2 (
Continued) Treatment outcome, response and hepatotoxicity in patients with TB (n = 34) depending on various factors.<18.5 kg/m 2 ; normal, 18.5-24.9kg/m 2 ; overweight, 25.0-29.9kg/m 2 .For categorical variables, data are presented as frequencies (%), equality of proportions between groups was tested using Chi-square test (or Fisher's exact test where Chi-square was not applicable) and is reported using p-values.Odds ratio (95% CI) was calculated.For continuous normally distributed variables, data are presented as means (± SD), comparison of means was performed by Student's t-test or one-way ANOVA between two or three groups, respectively; AcINH and INA was log-transformed before running the tests.Models with rank tranformations were used for non-normally distributed variables (AcINH/INH and INA/INH), and data are presented as medians (IQR); the comparison of medians was performed by Mann-Whitney-Wilcoxon or Kruskal-Wallis test between two or three groups, respectively.p-values ≤ 0.05 are highlighted in bold.BMI, body mass index; tSCC, time to sputum culture conversion; INH, isoniazid; AcINH, N-acetylisoniazid; INA, isonicotinic acid; AUC, area under the concentration-time curve; C max , maximum concentration; MR, metabolic ratio; NAT2, N-acetyltransferase 2; SA, slow acetylator; IA, intermediate acetylator; GSTM1, glutathione S-transferase mu 1 class; SD, standard deviation; IQR, interquartile range; OR, odds ratio.
*Two patients were lost to follow-up.**Sputum culture conversion was confirmed in 28 patients.***Body mass index groups: underweight,

TABLE 3
Pharmacokinetic parameters for isoniazid stratified by patient characteristics.
For normally distributed variables, data are presented as means (± SD); AcINH and INA AUC 0-6h values were log-transformed before running the tests.For normally distributed variables, comparison of means was performed by Student's t-test or one-way ANOVA between two or three groups, respectively.Models with rank tranformations were used for non-normally distributed variables (i.e.AcINH/INH and INA/INH ratios), and data are presented as medians (IQR).Comparison of medians was performed by Mann-Whitney-Wilcoxon or Kruskal-Wallis test between two or three groups, respectively.INH, isoniazid; AcINH, N-acetylisoniazid; INA, isonicotinic acid; AUC, area under the concentration-time curve; C max , maximum concentration; BMI, body mass index; SD, standard deviation; IQR, interquartile range.*Underweight <18.5 kg/m 2 ; normal 18.5-24.9kg/m 2 ; overweight 25.0-29.9kg/m 2 .

TABLE 4 (
Continued) Pharmacokinetic parameters for isoniazid stratified by NAT2 phenotype, GSTM1 genotype, and CYP2E1 SNP (rs) variants.For the analysis, six pharmacokinetic parameters of INH were adjusted for age, sex, and dose (mg/kg).For normally distributed variables, data are presented as marginal means; comparison of the means was performed by Student's t-test.AcINH and INA were log-transformed before running the tests.Models with rank tranformations were used for non-normally distributed variables (AcINH/INH and INA/INH ratios), however, the results did not differ from the unadjusted analysis, and therefore are presented as medians (IQR); the comparison of medians was performed by Mann-Whitney-Wilcoxon test between two groups.p-values < 0.05 are highlighted in bold.INH, isoniazid; AcINH, N-acetylisoniazid; INA, isonicotinic acid; AUC, area under the concentration-time curve; Cmax, maximum concentration; MR, metabolic ratio; NAT2, N-acetyltransferase 2; SA, slow acetylator; IA, intermediate acetylator; GSTM1, glutathione S-transferase mu 1 class; CYP2E1, cytochrome P450, 2E1 class; SNP, single nucleotide polymorphism; EMM, Estimated Marginal Means; SE, standard error; CI, Confidence Interval; IQR, interquartile range.

TABLE 5 (
Continued) Treatment outcome and hepatotoxicity in patients with TB (n = 34) stratified by NAT2 phenotype, GSTM1 genotype, and CYP2E1 SNP (rs) variants.Two patients were lost to follof up and were omitted from calculations.**Sputum culture conversion was observed in 28 patients.For categorical variables, data are presented as frequencies (% of the total), equality of proportions between groups was tested using Chi-square test (or Fisher's exact test where Chi-square was not applicable), and is reported using p-values.Odds ratio (95% CI) was calculated.OR 1 , NAT2 SA over IA; OR 2 , GSTM1 null over plus.NAT2, N-acetyltransferase 2; SA, slow acetylator; IA, intermediate acetylator; GSTM1, glutathione S-transferase mu 1 class; CYP2E1, cytochrome P450, 2E1 class; SNP, single nucleotide polymorphism.tSCC, time to sputum culture conversion; OR, odds ratio; na, not applicable.