Pharmacogenetics of tuberculosis treatment toxicity and effectiveness in a large Brazilian cohort

Gustavo Amorim, James Jaworski, Marcelo Cordeiro-Santos, Afrânio L. Kritski, Marina C. Figueiredo, Megan Turner, Bruno B. Andrade, Digna R. Velez Edwards, Adalberto R. Santos, Valeria C. Rolla, Timothy R. Sterling, David W. Haas, the Regional Prospective Observational Research in Tuberculosis (RePORT)-Brazil network
doi: https://doi.org/10.1101/2023.08.30.23294860

Abstract

Background Genetic polymorphisms have been associated with risk of anti-tuberculosis treatment toxicity. We characterized associations with adverse events and treatment failure/recurrence among adults treated for tuberculosis in Brazil.

Methods Participants were followed in Regional Prospective Observational Research in Tuberculosis (RePORT)-Brazil. We included persons with culture-confirmed drug-susceptible pulmonary tuberculosis who started treatment between 2015-2019, and who were evaluable for pharmacogenetics. Treatment included 2 months of isoniazid, rifampin or rifabutin, pyrazinamide, and ethambutol, then 4 months of isoniazid and rifampin or rifabutin, with 24 month follow-up. Analyses included 43 polymorphisms in 20 genes related to anti-tuberculosis drug hepatotoxicity or pharmacokinetics. Whole exome sequencing was done in a case-control toxicity subset.

Results Among 903 participants in multivariable genetic association analyses, NAT2 slow acetylator status was associated with increased risk of treatment-related grade 2 or greater adverse events, including hepatotoxicity. Treatment failure/recurrence was more likely among NAT2 rapid acetylators, but not statistically significant at the 5% level. A GSTM1 polymorphism (rs412543) was associated with increased risk of treatment-related adverse events, including hepatotoxicity. SLCO1B1 polymorphisms were associated with increased risk of treatment- related hepatoxicity and treatment failure/recurrence. Polymorphisms in NR1/2 were associated with decreased risk of adverse events and increased risk of failure/recurrence. In whole exome sequencing, hepatotoxicity was associated with a polymorphism in VTI1A, and the genes METTL17 and PRSS57, but none achieved genome-wide significance.

Conclusions In a clinical cohort representing three regions of Brazil, NAT2 acetylator status was associated with risk for treatment-related adverse events. Additional significant polymorphisms merit investigation in larger study populations.

Background

Tuberculosis is a leading cause of death from a single infectious agent worldwide.[1] Standard treatment for drug-sensitive tuberculosis involves two months of isoniazid, rifampin, pyrazinamide, and ethambutol, then four months of isoniazid and rifampin.[2, 3] Among patients who adhere to tuberculosis treatment in clinical trials, cure rates are as high as 97%.[4] However, in real-world settings only an estimated 86% have successful outcomes.[1] Factors that limit success include adverse drug reactions (ADR), acquired drug resistance during treatment, premature treatment discontinuation, and loss to follow-up.[5] Hepatotoxicity is the most frequent ADR with tuberculosis treatment, with reported rates from 2% to 28%,[6] with even higher rates among people with HIV (PWH).[7–9].

Treatment failure affects fewer than 5% of patients.[1, 2] While tuberculosis treatment is considered to be effective among PWH, death during treatment and loss to follow-up may be more frequent.[1, 10–12] Another concern is tuberculosis recurrence, whereby patients who complete treatment subsequently develop tuberculosis due to relapse or exogenous reinfection. Following standard therapy, approximately 4% of patients experience relapse or require retreatment within two years.[13, 14] Recurrence is reported to be more likely in PWH, particularly due to reinfection, and especially before availability of effective antiretroviral therapy (ART).[15–17]

Isoniazid is metabolized by N-acetyltransferase 2 [18], and NAT2 loss-of-function alleles are frequent [18–20]. One or two copies of such alleles confer intermediate or slow acetylator phenotypes, respectively, and progressively greater isoniazid exposure [18–20]. Slow acetylator alleles have been reported to increase the risk for isoniazid hepatotoxicity [21–23].

Polymorphisms in additional genes have been reported to affect isoniazid and other anti- tuberculosis drugs, including SLCO1B1 with rifampin [24], and GSTM1 and CYP2E1 with isoniazid [25], although less well replicated than for NAT2.

Improved understanding of relationships between human genetics and tuberculosis treatment toxicity and effectiveness could inform treatment strategies, which might include point- of-care pre-treatment genotyping. We characterized associations between selected genetic variants, treatment toxicity and effectiveness in a large, prospective cohort study in Brazil.

Methods

Study design and population

Participants were from Regional Prospective Observational Research in Tuberculosis (RePORT) Brazil, a cohort study of individuals with newly diagnosed, culture-confirmed, pulmonary tuberculosis. Between June 2015 and June 2019, participants were enrolled at five sites across three regions in Brazil - Rio de Janeiro, Salvador, and Manaus. RePORT-Brazil is broadly representative of tuberculosis cases in Brazil.[26] For this study, we included RePORT- Brazil participants with drug-susceptible tuberculosis on standard tuberculosis therapy, all of whom provided samples for genetics. Participants were followed for 24 months to assess treatment response and recurrence. RePORT-Brazil excluded individuals who previously received anti-tuberculosis therapy for ≥7 days, received >7 days of fluoroquinolone therapy within 30 days prior to enrollment, were pregnant or breastfeeding, or planned to leave the region during follow-up.

Data collection and definitions

Clinical, demographic, laboratory (including sputum mycobacterial culture), and outcome data were collected at four study visits per RePORT-Brazil protocol: at treatment initiation (baseline), month 1 and month 2 after treatment initiation, and at end of treatment, which was typically 6 months after initiation. Standard therapy included isoniazid, rifampicin or rifabutin, pyrazinamide, and ethambutol for two months, followed by isoniazid and rifampicin or rifabutin for four months. Telephone follow-up occurred every 6 months thereafter until 24 months.

Participants were tested for HIV at baseline unless already known to be positive. Diabetes status at baseline was based on self-report. Directly observed therapy (DOT) was prescribed at treatment initiation.

Outcomes of interest included tuberculosis treatment toxicity and effectiveness (i.e., treatment failure and recurrence). Treatment failure was defined as remaining sputum culture- positive or smear-positive at month 5 or later during treatment. Recurrence was defined as culture-confirmed tuberculosis or symptoms consistent with tuberculosis after apparent cure or treatment completion. Primary toxicity outcomes were any grade ≥2 ADR that was considered at least possibly related to tuberculosis treatment, and grade ≥2 hepatotoxicity that was considered at least possibly related to tuberculosis treatment. We also considered grade ≥3 ADR related to TB treatment. Categories of physician-assigned attribution of causality were “possibly”, “likely”, or “definitely” related to tuberculosis treatment. Secondary analyses considered any adverse event.

Cases selected for whole exome sequencing were defined as having grade ≥3 treatment-related toxicity, permanent drug discontinuation due to treatment toxicity, or grade ≥2 treatment-related hepatotoxicity. Controls were matched on disease risk score, which was derived from age, sex, self-reported race, HIV status, NAT2 acetylator group, and CYP2B6 metabolizer group (if on efavirenz). Models are adjusted for sex, age, and 4 PCs.

Genetic Polymorphisms

Human DNA was extracted from whole blood. We genotyped 45 selected polymorphisms in 21 genes relevant to anti-tuberculosis drug-induced hepatotoxicity (ABCB1, ABCC2, AGBL4, BACH1, CUX2, CYP2E1, CYP7A1, GSTM1, GSTP1, IL6, MAFK, NAT2, NFKB1, NOS2, NR1I2, RIPOR2, SOD1, TGFBRAP1, and XPO1), or pharmacokinetics of isoniazid (NAT2) and rifampin (SLCO1B1). A list of all genotyped polymorphisms, associated phenotypes, and selected references are provided in Supplemental Table S1). Genotyping was done at VANTAGE (Vanderbilt Technology for Advanced Genomics) using MassARRAY® iPLEX Gold (Agena Bioscience™, California, USA) and Taqman (ThermoFisher Scientific, Massachusetts, USA). Assay design is available upon request. We excluded from single variant analyses 2 polymorphisms with minor allele frequencies <5% (rs78872571 in NFKB1, rs2070401 in BACH1), as well as 4 NAT2 polymorphisms used to define NAT2 acetylator status (explained under Statistical analysis). Mean genotyping efficiency was 98.8%, and exceeded 95% for all but one polymorphism, which was 93.4%. In this ancestrally diverse cohort, Hardy- Weinberg P-values were not used to exclude polymorphisms, but rather considered in interpreting association results. Analyses ultimately characterized associations with 43 polymorphisms in 20 genes, including 4 polymorphisms which defined NAT2 acetylator status (Supplemental Table S1).

Ancestry-informative markers

A set of 46 autosomal ancestry-informative markers (AIMs), selected based on a previous analysis of a diverse Brazilian cohort,[27] were genotyped at the Laboratório de Biologia Molecular Aplicada a Micobactérias in Fiocruz, using xMAP multiplexing (Luminex Corporation). These AIMs were used to determine percentage ancestry as described elsewhere [28].

Whole exome sequencing

A subset of DNA specimens was whole exome sequenced on the Illumina NovaSeq 6000 platform and processed using the DRAGEN Germline application (v.3.8.4) within Illumina’s BaseSpace Sequence Hub (Illumina Inc., San Diego, CA). Gene boundaries were defined using Twist_Exome_plus_RefSeq_GenCode_targets_hg38.bed. The gVCF files were joint-called using the DRAGEN Joint Genotyping Pipeline (v.3.8.4). The joint called VCF file was hard- filtered with a quality score threshold of 10.4139 (error probability <0.01). Quality control of the joint called VCF was done using PLINK2 software [29]. Samples were filtered for call rate (98%) and checked for relatedness and sex assignment. A total of 203,337 bi-allelic single nucleotide variants and short indels were processed, of which 5,772 were novel. This covered 30,466 overlapped genes and 170,658 overlapped transcripts. Variants were annotated using the Ensembl variant effect predictor. To calculate principal components (PCs) using PLINK2, a set of variants present in 1000 Genomes Project phase3 were selected and pruned based on linkage disequilibrium.

Statistical analysis

NAT2 genotypes were categorized based on combinations of rs1801280 (NAT2*5), rs1799930 (NAT2*6), rs1799931 (NAT2*7), and rs1801279 (NAT2*14): as slow if homozygous for the variant allele at any locus or heterozygous at 2 or more loci; intermediate if heterozygous at a single locus; or rapid if no variant allele [30]. Other individual polymorphisms were analyzed as ordinal variables assuming additive genetic effect. CYP2B6 metabolizer group was defined based on combinations of three polymorphisms (rs3745274, rs28399499, rs4803419) as described elsewhere [31]. We did not correct for multiple comparisons since only functionally relevant candidate genes or polymorphisms were included. Two-sided tests were used.

Logistic regression analyses assessed associations between outcomes and NAT2. To decrease the likelihood of overfitting the models, we used two approaches to adjust for covariates: standard multivariable logistic regression and propensity score-adjusted logistic regression. In fully adjusted models, we adjusted for seven covariates: baseline age, sex, HIV status, hemoglobin A1c (HbA1c) percent, percent African ancestry, percent European ancestry, and prescribed DOT. For outcomes with relatively few events, we used logistic regression with propensity score weights to avoid overfitting [32]. Propensity score models, which are also used as variable reduction techniques in regression models,[33] were used to regress NAT2 on pre- specified covariates: baseline age, BMI, sex, HIV status, HbA1c percent, prescribed DOT, history of alcohol, tobacco, and drug use. We used multinomial regressions to estimate propensity scores for each patient and outcome. Propensity scores were later used as weights in primary outcome models that regressed outcomes on NAT2. Of note, some categories of NAT2 had very few or no events, a situation denoted as complete separation.[34] Standard logistic regression has numerical issues in such settings and cannot be applied. An alternative approach is logistic regression with Firth correction,[35, 36] which reduces the small sample bias of maximum likelihood estimates by maximizing a penalized log-likelihood while also providing a solution to separation.[37] We applied Firth correction to propensity score-weighted logistic regression to obtain estimates for parameters of interest. Since Wald-type confidence intervals may lead to poor coverage with small samples, we used profile likelihood to obtain 95% confidence intervals in Firth-corrected propensity score-weighted logistic regression. Similar analyses were performed using all grade ≥2 and all grade ≥3 ADR, as well as all grade ≥2 hepatic events, regardless of whether they were treatment-related.

We used statistical software R version 4.2.0 [38] and STATA version 17 (StataCorp, College Station, Texas, USA). Associations with the 43 polymorphisms were assessed using PLINK [29].

For whole exome sequencing, single variant and gene-based tests were performed using RVTESTS [39]. Single variant associations were tested on variants with call rates >98%, Hardy-Weinberg equilibrium p >10-7, and minor allele count >10. The model was adjusted for age, sex, and 4 PCs. Gene-based tests were performed with the option SKAT-O, which optimizes the generalized SKAT over a grid of values of Rho between 0 and 1, corresponding to a kernel SKAT test and a Burden test. The analysis was run using rare (minor allele frequency <0.05), coding, and loss-of-function variants. Gene boundaries were assigned using the Ensembl GRCh38.p13 genetic map. Tests were adjusted for age, sex, and 4 PCs.

Results

Characteristics of the cohort

A total of 963 RePORT-Brazil participants treated for culture-confirmed, drug-susceptible pulmonary tuberculosis were evaluable for ADR and microbiologic outcomes, of whom 937 received standard anti-tuberculosis therapy. After excluding participants in whom genotyping for NAT2 or AIMs failed, or who lacked baseline data on HIV status, HbA1c, or whether DOT was prescribed, 903 participants were included in genetic association analyses. Derivation of our cohort of 903 participants, stratified by HIV status, is presented in Figure 1.

Figure 1.
Figure 1. Derivation of the RePORT-Brazil genetic analysis cohort

Baseline characteristics of participants are presented in Table 1. Two-thirds were male, and 167 (18.5%) were PWH. Regarding NAT2, there were 77 (8.5%) rapid acetylators, 367 (40.6%) intermediate acetylators, and 459 (50.8%) slow acetylators. Median proportion genetic ancestry based on AIMs was 42% European, 28% African, and 21% Amerindian.

View this table:
Table 1. Baseline characteristics of RePORT-Brazil participants in genetic association analyses

Regarding adverse events, 125 (13.8%) had documented grade ≥2 adverse events during follow-up, including 83 (9.2%) considered tuberculosis treatment related, 49 (5.4%) had grade ≥3 adverse events, including 16 (1.8%) considered tuberculosis treatment related, and 25 (2.8%) had documented grade ≥2 hepatic adverse events. Regarding microbiologic outcomes, 37 (4.1%) experienced treatment failure/recurrence. The distribution of adverse event and microbiologic outcomes, stratified by HIV status, is provided in Table 2, and stratified by NAT2 group in Supplemental Table S2.

View this table:
Table 2. Tuberculosis treatment-related adverse drug reactions and treatment outcomes by HIV status

NAT2 acetylator status and adverse events

As noted in Methods, we used two approaches to assess the effect of NAT2 on outcomes of interest in a multivariable setting: standard multivariable logistic regression and Firth-corrected propensity score-adjusted models. Results for treatment-related adverse events (grade ≥2, grade ≥3, and hepatoxicity) are shown in Table 3. Associations with secondary outcomes (any grade ≥2 and grade ≥3 event) are shown in Supplemental Table S3.

View this table:
Table 3. Associations of NAT2 acetylator status with tuberculosis treatment-related adverse drug reactions and treatment outcome in multivariable models

Regarding grade ≥2 treatment-related adverse events, both Firth-corrected propensity score-weighted models and standard multivariable logistic regression showed that NAT2 slow acetylators had higher odds of adverse events compared to rapid acetylators: OR = 2.21 (95% CI = [1.01, 7.71]) and OR = 2.32 (95% CI = [0.79, 6.77]), respectively (Table 3). We did not detect differences between intermediate and rapid acetylators. Both approaches had similar point estimates, but different confidence intervals, highlighting differences between Likelihood- profile and Wald-type confidence intervals, and how the latter is affected by finite sample approximations.

Results were similar regarding grade ≥3 treatment-related adverse events. Firth- corrected propensity-score weighted models showed both NAT2 intermediate and slow acetylators had higher odds of ADR, compared to rapid acetylators: OR = 3.60, 95% CI = [1.92, 6.16] and OR = 3.38, 95% CI = [1.70, 5.66], respectively (Table 3). Because there were no events among rapid acetylators (i.e., complete separation; see Supplemental Table S2), standard multivariable logistic regression could not be computed.

Regarding treatment-related grade ≥2 hepatic adverse events, Firth-corrected propensity-score weighted models showed both NAT2 intermediate and slow acetylators had higher odds of ADR, compared to rapid acetylators: OR = 4.34, 95% CI = [2.12, 7.19] and OR = 7.03, 95% CI = [3.81, 11.25], respectively. Standard multivariable logistic regression could not be computed because there were no events in rapid acetylators (Table 3).

Similar results were observed when analyzing all ADRs, regardless of whether they were treatment-related. Compared to NAT2 rapid acetylators, slow acetylators had higher odds of any grade ≥2 ADR, while both intermediate and slow acetylators had higher odds of any grade ≥3 and any grade ≥2 ADR (Supplemental Table S3).

NAT2 acetylator status and treatment failure/recurrence

We hypothesized that risk of treatment failure/recurrence would be increased among NAT2 rapid acetylators due to lower isoniazid levels. We therefore combined NAT2 intermediate and slow acetylators as the reference group. In propensity score-weighted models, the odds ratio in rapid acetylators was approximately 1.51 (95% CI = [0.31; 3.55]). Similarly, in fully adjusted models for treatment failure/recurrence, the odds ratio in rapid acetylators was approximately 1.54 (95% CI = [0.57; 4.19]) (Table 3).

Genetic associations after adjusting for NAT2

To determine whether any of the 39 individual polymorphisms (not including the 4 polymorphisms that were used to define NAT2 acetylator group) in 20 genes were associated with ADR or treatment failure/recurrence outcomes, we performed multivariable logistic regression analyses of each polymorphism while adjusting for NAT2 acetylator group, baseline age, sex, HIV status, HbA1c percent, percent African ancestry, percent European ancestry, and prescription of DOT. For ADR we treated NAT2 as an ordinal variable with three levels (rapid, intermediate, and slow), while for failure/recurrence we treated it as two levels (rapid versus intermediate/slow). In these analyses, 4 polymorphisms in 4 genes (GSTM1, NAT2, NR1/2, and SLCO1B1) were associated with at least one treatment-related ADR category at p <0.05 (without correcting for multiple testing), while 3 polymorphisms in 2 genes (NR1I2 and SLCO1B1) and 1 long non-coding RNA (LINC01857) were associated with failure/recurrence. These associations are presented in Table 4.

View this table:
Table 4. Associations of selected polymorphisms with tuberculosis treatment-related adverse drug reactions and treatment outcome by multivariable models a

Whole exome associations

A total of 87 participants (43 cases and 44 controls) were included in whole exome sequence analyses. Cases had grade ≥3 toxicity at least possibly related to tuberculosis treatment, permanent drug discontinuation due to treatment toxicity, or grade ≥2 hepatotoxicity at least possibly related to treatment. One case that failed sex check quality control was excluded. In the single variant analysis, the lowest P-values was rs10991326 in ORC13C2 (p = 2.8 x 10-5). The three other lowest P-values were CPHXL rs882263 (p = 3.2 x 10-5), SULT1C3 rs6772745 (p = 4.6 x 10-5), and CHIA rs2820092 (p =8.7 x 10-5). The top gene-based result was WTAPP1 (p = 0.0008). In analysis limited to 37 cases with hepatotoxicity and 37 matched controls, the lowest p value was VTI1A rs17353359 (p =2.5 x 10-5), and the lowest p values for genes were METTL17 (p = 0.0007) and PRSS57 (p = 0.0008). None associations were significant after correcting for multiple testing.

Discussion

The present study involved 903 adults who received standard therapy for pulmonary tuberculosis during a prospective observational cohort study in Brazil. Compared to NAT2 rapid acetylators, slow acetylators were at increased risk for grade ≥2 treatment-related ADR, while both slow and intermediate acetylators were at increased risk for grade ≥3 treatment-related ADR, and for grade ≥2 treatment-related hepatic ADR. The association of NAT2 acetylator status with risk of hepatotoxicity during tuberculosis treatment was expected. Prior meta- analyses showed increased risk among slow acetylators [21–23, 40]. The largest such analysis, which included 38 studies with 2,225 patients and 4,906 controls, reported an odds ratio of 3.2 (95% CI: 2.5–4.1)[23]. In the present study, rapid acetylators had a higher odds ratio for failure/recurrence, but this was not statistically significant, possible due to the few events.

We determined NAT2 acetylator status based on four signature polymorphisms that identify NAT2 *5, *6, *7 and *14 loss-of-function alleles. We considered including rs1041983 based on a previous report suggesting its added utility in inferring slow acetylator alleles [41]. However, in that report, a four-polymorphism panel of rs1041983 plus *5, *6, and *7 signature polymorphisms performed no better than the three signature polymorphisms. Only with addition of the *14 polymorphism did performance improve [41]. In the present study, NAT2 polymorphisms that were not used to determine acetylator status (rs1041983, rs1208, and rs1799929) were analyzed individually in models that controlled for acetylator status.

In our cohort, standard tuberculosis therapy was generally effective and well-tolerated. Among the 903 participants, 4.1% had treatment failure/recurrence, 5.4% had grade ≥3 ADR (of which 32% were treatment-related), and 2.8% had grade ≥2 hepatic ADR (of which 88% were treatment-related). A recent study from Brazil reported that at least 11% of participants had a grade ≥3 ADR and 9% had liver and biliary system disorders.[42] A previous retrospective study in Brazil reported that 5-20% of individuals receiving standard tuberculosis therapy experienced minor/mild ADR, while 5% on isoniazid, rifampicin and pyrazinamide had hepatic transaminase elevations,[43] consistent with our study. In a recent tuberculosis treatment shorting trial, the standard treatment arm was associated with 1% failure, <1% recurrence, and 1% changed treatment due to ADRs, although nearly 20% experienced grade ≥3 ADRs, of which 3% were hepatic.[4]

In analyses that controlled for NAT2 acetylator status, as well as clinical covariates, eight additional polymorphisms in NAT2, GSTM1, NR1I2, SLCO1B1, as well as in a long non-coding RNAs (lncRNA), were associated with treatment-related ADR or failure/recurrence at P <0.05. Regarding ADR, the NR1I2 rs2461823 A allele, with a MAF of 47%, was associated with increased grade ≥3 ADR, which is generally consistent with a previous report that associated this allele with increased severity of nonalcoholic fatty liver disease (NAFLD) [44]. The NAT2 rs1799929 T allele, a tagging polymorphism with a MAF of 35%, was associated with decreased grade ≥2 ADR, but has not been previously associated with tuberculosis treatment outcomes. Similarly, the SLCO1B1 rs11045819 A allele, a tagging polymorphism with a MAF of 10%, was associated with increased grade ≥2 ADR, has not been previously associated with tuberculosis treatment outcomes. The GSTM1 rs412543 G allele, with a MAF of 7%, was associated with both grade ≥3 ADR and grade ≥2 hepatic ADR.

Regarding treatment failure/recurrence, the lncRNA (LINC01875 rs1055229) T allele, with a MAF of 24%, was associated with decreased treatment failure/recurrence, but a previous study associated this allele with increased tuberculosis drug-induced hepatotoxicity [45]. The NR1I2 rs7643645 G allele was associated with treatment failure/recurrence, but a previous report associated this with increased NAFLD severity [44]. The SLCO1B1 rs4149056 C allele, with a MAF of 13%, was associated with increased failure/recurrence (as was the linked SLCO1B1 rs2900478). While rs4149056 C allele increases plasma exposure of organic anion transporting polypeptide 1B1 substrates (e.g., statins), it is not known to affect anti-tuberculosis medications such as rifampin.

Whole exome sequencing provided interesting findings of nominal statistical significance. In single variant analyses, SULT1C3 rs6772745 was a top hit in both primary and secondary analyses. SULT1C3 encodes sulfotransferase 1C3, which has activity towards various substrates including hormones, neurotransmitters, drugs, and xenobiotics [46], although its mRNA has been detected only in human intestine, not in liver [47]. CPHXL rs882263 was the only top result with an effect in the negative direction. CPHXL is predicted to regulate transcription by RNA polymerase II. The top gene-based result, WTAPP1, is a pseudogene reported to affect cancer cell behavior and disease progression [48, 49].

This study had limitations. Only 1.8% of participants had grade ≥3 ADR, 2.4% had grade ≥2 hepatic ADR, and 4.1% had failure/recurrence. More cases would have increased power to detect associations. For most participants, we only had aggregated monthly DOT data, so could not reliably assess the impact of adherence. Because participants were receiving multiple medications, and we cannot attribute causation to specific medications.

In summary, NAT2 slow acetylator status was associated with increased risk for tuberculosis treatment-related ADR, including hepatic ADR, in a clinical cohort representing diverse ancestries from several regions of Brazil. The odds ratio for treatment failure/recurrence was higher among rapid acetylators. Polymorphisms in other genes (GSTM1, NR1I2, SLCO1B1) and in a lncRNA were independently associated with ADR and failure/recurrence. This study extends our understanding of the pharmacogenetics of tuberculosis therapy, and highlights the need for studies in larger cohorts. Studies should evaluate the effect drug dosing informed by host genetics on treatment outcomes.

Data Availability

All data produced in the present study are available upon reasonable request to the authors

Acknowledgments

We are grateful to Lauren S. Peetluk, Brian C. Hachey, and Leticia Linhares at Vanderbilt University Medical Center, and Edward P. Acosta at University of Alabama at Birmingham, for their work in support of R01 AI120790.

Footnotes

  • ↵17 Valeria C. Rolla, Maria Cristina S. Lourenço, Aline Benjamin, Quezia Medeiros de Oliveira, Felipe Ridolfi, Adriano Gomes da Silva, and Marcus Rockembach at Instituto Nacional de Infectologia Evandro Chagas, Fiocruz, Rio de Janeiro, Brazil; Solange Cavalcante, Betina Durovni, Jamile Garcia, Martinelle Godinho, and João Marine at Secretaria Municipal de Saúde do Rio de Janeiro e Clínica da Família Rinaldo Delamare, Rocinha, Rio de Janeiro, Brazil; Afranio Kritski, Adriana Moreira, André Luiz Bezerra, Anna Cristina Carvalho, and Elisangela Silva at Centro Municipal de Saúde de Duque de Caxias and Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil; Marcelo Cordeiro dos Santos, Alexandra Brito, Leandro Garcia, Renata Cordeiro dos Santos, Allyson Souza, and Jaquelane Silva at Fundação de Medicina Tropical, Manaus, Brazil; Bruno de Bezerril Andrade, Eduardo M. Netto, Michael Rocha, Vanessa Nascimento, Betânia Nogueira, Saulo R. N. Santos, and André Ramos at Instituto Brasileiro para Investigação da Tuberculose, Fundação José Silveira, Salvador, Brazil; Alice Andrade, Hayna Malta, Juan Cubillos, and Laise de Moraes at Laboratório de Inflamação e Biomarcadores, Instituto Gonçalo Moniz, Fundação Oswaldo Cruz, Salvador, Brazil; Bruno de Bezerril Andrade, Artur Trancoso, Kiyosh Fukutani, Mariana Araujo Pereira, Maria Arriaga, and Elze Leite at Multinational Organization Network Sponsoring Translational and Epidemiological Research (MONSTER) Initiative, Salvador, Brazil.

  • Funding: This work was supported by the Departamento de Ciência e Tecnologia–Secretaria de Ciência e Tecnologia–Ministério da Saúde, Brazil (25029.000507/2013-07); National Institute of Allergy and Infectious Diseases (NIAID) (U01 AI069923, R01 AI120790, U01 AI172064, R01 AI077505, P30 AI110527) and the National Center for Advancing Translational Science (UL1 TR000445) at the National Institutes of Health.

References

  1. 1.↵
    World Health Organization. Global Tuberculosis Report 2022. 2022. Available at: https://www.who.int/teams/global-tuberculosis-programme/tb-reports/global-tuberculosis-report-2022. Accessed January 30, 2023.
  2. 2.↵
    Nahid P, Dorman SE, Alipanah N, Barry PM, Brozek JL, Cattamanchi A, et al. Official American Thoracic Society/Centers for Disease Control and Prevention/Infectious Diseases Society of America Clinical Practice Guidelines: Treatment of Drug-Susceptible Tuberculosis. Clin Infect Dis. 2016;63(7):e147–e95.
    OpenUrlCrossRefPubMed
  3. 3.↵
    World Health Organization. Guidelines for treatment of drug-susceptible tuberculosis and patient care: 2017 update. Geneva, Switzerland. Available at: https://apps.who.int/iris/bitstream/handle/10665/255052/9789241550000-eng.pdf. Accessed September 20, 2022.
  4. 4.↵
    Dorman SE, Nahid P, Kurbatova EV, Phillips PPJ, Bryant K, Dooley KE, et al. Four-Month Rifapentine Regimens with or without Moxifloxacin for Tuberculosis. N Engl J Med. 2021;384(18):1705–18.
    OpenUrlCrossRefPubMed
  5. 5.↵
    Verbeeck RK, Gunther G, Kibuule D, Hunter C, Rennie TW. Optimizing treatment outcome of first-line anti-tuberculosis drugs: the role of therapeutic drug monitoring. Eur J Clin Pharmacol. 2016;72(8):905–16.
    OpenUrl
  6. 6.↵
    Tostmann A, Boeree MJ, Aarnoutse RE, de Lange WC, van der Ven AJ, Dekhuijzen R. Antituberculosis drug-induced hepatotoxicity: concise up-to-date review. J Gastroenterol Hepatol. 2008;23(2):192–202.
    OpenUrlCrossRefPubMedWeb of Science
  7. 7.↵
    Yimer G, Gry M, Amogne W, Makonnen E, Habtewold A, Petros Z, et al. Evaluation of patterns of liver toxicity in patients on antiretroviral and anti-tuberculosis drugs: a prospective four arm observational study in ethiopian patients. PLoS One. 2014;9(4):e94271.
    OpenUrl
  8. 8.
    Zaverucha-do-Valle C, Monteiro SP, El-Jaick KB, Rosadas LA, Costa MJ, Quintana MS, et al. The role of cigarette smoking and liver enzymes polymorphisms in anti-tuberculosis drug-induced hepatotoxicity in Brazilian patients. Tuberculosis (Edinb). 2014;94(3):299–305.
    OpenUrl
  9. 9.↵
    Araujo-Mariz C, Lopes EP, Acioli-Santos B, Maruza M, Montarroyos UR, Ximenes RA, et al. Hepatotoxicity during Treatment for Tuberculosis in People Living with HIV/AIDS. PLoS One. 2016;11(6):e0157725.
    OpenUrlPubMed
  10. 10.↵
    Sanchez M, Bartholomay P, Arakaki-Sanchez D, Enarson D, Bissell K, Barreira D, et al. Outcomes of TB treatment by HIV status in national recording systems in Brazil, 2003- 2008. PLoS One. 2012;7(3):e33129.
    OpenUrlCrossRefPubMed
  11. 11.
    Tweya H, Feldacker C, Phiri S, Ben-Smith A, Fenner L, Jahn A, et al. Comparison of treatment outcomes of new smear-positive pulmonary tuberculosis patients by HIV and antiretroviral status in a TB/HIV clinic, Malawi. PLoS One. 2013;8(2):e56248.
    OpenUrl
  12. 12.↵
    Karo B, Krause G, Hollo V, van der Werf MJ, Castell S, Hamouda O, et al. Impact of HIV infection on treatment outcome of tuberculosis in Europe. AIDS. 2016;30(7):1089–98.
    OpenUrl
  13. 13.↵
    Menzies D, Benedetti A, Paydar A, Martin I, Royce S, Pai M, et al. Effect of duration and intermittency of rifampin on tuberculosis treatment outcomes: a systematic review and meta-analysis. PLoS Med. 2009;6(9):e1000146.
    OpenUrlCrossRefPubMed
  14. 14.↵
    Johnston JC, Khan FA, Dowdy DW. Reducing relapse in tuberculosis treatment: is it time to reassess WHO treatment guidelines? Int J Tuberc Lung Dis. 2015;19(6):624.
    OpenUrl
  15. 15.↵
    Khan FA, Minion J, Pai M, Royce S, Burman W, Harries AD, et al. Treatment of active tuberculosis in HIV-coinfected patients: a systematic review and meta-analysis. Clin Infect Dis. 2010;50(9):1288–99.
    OpenUrlCrossRefPubMedWeb of Science
  16. 16.
    Chaisson RE, Churchyard GJ. Recurrent tuberculosis: relapse, reinfection, and HIV. J Infect Dis. 2010;201(5):653–5.
    OpenUrlCrossRefPubMed
  17. 17.↵
    Romanowski K, Balshaw RF, Benedetti A, Campbell JR, Menzies D, Ahmad Khan F, et al. Predicting tuberculosis relapse in patients treated with the standard 6-month regimen: an individual patient data meta-analysis. Thorax. 2019;74(3):291–7.
    OpenUrlAbstract/FREE Full Text
  18. 18.↵
    Huang YS, Chern HD, Su WJ, Wu JC, Lai SL, Yang SY, et al. Polymorphism of the N- acetyltransferase 2 gene as a susceptibility risk factor for antituberculosis drug-induced hepatitis. Hepatology. 2002;35(4):883–9.
    OpenUrlCrossRefPubMedWeb of Science
  19. 19.
    Roy PD, Majumder M, Roy B. Pharmacogenomics of anti-TB drugs-related hepatotoxicity. Pharmacogenomics. 2008;9(3):311–21.
    OpenUrlCrossRefPubMed
  20. 20.↵
    Ramachandran G, Swaminathan S. Role of pharmacogenomics in the treatment of tuberculosis: a review. Pharmgenomics Pers Med. 2012;5:89–98.
    OpenUrlCrossRefPubMed
  21. 21.↵
    Sun F, Chen Y, Xiang Y, Zhan S. Drug-metabolising enzyme polymorphisms and predisposition to anti-tuberculosis drug-induced liver injury: a meta-analysis. Int J Tuberc Lung Dis. 2008;12(9):994–1002.
    OpenUrlPubMed
  22. 22.
    Wang PY, Xie SY, Hao Q, Zhang C, Jiang BF. NAT2 polymorphisms and susceptibility to anti-tuberculosis drug-induced liver injury: a meta-analysis. The international journal of tuberculosis and lung disease : the official journal of the International Union against Tuberculosis and Lung Disease. 2012;16(5):589–95.
    OpenUrl
  23. 23.↵
    Cai Y, Yi J, Zhou C, Shen X. Pharmacogenetic study of drug-metabolising enzyme polymorphisms on the risk of anti-tuberculosis drug-induced liver injury: a meta-analysis. PLoS One. 2012;7(10):e47769.
    OpenUrlCrossRefPubMed
  24. 24.↵
    Chigutsa E, Visser ME, Swart EC, Denti P, Pushpakom S, Egan D, et al. The SLCO1B1 rs4149032 Polymorphism Is Highly Prevalent in South Africans and Is Associated with Reduced Rifampin Concentrations: Dosing Implications. Antimicrobial agents and chemotherapy. 2011;55(9):4122–7.
    OpenUrlAbstract/FREE Full Text
  25. 25.↵
    Forestiero FJ, Cecon L, Hirata MH, de Melo FF, Cardoso RF, Cerda A, et al. Relationship of NAT2, CYP2E1 and GSTM1/GSTT1 polymorphisms with mild elevation of liver enzymes in Brazilian individuals under anti-tuberculosis drug therapy. Clinica chimica acta; international journal of clinical chemistry. 2013;415:215–9.
    OpenUrl
  26. 26.↵
    Arriaga MB, Amorim G, Queiroz ATL, Rodrigues MMS, Araujo-Pereira M, Nogueira BMF, et al. Novel stepwise approach to assess representativeness of a large multicenter observational cohort of tuberculosis patients: The example of RePORT Brazil. Int J Infect Dis. 2021;103:110–8.
    OpenUrlPubMed
  27. 27.↵
    Saloum de Neves Manta F, Pereira R, Vianna R, Rodolfo Beuttenmuller de Araujo A, Leite Goes Gitai D, Aparecida da Silva D, et al. Revisiting the genetic ancestry of Brazilians using autosomal AIM-Indels. PLoS One. 2013;8(9):e75145.
    OpenUrlCrossRefPubMed
  28. 28.↵
    Alexander DH, Novembre J, Lange K. Fast model-based estimation of ancestry in unrelated individuals. Genome Res. 2009;19(9):1655–64.
    OpenUrlAbstract/FREE Full Text
  29. 29.↵
    Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. 2015;4:7.
  30. 30.↵
    Boukouvala S. Database of arylamine N-acetyltransferases (NATs). Democritus University of Thrace. 2016. Available at: http://nat.mbg.duth.gr/. Accessed September 28, 2022.
  31. 31.↵
    Holzinger ER, Grady B, Ritchie MD, Ribaudo HJ, Acosta EP, Morse GD, et al. Genome-wide association study of plasma efavirenz pharmacokinetics in AIDS Clinical Trials Group protocols implicates several CYP2B6 variants. Pharmacogenet Genom. 2012;22(12):858–67.
    OpenUrlCrossRef
  32. 32.↵
    Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70(1):41–55.
    OpenUrlCrossRefWeb of Science
  33. 33.↵
    D’Agostino RB, Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17(19):2265–81.
    OpenUrlCrossRefPubMedWeb of Science
  34. 34.↵
    Nemes S, Jonasson JM, Genell A, Steineck G. Bias in odds ratios by logistic regression modelling and sample size. BMC Med Res Methodol. 2009;9:56.
  35. 35.↵
    Firth D. Bias Reduction of Maximum Likelihood Estimates. Biometrika. 1993;80(1):27–38.
    OpenUrlCrossRefWeb of Science
  36. 36.↵
    Wang X. Firth logistic regression for rare variant association tests. Front Genet. 2014;5:187.
  37. 37.↵
    Heinze G, Schemper M. A solution to the problem of separation in logistic regression. Stat Med. 2002;21(16):2409–19.
    OpenUrlCrossRefPubMedWeb of Science
  38. 38.↵
    R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. . 2022. Available at: https://www.R-project.org/. . Accessed May 17, 2023.
  39. 39.↵
    Zhan X, Hu Y, Li B, Abecasis GR, Liu DJ. RVTESTS: an efficient and comprehensive tool for rare variant association analysis using sequence data. Bioinformatics. 2016;32(9):1423–6.
    OpenUrlCrossRefPubMed
  40. 40.↵
    Du H, Chen X, Fang Y, Yan O, Xu H, Li L, et al. Slow N-acetyltransferase 2 genotype contributes to anti-tuberculosis drug-induced hepatotoxicity: a meta-analysis. Mol Biol Rep. 2013;40(5):3591–6.
    OpenUrlCrossRef
  41. 41.↵
    Verma R, Patil S, Zhang N, Moreira FMF, Vitorio MT, Santos ADS, et al. A Rapid Pharmacogenomic Assay to Detect NAT2 Polymorphisms and Guide Isoniazid Dosing for Tuberculosis Treatment. Am J Respir Crit Care Med. 2021;204(11):1317–26.
    OpenUrl
  42. 42.↵
    Sant’Anna FM, Araújo-Pereira M, Schmaltz CAS, Arriaga MB, de Oliveira RVC, Andrade BB, et al. Adverse Drug Reactions Related to Treatment of Drug-Susceptible Tuberculosis in Brazil: A Prospective Cohort Study. Frontiers in Tropical Diseases. 2022;2.
  43. 43.↵
    Arbex MA, Varella Mde C, Siqueira HR, Mello FA. Antituberculosis drugs: drug interactions, adverse effects, and use in special situations. Part 1: first-line drugs. J Bras Pneumol. 2010;36(5):626–40.
    OpenUrlCrossRefPubMed
  44. 44.↵
    Sookoian S, Castano GO, Burgueno AL, Gianotti TF, Rosselli MS, Pirola CJ. The nuclear receptor PXR gene variants are associated with liver injury in nonalcoholic fatty liver disease. Pharmacogenet Genomics. 2010;20(1):1–8.
    OpenUrlCrossRefPubMedWeb of Science
  45. 45.↵
    Zhao Z, Peng W, Wu L, Ying B. Correlation between lncRNA AC079767.4 variants and liver injury from antituberculosis treatment in West China. J Infect Chemother. 2020;26(1):63–8.
    OpenUrl
  46. 46.↵
    Kurogi K, Shimohira T, Kouriki-Nagatomo H, Zhang G, Miller ER, Sakakibara Y, et al. Human Cytosolic Sulphotransferase SULT1C3: genomic analysis and functional characterization of splice variant SULT1C3a and SULT1C3d. J Biochem. 2017;162(6):403–14.
    OpenUrl
  47. 47.↵
    Duniec-Dmuchowski Z, Rondini EA, Tibbs ZE, Falany CN, Runge-Morris M, Kocarek TA. Expression of the orphan cytosolic sulfotransferase SULT1C3 in human intestine: characterization of the transcript variant and implications for function. Drug Metab Dispos. 2014;42(3):352–60.
    OpenUrlAbstract/FREE Full Text
  48. 48.↵
    Zhang L, Jin C, Yang G, Wang B, Hua P, Zhang Y. LncRNA WTAPP1 promotes cancer cell invasion and migration in NSCLC by downregulating lncRNA HAND2-AS1. BMC Pulm Med. 2020;20(1):153.
    OpenUrl
  49. 49.↵
    Deng J, Zhang J, Ye Y, Liu K, Zeng L, Huang J, et al. N(6) -methyladenosine-Mediated Upregulation of WTAPP1 Promotes WTAP Translation and Wnt Signaling to Facilitate Pancreatic Cancer Progression. Cancer Res. 2021;81(20):5268–83.
    OpenUrlAbstract/FREE Full Text
Back to top
Posted September 01, 2023.
Download PDF

Supplementary Material

Data/Code
Email
Pharmacogenetics of tuberculosis treatment toxicity and effectiveness in a large Brazilian cohort
Gustavo Amorim, James Jaworski, Marcelo Cordeiro-Santos, Afrânio L. Kritski, Marina C. Figueiredo, Megan Turner, Bruno B. Andrade, Digna R. Velez Edwards, Adalberto R. Santos, Valeria C. Rolla, Timothy R. Sterling, David W. Haas, the Regional Prospective Observational Research in Tuberculosis (RePORT)-Brazil network
medRxiv 2023.08.30.23294860; doi: https://doi.org/10.1101/2023.08.30.23294860
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area