Pharmacogenetics of HIV therapy: State of the art in Latin American countries

Abstract The use of combined antiretroviral therapy (cART) has resulted in a remarkable reduction in morbidity and mortality of people living with HIV worldwide. Nevertheless, interindividual variations in drug response often impose a challenge to cART effectiveness. Although personalized therapeutic regimens may help overcome incidence of adverse reactions and therapeutic failure attributed to host factors, pharmacogenetic studies are often restricted to a few populations. Latin American countries accounted for 2.1 million people living with HIV and 1.4 million undergoing cART in 2020-21. The present review describes the state of art of HIV pharmacogenetics in this region and highlights that such analyses remain to be given the required relevance. A broad analysis of pharmacogenetic markers in Latin America could not only provide a better understanding of genetic structure of these populations, but might also be crucial to develop more informative dosing algorithms, applicable to non-European populations.


Introduction
Since HIV discovery in the early 1980`s, more than 20 antiretrovirals (ARVs) have been developed, and combined antiretroviral therapy (cART) has provided a remarkable decrease in mortality and morbidity of people living with HIV/Aids (PLHA), turning HIV infection into a chronic manageable condition (Palella et al., 1998;Tseng et al., 2015).
Along with the development of new ARVs, several studies have been conducted to investigate their pharmacokinetics and possible impacts of genetic variations on treatment efficacy and safety.Indeed, polymorphisms in genes from ADME class (absorption, distribution, metabolism and excretion) have been consistently associated to cART safety and efficacy on virological control and CD4 recovery (Michaud et al., 2012;Mattevi and Tagliari, 2017;Yu et al., 2021).In addition to ADME class, genes related to immunological response have also been investigated in the context of CD4 recovery and hypersensitivity reactions (Mallal et al., 2008;Chaponda and Pirmohamed 2011;Rajasuriar et al., 2012).However, as observed in the context of many human diseases, such studies are often restricted to a few populations, and this strategy also restricts the use of any dosing algorithm that might be developed.
According to UNAIDS data, 37.7 million people were globally living with HIV in 2020.Latin American countries accounted for 2.1 million PLHA, and about 1.4 million (65%) were accessing treatment (UNAIDS, 2021).Attempting to assess the state of art of HIV pharmacogenetics in Latin American countries, a detailed literature search on PubMed and EMBASE was conducted using non-specific keywords such as "HIV", "polymorphism" and "country name" or "Latin America" to find as many studies as possible.Although this initial search has retrieved 1,062 articles published until February, 2022, less than 40 articles were selected for complete analysis and discussion after discarding those unrelated to the main subject, reviews and editorials.
This review begins highlighting the association of HLA-B*57:01 allele and hypersensitivity reactions to the reverse transcriptase inhibitor abacavir (ABC), which is considered a model of successful implementation in clinical practice.In the following items, the impact of ADME variations on ARVs plasma levels and response to therapy is discussed.Polymorphisms in genes encoding cytokines and other genes indirectly related with response to cART are also described.

HLA-B*57:01 Screening and hypersensitivity reactions to Abacavir: A successful example of translation from basic research to clinical practice
The association between HLA-B*57:01 allele and hypersensitivity reactions (HSR) to ABC (Mallal et al., 2008) is a relevant example of implementation of pharmacogenetics in clinical routine.Although ABC is generally well tolerated, 5-8% of patients experience HSR during the first 6 weeks of treatment.Clinical manifestations include mild symptoms such as fever, rash and nausea, but may also include multiorgan failure and anaphylaxis (Hewitt, 2002), leading to hospitalization and even death.The predictive effect of HLA-B*57:01 for HSR to ABC has been observed and

Review Young Brazilian Geneticists -Special Issue
Send correspondence to Cynthia Chester Cardoso.Universidade Federal do Rio de Janeiro, Instituto de Biologia, Laboratório de Virologia Molecular, Av.Carlos Chagas Filho, 373, CCS, Bloco A, Sala 121, 21941-902, Ilha do Fundão, RJ, Brazil.E-mail: cynthiac@ biologia.ufrj.br.Velozo et al. 2 validated using association analyses and clinical trials, followed by cost-effectiveness analyses (Lucas et al., 2007;Mallal et al., 2008).Screening for the HLA-B*57:01 allele prior to ABC prescription is currently recommended by the US Food and Drug Administration, in addition to health and regulatory agencies from Europe, Canada and Japan (PharmGKB).
Studies from Latin American countries have focused on the description of allele frequencies or at least frequency of HLA-B*57:01 carriers.Since sequencing-based HLA genotyping is still expensive and laborious, different methods have been developed to detect HLA-B*57:01 presence, regardless the genotype.In such cases, the authors can only define frequency of HLA-B*57:01carriers, but not the exact allele frequency.In Brazil, a retrospective study of 96 HIV positive individuals treated with ABC and 234 HIV negative individuals from Pernambuco State, at Northeast region, has found allele frequencies of 1.5 and 1.7% respectively.The three patients carrying HLA-B*57:01 allele presented symptoms of HSR to ABC (Crovella et al., 2011).Later, a study of 517 individuals from Central West region has showed a frequency of 5.6% for HLA-B*57:01 carriers among HIV positive individuals.Although exact allele frequencies could not be determined in this study, HSR to ABC was investigated, and results showed a significantly higher incidence among HLA-B*57:01 carriers (Araújo et al., 2014).According to the Allele Frequency Net Database, frequencies of this allele vary across Brazilian regions, ranging from 0.005 -0.026 in Brazilian Southeast and 0.03 among Puyanawas, from the North region (Gonzalez-Galarza et al., 2020).
Although all studies acknowledge that HLA-B*57:01 frequency varies according to ethnicity, remaining higher among those with European ancestry, most of them highlight the importance of a screening for this allele before prescribing ABC.Despite the lack of well-structured and statistically powered association studies performed in each country, the results available from other populations have provided strong evidence of a clear genotype-phenotype correlation (Mallal et al., 2008;Ostrov et al., 2012;Norcross et al., 2012), suggesting that the predictive value of HLA-B*57:01 would be less prone to variations according to genetic structure of each population.
According to technical notes and information obtained directly from local authorities, the HLA-B*57:01 screening is mandatory for patients eligible for ABC use in several Latin American countries including Argentina, Brazil, Chile, Costa Rica, Ecuador, Mexico, and the test has also been implemented in public health systems from a few countries (Ministerio

Metabolism enzymes and related transcription factors
Among metabolism enzymes, the impact of CYP2B6 variations on efavirenz (EFV) pharmacokinetics and/or response to therapy has been widely investigated.EFV is a non-nucleoside reverse transcriptase inhibitor which frequently causes central nervous system (CNS) adverse effects such as dizziness, nightmares, anxiety and depression (Kenedi and Goforth, 2011).Since EFV is metabolized mostly by CYP2B6 (Figure 2), single nucleotide polymorphisms (SNPs) at CYP2B6 gene have been consistently associated to EFV exposure, and slow metabolizers have increased risk of adverse reactions.Although a major effect is suggested for CYP2B6 +516G>T (rs3745274), composite genotypes including two other variations are apparently better predictors of CYP2B6 metabolic profile.Notably, data from ethnically diverse populations have showed that this association varies according to genetic ancestry (Haas et al., 2004;Holzinger et al., 2012), reinforcing the idea that validation of the effect in each relevant population is still required before clinical implementation.
A study with a main cohort from Haiti, and a replication sample of African Americans, has confirmed the association between CYP2B6 +516G>T and increased EFV levels among individuals from African descent (Leger et al., 2009).A study of CYP2B6 SNPs in Chileans showed higher EFV levels among +516TT carriers, and composite genotypes including rs10403955, rs2279345 and rs8192719 as tag SNPs were even more informative for EFV levels above the minimum toxic concentration (Carr et al., 2010).Similar results were obtained in another cohort from the same country, where +516G>T polymorphism was also associated to EFV levels (Cortes et al., 2013).
In Brazil, a study with a cohort from the Amazon region has found an association between +516TT genotype and lower CD4 T cell counts, while no association was observed for viral loads (Queiroz et al., 2017).Notably, the impact of this variation on HIV viral loads was investigated in the whole cohort, which also included individuals using protease inhibitors instead of efavirenz.Results of two retrospective cohort studies have also showed no association between CYP2B6 +516G>T and virological response in cohorts from the Northeast (Coelho et al., 2013) and Southeast (De Almeida Velozo et al., 2021) regions.The latter has also investigated the effect of composite genotypes and no association was observed either before or after adjustment for covariates, including genetic ancestry.Similarly, functional CYP2B6 variants were not associated with virological response in a cohort from Haiti after analyses considering single SNPs, composite genotypes or CYP2B6 metabolic profile (Haas et al., 2014).
Regarding adverse reactions to EFV, results obtained to date are controversial in some aspects, suggesting an impact of population substructure or heterogeneity across the different study designs.A study from the Brazilian southeastern region has found no association between CYP2B6 polymorphisms and intolerance to EFV-containing regimens (Arruda et al., 2016).Similarly, +516G>T was not associated to self-reported CNS adverse effects to EFV in a small cohort (N = 50) from the Brazilian southern region (Müller et al., 2017).Despite the higher sample size, limitations of the study from the Southeast region included analysis by drug class, implying that EFV and nevirapine-containing regimens could not be dissociated, and the use of a non-specific outcome "intolerance" (Arruda et al., 2016).To overcome such limitations, an independent cohort from Rio de Janeiro was recruited and only patients undergoing EFV-containing regimens were investigated.In this study, the authors have found an association between CYP2B6 slow metabolizers and increased risk of adverse effects to EFV, including either all effects or CNS adverse effects (de Almeida et al., 2018).Data analyses were also adjusted for genetic ancestry to control for confounding.Recently, +516G>T was significantly associated to CNS adverse effects to EFV in a small (N=38) Chilean cohort (Poblete et al., 2021).Prevalence of +516T was also determined for a cohort from Argentina (38.2%), although no association analysis was performed (Galván et al., 2012).
Four studies have also performed broader characterizations of EFV pharmacogenetics, including not only CYP2B6, but also candidate genes related to secondary metabolism pathways (CYP2A6, CYP3A4, CYP3A5) and the transcription factors PXR and CAR, encoded by NR1I2 and NR1I3 genes.Among Haitians, variations at CYP2A6 and CYP3A4/5 enzymes were not associated to EFV exposure or virological response (Leger et al., 2009;Haas et al., 2014).In Brazil, these 5 additional genes were not associated with adverse reactions to EFV (de Almeida et al., 2018).Nevertheless, variations in NR1I2 and NR1I3 were clearly associated to virological response, and the most prominent effect was observed for SNP rs2307424 (NR1I3), which was associated with increased virologic response after 12 months of cART.Haplotypes carrying allele rs2307424A were associated to this outcome as well (De Almeida Velozo et al., 2021).Notably, SNP rs2307424 has been previously associated to EFV exposure among Chileans (Cortes et al., 2013).Figure 2 summarizes the pathways for EFV primary metabolism in addition to the main genetic associations observed in Latin American countries.
In addition to EFV, genes encoding metabolism enzymes have also been investigated in Latin American populations in the context of responses to protease inhibitors (PIs).A study of 98 Brazilian HIV+ men undergoing cART regimens showed that CYP3A5*3 and CYP3A5*6 alleles do not affect plasma concentrations of lopinavir and ritonavir (Estrela et al., 2008).Moreover, polymorphisms at NR1I2 were not associated to lopinavir levels in a cohort of 38 perinatally infected children from Argentina (Bellusci et al., 2013).Studies of adverse reactions to PIs have showed that UGT1A1*28 allele was associated with increased risk of atazanavir related hyperbilirubinemia among Brazilians (Turatti et al., 2012) and Chileans (Poblete et al., 2021).The study by Arruda et al. (2016) has also investigated the role of metabolism enzymes in intolerance to PIs, but no association was found.The main findings regarding association between ADME genes and response to antiretrovirals in Latin American countries were summarized in Table 1.

Drug transporters
Genes encoding drug transporters are also important targets for HIV pharmacogenetics due to their role in ARVs influx and efflux from different cells, regulating bioavailability and penetration in target cells and viral sanctuary sites.
Among ABC transporters (ATP-binding cassette proteins), the ABCB1 gene -which encodes P-glycoprotein -has been widely investigated in response to different ARVs, including EFV and PIs (Michaud et al., 2012).Variations at this gene were not associated with plasma concentrations or virological response to EFV among Haitians (Leger et al., 2009;Haas et al., 2014).In Brazil, ABCB1 variations were not associated with intolerance or CNS adverse reactions to EFV (Arruda et al., 2016;de Almeida et al., 2018) in cohorts from Southeast region, while 1236C>T was associated with decreased immunological response to this antiretroviral in patients from Northeast region (Coelho et al., 2018).
Conflicting results were also observed in response to protease inhibitors, suggesting that ABCB1 effect may be influenced by population substructure.ABCB1 genotypes for 1236C>T, 2667G>T/A and 3435C>T were not predictors of lopinavir and ritonavir concentrations in plasma, semen or saliva of 113 HIV positive individuals (Estrela et al., 2009).By contrast, SNP 3435C>T was associated with virological failure of PI containing regimens (Coelho et al., 2013).In Argentinian cohorts, ABCB1 polymorphisms were associated to lopinavir plasma levels in HIV-1 perinatally infected children (Bellusci et al., 2013).Association to porphyria cutanea tarda was also suggested, although it was not clear whether this condition was mostly influenced by HIV infection or treatment (Pagnotta et al., 2020).Other ABC transporters have also been investigated in response to cART.In Brazil, ABCC1 and ABCC2 genes were respectively associated to virological failure (Coelho et al., 2013) and intolerance to PI containing regimens (Arruda et al., 2016).
Among solute carrier transporters, a SLCO1B1 polymorphism was associated with the trough plasma concentration of lopinavir, considering a cohort composed of 99 individuals treated with lopinavir and ritonavir for at least 4 weeks (Kohlrausch et al., 2010).Later, the same gene was associated with intolerance to nucleoside reverse transcriptase inhibitors (Arruda et al., 2016).Both studies were conducted among HIV positive individuals from Brazilian Southeast.

Other associations
In addition to ADME genes, association studies have also investigated a role for genes related to specific phenotypes, such as metabolic outcomes.In Brazil, polymorphisms in SCAP, APOE, APOA5 and ADIPOR2 genes were associated with high levels of triglycerides and cholesterol in individuals using protease inhibitors (Lazzaretti et al., 2013;Castilhos et al., 2015), while a polymorphism in APM1 influenced adiponectin levels (Trinca et al., 2010).ESR1 and ESR2 genes were both associated with body mass index (BMI) and total subcutaneous fat, while ESR2 also conferred risk for lipoatrophy in women undergoing PI-containing regimens (Gasparotto et al., 2012).SNPs at APOA5, APOC3 and SIK3 were also implicated in risk of hypertriglyceridemia in a casecontrol study including 602 Mexicans patients receiving PIs (Bautista-Martínez et al., 2022).
Additional studies have also investigated the role of polymorphisms in genes encoding cytokines and restriction factors and response to cART.In Brazil, variations at CCR5Δ32 (Rigato et al., 2008), IL2 (Coelho et al., 2018) et al., 2014).A large study, including 873 participants from United States and Puerto Rico, have also showed an association between polymorphisms in genes encoding TNF-α, TRAIL, Bcl-2, IL-15, IL-15Rα and IFN-α were associated CD4 lymphocyte counts after long-term cART.However, data analyses were not stratified by country, and data specific from Puerto Rico subjects were not available (Haas et al., 2006).

Conclusions
Pharmacogenetic analysis is a promising path to enable implementation of personalized regimens that might prevent cART unfavorable outcomes such as adverse reactions to ARVs.However, universal application of genetic tests may be challenging not only due to budget limitations of each country, but also due to the lack of validation in ethnically diverse populations.Indeed, large scale pharmacogenetics studies including participants with varying genetic backgrounds are still scarce, and populations such as Latin Americans are underrepresented.
In this review, we summarize the main data regarding HIV pharmacogenetics obtained from studies focused on Latin American cohorts.The literature search clearly showed that most studies were conducted among Brazilians, with additional analyses from Haiti, Chile, Argentina, Costa Rica, Mexico and Puerto Rico.The scarcity of studies focusing on this region raises concern.Despite the geographical proximity, allele frequencies may be remarkably different within Latin America populations since most of them may exhibit complex genetic backgrounds.Unlike the example of HLA-B*57:01 screening to prevent HSR to ABC -which is probably informative for any population -genetic associations described for Europeans or clearly defined ancestry groups are rarely generalizable to genetically complex populations.Further analyses of Latin American cohorts along with other understudied populations are crucial to improve fine mapping of causative variations and to ultimately develop more effective dosing algorithms, applicable to diverse ethnicities.

Figure 1 .
Figure 1.Frequency of HLA-B*57:01 allele in Latin American countries.Frequencies were retrieved from the Allele Frequency Net Database (http:// www.allelefrequencies.net/).Data were available for Argentina, Brazil, Chile, Colombia, Costa Rica, Cuba, Guatemala, Mexico, Nicaragua and Panama.When more than one frequency was provided for a single country, mean values were determined.

Figure 2 .
Figure 2. Efavirenz hepatic metabolism and pharmacogenetic associations in Latin American populations.CYP2B6 acts on primary efavirenz (EFV) metabolism pathway, converting EFV to 8-hydroxy EFV.Variations at CYP2B6 gene have been extensively associated with EFV plasma levels and also to adverse reactions to this antiretroviral.Other enzymes (CYP3A4, CYP3A5, CYP2A6) can also generate 8-hydroxy-EFV.Transcription factors PXR and CAR regulate expression of CYP2B6 and CYP3A enzymes, and may indirectly influence EFV metabolism.Polymorphisms at genes encoding PXR and CAR were associated to EFV exposure and virological response.Alternative pathways include hydroxilation by CYP2A6 and direct glucoronidation by UGT2B7.Intermediate metabolites are then glucoronidated before excretion in urine.Genes associated with EFV pharmacokinetics and/or response to EFV in Latin American populations are shown in yellow.Created with BioRender.com.

Table 1 .
Associations between ADME genes and response to antiretroviral therapy in Latin American populations.