P450 Pharmacogenetics in Indigenous North American Populations

Indigenous North American populations, including American Indian and Alaska Native peoples in the United States, the First Nations, Métis and Inuit peoples in Canada and Amerindians in Mexico, are historically under-represented in biomedical research, including genomic research on drug disposition and response. Without adequate representation in pharmacogenetic studies establishing genotype-phenotype relationships, Indigenous populations may not benefit fully from new innovations in precision medicine testing to tailor and improve the safety and efficacy of drug treatment, resulting in health care disparities. The purpose of this review is to summarize and evaluate what is currently known about cytochrome P450 genetic variation in Indigenous populations in North America and to highlight the importance of including these groups in future pharmacogenetic studies for implementation of personalized drug therapy.


Introduction
Pharmacogenetics, a form of genomic medicine, aims to establish how genetic variation can affect an individual's response to drugs, guiding the selection of the best drug and dose for a patient to improve healthcare quality [1]. The field of pharmacogenetics has the potential to improve health outcomes and reduce the cost of care by maximizing therapeutic success and minimizing the risk of adverse drug reactions or therapeutic failure at the population and potentially the individual level. However, a major issue in the translation of pharmacogenetic research into clinical practice is that existing databases have been populated from studies that lack significant ethnic and racial diversity.
Although diversity in genomic research, including pharmacogenetics, has increased in recent years, oversampling of populations of European ancestry continues to be a problem in the field. Indeed, the latest analysis of genome-wide association studies found that 81% of samples were from individuals of European ancestry [2]. The non-European portion was comprised of mostly Asian ancestry, leaving just 5% for the rest of the world's populations [2]. Failure to include diverse populations in genomic studies leads to a biased understanding of the health implications of genetic variation and resulting medical findings may be preferentially beneficial to patients of European

Results
We identified twenty-seven studies that met our inclusion criteria. These studies reported P450 polymorphisms in Indigenous North American populations for CYP1A1, CYP1A2, CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, CYP3A5 and CYP4F2. Six studies were in AIAN people, seven studies were in Indigenous people of Canada and fourteen studies were in Amerindian populations of Mexico. Figure 2 shows the geographical locations of the study populations and Table  1 summarizes the study results. Full-text articles excluded because focus was on diseaseassociated phenotypes (n = 3) Excluded if not an original study or for not focusing on drug metabolizing enzymes (n = 84) The following data were abstracted from selected studies: number of individuals in the study, the study population, P450 enzymes, method for genotyping and phenotyping, allele frequencies and conclusions from the study. For some of these studies, a reference population (e.g., European or Mestizos descent) was included. Our focus was to review available data for Indigenous North American populations, but we note that the studies reported in this review contain inconsistent approaches to population description for comparator populations, which may be categorized by race, ethnicity, nationality, or geographic location. Summary tables (Tables 2-13) also include reference data for different racial groups abstracted from the 1000 Genomes database (SNVs) [20] or from Zhou et al. [21] (complex haplotypes or structural variation not readily obtained from 1000 Genomes).

Results
We identified twenty-seven studies that met our inclusion criteria. These studies reported P450 polymorphisms in Indigenous North American populations for CYP1A1, CYP1A2, CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP2E1, CYP3A4, CYP3A5 and CYP4F2. Six studies were in AIAN people, seven studies were in Indigenous people of Canada and fourteen studies were in Amerindian populations of Mexico. Figure 2 shows the geographical locations of the study populations and Table 1 summarizes the study results.

CYP1A1
The CYP1A1 enzyme plays a role in the metabolism of caffeine [49] as well as the bioactivation of polycyclic aromatic hydrocarbons [50]. The CYP1A1*2A variant located in the 3 non-coding region confers a restriction endonuclease site for cleavage by Msp1 (Figure 3) [51]. CYP1A1*2C is characterized by a nonsynonymous base change that results in an amino acid substitution associated with an increase in CYP1A1 gene inducibility ( Figure 3) [52,53].
The frequencies of CYP1A1*2A and *2C were determined in two Amerindian peoples, the Teenek and Mayos and compared to the Mestizo Mexican population [22]. In the Teenek population, the minor allele frequencies (MAFs) of CYP1A1*2A and *2C were 71.4 and 65.4%, respectively and in the Mayos the MAFs were 46.9 and 54.6%, respectively (Table 2) [22]. Both Amerindian populations had a significantly higher frequency of CYP1A1*2C compared to the Mexican Mestizo population (34.4%), while only the Teenek had a significantly higher frequency of CYP1A1*2A compared to the Mestizo population (40.1%) [22].
De Andrés et al. found that the CYP1A2*1F allele was present at a MAF of 66.6% in the Amerindian population (Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío) ( Table 3) [23]. Using 100 mg of caffeine, as part of a probe drug cocktail for phenotyping, no association between CYP1A2*1F and higher enzyme activity was observed [23]. However, the CYP1A2*1F variant does not confer high constitutive activity but rather increases enzyme inducibility with exposure to an inducer, so the lack of genotype-phenotype association could be due to the fact that subjects were not stratified by their level of intake of a CYP1A2 inducer, such as high dose caffeine.
De Andrés et al. found that the CYP1A2*1F allele was present at a MAF of 66.6% in the Amerindian population (Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío) ( Table 3) [23]. Using 100 mg of caffeine, as part of a probe drug cocktail for phenotyping, no association between CYP1A2*1F and higher enzyme activity was observed [23]. However, the CYP1A2*1F variant does not confer high constitutive activity but rather increases enzyme inducibility with exposure to an inducer, so the lack of genotype-phenotype association could be due to the fact that subjects were not stratified by their level of intake of a CYP1A2 inducer, such as high dose caffeine.
Tanner et al. compared variation in CYP2A6 and NMR in two different AI populations and assessed differences in relation to smoking behaviors and risks [25]. In Northern Plains (NP) AIs, the CYP2A6*2, *4, *9 and *12, the MAFs were 0.3%, 1.6%, 11.9% and 0.3%, respectively, while in AIs from the Southwest (SW) in Arizona the frequencies were 0.6, 0.3, 20.9 and 0.3%, respectively (Table 4) [25]. CYP2A6*7 and *17 were absent from the Yup'ik, NP and SW populations. While CYP2A6*35 was not found in the Yup'ik or NP, it had a MAF of 0.3% in the SW AI population ( Table 4). The NP AI population had a lower frequency of CYP2A6 decreased function alleles and a higher rate of nicotine metabolism, compared to SW smokers [25]. CYP2A6 genetic variants are important to consider clinically because there are negative outcomes associated with higher rates of nicotine metabolism, including increased tobacco consumption, more difficulty with smoking cessation, poorer success with nicotine replacement therapy and elevated risk of lung cancer [27,75,79].
Tanner et al. compared variation in CYP2A6 and NMR in two different AI populations and assessed differences in relation to smoking behaviors and risks [25]. In Northern Plains (NP) AIs, the CYP2A6*2, *4, *9 and *12, the MAFs were 0.3%, 1.6%, 11.9% and 0.3%, respectively, while in AIs from the Southwest (SW) in Arizona the frequencies were 0.6, 0.3, 20.9 and 0.3%, respectively (Table 4) [25]. CYP2A6*7 and *17 were absent from the Yup'ik, NP and SW populations. While CYP2A6*35 was not found in the Yup'ik or NP, it had a MAF of 0.3% in the SW AI population ( Table 4). The NP AI population had a lower frequency of CYP2A6 decreased function alleles and a higher rate of nicotine metabolism, compared to SW smokers [25]. CYP2A6 genetic variants are important to consider clinically because there are negative outcomes associated with higher rates of nicotine metabolism, including increased tobacco consumption, more difficulty with smoking cessation, poorer success with nicotine replacement therapy and elevated risk of lung cancer [27,75,79].    [20]. N represents the number of alleles. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL). For alleles not captured by 1000 Genomes (noted by ‡), the frequencies were extracted from Exome Aggregation Consortium [80] and for alleles defined by multiple variants, the frequencies reported in Zhou et al. [21] using LDLink software [81] were used. Global population data for CYP2A6*35 was not included because it is currently difficult to accurately genotype due to the high homology to CYP2A7, which can result in false positives and false negatives.

CYP2B6
The CYP2B6 enzyme metabolizes bupropion [82,83], cyclophosphamide [84], efavirenz [85], ketamine [86,87], methadone [88], as well as other drugs. It is also thought to contribute to nicotine metabolism when CYP2A6 activity is low [89]. CYP2B6*4 (K262R) variation confers increased enzyme activity, while CYP2B6*6 is a haplotype that includes both K262R and Q172H and confers reduced function ( Figure 6); it is common among different ethnic groups. For example, in the Yup'ik AN population, the reported CYP2B6*6, was 51.7% (Table 5) [24]. Unlike CYP2A6, CYP2B6 genotype was not found to be associated with nicotine metabolism in this population. Previous studies found weak linkage disequilibrium between the CYP2A6 and CYP2B6 genes (localized together on chromosome 19), however, Binnington et al. reported strong linkage disequilibrium between these two genes in the Yup'ik population [24,[90][91][92]. The authors proposed that the unique linkage disequilibrium observed between the reference CYP2A6*1B allele and the low activity CYP2B6*6 allele may be responsible for the higher nicotine metabolism in Yup'ik individuals with CYP2B6*6 genotype.

CYP2B6
The CYP2B6 enzyme metabolizes bupropion [82,83], cyclophosphamide [84], efavirenz [85], ketamine [86,87], methadone [88], as well as other drugs. It is also thought to contribute to nicotine metabolism when CYP2A6 activity is low [89]. CYP2B6*4 (K262R) variation confers increased enzyme activity, while CYP2B6*6 is a haplotype that includes both K262R and Q172H and confers reduced function ( Figure 6); it is common among different ethnic groups. For example, in the Yup'ik AN population, the reported CYP2B6*6, was 51.7% (Table 5) [24]. Unlike CYP2A6, CYP2B6 genotype was not found to be associated with nicotine metabolism in this population. Previous studies found weak linkage disequilibrium between the CYP2A6 and CYP2B6 genes (localized together on chromosome 19), however, Binnington et al. reported strong linkage disequilibrium between these two genes in the Yup'ik population [24,[90][91][92]. The authors proposed that the unique linkage disequilibrium observed between the reference CYP2A6*1B allele and the low activity CYP2B6*6 allele may be responsible for the higher nicotine metabolism in Yup'ik individuals with CYP2B6*6 genotype.   [20]. N represents the number of alleles. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL). For alleles not captured by 1000 Genomes (noted by ‡), the frequencies were extracted from Exome Aggregation Consortium [80] and for alleles defined by multiple variants, the frequencies reported in Zhou et al. [21] using LDLink software [81] were used.

CYP2C9
The CYP2C9 enzyme metabolizes medications across many therapeutic classes [93] including nonsteroidal anti-inflammatories (e.g., naproxen) [94][95][96][97], angiotensin II blockers (e.g., losartan) [98], as well as narrow therapeutic index drugs such as (S)-warfarin [94,99], tolbutamide [100] and phenytoin [101]. Warfarin dosing is challenging and regularly monitored due to its wide inter-individual variability and narrow therapeutic index, which affects both its pharmacokinetics and pharmacodynamic response. AN populations are reported to require a lower dose of warfarin to achieve a desired therapeutic effect, with the average daily dose for the AN population being 4.34 mg versus 5.19 mg for those of European descent [102]. This observed difference in warfarin dosing is clinically meaningful and thought to be due, in part, to genetic polymorphisms in the CYP2C9, VKORC1 and CYP4F2 genes [103]. For example, individuals with CYP2C9*2 or CYP2C9*3 variant alleles require a lower warfarin dose to achieve therapeutic anticoagulation [104].
In the Indigenous population of Canada, the previously studied CYP2C9 variant allele frequencies were found to be distinct from the European and Asian reference groups, particularly for the Inuit population, where CYP2C9*2, *3 and *4 were absent [28]. The MAF of CYP2C9*2, *3 and *4 were 3.0%, 6.0% and 0.0%, respectively in the FN population (Table 6) [28].
With respect to novel variation, one new novel coding variant, CYP2C9 K119T (Figure 7), was identified in the CSKT population at a frequency of 0.57% [32]. In addition, two novel coding-region CYP2C9 were identified in the Yup'ik and SCF populations: CYP2C9 M1L (M1L) and CYP2C9 N218I (N218I) (Figure 7) [33]. These two SNVs are of interest due to the fact that they are both coding variants and present at relatively high frequencies in the Yup'ik population; MAFs of M1L and N218I were 6.3% and 3.8%, respectively, whereas in the SCF population, the MAFs were 1.0% and 1.4%, respectively ( Table 7). The switch from a methionine start codon to leucine for M1L is predicted to confer a PM phenotype in vivo for carriers of the M1L variant by severely slowing or stopping RNA translation and protein production. The N218I variant is also expected to have reduced enzyme activity as it had a Grantham score of 149 [33], where a score greater than 100 indicates that the amino acid substitution is predicted to be damaging. Presently, such in silico predictions cannot be relied upon and additional functional studies are needed [105]. However, there is increased confidence that individuals who are heterozygous or homozygous for the M1L variant would have a lower warfarin dose requirement, among phenotypic changes for other CYP2C9 substrates.  [20]. N represents the number of alleles. Not shown in the table are CYP2C9*4 (rs56165425) and CYP2C9*6 (rs9332131), as these SNVs were either absent or not tested for in North American Indigenous populations. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL).
With regard to CYP2C9*2 in the Tepehuano and Tarahumara  De Andrés et al. performed genotyping and phenotyping, using losartan as a probe substrate as part of a cocktail, in Amerindian population including the Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío groups [23]. The ratio of losartan to losartan carboxylic acid was significantly greater in CYP2C9*2 or *3 homozygotes or carriers, compared to those with the reference genotype [23]. However, there were also three individuals whose high parent to metabolite could not be explained by their genotype, suggesting that there may be unidentified rare variants that confer PM phenotype in this Amerindian population [23].
Nowak et al. reported that in FNs, the CYP2C19*2 allelic variant was found at 19.1%, while CYP2C19*3 was not detected (Table 8) [26]. Jurima-Romet et al. conducted CYP2C19 genotyping and phenotyping, based on the mephenytoin S/R enantiomeric ratio, in Inuit peoples in Canada [35]. The CYP2C19*2 allele frequency was 12%, while CYP2C19*3 was not detected in this Inuit population (Table 8). Subjects ingested a single dose of 100 mg (R,S)-mephenytoin and collected urine for 12 h. Urinary S/R mephenytoin enantiomeric ratio was used to distinguish between EMs and PMs, where S/R ratio ≥ 1.0 indicated PM phenotype and S/R ratio ≤ 0.5 indicated EM phenotype. Genotype results were in agreement with phenotype results and as expected, individuals classified as PMs did not have detectable levels urinary 4 -hydroxymephenytoin [35].

CYP2C19
The CYP2C19 enzyme plays an important role in the metabolism of antiplatelet agents (e.g., clopidogrel) [106], proton pump inhibitors (e.g., omeprazole) [107], tricyclic antidepressants [108,109], selective serotonin reuptake inhibitors [110][111][112][113] and benzodiazepines [114]. CYP2C19*2 and *3 alleles define the current PM phenotype status (Figure 8), with markedly different allele frequencies across different ethnic populations. The CYP2C19*17 variant is a gain-of-function allele, resulting in higher enzyme activity (Figure 8) [115]. CYP2C19 catalyzes the 4′-hydroxylation of (S)mephenytoin, a probe substrate for CYP2C19 [116,117]. The CYP2C19 PM phenotype is associated with a reduced ability to metabolize (S)-mephenytoin [118]. Thus, PMs eliminate racemic mephenytoin more slowly than EMs and urinary 4′-hydroxymephenytoin and S/R mephenytoin enantiomeric ratio was used in the past to distinguish between CYP2C19 EMs and PMs [119].   Clopidogrel is a prodrug that is bioactivated by multiple P450s enzymes (including CYP2C19) to its active metabolite, which acts as an antiplatelet agent by irreversibly inhibiting the P2Y12 adenosine diphosphate receptor on platelets [106]. CYP2C19 genetic variation has been associated with the fraction metabolized and clinical response to clopidogrel [120]. In the Oglala Sioux Tribe of South Dakota, the MAFs for CYP2C19*2, *3 and *17 were 11.2%, 0.0% and 8.7%, respectively (Table 8) [36]. P2Y12 reaction units (PRU) were used to evaluate the pharmacodynamic effect of clopidogrel, with lower values representing reduced platelet aggregation and higher efficacy of clopidogrel. Although PRU was not found to be significantly associated with genotype, the median PRU of 194 (range 29-400) is similar to values reported for other groups [36].
In a study examining CYP2C19 variant alleles in AI children, the frequencies of CYP2C19*2, *4 and *17 were 11.5%, 1.3% and 11.1%, respectively (Table 8) [34]. CYP2C19*8 and CYP2C19*14 were not present in the AI population studied and CYP2C19*3 was not tested. Currently, nothing is known about genotype-drug disposition and response phenotype associations in AN populations.
With regards to CYP2C19 variation in Amerindians of Mexico, Salazar-Flores et al. performed genotyping in Tarahumaras from Chihuahua, Purepechas from Michoacán, Tojolabales, Tzotziles and Tzeltales from Chiapas and Tepehuanos from Durango [37]. The CYP2C19*2 MAF was reported as 31%, 5.4%, 3.6%, 5.6% and 0.0% in Tarahumaras, Purepechas, Tojolabales, Tzotziles and Tzeltales, respectively (Table 8) [37]. The high allele frequency in the Tarahumara population is of particular interest clinically, as this predicts that a substantial portion of the population would be CYP2C19 PMs. CYP2C19*3, *4 and *5 were not detected in any of the Amerindian populations studied [37]. In an Amerindian population including the Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío groups, de Andrés et al. reported the MAFs of CYP2C19*2, *3 and *17 as 12.0%, 0.2% and 2.2%, respectively (Table 8) [23]. CYP2C19*4 and *5 were not detected in the population studied [23]. CYP2C19 phenotype was determined using omeprazole as a probe substrate as part of a drug cocktail. The ratio of omeprazole to 5-hydroxyomeprazole was significantly greater in individuals with the PM conferring variants, CYP2C19*2 or *3, compared to those with the reference genotype [23]. There were some individuals whose genotype did not correspond to their phenotype, which the authors suggest may be due to population-specific alleles not identified in other populations [23].

CYP2D6
The CYP2D6 enzyme metabolizes many basic drugs including opioids [121,122], antidepressants (e.g., nortriptyline and fluoxetine) [108,111,123,124], antipsychotics (e.g., risperidone) [125,126] and ß-blockers (e.g., metoprolol) [127][128][129]. It is another example of a polymorphic enzyme with highly penetrant null (loss-of-function) allelic variants conferring a PM phenotype or copy number variation resulting in a UM phenotype. Individuals homozygous or compound heterozygous for the CYP2D6*3, *4, *5, or *6 variants exhibit no enzyme activity (Figure 9), while those with CYP2D6*1 or *2 gene duplication have more CYP2D6 protein with resulting higher enzymatic activity. This is clinically relevant for codeine, which is metabolized to morphine by CYP2D6. CYP2D6 PMs may experience an inadequate analgesic effect while UMs are at risk of morphine toxicity [130,131]. This is particularly important for the pediatric population for which codeine use was common and so in 2013 a US Food and Drug Administration Black Box Warning was issued for the drug. A structurally related drug, dextromethorphan (DEX), is O-demethylated by CYP2D6 to dextrorphan (DXO) and, therefore, the metabolic ratio of DEX/DXO can be used to differentiate between CYP2D6 phenotypes [132,133].
In Inuit peoples in Canada, the allele frequency of CYP2D6*4 was 6.7-8.3%; the CYP2D6*10 MAF was 2.2% and the CYP2D6*3 and *6 alleles were not detected (Table 9) [38]. Phenotyping, based on the urinary DEX/DXO metabolic ratio, was also performed. Study participants ingested a single dose of 30 mg dextromethorphan hydrobromide and collected urine overnight. Phenotype results were in agreement with genotype and furthermore, individuals classified as PMs had lower recoveries of DXO as well as other CYP2D6-mediated metabolites, compared to EMs. Notably, the frequency of the CYP2D6*4 allele in the Inuit population was significantly lower than the MAF reported in the European population (23%) [134,135] and significantly greater than in the Asian population (<1.0%) [136,137]. Table 9. Comparison of CYP2D6 allele frequencies Indigenous North American populations to global populations from the 1000 Genomes Project [20]. N represents the number of alleles. Not shown is CYP2D6*7 (rs503086), as it was not detected or not tested for in the Indigenous populations studied. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL). For alleles not captured by 1000 Genomes (noted by ‡), the frequencies were extracted from Exome Aggregation Consortium [80] and for alleles defined by multiple variants, the frequencies reported in Zhou et al. [21] using LDLink software [81] were used.    [23]. Regarding CYP2D6 multiplications, the frequency of wtxN, *2xN and *4xN were 4.7, 1.1 and 0.1%, respectively [23]. CYP2D6 phenotype was also determined using the DEX/DXO ratio, with a significantly greater parent to metabolite ratio for individuals with reduced activity or null CYP2D6 variants [23]. As with the other P450 drug metabolizing enzymes evaluated in this study, there was some discordance between genotype and phenotype. Further studies are necessary to identify and characterize variants that may impact the activity of important drug metabolizing enzymes including CYP2D6.  CYP2D6 genotype, as well as phenotype using the O-demethylation ratio of DEX [138], was determined in a FN population [39]. The MAFs of CYP2D6*3, *4 and *10 were 0.0, 3.0 and 3.0%, respectively (Table 9) [39]. Interestingly, the one individual identified as a CYP2D6 PM by phenotyping with DEX was not found to have a CYP2D6*4/*4 genotype. This suggests that the PM phenotype could be attributed to a null allele not tested for in this study or a novel loss-of-function variant.
The genotype results reported by de Andrés et al. in an Amerindian population including the Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío groups, were consistent with previously published findings that CYP2D6 variants conferring PM status are rare in the Amerindian population [23]. Regarding CYP2D6 multiplications, the frequency of wtxN, *2xN and *4xN were 4.7, 1.1 and 0.1%, respectively [23]. CYP2D6 phenotype was also determined using the DEX/DXO ratio, with a significantly greater parent to metabolite ratio for individuals with reduced activity or null CYP2D6 variants [23]. As with the other P450 drug metabolizing enzymes evaluated in this study, there was some discordance between genotype and phenotype. Further studies are necessary to identify and characterize variants that may impact the activity of important drug metabolizing enzymes including CYP2D6.
The frequency of CYP2E1*1D was reported to be 9.3% in the FN population (Table 10), which is significantly higher than that seen in European Canadians (2.1%) [46]. Furthermore, FN individuals dependent on alcohol (as defined by DSM-IV) had a higher frequency of the CYP2E1*1D allele, compared to non-alcohol dependent FNs [46]. This same trend was found in Europeans and Southeast Asians [46]. CYP2E1*1D genotype was also associated with nicotine dependence in FNs, however further studies are needed to elucidate potential mechanisms responsible for this relationship [46]. Table 10. Comparison of CYP2E1 allele frequencies Indigenous North American populations to global populations from the 1000 Genomes Project [20]. N represents the number of alleles. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL). The MAF of the CYP2E1 −1295G>C variant was determined to be 51.5% in Huichols, an Amerindian population of Western-Central Mexico (Table 10) [47]. This frequency is very high compared to the Mexicans from Western Mexico (16.1%) and Europeans (1.7%) [47,152]. The CYP2E1 −1295G>C variant is of interest when considering the metabolism of ethanol, as well as other CYP2E1 substrates and further studies would be useful to establish the clinical relevance of CYP2E1 variation across diverse populations. however further studies are needed to elucidate potential mechanisms responsible for this relationship [46].

Country
The MAF of the CYP2E1 −1295G > C variant was determined to be 51.5% in Huichols, an Amerindian population of Western-Central Mexico (Table 10) [47]. This frequency is very high compared to the Mexicans from Western Mexico (16.1%) and Europeans (1.7%) [47,152]. The CYP2E1 −1295G > C variant is of interest when considering the metabolism of ethanol, as well as other CYP2E1 substrates and further studies would be useful to establish the clinical relevance of CYP2E1 variation across diverse populations.   [20]. N represents the number of alleles. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL).

CYP3A4 and CYP3A5
The CYP3A4 and CYP3A5 enzymes have overlapping substrate specificity and together, they control the clearance of approximately 50% of all drugs eliminated primarily through P450-mediated biotransformation [15]. CYP3A4 protein is present in almost all adults, whereas CYP3A5 protein expression varies across different ethnic groups [153]; polymorphisms in the genes encoding these proteins are shown in Figures 11 and 12. The CYP3A5 enzyme is expressed in individuals having at least one CYP3A5*1 allele, while those with two inactive alleles, CYP3A5*3, *6, or *7, encode a nonfunctional protein ( Figure 12); the PM phenotype is most common in Europeans and less so in Asians and African Americans [154,155]. The CYP3A4*22 variant is also associated with reduced CYP3A4 protein levels and enzyme function ( Figure 11) [156][157][158], as are rare deleterious coding variants [159]. The CYP3A4*1B and CYP3A4*1G alleles ( Figure 11) reportedly affect gene transcription but functional effects are unclear, as the data are mixed and interpretation is complicated by high linkage disequilibrium with CYP3A5*1 [160][161][162].

CYP3A4 and CYP3A5
The CYP3A4 and CYP3A5 enzymes have overlapping substrate specificity and together, they control the clearance of approximately 50% of all drugs eliminated primarily through P450-mediated biotransformation [15]. CYP3A4 protein is present in almost all adults, whereas CYP3A5 protein expression varies across different ethnic groups [153]; polymorphisms in the genes encoding these proteins are shown in Figures 11 and 12. The CYP3A5 enzyme is expressed in individuals having at least one CYP3A5*1 allele, while those with two inactive alleles, CYP3A5*3, *6, or *7, encode a nonfunctional protein ( Figure 12); the PM phenotype is most common in Europeans and less so in Asians and African Americans [154,155]. The CYP3A4*22 variant is also associated with reduced CYP3A4 protein levels and enzyme function ( Figure 11) [156][157][158], as are rare deleterious coding variants [159]. The CYP3A4*1B and CYP3A4*1G alleles ( Figure 11) reportedly affect gene transcription but functional effects are unclear, as the data are mixed and interpretation is complicated by high linkage disequilibrium with CYP3A5*1 [160][161][162].
In the CSKT population, resequencing followed by subsequent genotyping identified four novel CYP3A4 SNVs-three intronic and one in the 5' UTR region. The known CYP3A4 variants, *1B, *22 and *1G were found at frequencies of 2.2%, 2.4% and 26.8%, respectively (Table 11) [32]. This combination of allele frequencies may result in haplotypes conferring altered enzyme activity, which remains to be tested. With regard to CYP3A5 in the CSKT population, CYP3A5*1 was detected at a frequency of 7.5%, CYP3A5*3 at 92.5%, while CYP3A5*6 and *7 were not detected (Table 12). These data suggest that 14.9% of CSKT individuals express CYP3A5, contributing to their total CYP3A metabolic activity.
Reyes-Hernández et al. determined the MAFs of CYP3A4 variants in the Tepehuano and Mestizo populations and found that the frequencies were not significantly different between these two group [48]. The MAFs of CYP3A4*1B was 8.0% in Tepehuanos, compared to 8.8% in Mestizos and CYP3A4*2 was not detected in the Tepehuanos but was found at a low frequency of 0.5% in Mestizos (Table 11) [48]. CYP3A4*4, *5 and *18 were not found in either population [48]. Although the variant allele frequencies were similar between Tepehuanos and Mestizos for CYP3A4, it is important to consider that these populations may share similar allele frequencies in other P450 genes. De Andrés et al. found that the CYP3A4*1B MAF was 4.8% in an Amerindian population including the Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío groups (Table 11) [23]. The CYP3A4*1B allele did not significantly affect the parent to metabolite ratio of dextromethorphan to 3-methoxymorphinan [23], a CYP3A4 mediated pathway [163].    In the CSKT population, resequencing followed by subsequent genotyping identified four novel CYP3A4 SNVs-three intronic and one in the 5' UTR region. The known CYP3A4 variants, *1B, *22 and *1G were found at frequencies of 2.2%, 2.4% and 26.8%, respectively (Table 11) [32]. This combination of allele frequencies may result in haplotypes conferring altered enzyme activity, which remains to be tested. With regard to CYP3A5 in the CSKT population, CYP3A5*1 was detected at a frequency of 7.5%, CYP3A5*3 at 92.5%, while CYP3A5*6 and *7 were not detected (Table 12). These data suggest that 14.9% of CSKT individuals express CYP3A5, contributing to their total CYP3A metabolic activity. Table 11. Comparison of CYP3A4 allele frequencies Indigenous North American populations to global populations from the 1000 Genomes Project [20]. N represents the number of alleles. Not shown are CYP3A4*2 (rs55785340), CYP3A4*4 (rs55951658) and CYP3A4*8 (72552799), as these variants were either not detected or not tested for in the Indigenous populations studied. Global populations are abbreviated as follows: African Ancestry in Southwest US (ASW); Utah residents with Northern and Western European ancestry (CEU); Han Chinese in Beijing, China (CHB); Mexican Ancestry in Los Angeles, California (MXL).

Country
Population Reyes-Hernández et al. determined the MAFs of CYP3A4 variants in the Tepehuano and Mestizo populations and found that the frequencies were not significantly different between these two group [48]. The MAFs of CYP3A4*1B was 8.0% in Tepehuanos, compared to 8.8% in Mestizos and CYP3A4*2 was not detected in the Tepehuanos but was found at a low frequency of 0.5% in Mestizos (Table 11) [48]. CYP3A4*4, *5 and *18 were not found in either population [48]. Although the variant allele frequencies were similar between Tepehuanos and Mestizos for CYP3A4, it is important to consider that these populations may share similar allele frequencies in other P450 genes. De Andrés et al. found that the CYP3A4*1B MAF was 4.8% in an Amerindian population including the Tarahumara, Tepehuano, Mexicanera, Huichol, Cora, Seri, Mayo and Guarijío groups (Table 11) [23]. The CYP3A4*1B allele did not significantly affect the parent to metabolite ratio of dextromethorphan to 3-methoxymorphinan [23], a CYP3A4 mediated pathway [163].

CYP4F2
CYP4F2 enzyme catabolizes vitamin K and, along with CYP2C9 and vitamin K oxidoreductase (VKOR), can affect the pharmacological response of warfarin, a VKOR antagonist. The Yup'ik AN population has a high frequency of CYP4F2*3 (Figure 13), which is a vitamin K sparing variant [164] and is associated with a higher warfarin dose requirement [165]. It is found at a frequency in the Yup'ik AN population higher than that seen elsewhere in the world, with one exception [166]. And is associated with a relatively high hepatic vitamin K status [33,167]. Selective pressure may have acted on the CYP4F2 gene in the Yup'ik population to conserve vitamin K due to the inconsistent access to tundra greens throughout the year [33,167,168]. The MAF of CYP4F2*3 is reported to be 50.9% and 31.5% in the Yup'ik and SCF populations, respectively (Table 13) [33]. While the contribution of the CYP4F2 variation to warfarin dose requirement is relatively low in European or African American populations, it may take on greater significance in AN populations due to the higher frequency of the CYP4F2*3 variant. acted on the CYP4F2 gene in the Yup'ik population to conserve vitamin K due to the inconsistent access to tundra greens throughout the year [33,167,168]. The MAF of CYP4F2*3 is reported to be 50.9% and 31.5% in the Yup'ik and SCF populations, respectively (Table 13) [33]. While the contribution of the CYP4F2 variation to warfarin dose requirement is relatively low in European or African American populations, it may take on greater significance in AN populations due to the higher frequency of the CYP4F2*3 variant. Fohner et al. also reported the MAFs of CYP4F2*2, L519M, G185V and spliceCG in the Yup'ik population as 3.7%, 0.0%, 0.3% and 0.7%, respectively, while in the AIAN cohort at SCF they were 11.0%, 2.7%, 2.2% and 1.4%, respectively (Table 13) [33]. The functional impact of these variants is unclear but some may be deleterious.

Discussion
Pharmacogenetics is a growing field that presents the opportunity to improve safety and clinical outcomes of currently available treatments using individual genomic data. The implementation of this practice in all populations requires the elucidation and comprehensive understanding of genetic variation and its impact on drug phenotypes. In order to optimize clinical therapy and minimize  Fohner et al. also reported the MAFs of CYP4F2*2, L519M, G185V and spliceCG in the Yup'ik population as 3.7%, 0.0%, 0.3% and 0.7%, respectively, while in the AIAN cohort at SCF they were 11.0%, 2.7%, 2.2% and 1.4%, respectively (Table 13) [33]. The functional impact of these variants is unclear but some may be deleterious.

Discussion
Pharmacogenetics is a growing field that presents the opportunity to improve safety and clinical outcomes of currently available treatments using individual genomic data. The implementation of this practice in all populations requires the elucidation and comprehensive understanding of genetic variation and its impact on drug phenotypes. In order to optimize clinical therapy and minimize adverse drug events in underserved Indigenous populations, these groups must be adequately represented in pharmacogenetic studies that identify genetic variation and inform on the dosing drugs that have established clinical associations with pharmacogene variation (e.g., abacavir, 6-mercaptopurine, warfarin, codeine), especially with narrow therapeutic index treatments.
Not only do Indigenous groups often have different allele frequencies compared to other global populations but marked differences in allele frequencies can also be found between subcultures within a given geographical region. Notable findings from the original studies highlighted in this review include the lower frequency of CYP2A6 loss-of-function alleles and a higher rate of nicotine metabolism in the NP AI, compared to a population of SW AI smokers [25]. With regards to CYP2C9, there was a lack of CYP2C9*2 in the Canadian Inuit population as well as many Amerindian groups including the Mexicanera, Huichol, Guarijío, Cora, Tarahumara and Purepecha [28,30,31]. CYP2C9*3 was also absent from the Canadian Inuit, Mexicanera, Seri, Purepecha and Huichol populations [28,30,31]. In contrast, CYP2C19*2 was found at a high MAF in Tarahumaras, compared to other Indigenous peoples of Mexico [37]. The relatively high frequencies for CYP2D6*4 and CYP2D6*41 in the CSKT and AI youth may be important to consider with CYP2D6 substrates such as codeine, tamoxifen and antidepressants [32,34]. Conversely, certain subgroups in the Amerindian population (Tepehuano, Purepecha, Tojolabal, Tzotzil, Mexicanera, Cora and Guarijío) had a low proportion of CYP2D6 variants conferring PM status [37,[41][42][43]. About 15% of the CSKT population would be expected to express CYP3A5, based on the CYP3A5*1 allele frequency [32]. In the Yup'ik AN population, the CYP4F2*3 variant is expressed at a frequency of 50.9%, one of the highest MAFs seen across global populations for this SNV [33].
Indigenous populations of North America may also have novel P450 gene variation not seen in other populations of the world that can potentially influence drug phenotype. The allele frequencies of known and recently reported CYP2C9 variants in the AN Yup'ik population illustrate this scenario. The CYP2C9*2 and *3 alleles that define the CYP2C9 PM phenotype in the European population are found at very low frequencies in the Yup'ik population. The novel and relatively common M1L and N218I variants found in the Yup'ik population, in addition to CYP2C9*2 and *3 alleles, that are predicted to confer a CYP2C9 PM phenotype. Importantly, if only the allele frequencies known to be clinically relevant in the European population are applied to the Yup'ik population, an individual homozygous or heterozygous for the M1L or N218I variants would be classified as a CYP2C9 EM and improper warfarin dosing could result in adverse events for these individuals. This potential for misclassification and inappropriate drug dosing has also been described of African populations, where the CYP2C9*8 allele contributes to the PM phenotype [169]. This was also suggested by the work of de Andrés et al. where the CYP2C9, CYP2C19, or CYP2D6 phenotype for some Amerindians could not be accurately predicted based on genotype, possibly due to the presence of novel rare variation in these pharmacogenes [23].
Further studies are needed to identify and establish the allele frequencies of both known and novel variants in Indigenous populations, particularly in all P450 genes that encode enzymes that have a clinically significant impact on drug disposition. For example, despite the fact that CYP1A2 is a highly polymorphic enzyme important for the metabolism of many clinical drugs, no studies to date have assessed the frequencies of CYP1A2 allelic variants in Indigenous peoples of Canada or AIAN populations. Moving forward, it will also be important to improve genotyping and sequencing quality as well as increase study sample size, as these will help improve imputation and haplotype estimation in Indigenous populations, which may lead to the discovery of additional P450 SNVs or structural variants. Furthermore, there are problematic inconsistencies in population description for genetic studies, which is a recognized problem in the field [170]. In addition, the studies reported in this review do not fully capture the diversity of AI and other Indigenous tribes in North America. As seen in Figure 1, we found no studies from the US that were east of the Mississippi, leaving significant uncertainty about P450 genetic diversity for these people and drug phenotype relationships. In 2008, Jaja et al. conducted a systemic review of P450 variation in Indigenous and Native American Populations that identified ten original studies, of which six of were from Canada, four from North, Central and South America and none in AIAN [9]. This review identified twenty-seven original studies, with six in AIAN, seven in Indigenous people of Canada and fourteen in Amerindian populations of Mexico.
One method for increasing representation of Indigenous people in genetic studies is first to form collaborative research partnerships, in which community partners share control of the research process and apply the values and procedures of community-based participatory research to establish research priorities and acceptable conditions under which the research will occur [10]. We formed a research network involving three tribal organizations and three universities in 2010, to address the dearth of information about pharmacogenomics in AIAN populations. The research network built on several years of research and partnership development at three research sites. (1) Investigators at the Center for Alaska Native Health Research had established research partnerships with the Yukon-Kuskokwim Health Corporation, serving 23,000 Yup'ik people in southwestern Alaska and with several communities in the Yukon-Kuskokwim River Delta [171]. (2) SCF, a tribally owned and operated healthcare organization had established a Research Department and developed collaborative projects with investigators from the University of Washington. SCF is based in Anchorage and provides healthcare services to 65,000 AIAN customer-owners, serving about 55% of the total AN population in Alaska [171,172]. (3) Investigators at University of Montana had established partnerships with the CSKT of the Flathead Indian Reservation in northwestern Montana to pursue research with the Bitterroot Salish, Upper Pend d'Oreille and Kootenai tribes. There are >7900 enrolled CSKT members, with a large number of descendants [171,173]. These research partnerships provided the foundation for the Northwest-Alaska Pharmacogenomic Research Network, academic-tribal partnerships initiated with support from the National Institutes of Health. This review highlights findings of research derived from collaborations we have initiated, as well as additional relevant publications identified in the literature search.
The limited data on P450 genetic variation in Indigenous North American populations translates to missed opportunities for optimizing care. Interestingly, the data that does exist for Indigenous North American people suggests that they have unique genetic variation profiles that may critically impact their response to drug therapy. Without a complete understanding of this population's unique pharmacogene variation profile, Indigenous people may not derive the same benefit from genomics-based precision medicine as the European population. The populations included in pharmacogenetic research stand to gain the most from clinical trials findings that establish test validity and utility. A better understanding of the unique P450 pharmacogenetic variation in Indigenous populations is needed if these communities are to be included in clinical decisions regarding personalized drug therapy and policies surrounding precision medicine.