Worldwide Distribution of Cytochrome P450 Alleles: A Meta‐analysis of Population‐scale Sequencing Projects

Abstract

Genetic polymorphisms in cytochrome P450 (CYP) genes can result in altered metabolic activity toward a plethora of clinically important medications. Thus, single nucleotide variants and copy number variations in CYP genes are major determinants of drug pharmacokinetics and toxicity and constitute pharmacogenetic biomarkers for drug dosing, efficacy, and safety. Strikingly, the distribution of CYP alleles differs considerably between populations with important implications for personalized drug therapy and healthcare programs. To provide a global distribution map of CYP alleles with clinical importance, we integrated whole‐genome and exome sequencing data from 56,945 unrelated individuals of five major human populations. By combining this dataset with population‐specific linkage information, we derive the frequencies of 176 CYP haplotypes, providing an extensive resource for major genetic determinants of drug metabolism. Furthermore, we aggregated this dataset into spectra of predicted functional variability in the respective populations and discuss the implications for population‐adjusted pharmacological treatment strategies.

1.

Study Highlights.

1.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

☑ The human CYP supergene family harbors a multitude of single nucleotide variants (SNVs) and copy number variations. The frequencies of these pharmacogenetically important polymorphisms have mostly been studied in relatively small populations.

WHAT QUESTION DID THIS STUDY ADDRESS?

☑ This study utilized next‐generation sequencing data from 56,945 unrelated individuals from five major populations to derive representative haplotype frequencies. In total, we report the frequencies of 176 alleles distributed over the 12 CYP genes with highest relevance for human drug metabolism and aggregate this dataset into worldwide patterns of predicted functional variability.

WHAT THIS STUDY ADDS TO OUR KNOWLEDGE

☑ Our analyses quantify the large extent of genetic variability between major populations on an unprecedented scale and reveal unexpectedly large interethnic differences. This genetic variability is expected to result in major functional differences, especially for the metabolism of CYP2C19, CYP2D6, and CYP3A substrates.

HOW THIS MIGHT CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

☑ This study reveals clinically interethnic differences in human drug metabolism that provide important information for the guidance of personalized drug therapy and the design of clinical trials.

2.

Drug response is highly variable between individuals, resulting in 40–70% of patients exhibiting a lack of efficacy of pharmacological treatment or adverse drug reactions.1, 2 Importantly, it is estimated that 15–30% of this variability is caused by genetic polymorphisms.2 Cytochrome P450 (CYP) enzymes and their roles in drug metabolism have been extensively studied and constitute major determinants of drug response, as they are responsible for 80% of phase 1 drug metabolism and 65–70% of drug clearance.3, 4 Furthermore, metabolism by polymorphic phase 1 enzymes, primarily CYPs, has been implicated in an increased risk of drugs to cause adverse drug reactions.5

The human CYP supergene family consists of 57 genes, however, 12 of the encoded enzymes are responsible for more than 75% of all phase 1 drug oxidation reactions.4 Importantly, these CYP genes are highly polymorphic and harbor a large repertoire of single nucleotide variants (SNVs) and copy number variations. This diversity is primarily based on low evolutionary constraints due to the lack of essential endogenous functions of the encoded gene products and genetic drift. Yet, although the vast majority of such polymorphisms are rare with minor allele frequencies (MAFs) <1%,6 pharmacogenetic testing in the clinics is currently restricted to validated and experimentally characterized variants for the derivation of qualified predictions about phenotypic consequences of the observed genetic variation patterns.7

Each variant is identified by its genomic coordinate in comparison to a curated reference sequence and is named according to the Human Genome Variation Society nomenclature, which represents the current standard in clinical diagnostics.8 Furthermore, each SNV, microdeletion, or insertion of <50 bp that has been observed in multiple genomes is assigned a unique rs number by database‐single‐nucleotide polymorphism, which identifies the respective variant. On the basis of this variant information, a unified and standardized CYP allele nomenclature system was established that integrates variant combinations into more accessible and human readable haplotypes designated by “star alleles.” This system is made available to the research community through a peer‐reviewed, regularly updated website9 (http://www.cypalleles.ki.se).

In the last decades, a plethora of studies analyzed the links between genetic variants in CYP genes and drug responses. Although these studies have provided important information about the frequencies of clinically important CYP allele variants and their interethnic differences,10, 11 their significance is limited due to their relatively small sample sizes and their focus on small, nonstandardized panels of specific variants of interest. Recently, the increasing dissemination of next‐generation sequencing technologies and the implementation of population‐scale sequencing projects presents a paradigm shift that allows to derive a comprehensive and consolidated overview of the genetic diversity and interethnic variability of CYP alleles across worldwide populations.

By integrating exome sequencing data from 56,945 unrelated individuals provided by the Exome Aggregation Consortium with whole genome and linkage information from the 1000 Genomes Project, we derive representative frequencies12, 13 of 176 alleles distributed over the 12 CYP genes with highest relevance for human drug metabolism in five major populations (Europeans, Africans, South Asians, East Asians, and admixed Americans). The resulting data, to our knowledge, is the most comprehensive overview of CYP allele distributions published to date and provides important information for the guidance of population‐specific genotyping strategies.

3. RESULTS

In this study, we analyzed the prevalence of 176 alleles distributed over the 12 CYP genes that jointly account for >75% of human phase 1 drug metabolism.4 In CYP2A6, we analyzed a total of 23 variants, of which 10 are rare in every population analyzed (Figure 1 a, Table 1). CYP2A6*17 (MAF = 11.2%), CYP2A6*23 (MAF = 1.5%), CYP2A6*25 (MAF = 1.4%), and CYP2A6*28 (MAF = 2%) are only identified in Africans, whereas East Asians are the only population to harbor CYP2A6*7 (MAF = 12.9%) and CYP2A6*19 (MAF = 1.2%). CYP2A6*9 is the most abundant allele in both Asian populations and admixed Americans accounting for 33.3%, 42%, and 49.2% of all variant alleles, respectively, whereas CYP2A6*35 is the most abundant variant in Europeans (42.1% of all variants). Frequencies of the deletion allele CYP2A6*4 could not be derived from exome sequencing data and, therefore, we used previously published data to approximate CYP2A6*4 frequencies in the respective populations.14, 15

Figure 1.

Distribution of the relative contributions of major cytochrome P450 (CYP) alleles in European, African, Asian, and American populations. Pie charts showing the relative contribution of common variants in CYP2A6 (a), CYP2B6 (b), CYP2C8 (c), CYP2C9 (d), CYP2C19 (e), CYP2D6 (f), CYP3A4 (g), and CYP3A5 (h) in five major populations. Only variants with a minor allele frequency in the respective population above 1% are shown, whereas all other variants analyzed in this study are summarized as “rare” (grey).

Table 1.

Important variant and allele frequencies of the human CYP2A6 gene

Allele	Defining variants	Variant type	Allele frequencies in indicated populations, %					Functional consequence
			EUR	AFR	EAS	SAS	AMR
*1	None		64.6	65.1	30.8	65.6	71.9	Normal
*2	rs1801272	Missense (L160H)	2.3	0.5	0	1.1	1.2	Inactive
*4	CYP2A6 deleted		1	1.5	17	7	4	Inactive
*5	rs5031017	Missense (G479V)	0	<0.1	0.1	<0.1	< 0.1	Inactive
*6	rs4986891	Missense (R128Q)	0	0	<0.1	0	0	Decreaseda
*7	rs5031016	Missense (I471T)	<0.1	<0.1	12.9	0.3	0.3	Decreaseda
*8	rs28399468	Missense (R485L)	0.3	0.3	0.3	0.3	0.3	Normal
*9	rs28399433	TATA box	11.1	8.3	23.0	14.4	13.8	Decreased
*10	rs5031016, rs28399468	Missense (I471T, R485L)	<0.1	<0.1	0.3	0.3	0.3	Decreasedb
*14	rs28399435	Missense (S29N)	1.4	0.8	<0.1	3.5	1.2
*16	rs56256500	Missense (R203S)	0	<0.1	0	<0.1	<0.1
*17	rs28399454	Missense (V365M)	0	11.2	0	0	0.6	Decreased
*18	rs1809810	Missense (Y392F)	1.5	0.6	1.2	1.4	1.1	Decreaseda
*19	rs5031016, rs1809810	Missense (I471T, Y392F)	<0.1	<0.1	1.2	0.3	0.3
*21	rs6413474	Missense (K476R)	2.8	0.2	<0.1	1.9	0.3	Decreasedb
*22	rs60605885, rs60563539	Missense (D158E, L160I)	0	0	0	0	<0.1
*23	rs56256500	Missense (R203C)	0	1.5	0	<0.1	<0.1	Decreased
*24	rs143731390, rs72549435	Missense (N438Y, V110L)	0	0.9	<0.1	<0.1	<0.1
*25	rs28399440	Missense (F118L)	0	1.4	0	0	0
*26	rs59552350	Missense (S131A)	0	0.4	0	0	0	Decreased
*28	rs28399463, rs8192730	Missense (N418D, E419D)	0	2	0.1	<0.1	0.2
*35	rs143731390	Missense (N438Y)	14.9	5.9	12.8	3.7	4.5	Decreased
*39	rs143690364	Missense (V68M)	0	<0.1	0	<0.1	0	Decreasedb
*41	rs140471703	Missense (R265Q)	0	0.2	<0.1	0	<0.1	Inactive

AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians.

For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

^aIndicates alleles whose functionality assessment is based solely on in vitro data.

^bIndicates alleles whose functionality assessment is based solely on in vivo data.

We assessed 25 CYP2B6 alleles of which 8 are common in at least one population analyzed (Figure 1 b, Table 2). CYP2B6*18 (MAF = 7.1%) and CYP2B6*16 (MAF = 6.5%) are restricted to Africans and CYP2B6*4 is only common in South Asian populations (MAF = 1.8%). CYP2B6*9 is the most abundant allele in all populations analyzed accounting for 40.1–68.6% of all variants in the major populations. CYP2B6*5 is highly abundant in Europeans (MAF = 12.8%), less prevalent in Africans (MAF = 2.6%), South Asians (MAF = 8.2%), and admixed Americans (MAF = 4.5%), and virtually absent in East Asians. Notably, the variant rs2279343, which defines CYP2B6*4 and is part of CYP2B6*6, was not called in the 1000 Genome Project and, therefore, no linkages with this variant could be computed. Thus, to derive information about the CYP2B6*4 and *6 haplotype frequencies, we estimated the co‐occurrence of the two variants based on previous studies (90% of rs2279343 in combination with rs3745274 as CYP2B6*6 and 10% without rs3745274 as CYP2B6*4).16

Table 2.

Important variant and allele frequencies of the human CYP2B6 gene

Allele	Defining variants	Variant type	Allele frequencies in indicated populations, %					Functional consequence
			EUR	AFR	EAS	SAS	AMR
*1	None		61.1	37.9	75.5	45	57.8	Normal
*2	rs8192709	Missense (R22C)	4.9	3.7	4.5	3.0	2.9
*3	rs45482602	Missense (S259R)	<0.1	<0.1	0	0.3	0.1
*4	rs2279343	Missense (K262R)	0.4	0.6	0.3	1.8	0.3	Increasedb
*5	rs3211371	Missense (R487C)	12.8	2.6	0.1	8.2	4.5
*6	rs2279343, rs3745274	Missense (K262R, Q172H)	3.4	5.8	2.7	15.8	3	Decreased
*7	rs2279343, rs3745274, rs3211371	Missense (K262R, Q172H, R487C)	0	0	0	0	0
*8	rs12721655	Missense (K139E)	<0.1	<0.1	0	<0.1	<0.1	Decreaseda
*9	rs3745274	Missense (Q172H)	15.6	30.9	16	23.2	29
*10	rs8192709, rs34883432	Missense (R22C, Q21L)	<0.1	<0.1	0	0.1	0.3
*11	rs35303484	Missense (M46V)	<0.1	0.1	0	0.2	0.1	Decreaseda
*12	rs36060847	Missense (G99E)	0	<0.1	0	0	<0.1	Decreaseda
*13	rs2279343, rs3745274, rs12721655	Missense (K262R, Q172H, K139E)	<0.1	<0.1	0	<0.1	<0.1
*14	rs35773040	Missense (R140Q)	0.5	<0.1	<0.1	0.3	<0.1	Decreaseda
*15	rs35979566	Missense (I391N)	0.2	0.2	0	<0.1	0.2	Decreaseda
*16	rs2279343 & rs28399499	Missense (I328T)	0	6.5	0	<0.1	0.3	Decreased
*17	rs33973337, rs33980385, rs33926104	Missense (T26S, D28G, R29T)	0	<0.1	0	0	0
*18	rs28399499	Missense (I328T)	0	7.1	0	<0.1	0.3	Decreaseda
*19	rs34826503	Missense (R336C)	0	0.3	<0.1	0	<0.1	Decreaseda
*20	rs36056539	Missense (T168I)	0	0.1	0	0	<0.1	Decreaseda
*21	rs35010098	Missense (P428T)	0	<0.1	0	0	0	Decreaseda
*22	rs34223104	Regulatory	0.9	3.6	0.2	1.8	0.7	Increaseda
*23	rs3211369	Missense (M459V)	0	0	<0.1	<0.1	0
*26	rs2279343, rs3745274, rs3826711	Missense (K262R, Q172H, P167A)	0	0	0.5	0	0	Decreasedb
*27	rs36079186	Missense (M198T)	0	0.2	0	0.1	<0.1	Decreaseda
*28	rs34097093	Stop‐gain (R378X)	0	<0.1	0	0	<0.1	Inactive

AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians.

For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

^aIndicates alleles whose functionality assessment is based solely on in vitro data.

^bIndicates alleles whose functionality assessment is based solely on in vivo data.

Only three of nine CYP2C8 alleles are common in at least one of the super populations assessed (Figure 1 c, Table 3). In Europeans and admixed American populations, only CYP2C8*3 (MAF = 11.2% and 6.7% in Europeans and Americans, respectively) and CYP2C8*4 (MAF = 6% and 2.3% in Europeans and Americans, respectively) are common. In contrast, CYP2C8*2 constitutes the most prevalent allele in Africans (MAF = 15.9%, corresponding to 82.9% of all variant CYP2C8 alleles in Africans) but is not detected in Europeans and East Asians. Surprisingly, CYP2C8 is largely invariant in populations in East Asia and no variant with an allele frequency >1% was detected. This is in pronounced contrast to South Asian populations, in which the *2 (MAF = 1.9%), *3 (MAF = 4%), and *4 (MAF = 1.5%) alleles are common.

Table 3.

Important variant and allele frequencies of important pharmacogenes in the human CYP2C family

Allele	Defining variants	Variant type	Allele frequencies in indicated populations, %					Functional consequence
			EUR	AFR	EAS	SAS	AMR
CYP2C8
*1	None		82.8	80.8	98.7	92.6	90.1	Normal
*2	rs11572103	Missense (I269F)	0	15.9	0	1.9	0.9	Decreaseda
*3	rs10509681, rs11572080	Missense (K399R, R139K)	11.2	2.1	<0.1	4	6.7	Decreaseda
*4	rs1058930	Missense (I264M)	6.0	1.1	0	1.5	2.3	Decreaseda
*5	rs72558196	Frameshift	0	0	0.2	0	0	Inactive
*6	rs142886225	Missense (G171S)	0	0	0.5	0	0	Normal
*7	rs72558195	Stop‐gain (R186X)	0	<0.1	<0.1	<0.1	0	Inactive
*11	rs78637571	Stop‐gain (E274X)	0	0	0.4	0	<0.1	Inactive
*12	rs3832694	Inframe deletion (461delV)	0	0	<0.1	0	0
*14	rs188934928	Missense (A238P)	0	0	0.1	<0.1	0	Decreaseda
CYP2C9
*1	None		81.8	78.4	95.7	81.3	88.9	Normal
*2	rs1799853	Missense (R144C)	11.7	2.4	<0.1	4.6	6.6	Decreaseda
*3	rs1057910	Missense (I359L)	5.6	1.3	3.4	11.3	3.6	Decreased
*4	rs56165452	Missense (I359T)	0	<0.1	0	0	0
*5	rs28371686	Missense (D360E)	0	1.3	0	0	<0.1	Decreased
*6	rs9332131	Frameshift	0	1	0	0	<0.1	Inactive
*7	rs67807361	Missense (L19I)	0	0	0	0.3	0
*8	rs7900194	Missense (R150H)	0	5.6	<0.1	<0.1	0.2	Decreased
*9	rs2256871	Missense (H251R)	0	7.5	<0.1	<0.1	0.2
*11	rs28371685	Missense (R335W)	0.5	2.1	<0.1	0.2	0.2	Decreased
*12	rs9332239	Missense (P489S)	0.2	<0.1	0	<0.1	0.1
*13	rs72558187	Missense (L90P)	0	0	0.2	0	0	Decreased
*14	rs72558189	Missense (R125H)	0	<0.1	<0.1	2.0	<0.1
*15	rs72558190	Stop‐gain (S162X)	0	0	<0.1	0	0
*16	rs72558192	Missense (T299A)	0	0	0.3	0	0
*29	rs182132442	Missense (P279T)	0.1	0	0.2	<0.1	<0.1
*30	rs781583846	Missense (A477T)	<0.1	<0.1	<0.1	<0.1	<0.1
*31	rs57505750	Missense (I327T)	0	0.2	0	0	0
*33	rs200183364	Missense (R132Q)	0	<0.1	0	<0.1	0
*36	rs114071557	Start lost	0	0.2	<0.1	<0.1	<0.1
*42	rs12414460	Missense (R124Q)	<0.1	<0.1	<0.1	0	<0.1
*44	rs200965026	Missense (T130M)	0	0	<0.1	0	<0.1
*45	rs199523631	Missense (R132W)	<0.1	<0.1	0	< 0.1	<0.1
CYP2C19
*1	None		59.2	44.5	60.5	51.9	77	Normal
*2	rs4244285	Splicing defect	18.3	18.1	31.0	34.0	10.1	Inactive
*3	rs4986893	Stop‐gain (W212X)	<0.1	<0.1	6.7	0.4	<0.1	Inactive
*4	rs28399504	Start lost	0	<0.1	<0.1	<0.1	0.2	Inactive
*5	rs56337013	Missense (R433W)	0	0	0	<0.1	0	Inactive
*6	rs72552267	Missense (R132Q)	0	0	<0.1	0	<0.1	Inactive
*7	rs72558186	Splicing defect	0	0	0	<0.1	0	Inactiveb
*8	rs41291556	Missense (W120R)	<0.1	<0.1	0	<0.1	<0.1	Inactive
*9	rs17884712	Missense (R144H)	0	1.2	0	<0.1	<0.1
*10	rs6413438	Missense (P227L)	0	0.4	<0.1	0	<0.1	Decreaseda
*12	rs55640102	Stop‐lost (X491C)	0	<0.1	0	0	0	Decreaseda
*13	rs17879685	Missense (R410C)	0	1.6	0	<0.1	0.1
*15	rs17882687	Missense (I19L)	0	2.0	0	<0.1	<0.1
*16	rs192154563	Missense (R442C)	0	<0.1	0	<0.1	0
*17	rs12248560	Regulatory	22.4	23.5	1.5	13.6	12.0	Increased
*22	rs140278421	Missense (R186P)	0	0.1	0	0	<0.1
*23	rs118203756	Missense (G91R)	0	0	<0.1	0	0
*24	rs118203757	Missense (R335Q)	0	<0.1	0	<0.1	<0.1
*25	rs118203759	Missense (F448L)	0	0	0	0	0
*27	rs7902257	Regulatory	0.1	8.3	0.1	0	0.3	Decreaseda

AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians.

For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

^aIndicates alleles whose functionality assessment is based solely on in vitro data.

^bIndicates alleles whose functionality assessment is based solely on in vivo data.

Genetic variability of CYP2C9 is dominated by the *2 allele in Europeans (MAF = 11.7%) and admixed Americans (MAF = 6.6%), whereas the major alleles in Asian populations and Africans are *3 (MAF = 3.4% in East Asians and 11.3% in South Asians) and *9 (MAF = 7.5%), respectively (Figure 1 d, Table 3). The spectrum of CYP2C9 variants is highly population‐specific with CYP2C9*5, *6, *8, *9, and *11 being restricted to African populations and CYP2C9*14 being almost uniquely identified in South Asians.

Overall, we analyzed 19 variants in CYP2C19, of which CYP2C19*17, and CYP2C19*2 are the major variants (Figure 1 e, Table 3). In Europeans, Africans, and admixed Americans, the *17 promoter variant, which results in increased CYP2C19 transcription most probably due to the modulation of GATA4 binding,17, 18 is most abundant, accounting for 42–55% of all variant alleles, whereas the *2 loss‐of‐function variant that causes aberrant splicing is most prevalent in Asian populations (78.5% and 70.7% of all variant alleles in East and South Asians, respectively). Further CYP2C19 alleles are highly specific to individual populations. The stop‐gain variant CYP2C19*3 is abundant exclusively in East Asians (MAF = 6.7%), whereas CYP2C19*9 (MAF = 1.2%), *13 (MAF = 1.6%), *15 (MAF = 2%), and *27 (MAF = 8.3%) are restricted to Africans.

The CYP2D6 gene locus is complex and highly polymorphic, harboring a multitude of common genetic variants with clinical importance (Figure 1 f, Table 4). The CYP2D6*2 allele is most abundant in European, African, South Asian, and admixed American populations and the most prevalent haplotype worldwide. The variant defining CYP2D6*4 (rs3892097), which causes a splicing defect and inactivation of the CYP2D6 gene product, is present in all analyzed populations with MAFs between 11.6% and 15.7%, except in East Asians in which the frequency of the allele was <1%. CYP2D6*10 is almost exclusively found in African, South Asian, and, most notably, East Asian populations in which *10 constitutes the most common CYP2D6 allele (MAF = 58.7%). CYP2D6*3 (MAF = 4.1%) and CYP2D6*6 (MAF = 2.2%) are only found in individuals of European ancestry. CYP2D6*17 (MAF = 19.7%), CYP2D6*29 (MAF = 9.2%), and CYP2D6*43 (MAF = 2%) are distinctive haplotypes for Africans and CYP2D6*14 was only found in East Asian populations (MAF = 1.6%). As quantification of certain genetic rearrangements can be difficult to assess with short‐read sequencing based methodologies, we included frequency information for the duplicated alleles CYP2D6*1xN and CYP2D6*2xN as well as for the deletion CYP2D6*5 from the published literature. Duplications of CYP2D6 occur with frequencies of 1–2% in whites and Asians but are more common in certain African populations, in which their frequency can be up to 29%, as previously reported.19, 20 For deletions of CYP2D6, we assumed frequencies between 1% and 7% for the analyzed populations based on previous extensive meta‐analyses.20

Table 4.

Important variant and allele frequencies of the human CYP2D6 gene

Allele	Defining variants	Variant type	Allele frequencies in indicated populations, %					Functional consequence
			EUR	AFR	EAS	SAS	AMR
*1	None		33.1	9.3	13.6	25.8	40.2	Normal
*1xN	Amplification of *1		1	3.3	1	0.5	0.5	Increased
*2	rs16947, rs1135840	Missense (R296C, S486T)	34.3	26.7	14	36.2	32.7	Normal
*2xN	Amplification of *2		1.3	6	1	1	0.5	Increased
*3	rs35742686	Frameshift	4.1	0.3	0	0.1	0.3	Inactive
*4	rs3892097	Splicing defect	15.5	11.9	0.4	11.6	15.7	Inactive
*5	CYP2D6 deleted		3	4	6.5	2	3	Inactive
*6	rs5030655	Frameshift	2.2	0.3	0	0.1	0.4	Inactive
*7	rs5030867	Missense (H324P)	0	<0.1	0	0.8	<0.1	Inactiveb
*8	rs5030865	Stop‐gain (G169X)	0	<0.1	0	<0.1	0	Inactive
*9	rs5030656	Inframe deletion (K281del)	1.6	0.4	0	0.2	1.3	Decreased
*10	rs1065852, rs1135840	Missense (P34S, S486T)	0.2	3.2	58.7	6.5	0	Decreased
*11	rs201377835	Splicing defect	0	<0.1	0	0	0	Inactiveb
*12	rs5030862	Missense (G42R)	0	<0.1	0	0	0	Inactive
*14	rs5030865	Missense (G169R)	0	0	1.6	<0.1	0	Inactive
*17	rs16947, rs28371706	Missense (R296C, T107I)	<0.1	19.7	0	0.1	0.7	Decreased
*29	rs16947, rs1135840, rs61736512, rs59421388	Missense (R296C, S486T, V136I, V338M)	0	9.2	<0.1	<0.1	0.4	Decreased
*33	rs28371717	Missense (A237S)	0.7	0.2	0	0.7	0.1	Normal
*41	rs28371725	Splicing defect	3.0	3.0	3.0	13.5	3.5	Decreased
*42	rs72549346	Frameshift	0	0.2	0	0	<0.1	Inactive
*43	rs28371696	Missense (R26H)	<0.1	2.0	<0.1	0.8	0.2
*53	rs1135822, rs1135823	Missense (F120I, A122S)	0	<0.1	<0.1	<0.1	0.5	Increaseda
*62	rs730882171	Missense (R441C)	<0.1	<0.1	<0.1	<0.1	<0.1	Inactive

AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians.

For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

^aIndicates alleles whose functionality assessment is based solely on in vitro data.

^bIndicates alleles whose functionality assessment is based solely on in vivo data.

CYP3A4, which metabolizes around one‐third of all medications4 as well as endogenous steroid hormones, harbors only few common genetic variants (Figure 1 g, Table 5). Of the 19 CYP3A4 variants we assessed, only 5 were common in at least one of the major populations. In Europeans and admixed Americans, CYP3A4*22 is the major allele (MAF = 5% and 2.6%, respectively) with *3 and *2 contributing to the genetic variability in the former. In contrast, CYP3A4*15 (MAF = 2.5%) and CYP3A4*18 (MAF = 1.9%) constitute the only common CYP3A4 alleles in Africans and East Asians, respectively. In South Asian populations, no common CYP3A4 variant with MAFs >1% are present.

Table 5.

Important variant and allele frequencies of important pharmacogenes in the human CYP3A family

Allele	Defining variants	Variant type	Allele frequencies in indicated populations (in %)					Functional consequence
			EUR	AFR	EAS	SAS	AMR
CYP3A4
*1	None		91.5	96.6	97	99.1	96.9	Normal
*2	rs55785340	Missense (S222P)	1.1	0	0	0	0
*3	rs4986910	Missense (M445T)	2.1	0.1	0	0	0.2
*4	rs55951658	Missense (I118V)	0	0	0.6	<0.1	<0.1
*5	rs55901263	Missense (P218R)	0	0	<0.1	0	0
*6	rs4646438	Frameshift	0	0	0.2	<0.1	<0.1
*7	rs56324128	Missense (G56D)	0.1	0	0	0	0
*8	rs72552799	Missense (R130Q)	0.1	0	0	<0.1	<0.1	Decreaseda
*9	rs72552798	Missense (V170I)	0	0	0	0	<0.1
*10	rs4986908	Missense (D174H)	<0.1	0.2	<0.1	0.1	<0.1
*11	rs67784355	Missense (T363M)	0	<0.1	<0.1	<0.1	0	Decreaseda
*12	rs12721629	Missense (L373F)	0	0.3	0	<0.1	<0.1
*13	rs4986909	Missense (P416L)	0	0	0	0	<0.1	Decreaseda
*15	rs4986907	Missense (R162Q)	0	2.5	0	<0.1	0.2
*16	rs12721627	Missense (T185S)	0	0	0.1	0	0	Decreaseda
*18	rs28371759	Missense (L293P)	0	0.2	1.9	0	<0.1	Decreased
*19	rs4986913	Missense (P467S)	0	0	0	<0.1	0
*20	rs67666821	Frameshift	0	<0.1	0	0	<0.1	Inactive
*22	rs35599367	Intronic	5.0	<0.1	0	0.6	2.6	Decreasedb
*26	rs138105638	Stop‐gain (R268X)	0	<0.1	0	<0.1	<0.1	Inactive
CYP3A5
*1	None		5.3	56.3	27.7	33.2	17.5	Normal
*2	rs28365083	Missense (T398N)	0.1	<0.1	0	0	<0.1
*3	rs776746	Splicing defect	94.3	18.0	71.3	66.8	79.7	Inactive
*4	rs56411402	Missense (Q200R)	0	0	0.3	0	0
*5	rs55965422	Splicing defect	0	0	0.7	<0.1	0
*6	rs10264272	Splicing defect	0.3	15.4	0	0	2.3	Inactive
*7	rs41303343	Frameshift	0	10.3	<0.1	<0.1	0.4
*8	rs55817950	Missense (R28C)	0	0	<0.1	0	0	Decreaseda

AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians.

For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

^aIndicates alleles whose functionality assessment is based solely on in vitro data.

^bIndicates alleles whose functionality assessment is based solely on in vivo data.

The CYP3A5 locus is highly variable across all human populations, yet harbors only few common genetic variants (Figure 1 h, Table 5). CYP3A5*3, an SNV causing alternative splicing and protein truncation, results in the almost complete abrogation of CYP3A5 activity.21 This allele is highly abundant in South Asians, East Asians, admixed Americans, and Europeans and constitutes the major allele with frequencies of 66.8%, 71.3%, 79.7%, and 94.3%, respectively. In contrast, the variant spectrum of CYP3A5 in Africans is distinctly different. Whereas the frequency of CYP3A5*3 is much lower (MAF = 18%), Africans also harbor the CYP3A5*6 (MAF = 15.4%) and *7 (MAF = 10.3%) alleles with high frequencies.

Besides those CYP genes with the highest pharmacogenetic importance, we also analyzed the allelic variant profiles of CYP1A1, CYP1A2, CYP2E1, and CYP4F2, genes with a lower extent of functionally important polymorphisms. The frequencies of the respective alleles for these genes can be found in Supplementary Tables S1–S4.

Among the pharmacogenetically most important CYPs, CYP2D6 is the most variable with cumulative allele frequencies between 59.8% in admixed Americans and 90.7% in Africans (Figure 2 a). In contrast, CYP3A4 is the most conserved gene with <10% of alleles harboring identified variant haplotypes. Notably, the largest differences in variability between populations are observed for CYP3A5 (cumulative allele frequencies 44–95%), CYP2A6 (28–69%), and CYP2B6 (24–62%).

Figure 2.

The genetic variability and their functional consequences differ drastically between major cytochrome P450 (CYP) genes and across populations. (a) Cumulative frequencies of all major variant alleles in Europeans (EUR; blue), Africans (AFR; red), East Asians (EAS; green), South Asians (SAS; purple), and admixed Americans (AMR; turquoise) are shown for each major CYP gene. Overall, CYP2D6 constitutes the most variable gene, whereas CYP3A4 is most conserved. (b) The expected functional consequences of allelic distributions across world populations are shown. Frequencies of haplotypes with decreased (blue), increased (green), and normal (red) functionality, as defined in Tables 1, 2, 3, 4, 5, were aggregated for each gene and population, revealing the spectrum of functional variability in major world populations.

Importantly, aggregation of allelic frequency data by their functional consequences revealed major differences in predicted functionality patterns across the populations. Functionally impaired CYP2A6 alleles are primarily found in East Asian populations where 67.5% of alleles are defective, whereas the frequencies of such alleles pivot around 30% in the other populations analyzed (Figure 2 b). Although the CYP2B6 locus is highly variable, the fraction of alleles that result in functional consequences is considerably smaller. In Africans and South Asians, around 20.5% and 16.7% of CYP2B6 alleles are expected to result in decreased activity, respectively. Between 7% and 20% of CYP2C8 and CYP2C9 alleles exhibit decreased functionality across major world populations, with the exception of East Asians for which <1% and 4% of functional CYP2C8 and CYP2C9 variants are observed, respectively. Of the genes analyzed, CYP2C19 harbors the highest frequency of increased activity alleles (CYP2C19*17), ranging from 1.5% in East Asians to 22.4% and 23.5% in Europeans and Africans, respectively.

CYP2D6 constitutes the most complex CYP locus known to harbor a large number of distinctly different common haplotypes with important clinical implications. In this study, we found that between 25.3% and 70.3% of analyzed alleles contained variant combinations with no or reduced functional activity. Furthermore, the CYP2D6 gene is commonly duplicated in 1.5–9.3% of alleles, causing increased metabolic activity toward the respective substrates. The enzymes encoded by the two major genes in the CYP3A family, CYP3A4 and CYP3A5, exhibit similar metabolic capabilities22 but drastically different variation profiles. Although CYP3A4 harbored only few common functional variants, the high prevalence of CYP3A5*3 causing alternative splicing resulted in the major fraction of alleles showing severely reduced functional activity (Figure 2 b).

4. DISCUSSION

To date, a multitude of studies have analyzed the frequencies of CYP alleles in populations around the world. Yet, most studies analyzed only the prevalence of a selected subset of SNVs in a particular gene and only in one or few subpopulations of relatively small sample size. To overcome these limitations, a variety of studies consolidated frequency information from multiple studies to allow a broader overview of the true frequencies between populations. Fricke–Galindo et al.23 performed a meta‐analysis integrating variant frequency data of seven CYP2C19 alleles from 138 studies. Furthermore, a recent extensive meta‐analysis of 173 reports revealed the spectrum of allele frequencies and predicted functional consequences across major populations for CYP2D6.20 However, meta‐analyses are limited to a small number of haplotypes for which several reports are available and integration of multiple studies can be problematic due to differences in underlying genotyping strategies and differentially designed assay panels.24 For instance, in a study by Griese et al.,25 which was included in the highlighted CYP2D6 meta‐analysis, the prevalence of 16 CYP2D6 haplotypes were analyzed in a Ghanaian population by allele‐specific polymerase chain reaction, including the normactive CYP2D6*2 haplotype, which is defined by two variants (rs16947 and rs1135840).25 Yet, in combination with two additional SNVs (rs61736512 and rs59421388), which were not assessed in the respective study, these two variants constitute CYP2D6*29, a haplotype that exhibits reduced activity. Thus, the *2 haplotype, as designated by Griese et al.,25 likely encompasses the less active *29 allele, confounding the reported haplotype and activity frequencies.

To overcome these limitations, in this study, we analyzed uniform next‐generation sequencing data ensuring that variant information is fully consistent and compatible. Our analyses reveal the drastic extent of genetic variability between major populations, which entail phenotypic consequences on the level of differential metabolic activity profiles and incentivizes population‐adjusted pharmacogenetic genotyping strategies. Notably though, there are additional layers of genetic variability between specific subpopulations within the aggregated superpopulations that we analyzed, as has been shown, for instance, for CYP2D6 gene duplications and the distribution of the CYP2C19*17 allele within Europe, which occur with a strong south‐north and west‐east gradient, respectively.20, 23 Overall, Africans constitute the most heterogeneous superpopulation, as exemplified by large differences in CYP2D6*5, CYP2D6*29, and CYP3A5*3 frequencies that vary among 3–17%, 2–20%, and 4–95% between subpopulations, respectively.26, 27 Therefore, higher resolution maps across the subpopulations will add clinically useful information and we expect that such datasets will be available in the near future when the many current whole genome analyses projects have been published.

CYP2A6 is the main enzyme involved in nicotine metabolism and is also involved in the bioactivation of procarcinogens from tobacco smoke. Previous studies showed that CYP2A6 deficiency resulted in reduced nicotine dependence28 as well as a markedly reduced lung cancer risk (odds ratio = 0.23).29 Given the high prevalence of functionally deficient CYP2A6 alleles in East Asian populations (67.5%; Figure 2 b), these data corroborate that the genetic contribution to nicotine addiction as well as the risk to develop pulmonary lesions is reduced in these populations.

CYP2B6 and CYP2A6 are the key enzymes responsible for metabolism of the antiretroviral efavirenz, and deficiency of these enzymes can result in increased systemic exposure.30, 31, 32 Importantly, significant associations between reduced CYP2B6 activity and neurological symptoms were described for pediatric as well as adult patients, indicating that CYP2B6 genotyping might reveal biomarkers for increased risk of efavirenz toxicity particularly in African populations in which CYP2B6 deficient alleles are most frequent.33, 34 Notably, no clear effect on CYP2B6 functionality has been demonstrated for CYP2B6*5 and CYP2B6*9, the most abundant CYP2B6 variant alleles in Europeans and admixed Americans. Interestingly, CYP2B6*5 apparently results in significantly reduced expression levels of CYP2B6 in vitro, however, without altering efavirenz metabolism suggesting increased specific activity of the gene product, which compensates for the reduced expression levels.35 In agreement with these in vitro findings, no effect of CYP2B6*5 on efavirenz pharmacokinetics was observed in vivo.36 Thus, interrogating the functional effects of these alleles using a variety of CYP2B6 substrates will be an important task for future studies to allow for a more accurate translation of CYP2B6 genetic variability into functional activity profiles.

CYP2C8 is involved in the metabolism of a multitude of chemically diverse medications, including nonsteroidal anti‐inflammatory drugs, thiazolidinediones, and chemotherapeutic agents. Importantly, CYP2C8 alleles causing impaired activity have been implicated in reduced clearance and increased exposure to paclitaxel, which correlates with higher incidences of paclitaxel‐induced neuropathies.37, 38 In light of the distribution of CYP2C8 deficient alleles, CYP2C8 genotyping might be a viable option in Africans and Europeans in which 19.2% and 17.2% of CYP2C8 alleles are expected to exhibit reduced functionality, respectively. In contrast, genotyping in East Asians cannot be recommended due to the low frequencies of deficient CYP2C8 variants (<1%).

Extremely slow clearance of phenytoin and increased risk of neurotoxicity has been observed in patients homozygous for CYP2C9‐deficient alleles.39 Furthermore, CYP2C9 deficiency is linked to bleeding complications upon treatment with warfarin and other anticoagulants. Together with age, body‐surface area, and polymorphisms in VKORC1, CYP2C9 variants explain 50% of the interindividual variability in dose requirements40 and a recent prospective trial showed that dosing guided by CYP2C9 and VKORC1 genotypes can increase the time in the therapeutic range and decrease the number of adverse events with atrial fibrillation or venous thromboembolism.41 These findings suggest that due to the distribution of CYP2C9 deficiencies in major populations, Europeans, Africans, and South Asians might be at increased risks of phenytoin toxicity as well as complications due to warfarin treatment.

The arguably most extensively studied CYP gene is CYP2D6, which metabolizes around 25% of all drugs in clinical use. Genetic polymorphisms that result in increased CYP2D6 metabolic capacities, primarily observed in Africans, have been linked to decreased treatment response observed for the treatment with tricyclic antidepressants,42 increased incidences of respiratory depression after tramadol treatment,43 opioid intoxication after codeine treatment with concomitant CYP3A4 inhibition, and transient reduction in renal functionality.44 In contrast, patients with reduced functionality CYP2D6 alleles, which are found with highest frequencies in East Asian (70.3%) and African populations (52.5%), bear an increased risk to develop hepatotoxicity upon treatment with the antianginal agent perhexiline,45 metoclopramide‐induced acute dystonic reactions,46 and adverse drug reactions caused by the antipsychotic risperidone.47

The functional CYP3A5*1 variant has recently been implicated in decreased safety of sunitinib treatment in patients with metastatic renal cell carcinoma.48 The splice variant rs776746 that results in CYP3A5 deficiency (CYP3A5*3) was found at very high frequencies in European populations (94.3%), whereas the combined prevalence of the inactive *3 and *6 alleles in Africans was only 33.4%. These results suggest that Africans might be at increased risks of sunitinib toxicity, incentivizing the clinical investigation of outcomes of treatment with lower sunitinib doses in these populations.

The CYP star allele nomenclature system provides a standardized nomenclature system, which facilitates scientific exchange and promotes the understanding of pharmacogenetic variability by nonpharmacogeneticists, such as clinicians. In this nomenclature system, genetic variants within the CYP genes that exert functional effects or cause unique amino acid substitutions are designated with a star (*) number. Yet, for some complex loci, interpretations of the star allele nomenclature can be impeded by intricate haplotype structures and the inclusion of different nonfunctional variants in the allelic definitions, frequently resulting in the use of different genotyping strategies for the analyses of CYP allelic variants. However, it is of importance to interrogate a sufficiently large number of single‐nucleotide polymorphisms in order to correctly assign a specific star allele. For instance, CYP2D6*2 cannot be called solely based on the presence of the defining variants rs16947 and rs1135840. Rather, other variants that are commonly detected in linkage with this variant combination, such as rs3892097 (CYP2D6*4) and rs5030865 (CYP2D6*14), also need to be excluded. Furthermore, due to the large number of extensive next‐generation sequencing projects, many novel rare variants are detected, which are important modulators of patients' response to pharmacological treatment but cannot be subsumed under existing star alleles.49 Thus, although the star allele nomenclature has considerable advantages for simple descriptions of the variant CYP alleles, it cannot be used to define the true global variability of all different CYP alleles.

In summary, our analyses reveal the large extent of genetic variability between major populations and can serve as a powerful resource for the worldwide distribution of CYP allele frequencies. Furthermore, these findings underscore the need to consider population‐specific genetic backgrounds when conducting pharmacogenetic analyses and clinical trials.

5. METHODS

5.1. Allele frequency data

Frequency information of exonic variants was obtained from Exome Aggregation Consortium, which integrated 14 large‐scale sequencing efforts encompassing in total sequences of 56,945 unrelated individuals from five major populations, encompassing 33,370 Europeans, 5,203 Africans and African Americans, 4,327 East Asians, 8,256 South Asians, and 5,789 admixed Americans.13 Frequencies of variants outside of coding regions were extracted from 1,851 genomes provided by the 1000 Genomes Project.12 For alleles that were defined by a set of variants, as specified in Tables 1, 2, 3, 4, 5 and Supplementary Tables S1–S4, we calculated linkage of the respective SNVs using the LDhap feature of the software package LDLink (https://analysistools.nci.nih.gov/LDlink)50 based on population‐specific linkage information provided by the 1000 Genome Project. Frequencies of gene deletions (CYP2A6*4 and CYP2D6*5) or amplifications (CYP2D6*1xN and CYP2D6*2xN) were obtained from the literature, as indicated due to technical limitations regarding the reliable quantification of their frequencies from synthesized short‐read sequencing projects. Cumulative allele frequencies (F_cum,i,p) for each gene i were calculated for each population p by summing up the individual haplotype frequencies of all major alleles included in Tables 1, 2, 3, 4, 5. Average *1 allele frequencies were determined as:

$$F_{*1,i}=1-{\displaystyle \sum }_p{F}_{cum,i,p}$$

5.2. Allele nomenclature and definitions

CYP star (*) alleles were defined according to the Human CYP Allele Nomenclature Database (http://www.cypalleles.ki.se) and we refer to this source for references describing the functional characterization of the respective alleles. In this study, we pooled suballeles and only considered the variants that effectuate the functional effects of the allele, as indicated in Tables 1, 2, 3, 4, 5 and Supplementary Tables S1–S4. Throughout this paper, we defined common alleles as having MAF >1%, whereas variants or alleles with an allelic frequency <1% were defined as rare.

6. SOURCE OF FUNDING

This study was supported by the European Union's Horizon 2020 research and innovation program U‐PGx under grant agreement no. 668353 and by the Swedish Research Council (grant agreement numbers: 2015‐02760, 2016‐01153, and 2016‐01154).

7. AUTHOR CONTRIBUTIONS

V.M.L., Y.Z., and M.I.‐S. wrote the manuscript. V.M.L. designed the research. V.M.L. and Y.Z. performed the research. V.M.L., Y.Z., and M.I.‐S. analyzed the data.

8. CONFLICT OF INTEREST

V.M.L. and M.I.‐S. are co‐founders and owners of HepaPredict AB.

Supporting information

Supplementary Table S1 Important variant and allele frequencies of the human CYP1A1 gene. AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians. For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.

Supplementary Table S2 Important variant and allele frequencies of the human CYP1A2 gene. AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians. For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se. ^# Indicates alleles whose functionality assessment is based solely on in vitro data.

Supplementary Table S3 Important variant and allele frequencies of the human CYP2E1 gene. AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians. For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se. ^# Indicates alleles whose functionality assessment is based solely on in vitro data.

Supplementary Table S4 Important variant and allele frequencies of the human CYP4F2 gene. AFR, Africans; AMR, admixed Americans; CYP, cytochrome P450; EAS, East Asians; EUR, Europeans; SAS, South Asians. For references describing the functional characterization of the indicated alleles, see http://www.cypalleles.ki.se.