CYP2D6 and CYP2C19 in Papua New Guinea: High frequency of previously uncharacterized CYP2D6 alleles and heterozygote excess

Abstract

Purpose: A high frequency of previously unknown CYP2D6 alleles have been reported in Oceania populations. Genetic and functional properties of these alleles remain unknown. Methods: We performed analyses of the genetic variability of CYP2D6 and CYP2C19 genes using AmpliChip genotyping in cohorts from two distinct Papua New Guinea (PNG) populations (Kunjingini, n=88; Alexishafen, n=84) focussing on the genetic characterisation of PNG-specific alleles by re-sequencing. Results: Previously unknown CYP2D6 alleles have population frequencies of 24% (Kunjingini) and 12% (Alexishafen). An allele similar to CYP2D6*1, but carrying the 1661G>C substitution, was the second most frequent CYP2D6 allele (20% Kunjingini and 10% Alexishafen population frequency). Sequencing suggests the CYP2D6* 1661G>C allele originated from a cross-over between CYP2D6*1 and *2 and thus is predicted to confer fully active CYP2D6 enzyme. Two additional predicted full activity alleles [1661G>C;4180G>C] and 31G>A were found in the Kunjingini cohort (frequencies 3 c/c and 1%, respectively) and a novel predicted reduced activity allele [100C>T;1039C>T] was found in the Alexishafen cohort (frequency 2%). A high frequency of ultra-rapid (15%) and notably low frequencies of intermediate and poor CYP2D6 metabolizers (<5%) and a high frequency of poor CYP2C19 metabolizers were observed in PNG. Both CYP2D6 and CYP2C19 showed heterozygote excess that may be explained by exogamy and recent introduction of alleles by migration that are yet to reach HWE in relatively isolated populations. Conclusion: The CYP2D6*1661 allele common in Oceania may be regarded as functionally equivalent to the full activity CYP2D6*1 allele.

Introduction

The Cytochrome P450 enzymes (CYP) are the major phase I enzymes involved in drug metabolism. Among CYPs, CYP2C9, CYP2D6 and CYP2C19 are highly polymorphic and together account for ∼40% of hepatic phase I metabolism [1]. CYP2D6 contributes to the biotransfor-mation of 20-25% of drugs in clinical use [2]. The aim of pharmacogenetics is to study the different responses to drugs based on inherited factors and to use this information to develop individualized therapy [3]. However, there are considerable regional and inter-ethnic differences in CYP2D6 variant allele frequencies [4,5]. In a group with European ancestry, the most common CYP2D6 variant allele, CYP2D6*4, is associated with absent enzyme activity. In Asian populations, the CYP2D6*10 variant and, in African populations, the CYP2D6*17 variant are more relevant for reduced CYP2D6 activity. Ethiopian/Saudi Arabian populations have a particularly high frequency of CYP2D6 gene duplications that lead to ultra-rapid metabolism of CYP2D6 substrates [2]. Sistonen et al. reported a ∼30% frequency of a CYP2D6* 1661 allele in a small PNG cohort (n=17) but further characterisation of this allele has not been carried out [4]. CYP2C19*2 and *3 deficiency alleles are known to be exceptionally frequent in Asian and Pacific populations including Papua New Guinea (PNG) [6].

Since both CYP2D6 and CYP2C19 are involved in the metabolism of antimicrobial (including antiparasitic) agents that are in common use in developing countries such as PNG [7], we have assessed the genetic variation at these loci in two PNG populations. These findings should facilitate an understanding of CYP2D6 genetic variation in populations from other parts of the world.

Patients and Methods

Study sites

The present study utilised baseline blood samples taken as part of a large clinical trial comparing the efficacy of different antimalarial treatments in children living in Kunjingini (East Sepik Province) and Alexishafen (Madang Province) (Figure 1) [8]. Both study sites are situated in the northern part of mainland PNG at latitude approximately 4° and have an equatorial climate with very high rainfall and humidity. However, they differ significantly with regard to migration history, lifestyle and accessibility.

Figure 1.

Location of studied populations. Map of Papua New Guinea with northern mainland sites Kunjingini (East Sepik Province) and Alexishafen (Madang Province) shown.

Alexishafen is a coastal settlement situated 20km from the provincial capital Madang (population 30,000). The Madang north coast area was initially settled by people speaking both Austronesian and Papuan languages, with the major local language in the Alexishafen area being “Bel”, a member of the Austronesian language family [8]. The development of Madang as a significant administrative centre and major port serving local industry (tuna fishing, copra and cocoa plantations) has resulted in significant transmigration of other Melanesian and Papuan ethnic groups into the Alexishafen area and subsequent inter-marriage. Most of the population around Alexishafen are subsistence farmers and fishermen.

Kunjingini is part of Wosera sub-district, East Sepik Province, situated inland on the alluvial plains between the Torricelli Range and the Sepik River [9] 75 km south-west of the provincial capital Wewak. The Wosera was settled by members of the Abelam ethnolinguistic group that speak Papuan Sepik languages. Most of the Wosera population is engaged in subsistence farming. Due to the high population density, the area has seen significant recent out -migration, while in-migration is rare. The Wosera is situated 340km NW of Alexishafen with two major rivers (Sepik and Ramu River) separating the two areas. Because there is no road link, migration between the two populations has been limited.

Subjects

The comparative treatment trial [8] was approved by the institutional review boards of the University of Western Australia (Approval no. 1060) and of the PNG Institute of Medical Research (MRAC 04.14). Parental consent was given. From the first patients enrolled at each site, a total of 84 blood samples from Alexishafen and 88 from Kunjingini were available and these were genotyped under supplementary ethical approval (MRAC 07/08). The median age of children from whom the samples were taken was 3.0 (range 0.5-5.3) years and 57% were males. There were no significant differences in age or sex between children from the two study sites.

Amplichip genotyping and allele scoring

The AmpliChip® (Roche, Mannheim, Germany) is the first FDA-approved pharmacogenetic gene chip for comprehensive CYP2D6 and CYP2C19 genotyping [10, 11]. The AmpliChip holds ∼15,000 probe cells that allow comprehensive genotyping of more than 20 CYP2D6 alleles including gene duplication and deletion. In addition, the CYP2C19 *2 and *3 alleles can be genotyped. Genotyping was performed according to the manufacturer's instructions. In brief, PCR products encompassing CYP2D6 and CYP2C19 gene loci were generated, fragmented and labelled. They were further processed (hybridization, staining) on an Affymetrix GeneChip Fluidics Station 450Dx and scanned on a Affymetrix GeneChip Scanner 3000Dx. Data were processed using the AmpliChip CYP450 Data Analysis Software V2.0.0 Beta to generate genotype calls. The following CYP2D6 SNPs are tested as part of the AmpliChip P450 procedure: -1584C>G, 31G>A; 100C>T; 138insT; 883G>C; 1023C>T; 1039C>T; 1659G>A; 1661G>C; 1707delT; 1758G>T; 1758G>A; 1846G>A; 1976G>A; *20 cluster (1973insG, 1978C>T, and 1979T>C), 2539-2542delAACT; 2549delA; 2613-2615delAGA; 2850C>T; 2935A>C; 3183G>A; 3198C>G; 3277T>C; 4042G>A; *36 gene conversion; 4180G>C. The SNPs positions are numbered according the distance from the A of the ATG start codon in the reference sequence M33388. Based on these SNPs the alleles CYP2D6*1 to *4, *6 to *11, *14, *15, *17, *19, *20, *25, *26, *29-*31, *35, *36, *40 and *41 are assigned. For CYP2C19 the *2 and *3 alleles are genotyped. Putative phenotypes are assigned by the AmpliChip software e.g. CYP2D6*1/*4 results in extensive metabolizer (EM) and CYP2D6*4/*41 results in intermediate metabolizer (IM). In addition we used the Gaedigk et al. activity score system with its finer scale [12].

CYP2D6 sequencing

A 5092 bp fragment including the CYP2D6 gene was amplified in a long-range PCR reaction using primers 5′-CCA GAA GGC TTT GCA GGC TTC A -3′ and 5′-ACT GAG CCC TGG GAG GTA GGT A-3′ that were previously shown to amplify specifically CYP2D6, but not CYP2D7P and CYP2D8P [13]. The unused primers and ddNTPs were removed by incubation with exonuclease I and shrimp alkaline phosphatase (both from USB, Staufen, Germany). The PCR products were se-quenced using BigDyeTerminator v1.1 Cycle sequencing kit (Applied Biosystems, Darmstadt, Germany) with primers CYP2D6_seq1 (5′- AAA CGG CAC TCA GGA CTA AC), CYP2D6_seq2 (5′-CTG GAG GTG GAG CTG GAC TT), CYP2D6_seq3 (5′- GGG TCT TCC CTG AGT GCA AA), CYP2D6_seq4 (5′ GGA GGG GAC ATC TCA GAC ATG), CYP2D6_seq5 (5′- GAG GGG AGG CTG GGC AAA AG) and CYP2D6_seq6 (5′- TCG GGC GCA GGA CTA GTT GA).

Statistical analysis

Genetic marker analyses were performed as implemented in the PowerMarker 3.25 (exact test for Hardy-Weinberg equilibrium (HWE), gene/allele summary statistics) [14], Genepop 4.0 (F_IS estimates [15], heterozygote deficiency and excess) [16] and Popgene 1.32 software (observed/expected hetero-/homozygosities). The PHASE 2.1 software [17] was used to reconstruct haplotypes for each individual. Ten independent runs with different seed numbers (used seeds 2, 1536, 2936, 3123, 4957, 5283, 6757, 7992, 8633, 9045) were performed in order to avoid seed-biased allele assignments. Frequencies were compared by Fisher's exact test (two-sided mid-p values) or the Mann-Whitney test as appropriate.

Results

CYP2D6

Genotype data are shown in Table 1 as inferred phenotypes. The total no call rate for CYDP2D6 in these patient samples was 33%, and significantly higher in the Kunjingini (45%) than in the Alexishafen (20%) samples. “No call” signifies that automated genotype assignment was impossible due to presence of unknown alleles. Both cohorts had a high rate of UM (∼15%). The rate of extensive metabolizers (EM) was significantly higher in the Alexishafen (58%) than the Kunjingini (41%) sample. It was inappropriate to test for significant differences between Kunjingini and Alexishafen populations in CYP2D6 poor metabolizers (PM), intermediate metabolizers (IM) and ultra-rapid metabolizers (UM) because of sample numbers and presence of unknown SNPs. PM and IM were not detected at all in the Kunjingini cohort and were rare in the Alexishafen cohort.

Table 1.

Predicted phenotypes for CYP2D6 and CYP2C19

CYP2D6	Total		Kunjingini		Alexishafen		p
no call	57	(33%)	40	(45%)	17	(20%)	<0.001
Classical phenotype acc. to AmpliChip software
PM	1	(1%)	0	(0%)	1	(1%)	0.244
IM	3	(2%)	0	(0%)	3	(4%)	0.057
EM	85	(49%)	36	(41%)	49	(58%)	0.028
UM	26	(15%)	12	(14%)	14	(17%)	0.600
Activity score
0.0	1	(1%)	0	(0%)	1	(1%)	0.244
0.5	2	(1%)	0	(0%)	2	(2%)	0.120
1.0	15	(9%)	4	(5%)	11	(13%)	0.044
1.5	6	(3%)	0	(0%)	6	(7%)	0.006
2.0	113	(66%)	66	(75%)	47	(56%)	0.008
2.5	3	(2%)	1	(1%)	2	(2%)	0.430
3.0	32	(19%)	17	(19%)	15	(18%)	0.771
Sum	172		88		84
CYP2C19
PM	68	(40%)	35	(40%)	33	(39%)	0.938
EM	104	(60%)	53	(60%)	51	(61%)	0.938
Sum	172		88		84

CYP2D6 and CYP2C19 inferred phenotypes from two Papua New Guinean populations. p-values compare Kunjingini and Alexishafen cohorts. The activity score according to model A [12] is given for all observed CYP2D6 genotypes. The p-value for AS score between group comparison is 0.036 (Mann-Whitney test). The assignment is putative for novel alleles (see results). p-values are given as two-sided Fisher's exact test mid-p values.

To analyse the dataset including “no call” alleles further the AmpliChip raw data SNP report was used and alleles were determined by Bayesian forecasting with PHASE 2.1. Table 2 gives a more detailed analysis of the observed CYP2D6 alleles. The CYP2D6*1 allele was most prevalent (∼60% in both cohorts). The CYP2D6*1XN gene duplication was also prevalent (∼12% in both cohorts). Overall, the CYP2D6 genotype distribution was highly significant different between the two cohorts (p<0.001). The frequencies of the gene deletion (CYP2D6*5), and CYP2D6*4 and *10 alleles were significantly higher in the Alexishafen compared to the Kunjingini cohort. The Kunjingini cohort has only one non-functional allele (*5 [frequency 3%]). The Alexishafen cohort carries more non- and reduced-activity alleles (CYP2D6*4 [3%]; *5 [8%]; *10 [3%]) and the novel reduced-activity allele [100C>T;1039C>T] (2%; Table 2, see also below). The Kunjingini cohort had a higher mean AS (2.15) than the Alexishafen cohort (1.96, p=0.036).

Table 2.

CYP2D6 and CYP2C19 allele frequencies

CYP2D6 alleles	AS	Total		Kunjingini		Alexishafen		p
*1	1	210	(61%)	110	(63%)	100	(60%)	0.545
*2	1	1	(0%)	0	(0%)	1	(1%)	0.244
Duplication (*1XN)	2	40	(12%)	19	(11%)	21	(13%)	0.678
Deletion (*5)	0	18	(5%)	5	(3%)	13	(8%)	0.040
*4	0	5	(1%)	0	(0%)	5	(3%)	0.013
*10	0.5	5	(1%)	0	(0%)	5	(3%)	0.013
*41	0.5	5	(1%)	1	(1%)	4	(2%)	0.133
31G>A	1	1	(0%)	1	(1%)	0	(0%)	0.744
[100C>T;1039C>T]	0.5	3	(1%)	0	(0%)	3	(2%)	0.058
1661G>C	1	51	(15%)	35	(20%)	16	(10%)	0.008
[1661G>C;4180G>C]	1	5	(1%)	5	(3%)	0	(0%)	0.044
Sum			344		176		168

Comparison of CYP2D6 alleles/SNPs called by AmpliChip P450 analysis from two Papua New Guinea populations. Allele frequencies are given in brackets. If possible alleles are given in the star nomenclature. Novel alleles are indicated by their variant sites. Haplotypes were reconstructed using ten independent runs of the PHASE software. The haplotype assignment was consistent in all ten PHASE runs. p-values compare Kunjingini and Alexishafen cohorts. AS; activity score accord ing to model A [12], putative assignment for novel alleles. The CYP2D6*41 allele may also have occurred as duplicated allele (PHASE probability 13%).The 31G>A haplotype may also have occurred as [31G>A;1661G>C] (PHASE probability 50%). The [100C>T; 1039C>T] haplotype may also have occurred as [100C>T;1039C>T; 1661G>C] (PHASE probability 16%). The [1661G>C;4180G>C] haplotype may also have occurred as 1661G>C and 4180G>C in trans (PHASE probability 2%).

p-values are given as two-sided Fisher's exact test mid-p values.

CYP2D6 showed an excess of heterozygotes (Table 3). In accordance with this finding, a negative FIS estimate (within-population inbreeding coefficient) was seen [15]. The Kunjingini cohort showed a tendency to violate HWE (see below) and this was significant in the analysis of the combined cohort (Table 3).

Table 3.

Population genetic descriptive statistics

	Observed heterozygosity	Expected heterozygosity	Observed homozygosity	Expected homozygosity	Fis	HWE, p
CYP2C19
Kunjingini	0.602	0.580	0.398	0.420	−0.026	0.299
Alexishafen	0.750	0.667	0.250	0.333	−0.112	0.407
All samples	0.674	0.636	0.326	0.364	−0.070	0.739
CYP2D6
Kunjingini	0.659	0.560	0.341	0.440	−0.133	0.076
Alexishafen	0.620	0.620	0.381	0.384	−0.005	0.117
All samples	0.640	0.590	0.361	0.0	−0.068	0.003

Observed and expected hetero- and homozygosities for CYP2C19 and CYP2D6 and resulting F_IS according to [15]. Exact test for departure from HWE given as p-value.

Characterization of novel CYP2D6 alleles

Several haplotypes were identified that could not be assigned to any previously known CYP2D6 allele, either by the AmpliChip software or by direct comparison with the cytochrome P450 allelic database (http://www.cypalleles.ki.se). A previously undefined CYP2D6 haplotype, referred to here as CYP2D6*1661 allele, was observed with a frequency of 15% in the sample (Table 2). The CYP2D6*1661 allele carried a 1661G>C substitution but lacked 2850C>T or 4180G>C substitutions. We re-sequenced 3961 bp of the CYP2D6 gene in three heterozygous and one hemizygous carrier of the 1661G>C allele. This covered the whole CYP2D6 coding region, 2064 bp of intron sequence and 251 bp of the proximal promoter region. Genetic variants typical for the CYP2D6*2 allele were observed in the 5′ part but were missing in the 3′ part of the 1661G>C allele (see Figure 2). The CYP2D6*1661 allele may have originated from a crossing-over event between alleles CYP2D6*1 and *2. There were no additional variants in the 1661C>G allele that differed from those known in the fully active CYP2D6*1 allele. Therefore we assigned a putative activity score of 1 according to model A [12]. The CYP2D6*1661 allele was not in HWE in the Kunjingini population (p<0.001) but in the Alex-ishafen population (p=1).

Figure 2.

Illustration of the novel CYP2D6*1661 allele (CYP2D6*1661C>G) with high prevalence in the Papua New Guinean population. Exon-intron structure of the gene and the localisation of single polymorphic loci characteristic for CYP2D6 *2 are shown. The regions of CYP2D6 gene depicted in grey were analysed by re-sequencing in carriers of the 1661G>C allele. The combination of alleles of single polymorphic loci within the gene in the CYP2D6 allele 1661G>C is compared with corresponding combinations of CYP2D6 alleles 1 and 2 (obtained from Kimura et al. ([34]) and the genome reference sequence respectively). Suggested crossover between CYP2D6*1 and *2 is illustrated as a combination of grey and black shading for the CYP2D6*1661 allele.

Re-sequencing of a heterozygous [100C>T; 1039C>T] allele carrier showed a combination of single variants 100C>T, 843T>G, 1039C>T, 2097G>A, 3384A>C, 4180G>C. This allele differed from the previously known allele CYP2D6*10B by an absence of the G>C exchange at position 1661, a variant known not to cause functional changes in enzyme expression or activity. Therefore we assigned an activity score of 0.5 due to presence of the 100C>T as for the CYP2D6*10 reduced-activity allele.

CYP2C19

All CYP2C19 blood samples (n=172) had successful genotype calls. There was a high proportion (40%) of predicted CYP2C19 PMs that was equal in both patient cohorts. Despite the comparable frequencies of predicted CYP2C19 PMs (Table 1), we observed a different genetic basis for non-functional alleles (Table 4). The CYP2C19*2 allele frequency was significantly higher in the Kunjingini cohort while the frequency of the CYP2C19*3 allele was significantly higher in the Alexishafen cohort. The CYP2C19*1 allele was of comparable frequency in both cohorts. The CYP2C19 genotype distribution was not different between the two cohorts (p=0.815) and both were in HWE. CYP2C19 also showed an excess of heterozygotes (Table 3) and, therefore, a negative F_IS estimate was seen.

Table 4.

CYP2C19 genotype and allele data

CYP2C19	Total		Kunjingini		Alexishafen		p
*1/*1	19	(11%)	11	(13%)	8	(10%)	0.550
*1/*2	61	(35%)	38	(43%)	23	(27%)	0.032
*1/*3	24	(14%)	4	(5%)	20	(24%)	<0.001
*2/*2	31	(18%)	22	(25%)	9	(11%)	0.013
*3/*3	6	(3%)	2	(2%)	4	(5%)	0.323
*2/*3	31	(18%)	11	(13%)	20	(24%)	0.061
Sum	172		88		84
*1	123	(36%)	64	(36%)	59	(35%)	0.779
*2	154	(45%)	93	(53%)	61	(36%)	0.002
*3	67	(19%)	19	(11%)	48	(29%)	<0.001
Sum	344		176		168

Comparison of CYP2C19genotypes and alleles called byAmpliChip P450 analysis in two Papua NewGuinean populations, p-values compare Kunjingini and Alexishafen cohorts.

p-values are given as two-sided Fisher's exact test mid-p values.

Discussion

Pharmacogenetics has great potential to allow pharmacotherapy to be tailored to the individual based on drug metabolism and action. Ethnicity has generally proven to be non-specific in predicting pharmacogenetically relevant features [18, 19, 20]. It is accepted that the genetic variability between any two individuals within any one group is almost as high as the genetic variability between any two individuals anywhere in the world [21]. Therefore the knowledge of population specific variants in pharmacogenetically relevant genes such as CYP2D6 will help to improve prediction overall.

This study undertook genotyping of CYP2D6 and CYP2C19 using the AmpliChip P450 gene chip in cohorts from two Melanesian populations in PNG. CYP2D6 genotyping revealed at least 3 novel alleles in these cohorts. The most frequent allele CYP2D6*1661 (15% in the total sample) was previously assigned on a SNP basis by Sistonen et al. [4] who also found this allele prevalent in Papuan populations. The CYP2D6*1661 allele was the second most frequent allele in both populations studied while the CYP2D6*2 was present in only one child (0.2%; Table 2). This is in concordance with previous data [4] and indicates high relevance of CYP2D6*1661 in the PNG populations. Of note the CYP2D6*1661 alleles have also been observed in Eastern Europe, the Middle East and Subsaharan Africa [4]. Therefore we completely re-sequenced the CYP2D6*1661 allele. Our data suggest that CYP2D6*1661 originates from a crossing over event between of the CYP2D6*1 and CYP2D6*2 alleles within the regions between exon 3 and 6 (Figure 2). Both alleles confer full CYP2D6 activity and the CYP2D6*1661 allele does not contain any additional polymorphisms. Therefore, full activity is likely for the CYP2D6*1661 allele. Interestingly, the 1661G>C SNP violated HWE in the Kunjingini but not in the Alexishafen cohort; the observed absence of homozygote carriers and excess of heterozygote carriers may be explained by introduction of this allele by migration of a single heterozygote into the population. This results in a relative homozygote deficiency for several generations [22].

In contrast to Sistonen et al. [4], we found no isolated 4180G>C haplotype (PHASE probability only 0.02) but a [1661G>C;4180G>C] haplotype (PHASE probability 0.98). Another allele with putative full activity was characterized by 31G>A. This allele occurred only once and could also be [31G>A;1661G>C](PHASE probability 0.50). Presence of 1661G>C (PHASE probability 0.08) as well as CYP2D6*41 (PHASE probability 0.13) in the duplicated gene could not be confirmed but is theoretically possible based on haplotype assignments. The duplicated allele CYP2D6*1XN frequencies (12%) in both cohorts are high compared to other populations [2] but are comparable to or below those reported in Saudi Arabians [23] and Ethiopians [24].

The high frequency of duplicated active alleles in the PNG population is mirrored by a low frequency of low activity alleles. In the Kunjingini cohort, all cases had an activity score according to Gaedigk et al. [12] of ≥1 (Table 1). The Alexishafen cohort had more known (CYP2D6*4, *5, *10) and novel putative reduced-activity alleles [100C>T;1039C>T] but still the frequency of cases with an activity score <1 was only 3%. We were not in a position to corroborate these findings by investigation of the phe-notype but, based on the known genotype-phenotype correlations, we expect that medications or CYP2D6 substrates in general undergo extensive to even ultra-rapid metabolism in the studied populations. Toxicity associated with PM should be almost absent.

The CYP2D6 genetic diversity in the Alexishafen cohort was higher than in Kunjingini. The presence of both Austronesian and Papua speakers as well as increased transmigration into the Alexishafen area may have contributed to increased genetic diversity by admixture of alleles of different origins. By contrast, the Kunjingini area is geographically more isolated and linguistically more uniform. Minority (deficient) alleles may thus have been selected out or lost through genetic drift, consistent with observations in Pacific Islanders of a lower genetic variability in large island interiors versus more intermixed Austronesian speaking groups living along coastlines [25]. CYP2D6 was not in HWE in the total sample and showed a similar tendency in the Kunjingini sub-group (Table 3). We are not aware of formal marriage restrictions in PNG but observed homozygosities were always smaller than expected. This favours outbreeding, including through arranged marriages between different clans of the same tribe, due to strong taboos against inbreeding causing exogamy [26]. The FIS was negative in all cohorts and genes which indicates a departure from HWE towards a heterozygote excess (Table 3). Such negative FIS values are to be expected in small populations [27]. In general, heterozygote excess has been much less discussed in the literature than heterozygote deficiency [28], especially in the multiple allele case as for CYP2D6.

For CYP2C19 we observed an exceptional high rate of 40% predicted PM. This was due to CYP2C19*2 and *3 deficiency alleles. Our findings are in line with the previously-reported high frequencies in the Melanesian people from Vanuatu [29] and East Sepik [6]. Interestingly, while the Kunjingini cohort showed comparably high frequencies for CYP2C19*2 and CYP2C19*3 as reported for the other geographically (and linguistically) close East Sepik population [6], the CYP2C19*3 allele was significantly more common in the Alexishafen cohort. Whether this is due to different migration histories, genetic drift or simply by chance in the studied samples is unclear. However, CYP2C19*3 is significantly more common in the Austronesian speaking Madang population, while the Papuan speaking non-Austronesian populations in the Sepik and the also Austronesian speaking populations in Vanuatu have similar frequencies.

Deviations from HWE with global heterozygote excess and homozygote deficiency as observed for both CYP2D6 and CYP2C19 in this study can be caused by viability selection [30]. However, given the PNG population structure, it would be difficult or impossible to determine whether heterozygosity is maintained due to viability selection of certain alleles [31]. In addition to a heterozygote advantage, the population size, differential fertility (including confounding by genotype) and eventually differing allele frequencies in males and females may contribute to heterozygote excess [22, 32]. Analysis of genotypes classified by activity scores did not suggest violation of HWE (data not shown) which argues against a strong selection of certain phenotypes. Further factors include genetic drift, lack of panmixia, admixture and migration effects may play important roles in shaping CYP2D6 allelic diversity in small populations such as those studied here. A bottleneck analysis was not feasible due to the small number of polymorphic loci studied [33].

In conclusion, we confirmed a high frequency of the CYP2D6*1661 allele in PNG. The CYP2D6* 1661 allele was classified as fully active allele by sequence based characterization. We found a very low frequency of PM and a high frequency of UM in samples from the studied PNG populations compared to European populations. This study also confirms the high frequency of CYP2C19 PM which is typical for Oceanian populations. The observed pattern of violation of HWE with heterozygote excess and homozygote deficiency could be explained by peculiarities in relatively isolated PNG populations with exogamy and recent introduction of alleles by migration. The characterization of these common PNG alleles has the potential to also improve the CYP2D6 phenotype prediction in other world populations.