Introduction

CYP2D6 is one of the best-investigated drug-metabolising enzymes and has been scrutinised extensively in all major ethnic populations [1]. In contrast, genotyping data are sparse or only emerging for many indigenous or geographically defined populations [2], and there is a void of information for unique admixed populations, including the Coloureds of South Africa. The population at Africa’s southern tip is very diverse and consists of at least 20 ethnic groups and cultures, including the indigenous Khoi and San people. Four major racial groups are recognised (http://www.statsonline.gov.za/publications/P0302/P03022006.pdf). South Africa has a population of approximately 23.5 million and is composed of Africans (79.6%), whites (9.2%), Indian/Asians (2.4%) and Coloureds (8.8%). The African group encompasses an extraordinary variety of cultural and tribal identities, including the Nguni, Sotho and Khoi-Khoi groups. The white population group is composed of descendants of European Caucasian immigrants who came to South Africa in previous centuries. The term Coloureds refers to a racially heterogeneous group of people who possess some degree of sub-Saharan ancestry in addition to substantial ancestry from Europe, Indonesia, South India, Ceylon, Madagascar, Mozambique, Mauritius, St. Helena, and western and southern Africa. People of this particular heritage are also referred to as Cape Coloureds, whereas Coloureds in KwaZulu-Natal are predominantly of British and Zulu heritage.

The impact of polymorphic CYP2D6 expression on a subject’s capability to metabolise numerous clinically used drugs is dramatic [1, 3, 4]. Depending on which alleles are present in an individual, a wide range of metabolic activity is observed among subjects: poor (PM), intermediate (IM), extensive (EM) and ultrarapid. Consequently, dose-related adverse events include toxicity in PMs due to higher than normal drug levels, to therapeutic failure due to extremely fast drug metabolism in subjects with a ultrarapid metabolism [5]. In Caucasians, four major nonfunctional alleles (CYP2D6*3, *4, *5, *6) with compiled frequencies of approximately 20–30% explain the majority of PM, whereas a number of reduced function alleles (CYP2D6*9, *10, *41) and alleles carrying functional gene duplications or multiplications (CYP2D6*1xN, *2xN) contribute to IM and ultrarapid metaboliser phenotypes, respectively. Other alleles appear to be restricted to certain ethnic groups or admixed populations. For example, the CYP2D6*17 and *29 alleles have African origins and are found predominantly or exclusively in black Africans, African Americans and their descendants [2, 6, 7]. Yet other alleles are characterised by large differences in allele frequencies, such as CYP2D6*10, which is observed at a frequency of <5% in most populations but can reach frequencies over 50% in Asians [2].

Nonfunctional hybrid genes, such as CYP2D6*13 and *16 [8, 9], have been known for a long time, but only limited information regarding their structure and frequency is available. Hybrids are composed of a 5′-CYP2D7 portion and a 3′-CYP2D6 portion (Fig. 1) and are believed to be products of large deletion events that fused respective gene regions [8, 9]. Because we detected an unusually high frequency of the CYP2D6*5 deletion allele, the genotyping analysis was extended to include hybrid gene detection and characterisation.

Fig. 1
figure 1

Overview of the structure of the CYP2D6*5 gene deletion and CYP2D7/2D6 hybrid genes. The CYP2D6*5 gene deletion and all CYP2D7/2D6 hybrid structures are characterised by large deletions fusing proximal and distal regions within the locus. In CYP2D6*5, the entire CYP2D6 gene is deleted, whereas the CYP2D7 gene is completely retained. In the hybrid genes, 5′ portions of CYP2D7 and 3′ portions of CYP2D6 are fused together to create a number of hybrid genes containing different parts of each gene. Respective gene origins are given in red (CYP2D7) and black (CYP2D6). Sequences that could be derived from either gene or are not fully characterised are shown in green. Arrows denote approximate primer binding locations; primer names correspond to those given in Table 1. Red and black arrows denote CYP2D7- and CYP2D6-specific primers, respectively. Polymerase chain reaction product lengths are as indicated. The structures shown for CYP2D6*13 and *16 are according to those described by Panserat et al. [9] and Daly et al. [8], respectively. Brackets indicate the region within recombination likely has occurred; red and black lines represent the most 5′ and 3′ position for which CYP2D7 and CYP2D6 sequences could unequivocally be determined; green indicates that the sequence could have been retained from either

Full size image

Due to the unique and complex admixtue of South African Coloureds, it is impossible to predict their CYP2D6 allele frequencies. The goal of this study was to characterise the CYP2D6 gene locus in this unique population to generate basic knowledge about allele presence and genotype frequencies and use this information to predict their phenotype profile.

Methods

Subjects and source of DNA

Blood samples from 100 subjects were obtained from the archives of the Department of Hematology and Cell Biology of the University of the Free State, Bloemfontain, South Africa. These samples were collected for genetic testing and were from individuals who had voluntarily disclosed their ethnicity. The samples were anonymised to conform to the study protocol that was approved by the Ethics Committee of the Faculty of Health Sciences of the Free State. Blood was collected in ethylenediaminetetraacetate (EDTA) tubes and stored at 4°C. DNA was isolated from 0.4 ml of blood using a chloroform/phenol extraction method adapted from the Promega Wizard Genomic DNA extraction kit (Promega, Madison, WI, USA). DNA quality was assured by agarose gel electrophoresis. DNA was obtained for 99 samples.

CYP2D6 genotyping

Genotype analysis was performed by long-range polymerase chain reaction (PCR) in combination with PCR restriction fragment length polymorphism (RFLP), allele-specific PCR and diagnostic long-range PCR reactions, as previously described in detail [10, 11] and references therein. Testing comprised CYP2D6*2, *2A, *3, *4, *5, *6, *7, *8, *9, *10, *11, *17, *29, *36, *40, *41, *42, *45/46 and *56 allelic variants. Gene duplications were detected and characterised as described previously [10]. In addition, all subjects with homozygous genotypes were tested for the presence of hybrid genes such as CYP2D6*13 and *16 [8, 9].

Detection and characterisation of CYP2D7/2D6 hybrid genes

Identification of hybrid genes by a 8 kb amplification product

Hybrid genes can be detected by amplifying the entire gene using a CYP2D7-specific primer located upstream of exon 1 (hybrid-F1) and a CYP2D6-specific primer (*5-R) binding downstream of the CYP2D6 gene locus [8]. This assay (Table 1), adapted from Daly et al. [8] produces an 8-kb-long product from hybrid genes (Fig. 1). Of note, these primers also allow amplification of a 9-kb fragment from the CYP2D6*5 deletion allele (Fig. 1). Even though the reaction conditions have extensively been optimised, assay results remain highly variable (i.e. amount of PCR product generated, presence of unspecific PCR products) and are sensitive to the quality of the genomic DNA. Alleles carrying hybrid genes do not produce the 6.6-kb-long product that is utilised for genotyping (i.e. the 6.6-kb genotyping product which was previously described in detail [11, 12]). Therefore, subjects carrying two hybrid alleles or a hybrid allele in combination with a CYP2D6*5 are negative in the 6.6-kb PCR reaction, whereas subjects carrying any 6.6-kb-producing allele such as CYP2D6*1, *2 in combination with a hybrid or CYP2D6*5 allele will produce a genotype that reflects only one allele and consequently appears to be homozygous (e.g. a CYP2D6 *1/*1 DNA may indeed be a *1/*5 or *1/hybrid).

Table 1 Polymerase chain reaction (PCR) primers and reaction conditions
Full size table

Identification of CYP2D6*16 and *66 hybrid genes by a 1.1 kb amplification product

To specifically detect CYP2D6*16, *66 or similar hybrid structures, an assay amplifying a 1.077-kb (1.1 kb)-long fragment was developed. PCR reactions were carried out with primers *16/66-F and *16/66-R in 8-μl volumes in the presence of 10–20 ng genomic DNA, and typically, 1–2 μl were analysed by agarose gel electrophoresis. Assay details are summarised in Table 1. The hybrid gene of a Caucasian individual previously identified to carry a hybrid gene was sequenced and deposited into GenBank (EU093102). This gene was initially called CYP2D*16 [13] and was revised to CYP2D6*66.

Sequence analysis of hybrid genes

For DNA sequencing, a nested PCR product of 5 kb encompassing the hybrid gene was generated from the initial 8-kb PCR amplicon with primers hybrid-F2 and 2D6-R, as detailed in Table 1. Alternatively, this product was generated directly from genomic DNA. The PCR products were treated with EXOSAP-IT (USB, Cleveland, OH, USA) and subsequently sequenced with primers that did not discriminate between CYP2D6 and CYP2D7 to determine the hybrid structure. Sequence gaps were filled in with CYP2D6- or 2D7-specific primers to obtain consecutive full-length sequence. Sequencing was performed with DYEnamic ET dye terminator chemistry and a MegaBACE 500 capillary sequencer (GE, Piscataway, NJ, USA).

Improved strategy for hybrid gene detection and genotyping

The size of the hybrid-specific amplicon was decreased from 8 kb to 5 kb. This was achieved by replacing the reverse primer (*5-R) with the primer also used to generate the CYP2D6-specific 6.6-kb genotyping fragment (2D6-R) and pairing it with a CYP2D7-specific forward primer (hybrid-F1). Assay details are given in Table 1. DNA samples with CYP2D6*5/*5 and CYP2D6*1/*1 genotypes served as negative controls and the previously characterized CYP2D6*5/*66 DNA as positive control. To further characterise the 5-kb amplicon and demonstrate that it contained a hybrid gene arrangement, this fragment was used as a template to genotype for key sequence variations in exon 1 and exon 9. For the nested reamplification reactions, the 5-kb amplicon was diluted 1,000- to 2,000-fold, with 10 mM Tris ph 8 and 0.8 μl used for PCR. Specifically, the CYP2D7-specific T insertion in exon 1 (137insT) was detected with an assay modified from that described previously [14]. PCR was carried out with a forward primer that contained a partial HindIII restriction site (Tins-F) and a reverse primer (Tins-R) that bound to intron 1. PCR products containing 137insT were cut with HindIII [14]. The nature of exon 9 was determined by gene-specific PCR that utilised a CYP2D6-specific forward primer (L × 2F) and the reverse primer also used to generate the 5-kb amplicon (2D6-R). A product was only formed from 5-kb templates containing CYP2D6 exon 9 sequences. Specificity of this assay was demonstrated on a CYP2D6*5/*5 DNA. Further details are given in Table 1.

Detection and characterisation of CYP2D6*64 and CYP2D6*65

For CYP2D6*64, the 6.6-kb-long PCR product encompassing the CYP2D6 gene and 1.5 kb of upstream region was cloned with the TOPO XL PCR cloning kit (Invitrogen, Carlsbad, CA, USA). For CYP2D6*65, a 5-kb-long nested PCR product was generated due to poor yield of the 6.6-kb amplicon and cloned. Clones were genotyped for 100C>T and 1023C>T (CYP2D6*64) and 100C>T and 2850C>T (CYP2D6*65), respectively, to identify those carrying inserts of genes with the novel single nucleotide polymorphism (SNP) haplotypes. Subsequently, clones were entirely sequenced with appropriately spaced primers, DYEnamic ET dye terminator chemistry and a MegaBACE 500 capillary sequencer (GE). The sequences were compared with previously analysed allelic variants and allele designation obtained from the nomenclature committee (http://www.cypalleles.ki.se/).

To directly genotype genomic DNA for CYP2D6*64 (100T, 1023T), allele-specific amplification was carried out with primers *64-F and *64-R. In order to control for PCR performance, a second primer pair was included in the reaction [internal control (IC) primers *64 IC-F and *64 IC-R]. CYP2D6*64-negative samples produced only the IC amplicon of 775 bp, whereas CYP2D6*64-positive samples produced two bands, the 775-bp IC and the 956-bp-long diagnostic CYP2D6*64 amplicon. Assay details including primer sequences and PCR reaction conditions are given in Table 1. Of note, addition of the IC amplicon enhanced the specificity of the assay as it suppressed weak, unspecific amplification of the CYP2D6*64 amplicon, which otherwise occurred in samples positive for either or both CYP2D6*10 (100T, 1023C) and CYP2D6*17 (100C, 1023T). Assay performance was validated on appropriate positive and negative samples.

Activity score assignment

The activity score (AS) is the sum of two values assigned to each allele reflecting their relative activity. The AS system has recently been described in detail elsewhere [11]. In this study, values for AS model A have been applied. Briefly, a value of 1 was given to the fully functional reference CYP2D6*1 allele and 0 to nonfunctional alleles. Alleles carrying gene duplications or multiplications received double the value compared with that assigned to an allele with a single gene copy (e.g. CYP2D6*2xN received a value of 2). Reduced-activity alleles received a value of 0.5 to reflect reduced activity. For example, subjects with CYP2D6*1/*29 and CYP2D6*2xN/*5 genotypes received an AS of 1.5 and 2, respectively.

Results

CYP2D6 allele frequencies

DNA samples from 99 subjects who identified themselves as Coloureds were tested for over 30 allelic CYP2D6 variants to provide a detailed genetic analysis and allow for the most accurate phenotype prediction possible with current knowledge. Table 2 summarises the allele frequencies determined in Coloureds and gives a comparison to those found in Caucasian and African American populations previously analysed [11]. Specifically, some alleles, such as CYP2D6*4 and *41, were lower compared with those in Caucasians but higher than those in African Americans. A number of alleles present in Caucasians and/or African Americans were absent (e.g. CYP2D6*3, *6, *9, etc.), while the CYP2D6*5 gene deletion allele was detected at a relatively high frequency of 17.2%.

Table 2 Allele frequencies in Caucasians, African Americans and South African Coloureds
Full size table

Detailed analysis of the CYP2D6*5 gene deletion allele

To confirm the CYP2D6*5 results obtained with a modified assay [11], all samples identified as either homozygous or heterozygous for a CYP2D6*5 allele were reanalysed by amplifying a larger, approximately 6-kb-long, fragment, including the exon 9 portion of CYP2D7 (Fig. 1). This assay was similar to that described by Fukuda et al. [15] which detects not only CYP2D6*5 but also CYP2D6 genes harbouring CYP2D7-derived downstream sequences. Such CYP2D6 genes would appear as false-positives in CYP2D6*5 assays targeting the intergenic region and producing amplicons ranging between 2.9 and 3.5 kb in length (i.e. assays used by many investigators for routine testing). However, no such CYP2D6 gene structures were detected. All samples were 100% concordant between the CYP2D6*5 assay utilised by Fukuda et al. [15] and our own modified assay. The observed CYP2D6*5 genotype frequencies were in Hardy–Weinberg equilibrium.

Characterisation of CYP2D6*66 and strategy for hybrid gene detection

Samples presenting with a homozygous genotype (e.g. CYP2D6*1/*1 or *2/*2) and negative for CYP2D6*5 were subsequently scrutinised for the presence of hybrid genes such as CYP2D6*13 or *16 (Fig. 1). The gold standard for their detection is long-range PCR that produces an 8-kb amplicon. In four samples, 8-kb PCR product was present, suggesting the presence of a hybrid gene. One of those samples produced a 1.1-kb product with a CYP2D7-specific primer binding to intron 6 and a CYP2D6-specific primer binding to exon 9 (Figs. 1 and 2). This assay detects hybrids with breakpoints upstream and downstream of respective primer binding sites (i.e. CYP2D7 intron 6 and CYP2D6 exon), including CYP2D6*16. The PCR product of the positive sample was partially sequenced and compared with a Caucasian reference allele, which has been designated CYP2D6*66 by the nomenclature committee and deposited in GenBank under accession number EU093102. Both sequences had a CYP2D7 intron 6, which is clearly distinguishable from that of CYP2D6. Also, the first nucleotide of exon 7 (T) corresponded to CYP2D7 (CYP2D6 exhibits a G at the corresponding position 4776 in M33388), suggesting that the CYP2D7 sequence extends into that exon. Beyond M33388 position 4873, however, the sequence was identified as CYP2D6. This narrows the region of hybridisation to 96 nucleotides between positions 4776 (CYP2D7) and 4873 (CYP2D6). The remaining three samples were negative in the CYP2D6*16-specific assay.

Fig. 2
figure 2

Allele-specific amplification of CYP2D6*16 and *66. Primers binding to CYP2D7 intron 6 and CYP2D6 exon 9 were used to amplify a 1077-bp-long diagnostic fragment. As depicted in Fig. 1, this assay amplifies all hybrid structures that were created by recombination within the primer binding sites, including CYP2D6*16 and *66. No amplification was observed for CYP2D6*1/*1 (neg negative control) and CYP2D6*5/*5 (not shown) DNA samples. Lane 1: this sample produced hybrid-specific 8-kb- and 5-kb-long amplicons and produced the 1,077-bp polymerase chain reaction (PCR) product. CYP2D6*66 was confirmed by partial sequence analysis and comparison with the fully sequenced reference DNA (CYP2D6*5/*66; pos, positive control). Lanes 2–4: these samples produced hybrid-specific 8-kb- and 5-kb-long amplicons, but not the 1,077-bp PCR product. The genotypes of all samples are as indicated. M 100-bp ladder

Full size image

To improve hybrid gene detection and characterisation, a 5-kb-long PCR product was amplified from all samples that produced an 8-kb amplicon. This product was then further genotyped for the presence of the CYP2D7 hallmark T insertion in exon 1 and a series of key nucleotides in exon 9 that identify CYP2D6. As shown in Fig. 3, the 5-kb template of all four samples with hybrid structures and the CYP2D6*66 reference control had the CYP2D7 137insT and produced the CYP2D6-specific primers from exon 9.

Fig. 3
figure 3

Improved strategy to identify and characterise CYP2D7/2D6 hybrid genes. The graph depicts the generic hybrid gene structure, i.e. CYP2D7- and CYP2D6-derived sequences in exons 1 and 9, respectively (indicated in black and white boxes) and CYP2D7 or CYP2D6 regions in between (gray box). Amplification of a 5-kb-long fragment with CYP2D7- and CYP2D6-specific primers encompasses the entire hybrid gene and serves as template to genotype key sequence variations in CYP2D7 exon 1 [137 T insertion (insT)] and CYP2D6 exon 9 (series of nucleotide variations). a Polymerase chain reaction restriction fragment length polymorphism (PCR-RFLP) assay detecting 137insT. CYP2D7-derived PCR fragments carrying the T insertion are cut by HindIII, whereas CYP2D6-derived sequences remain intact. b CYP2D6 exon-9-specific PCR. PCR and PCR-RFLP fragment lengths are as indicated. 2D6 pos 6.6 kb CYP2D6 long-range amplicon of a CYP2D6*1/*1 subject was used as positive control; 2D7 pos plasmid containing the entire CYP2D7 gene was used as positive control; *5/*5 neg genomic DNA of a CYP2D6*5/*5 subject was used as negative control; *5/*66 reference the CYP2D6*66 gene was entirely sequenced and subject’s DNA used as positive control. CYP2D7 exon 1 and CYP2D6 exon 9 sequences were detected in all subjects that produced the 8-kb and 5-kb hybrid-specific amplicons. Amplification of a smaller, unspecific PCR product was only observed in the four study samples

Full size image

Detection and characterisation of CYP2D6*64 and *65

Genotyping results also revealed deviation from known SNP haplotypes in three samples. These cases genotyped as following for three key SNPs. Cases 1 and 2: 100C/T, 1023C/T, 2850C/C; case 3: 100C/T, 1023C/C, 2850T/T. For the first two cases, 2850T was absent in the DNA carrying the CYP2D6*17 SNP. In the third case, 2850T appeared to be located on a CYP2D6*10 allele. To characterise these alleles, the CYP2D6-specific 6.6-kb-long range PCR product was cloned and respective alleles of interest identified by genotyping and entirely sequenced. Complete allele information and a comparison with other alleles is given in Fig. 4. The novel allele found in cases 1 and 2 was not a CYP2D6*17 allele that lacked 2850T but had 100T and 1023T in a new haplotype. This allele was designated CYP2D6*64 by the nomenclature committee. The novel allele in case 3 was confirmed to be a CYP2D6*10 that had acquired 2850T and was designated CYP2D6*65.

Fig. 4
figure 4

CYP2D6*64 and CYP*65 summary and comparison with other alleles. A 6.6-kb-long polymerase chain reaction (PCR) product encompassing the entire gene region (CYP2D6*64) and a ∼5-kb nested product thereof (CYP2D6*65) were cloned and sequenced. M33388 and AY545216 represent reference sequences. Sequences for CYP2D6*2, *10B and *17 were previously sequenced and are shown to demonstrate the presence of single nucleotide polymorphisms (SNPs) that are shared among respective alleles. Black boxes indicate the presence of a SNP compared with the reference sequences. The top panels provide the sequence context for each SNP and their respective positions in the reference sequences. Gene regions [5′-UTR, 3′-UTR, exons (Ex) and introns (In)] and amino acid changes for nonsynonymous SNPs are as indicated in the bottom panels. X denotes the region for which no sequence was obtained for CYP2D6*65, and n/d indicates that the number of As could not be resolved

Full size image

Subsequently, an assay based on allele-specific amplification amplifying the 100T/1023T haplotype was developed. As shown in Fig. 5, a CYP2D6*64-specific PCR product and the internal control amplicon were generated only in the genomic DNA of the index case and clones derived thereof. DNAs carrying CYP2D6*10 and/or *17 produced only the internal control amplicon. It should be noted that amplification specificity was achieved in the presence of the internal control amplicon and that some unspecific product is produced from CYP2D6*10 and *17 in the absence of the internal control amplicon.

Fig. 5
figure 5

Allele-specific polymerase chain reaction (PCR) assay to detect CYP2D6*64. Allele-specific PCR amplified from alleles carrying the CYP2D6*64-defining 100T/1023T haplotype. An internal control (IC) PCR product was coamplified to ensure assay performance, i.e. all samples produced the IC, whereas only CYP2D6*64 carriers generated the diagnostic 956-bp-long fragment. The diagnostic CYP2D6*64 amplification product was generated from the 6.6-kb-long genotyping product (XL PCR) as well as genomic DNA of the index case and the second carrier (not shown). Clones representing the two alleles of the index case amplified as expected, further demonstrating the linkage between 100T and 1023T

Full size image

CYP2D6 genotypes and phenotype prediction

Of the 198 alleles scrutinised, 50 (26.8%) were genotyped as nonfunctional (CYP2D6*4, *5, *14, *66 or other hybrids, Table 2), and three subjects were identified to be homozygous for two nonfunctional alleles. Even though the observed incidence of the PM phenotype was lower at 3.03% (n = 3) than predicted by allele frequencies (6.7%, n = 6.6) (Table 3), there was no statistically significant deviation from the Hardy–Weinberg equilibrium for genotypes with two, one or no nonfunctional alleles (p = 0.17). Five alleles (2.53%) had duplications or multiplications of functional genes (CYP2D6*1 × N and *2 × N), predicting ultrarapid metabolism in 2.3% subjects, i.e. those also having a fully functional second allele (CYP2D6*1 or *2) or having two CYP2D6*1 × N or *2 × N alleles.

Table 3 Observed and expected allele and genotype frequencies
Full size table

Phenotype prediction using the activity score

Figure 6 presents the distribution of AS groups in the study population in comparison with those observed in Caucasians and African Americans. A high proportion of subjects in the AS groups of 0.5 and 1 was noted for Coloureds. Whereas the number of subjects with an AS of 0.5 corresponded well with the number predicted, the number of subjects in the AS-1 group was higher than expected (41% vs. 33%).

Fig. 6
figure 6

Activity score (AS) distribution in Coloureds, Caucasians and African Americans. AS is the sum of the value assigned to each allele reflecting its activity [11]. Each AS group contains subjects with genotypes that confer similar CYP2D6 activity. Subjects with an AS of 0 (no functional alleles) are poor metabolisers, whereas an AS >2 on the other end of the distribution contains genotypes with one functional allele and one allele carrying a functional gene duplication. A subject with an AS ≥0.5 has a certain probability of presenting with an ultrarapid, extensive or intermediate metaboliser phenotype, as described in detail elsewhere [11]. Coloureds are predicted to have overall lower CYP2D6 activity (also referred to as a right shift) compared with Caucasians due to a higher frequency of reduced activity alleles. Coloureds, n = 99 (open bars); Caucasians, n = 347 (gray bars); African Americans, n = 272 (black bars)

Full size image

Discussion

This is the first investigation of the CYP2D6 locus in Coloureds of South Africa. Considering the unique heritage of this population group, along with the high degree of diversity among Africans in general [2], finding a unique CYP2D6 allele composition, including some rare as well as novel alleles, was not surprising. Notable is the relatively high frequency of the CYP2D6*5 allele of 17.2%, one of the highest ever observed in any population [2]. Comparably high frequencies, albeit observed in small numbers of subjects, were recently also reported by Sistonen et al. [2] for the San (n = 14, 14.3%), an indigenous people living in South Africa; the southeastern and southwestern Bantu (n = 16, 18.3%), but interestingly not the northeastern Bantu (n = 24, 4.2%). Frequencies of ≥10% for CYP2D6*5 were also reported for the sub-Saharan Mbuti, but also other central/south Asian populations. This high frequency of CYP2D6*5 in Coloureds, however, does not lead to an exceptionally high frequency of PM, as the cumulative frequency of all other nonfunctional allelic variants is low, at 9.6%. In fact, the only other loss-of-function alleles observed were CYP2D6*4 (7.1%), CYP2D6*14 (0.5%) and hybrid genes (2.0%).

Given the unexpectedly high frequency of CYP2D6*5 alleles, we systematically reviewed our procedures to rule out any experimental artifacts. First, the majority of CYP2D6*5 carriers also produced a CYP2D6-derived 6.6-kb fragment indicating not only heterozygosity but also demonstrating that the DNA sample is of good quality and capable of supporting XL-PCR amplification. Secondly, all potential homozygous CYP2D6*5/*5 subjects were tested for hybrid genes to discriminate between true homozygous CYP2D6*5/*5 and CYP2D6*5/*hybrid genotype assignments. Third, all DNA samples that produced only faint XL-PCR products or failed to amplify CYP2D6 products and/or internal control products were further evaluated by CYP2D6-specific assays, which do not rely on XL-PCR amplification. Those samples interpreted as CYP2D6*5/*5 were consistently negative for any CYP2D6-derived amplification, whereas heterozygous CYP2D6*5 samples produced consistent results for assays performed on both the XL-PCR template and directly from genomic DNA. Finally, all potential CYP2D6*5 carriers had concordant results in two independent CYP2D6*5 assays.

In the CYP2D6*5 allele, the entire CYP2D6 gene, along with flanking upstream and downstream regions, has been lost, whereas hybrid genes are characterised by a large deletion that removes partial gene sequences along intergenic sequences and fuses the remaining 5′-CYP2D7 and 3′-CYP2D6 portions [8]. Frequency data for hybrid genes, such as CYP2D6*13 and *16, are scarce, which is likely due to the cumbersome task of generating an approximately 8-kb-long PCR product [8]. Because the CYP2D6*5 deletion was relatively abundant in Coloureds, we suspected that hybrid genes may also be present. Indeed, four potential carriers were identified, but a definitive allele call could not be made due to inconsistent amplification of the 8-kb product. However, the resequenced control DNA performed well under the assay conditions, suggesting that amplification difficulties may be due to other factors, e.g. residual impurities left behind during the phenol-based DNA extraction procedure. To overcome these problems and simplify hybrid gene characterisation, a shorter 5-kb PCR product was generated and subsequently genotyped for key sequence variations in exon 1 (CYP2D7) and exon 9 (CYP2D6). This approach confirmed all four subjects as hybrid carriers. One was subsequently assigned as CYP2D6*66, whereas the remaining three samples defied further characterisation by sequence analysis (i.e. sequence traces appeared to be composites of multiple sequence templates present in the reaction, which are suspected to be nonsensical/artificial amplicons formed during the PCR process). Nonetheless, the combined approach of amplifying a hybrid-specific 5-kb product and subsequent genotype analysis proved to be a valid approach, as demonstrated on the fully characterised CYP2D6*66 DNA (Fig. 3).

The DNA now defined as CYP2D6*5/*66 was initially consistent with a CYP2D6*5/*16 genotype [8] (also see http://www.cypalleles.ki.se/). It produced the 8-kb-long hybrid-specific PCR product off the hybrid allele, a 9.5-kb-long CYP2D6*5-specific product and was also positive in an assay that detects hybrid structures with recombination points between CYP2D7 exon 6 and CYP2D6 exon 9 (Fig. 1). Resequencing of the entire hybrid gene of this phenotypic PM [13] revealed, however, that the recombination point is located within the first half of exon 7 (the first nucleotide in exon 7 corresponds to CYP2D7 exon 7, nucleotide 98 to CYP2D6, whereas the nucleotides in between match both genes) and not within intron 7, exon 8 or intron 8, as postulated by Daly et al. for CYP2D6*16 [8]. One of our study subjects perfectly matched this hybrid sequence and was assigned CYP2D6*66. Additional hybrid genes previously genotyped as CYP2D6*16 are being investigated to determine whether they also harbour CYP2D6*66-like sequence or indeed are CYP2D6*16. Unfortunately, no sequence was deposited for CYP2D6*16 for more detailed comparisons.

Since both CYP2D6*5/*5 and CYP2D6*5/hybrid genotypes cause poor metabolism, such subjects would not benefit from additional hybrid testing. However, any subjects presenting with a homozygous genotype for functional or partially functional alleles, such as CYP*1*1, *10/*10 or *17*/17, etc., could conceivably carry a nonfunctional hybrid and their phenotype prediction would change accordingly. Also, depending on the assay methodology, hybrid genes could cause unusual or weird genotyping results or allele calls. Consequently, knowledge and awareness about hybrid genes and their structures in general will aid in assay result interpretation and allow initiation of targeted additional testing if necessary.

The novel CYP2D6*64 and *65 alleles may also be products of recombination, as they are hybrids between certain CYP2D6 alleles or haplotypes. As shown in Fig. 4, the former is a hybrid between CYP2D6*10 and *17 and the latter between CYP2D6*10 and *2. The impact of the novel haplotypes on function remains unknown. Since 2850C > T (R296C) does not appear to diminish activity in the CYP2D6.2 isoenzyme when compared with CYP2D6.1 [16], we would not anticipate this SNP to diminish the activity of CYP2D6.65 but expect characteristics similar to that seen for CYP2D6.10.

We have not found any evidence of CYP2D6*64 or CYP2D6*65 (i.e. presence of 100T and 1023T in the absence of 2850T or with any other SNP) in any Caucasian or African American subjects previously studied. Regarding CYP2D6*64, a sample heterozygous for 100C>T, 1023C>T and 2850C>T would be interpreted as CYP2D6*10/*17 (100T assigned to CYP2D6*10, 1023T and 2850T assigned to CYP2D6*17) according to current nomenclature. Heterozygosity for these SNPs is also compatible with a CYP2D6*2/*64 genotype assignment, and only CYP2D6*64-specific testing would allow discrimination. The function of CYP2D6*64 remains to be further investigated, but one may speculate that it may have properties of both CYP2D6*10 (reduced activity due to unstable protein) and CYP2D6*17 (level of reduction appears to be substrate dependent) [17]. Its impact on phenotype prediction may be limited, as we found no CYP2D6*10/*17 candidate subjects among African Americans, suggesting that CYP2D6*64 is rare in this population (>0.001). This allele may predominantly be found in Coloureds and possibly population(s) that contributed to their ethnic admixture. Therefore, routine testing for the CYP2D6*64 allele does not appear to be warranted. Since we encountered only a single CYP2D6*65 allele, the challenge of testing for linkage of SNPs that are distant, and lack of evidence that 2850T would alter protein function, we did not further pursue this allele in our population samples. Understanding CYP2D6*64 and CYP2D6*65 may, however, be helpful in interpreting genotyping results where the presence of respective SNPs cannot be reconciled with known haplotypes.

Based on the frequency of nonfunctional and reduced-function alleles and corresponding AS group distribution in Coloureds, one would expect a mean dextromethorphan/dextrorphan urinary metabolite ratio that is higher (i.e. lower CYP2D6 activity) compared with that observed in Caucasians and African Americans. However, whether this prediction, as suggested by the observed number of subjects in the AS groups, is accurate requires confirmation in an independent population sample that is both genotyped and phenotyped. Observed numbers of genotypes and consequently the number of subjects in AS groups 1 and 2 that deviated from those expected may be under- and overestimated, respectively. One possible explanation may be the relatively small number of 99 subjects in the study, or the sampled population may comprise subjects of different lines of heritage (e.g. Cape Coloureds and Coloureds of KwaZulu-Natal). No information about the heritage was collected under the approved protocol, as all samples were anonymised to protect personal information, and collection of detailed demographic data would benefit future studies.

The incidence of HIV infection in South Africa is one among the highest in the world. In 2006, the national estimated HIV prevalence was 10.8% across all ethnicities and a staggering 29.1% among pregnant woman http://www.avert.org/safricastats.htm, accessed Aug 3, 2007). It appears that HIV infection can impact CYP2D6 activity on a population basis as well as have intraindividual variability. O’Neil et al. [18] reported an overall right shift towards lower activity in a mostly Caucasian HIV-positive cohort as well as phenotype switching events in a number of subjects. This phenomenon was also observed by Werner et al. (BJCP, manuscript submitted). Such a shift may have dramatic consequences in populations with high proportions of IMs, including Coloureds (HIV prevalence 1.9%) and other black South African populations. HIV patients with genotypes leading to reduced baseline CYP2D6 activity may have a higher risk of switching to a PM phenotype and encounter dose-dependent adverse drug reactions. However, the impact of HIV goes beyond CYP2D6, as this is only one of many drug-metabolising enzymes whose activity may be altered via cytokines such as interleukin (IL)-1β, IL-6, tumor necrosis factor (TNF)-α and interferon (IFN)-α or γ [19].

To date little is known about phenotype and genotype and their relationship of phase I and II drug-metabolising enzymes in South Africans of any ethnicity. This study is a first step in that direction. The data presented not only demonstrate that Coloureds are unique in respect to the complement of CYP2D6 alleles present and their frequencies, but also underpin the importance of characterising unique admixed populations. Clearly, personalised medicine in a country such as South Africa faces many challenges but also presents many opportunities to improve drug therapy.