Phylogenetic analysis of the precore/core gene of hepatitis B virus genotypes E and A in West Africa: new subtypes, mixed infections and recombinations

Infection with Hepatitis B virus (HBV) is a major cause of acute and chronic liver disease. More than 350 million people are chronic carriers of HBV and many of them develop progressive diseases, including cirrhosis and hepatocellular carcinoma (Kane, 1995). Worldwide, HBV infections account for 1 million deaths per year, most of which occur in the developing world. In sub-Saharan Africa, the seroprevalence of anti-HBV antibodies is particularly high (Olubuyide et al., 1997) and excessive incidence rates of chronic carriers have been reported from these populations. In different cohorts of otherwise healthy children or adults from seven West African countries, we observed between 9 and 65 % chronic carriers and up to 100 % in human immunodeficiency virus (HIV) patients from Cameroon (Mulders et al., 2004; Olubuyide et al., 1997). Genomic mutations introduced during reverse transcription (Girones & Miller, 1989) explain the 15 % genetic diversity of human HBV and account for genetically distinct genotypes, designated AH. Most genotypes show a more or less distinct geographical distribution (Arauz-Ruiz et al., 1997; Gray et al., 1997; Kato et al., 2002; Magnius & Norder, 1995; Mangia et al., 1996; Minami et al., 1996; Robertson & Margolis, 2002; Stuyver et al., 2000). Genotype E (serotype ayw4) is the most prevalent genotype throughout a crescent covering almost 6000 km from Senegal to Angola (Baptista et al., 1999; Liu et al., 2001; Norder et al., 1994; Odemuyiwa et al., 2001; Owiredu et al., 2001; Suzuki et al., 2003; van Steenbergen et al., 2002) and this genotype has essentially not been found anywhere else in the world. Several studies by us and others found that, only in Cameroon, this HBV/E prevalence was interrupted by a high prevalence of genotype A in Pygmies, Bantus and particularly in HIV-infected individuals (Kurbanov et al., 2005; Mulders et al., 2004). Most of these studies were based on genetic analysis of the preS/S gene. An intriguing finding of our earlier studies, also confirmed since by others, was the conspicuously low sequence diversity of HBV/E of 1·67 % in the preS/S gene (Kramvis et al., 2005; Mulders et al., 2004; Odemuyiwa et al., 2001). Only relatively few preC/C gene sequences of genotype E have been analysed and genotype E did not seem to separate from genotype D in the X and C open reading frames (ORFs) (Bowyer & Sim, 2000), raising some initial questions about its existence (Kidd-Ljunggren et al., 1995; Kramvis et al., 2005). Therefore, further studies and sequencing of larger numbers of, in particular, the C gene have been recommended (Kramvis et al., 2005). Here, we report 122 new preC/C sequences from three West African countries in order to determine whether the low sequence diversity found in the preS/S gene can also be confirmed for this gene and how these sequences compare phylogenetically with the other genotypes. Cameroon was included in this study to investigate whether the high genotype A prevalence within the genotype E-endemic crescent may have led to mixed infections and recombinations. To distinguish between genotypes A and non-A, a 6 nt insertion in the former was exploited (Hannoun et al., 2002). Considering the very high prevalence of chronic carriers of HBV/E throughout West Africa (Ahmed et al., 1998; Allain et al., 2003), genetic analysis of the preC/C gene may also provide clues to explain the excessive rate of chronicity associated with this genotype. Clinical samples.
Serum samples were collected from 110 HBsAg-positive donors between 1998 and 2004 in three West African countries: Nigeria (Lagos, Ibadan), Mali (Bamako) and Cameroon (North, East, West, South and Central Provinces). The adults from Nigeria were patients admitted to local hospitals, many of them with liver disease. The donors from Mali were otherwise healthy students. About two-thirds of patients were known to be HIV-positive or became known during hospitalization. Blood was drawn after informed consent of donors or their parents or guardians in the case of children. The larger part of the described cohorts had already been included in an earlier study (Mulders et al., 2004).

Sera were tested for HBs and HBe antigens by using Murex kits (Abbott Laboratories). HIV infections were confirmed by using a Murex HIV-1.2.0 kit (Abbott Laboratories). Serum samples were stored at 80 °C until use.

Amplification of the preC/C region.
A first PCR served as detection PCR for mixed infections (Fig. 1b). Whenever both genotype-specific reverse primers amplified a PCR product, a mixed infection was assumed. To confirm this and to obtain a larger PCR fragment for sequencing, the second round of the detection PCR was repeated with C1 as the forward primer (Fig. 1c). The PCR products obtained (C1/rvA and C1/rvnonA) were purified in a 1 % agarose gel and sequenced by using the same primers. When, after the detection PCR, only one of both genotype-specific reactions was positive, the sequencing was done with the corresponding reverse primer and the C1 forward primer. In parallel, a genotype-insensitive PCR (Fig. 1a) was run on all samples. Technical details of the different PCRs are described below.

(18K): Fig. 1. PCRs used to detect mixed infections (b) and to generate genotype-specific (c) and genotype-unspecific (a) products for sequencing. All PCRs are semi-nested with different forward primers. Positions are given according to the HBV/E reference strain (GenBank accession no. X75657[GenBank] ).

HBV DNA was isolated from 200 µl serum by using a QIAamp DNA Blood Mini kit (Qiagen) and eluted in a 200 µl volume. Genomic amplification of the preC/C region was performed by PCR in a semi-nested format using primers PC1 (5'-GGAGACCACCGTGAACGC-3', positions 16101627) and C2 for the first round and C1 (5'-CTGGGAGGAGTTGGGGGA-3', positions 17301747) and C2 (5'-GTAGAAGAATAAAGCCC-3', positions 24872503) (Kao et al., 2002) for the second round (Fig. 1a). Nucleotides are numbered according to the HBV/E reference strain with GenBank accession no. X75657[GenBank] (Norder et al., 1994). The first round was performed for 40 cycles (95 °C for 1 min; 50 °C, 1 min; 72 °C, 2 min and a final extension step at 72 °C for 10 min) in a 50 µl reaction volume containing 5 µl extracted DNA, 1x PCR buffer, 0·2 mM dNTPs, 1·8 mM MgCl₂, 0·3 µM each primer and 2 U Platinum Taq (Invitrogen). After the first amplification, 1 µl PCR product was reamplified with primers C1 and C2 for another 40 cycles and using the same PCR conditions (Fig. 1a).

Genotype mixtures were analysed by amplification of the C gene with A-specific and non-A-specific primers as described by Hannoun et al. (2002). In order to increase PCR sensitivity, each sample was tested first in parallel by using the forward primer PC1 in combination with either rvA (5'-TTCTTCTTCTAGGGGACCTGCCTCAGTCC-3', positions 23562384) or rvnonA (5'-TTCTTCTTCTAGGGGACCTGCCTCATCGT-3', positions 23502378). In the second round, PC1 was replaced by fw1865 (5'-CAAGCCTCCAAGCTGTGCCTTGGGTGGCCTT-3', positions 18651895), giving a fragment of a maximum of 560 bp (Fig. 1b). PCR conditions were as described above except for 30 cycles and an annealing temperature of 58 °C in the first round and 62 °C in the second round. The amplified products were separated in a 1 % agarose gel stained with ethidium bromide. In the case of a positive reaction, the second round was repeated under the same PCR conditions, using C1 as forward primer (Fig. 1c).

Amplification of the complete genome.
The complete genome was amplified as four overlapping fragments obtained with four two-round PCRs. The fragments were designated preS, S, X and C. Primers and conditions used are summarized in Supplementary Table S1 (available in JGV Online). All PCRs were done in 25 µl with 1x PCR buffer, 200 nM dNTPs, 200 nM each primer, variable MgCl₂ concentrations, 1 U Platinum Taq and 5 µl template DNA. For the second-round PCRs, 2 µl 1 : 100-diluted first-round product was used. PCR conditions were as follows: 5 min denaturation at 95 °C followed by 40 cycles (30 cycles for the second round) of 20 s at 95 °C, 20 s annealing at variable temperature (see Supplementary Table S1, available in JGV Online) and 90 s elongation at 72 °C.

Cloning.
One sample was further analysed for mixed infection by cloning the product of the S fragment PCR into the pCR4-TOPO vector (Invitrogen), which was used to transform Top10 electrocompetent Escherichia coli cells (Invitrogen) by electroporation. Colonies were screened by using M13 universal primers. M13 PCR products of the correct size were sequenced as described above, again by using M13 primers.

Phylogenetic analysis.
PCR products were purified by using a JetQuick Purification Spin kit (Genomed GmbH). Purified DNA was quantified with Picogreen (Invitrogen) by using a Genios Plus fluorescence reader (Tecan). Purified DNA (50 ng) was then sequenced in both directions on an ABI Prism 377 sequencer (Applied Biosystems). Briefly, 5 µl DNA was amplified in a 10 µl reaction volume containing 4 µl premix (BigDye Terminator Cycle Sequencing Ready Reaction kit; Applied Biosystems) and 1 µl of each sequencing primer (0·5 µM final concentration).

One hundred and eighty sequences were obtained either with primers C1/C2 (Fig. 1a) or C1/rvA or C1/rvnonA (Fig. 1c) and included the entire preC/C gene with a total length of 517584 bp, depending on the genotype (genome positions 18142331; numbering according to GenBank accession no. X75657[GenBank] ). In addition, three complete genomes and four preS fragments (positions 2455159) were sequenced. Nucleotide sequences were analysed by using ABI Sequencing Analysis (version 3.4.1) and Sequence Navigator (version 1.0.1), then aligned with CLUSTAL W software and checked by visual inspection. Phylogenetic trees were constructed with the MEGA 3.0 software (Kumar et al., 2004), using the neighbour-joining and Kimura two-parameter method and including reference strains of genotypes AG and all known A and D subtypes. Sequences were submitted to GenBank/ EMBL/DDBJ under accession nos AM110794[GenBank] AM110915[GenBank] for the preC/C gene and AM180623[GenBank] AM180628[GenBank] for the complete-genome and preS fragment sequences.

Screening for mixed infection
All serum samples were screened for mixed infections by genotype A and non-A strains, using two genotype-specific reverse primers (rvA/rvnonA), exploiting the 6 nt insertion in genotype A (Fig. 1b) (Hannoun et al., 2002). All samples were sequenced with the C1 forward primer and either the reverse primer rvA or rvnonA, depending on which reverse primer gave a positive fragment in the genotype-specific PCR (Fig. 1c). In total, 49 samples were only positive in either the rvA (20 of 49)- or the rvnonA (29 of 49)-detection PCR. Phylogenetic analysis showed that 12 rvA products obtained were of genotype A and 28 rvnonA products were of genotype E. Surprisingly, eight rvA products were of genotype E and one rvnonA product was of genotype A. The C1/C2 sequences showed that, in the latter nine samples, binding of the rvA and rvnonA primers was non-specific (Table 1).

Table 1. Comparison of the results of the genotype A (rvA)- and nonA (rvnonA)-specific PCR Sequences in bold were used for phylogenetic analysis.

Another 33 samples were positive in both genotype-specific detection PCRs after two rounds (Fig. 1b), giving 66 fragments. Sequencing assigned 13 of 33 rvA products to genotype A and 31 of 33 rvnonA products to genotype E. Twenty-two of the 66 sequences did not correspond to the expected genotype: two of the 33 rvnonA products were of genotype A and 20 of the 33 rvA products were identified as genotype E. Thus, only 84 of 115 sequences that were obtained by genotype-specific detection PCR were confirmed by sequencing (Table 1). In 11 patients among the above 110, both rvA and rvnonA sequences were of the predicted genotype and, for one, the C1/C2 sequence was of genotype A, whereas the rvnonA sequence was of genotype E, thus giving a total of 12 mixed-infected samples with two distinct sequences available.

One sample (NIE24072) was further analysed by cloning the product of the S fragment PCR into a vector. Five colonies were selected randomly and the plasmid was sequenced by using M13 primers. One clone did not contain an insert, three clones contained sequences identified as genotype A3 with a divergence of 2·5 % among the three sequences and one clone contained a genotype E sequence.

The genotype-independent PCR (Fig. 1a) was positive for 65 samples. For 27, the sequence of this PCR confirmed the sequence obtained by the genotype-specific PCR. For 28 patients, sequences were obtained by the genotype-independent PCR only, whilst the genotype-specific PCR was negative for both fragments, probably because of its lower sensitivity. For nine samples, the genotype-independent PCR gave the same sequence as the genotype-specific PCR. However, in these cases, the genotype-specific sequence was obtained by binding of the false reverse primer (Table 1).

The robustness of the genotype-sensitive PCR was tested by using two plasmids containing the C fragment (positions 16442661; see Supplementary Table S1, available in JGV Online) of either genotype A or E. The plasmid concentration of one genotype was kept constant (10³ copies µl¹) and increasing concentrations of the other genotype (10³10⁹ copies µl¹, i.e. ratios of 1 : 11000 : 1) were added. When the irrelevant genotype was present in a sufficient excess (>100 times) over the specific genotype, binding and amplification of the false template occurred, explaining most of the false-positive PCR products.

The electropherograms of seven of the forward sequences obtained in the C1/rvA/rvnonA (Fig. 1c) and the C1/C2 (Fig. 1a) PCRs showed distinct peaks in at least five genotype-specific positions, supposedly because of the presence of both genotypes. The reverse reaction of these seven sequences showed a high background or was not readable, an expected result of a frame shift between the two reverse fragments of A and non-A (i.e. E), starting at the genotype A insertion. These seven patients were therefore also considered as having mixed infections. Thus, the sole detection of both sequences may lead to an underestimation of the occurrence of mixed infections (Table 1).

Genotype prevalence and mixed infections
One hundred and twenty-two sequences from 110 patients from HIV-negative and HIV-positive children and adults were further analysed (Table 2). Ninety-two of 110 patients were (co)infected with genotype E and this genotype was also the most prevalent in each of the three countries. Genotype A was the only other genotype detected in the three countries (Nigeria, 8/49; Mali, 3/20; Cameroon, 26/41 patients). This genotype was most dominant in HIV-positive donors from Cameroon. Mixed infections of genotype E with genotype A were essentially limited to Cameroon and Nigerian HIV-positive donors (18/19).

Table 2. Prevalence of single- and mixed-genotype infections

Genetic variability of genotypes and subtypes
With a maximal diversity of 9·43 %, genotype A sequences found in the three countries were highly diverse. Phylogenetic analysis differentiated three distinct subtypes. The A sequences from Cameroon clustered with the recently defined subgroup A''/Ac/A3 (Fig. 2a). The bootstrap value separating this A3 lineage from the A1 and A2 subtypes was 82 %. The genetic distance between our A3 strains and the European/American A2 subtype (5·88 %) is smaller than the distance to the African/Asian A1 subtype (5·97 %), although this was not the case for the complete-genome sequences (Fig. 3a). The three genotype A sequences from Mali did not cluster with subtypes A1A3, but formed a cluster of their own, tentatively designated subtype A4, with a mean distance of 5·41 % (4·93 % to A2 and 5·86 % to A3) to the other A subtypes, although the bootstrap value was relatively low (42 % of calculated trees). The diversity among these Mali sequences was large (2·32 %) in comparison to that among Mali E sequences (1·47 %). For two A4 strains from Mali (GenBank nos AM110795[GenBank] and AM180623[GenBank] ) and one A3 strain from Cameroon (AM180624[GenBank] ), complete-genome sequences were obtained (Fig. 3a). The node separating the A3 and A4 groups had a bootstrap value of 97 %. The genetic distance of the A4 complete genome to the A3 genome was 6 % (divergence of 0·7 %). Four Nigerian genotype A sequences formed another group of their own (tentatively designated A5), with a mean diversity of 2·37 % and a mean distance to the other A subtypes of 5·11 % (4·47 % to A3 and 5·32 % to A2). A4 and A5 were separated by a mean genetic distance of 5·05 %. For two A5 (GenBank nos AM180625[GenBank] and AM180626[GenBank] ) and two A3 (AM180627[GenBank] and AM180628[GenBank] ) strains, only the preS fragment (in addition to the preC/C fragment) was obtained. The bootstrap value separating the A5 from the A3 group was 96 % (Fig. 3b). For the A5 preS sequences, a distance to the A3 sequences of 4·3 % was calculated. The diversity between A4 and A5 was 5·2 %. The divergence within A4 and A5 was 0·7 and 2·4 %, respectively.

(22K): Fig. 2. Phylogenetic tree of sequences clustering with genotype A (a) or E (b) reference strains. Sequences clustering with subtype A3 are marked , the Mali subtype A4 , the Nigerian subtype A5 and the Nigerian recombinant . Genotype E sequences are labelled . Bootstrap values of important nodes are indicated, as well as the diversities of subgroups. The following abbreviations have been used: MAL, Mali adults; CAE, Cameroon adults; CAEch, Cameroon children; NIE24xxx and NIElagxxx, Nigerian adults; NIE12xxx, Nigerian children. The insert (c) shows a tree calculated by using complete-genome sequences of references.

(13K): Fig. 3. Phylogenetic tree of genotype A complete-genome sequences (a) and preS fragment sequences (b) (positions 2455159). Sequences clustering with subtype A3 are marked , the Mali subtype A4 and the Nigerian subtype A5 . Bootstrap values of important nodes are indicated. The following abbreviations have been used: MAL, Mali adults; CAE, Cameroon adults; NIE24xxx, Nigerian adults.

The diversity of the genotype E sequences was 1·41 %, which was significantly lower (P<0·001; Fisher's exact test) than the mean diversity of 2·85 % found within the genotype A subtypes A3A5. The topology of the tree, as well as the diversity calculations, showed that the HBV/E preC/C sequences are related most closely to HBV/D1 and D2, with a mean distance between E and D1D4 of 4·85·9 % versus 9·2 % to the A3 subtype (Fig. 2b).

Identical genotype E sequences were found in patients from distant provinces in Nigeria and Cameroon (e.g. Abia, NIE24072; Ogun, NIE24008; Imo, NIE24146; Kogi, NIE24142; Northern Province, CAE12; Central Province, CAE151). Identical sequences were also found in different countries, such as east Cameroon (CAE388) and the province of Oyo in western Nigeria (NIE24233). In addition, seven genotype A strains obtained from the Cameroon children cohort showed identical sequences. When the patients were divided into three age groups (<14, 1430 and >30 years), the ratio of E to A sequences in the three groups decreased (3·3, 3 and 2·4, respectively), at least partially explaining the significant (P<0·001; Fisher's exact test) increase in genetic diversity when all sequences were combined (0·67, 2·1 and 3·2 %, respectively). Also, within both genotypes, the diversity tended to increase with age, but significance was lost.

Mutations in the preC/C gene
A summary of specific mutations is shown in Table 3. Four patients, all from Cameroon, were infected with genotype A strains displaying a mutation in the start codon (A1814C or A1814T) of the preC ORF. Mutations in positions G1896A or G1897A were identified in 34 patients, primarily in genotype E sequences (32/34). In 18 cases, both the wild type and the mutation were found to coexist at position 1896 (10 quasispecies and eight mixed-infected patients). The latter mutation was frequently found (7/14) together with G1899A. These mutations introduce a new stop codon (TGG to TAG or TGA) that causes a premature, non-functional HBeAg. Whenever G1896A is associated with the double mutation A1850T and C1858T, it causes an increase in stability of the pregenomic RNA-encapsidation signal. This was the case in 31 of 88 sequences, all of genotype E. In two genotype A sequences, a C1857T mutation was found and the same two sequences had G1897A. The A1850T mutation, normally restricted to non-A genotypes, was also found in a genotype A strain from Mali. Interestingly, the G1896A mutation was three times more frequent in HIV-infected patients (34·2 versus 11 %). Whilst in Nigeria, this mutation was restricted to adults, in Cameroon, it was four times more frequent in children than in adults (48 versus 11 %). In two patients, additional, rare stop codons were identified: C1817T in the preC and G2262T in the core region. Finally, all identified A genotypes except two (NIE24063 and MAL42) presented C1862, found to be very rare in any of the genotypes AH. No significant difference in the mutation pattern was found between sequences obtained from mixed- or single-infected patients.

Table 3. Specific mutations in the preC/C gene of 29 genotypeA and 89 genotype E sequences The preC sequences of one genotype A and three genotype E strains were not entirely readable and are therefore excluded.

Insertions and deletions in the preC/C gene
The rvA sequence of a mixed-infected patient from Nigeria (GenBank no. AM110794[GenBank] ) had an insertion of 36 bp at the beginning of the C ORF (position 1903) that differed only by three non-synonymous nucleotides from a 36 bp long C gene fragment of genotype G located in the same nucleotide positions in both the genotype G reference sequence and NIE24072 (Fig. 4). A BLAST search of this insert showed that a similar sequence was not found in any other sequence in GenBank and that the closest other match has only 19 of 36 bp identity. The first 200 bp of this strain clustered with genotype E as well as genotypes A and D, whereas the last 356 bp clustered with the HBV/A3 sequences; all other genotypes were grouped correctly (Fig. 4). However, the first 200 bp upstream of the insert contains two mutations specific to non-A genotypes (positions 1850 and 1858), whilst further downstream of the insertion, many typical genotype A mutations were identified. Five patients presented deletions in the C ORF (Table 3): one from Mali (20082054), two from Nigeria (2121 and 21722307), the latter being mixed-infected, and two with single-nucleotide deletions from Cameroon (nt 2170 and 2250). All of these deletions cause a disruption of the reading frames of both HBeAg and the core protein.

(15K): Fig. 4. Phylogenetic analysis of the NIE24072 outlier based on the first 200 bp (left), the complete preC/C gene (centre) and the last 317 bp (right). The scale is the same as for the trees in Fig. 2. Genotypes A, D and E group on the same node when the phylogenetic tree is constructed by using only the first 200 bp.

HBeAg status and mutations
Forty-nine patients for whom there was enough serum were tested for HBeAg, including three mixed-infected with two sequences and two with only one sequence available. One of two sequences from a mixed-infected, HBeAg-negative patient had the 1814 start-codon mutation, whereas the other sequence had no mutation affecting HBeAg expression. Five sequences from single-infected patients had mutations at 1817, 1896 or 1897 (quasispecies at these positions was considered as wild type). Two of the five patients with mutations at 1896/1897 and the patient with the C1817T mutation were HBeAg-negative. Of the patients with wild-type nucleotides in these three positions, 46 were HBeAg-positive (including four mixed infections) and six were HBeAg-negative (including one mixed infection). In all of the mixed-infected cases, the second sequence, when available, was also of wild type. Three patients had deletions: the one with the single-nucleotide deletion at nt 2170 was HBeAg-negative, whereas the patients with the deletions at 2121 and 21722307 were associated with an HBeAg-positive status. The present survey presents 122 new preC/C sequences from three major countries in West Africa, as well as three complete-genome sequences and four preS regions of HBV. As expected from previous studies by ourselves and others (Baptista et al., 1999; Liu et al., 2001; Norder et al., 1994; Odemuyiwa et al., 2001; Owiredu et al., 2001; Suzuki et al., 2003; van Steenbergen et al., 2002), the majority of the sequences from this region were of genotype E. Genotype A was the only other genotype found in the region. Although in our cohorts, HIV-positive donors dominated, the high prevalence of genotype A in Cameroon does not seem to be limited to HIV patients (Kurbanov et al., 2005; Mulders et al., 2004). Despite the large distances between the different collection sites (up to 3000 km), the 92 genotype E preC/C sequences showed very little genetic diversity, irrespective of origin and other patient characteristics. Similar to the genetic diversity reported earlier for the preS/S gene (1·67 %; Mulders et al., 2004) (1·54 % with all preS/S sequences included), the diversity of our 92 sequences was 1·41 or 1·59 % when all preC/C sequences from GenBank were included. When the preS/S and preC/C genes of the same 45 strains were compared, the C gene sequences were 1·5 and 2·4 times more diverse than the S genes for genotypes A and E, respectively. Similarity plots showed that, similar to the preS/S gene, the preC/C gene was related most closely to genotype D, in particular the subgenotypes D1 and D2 (about 5 %, compared with >9 % for the other genotypes). Two single-infected samples seemed to be the result of an AE recombination, as the preS/S sequence was of genotype A (Mulders et al., 2004) and the preC/C sequence was of genotype E (e.g. GenBank nos AJ605037[GenBank] and AM110899[GenBank] ).

All genotype A sequences from Cameroon were identified as A3 and none grouped with the African/Asian A1 or European/American A2 subtype (Bowyer et al., 1997; Hasegawa et al., 2004). In our earlier study, the preS/S sequences of several of these strains had been assigned to group A'' (Mulders et al., 2004), which has now been renamed A3 (Hannoun et al., 2005; Kurbanov et al., 2005). Our results confirm that subtype A3 is somewhat closer to the European A2 subtype (mean genetic distance, 5·88 %) than to the African/Asian A1 (mean genetic distance, 5·97 %; Hannoun et al., 2005). Instead of the G1862T mutation in the precore gene, thought to be characteristic for A3 strains (Hannoun et al., 2005), G1862C was predominant in our sequences. All strains were of the A2 type at position 1888. The genetic diversity of subtype A3 is 3·85 %, compared with 2·19 % for subtype A2. Thus, A2 may have a shorter evolutionary history than A3 and may be a more recent progeny of A3.

The Mali genotype A strains are phylogenetically distinct from the other A subtypes (mean distance of 5·4 %). With a genetic distance of 4·93 %, they are closest to the A2/Ae subtype, but warrant a new subgenotype, tentatively designated A4, according to the recommendation of >4 % genetic distance between subgenotypes. Two complete-genome sequences of this subtype corroborated the phylogenetic difference from the other subtypes, with a bootstrap value of 97 %.

The A sequences from Nigeria formed a cluster of their own, with a mean genetic distance of 4·47 % and a maximum distance of 6·5 % from the Cameroon A3 strains, complying with the proposed criteria (genetic distance >4 %) for a new genotype subtype, tentatively designated A5. Phylogenetic analysis of two preS sequences further corroborated the definition of this new subtype. Assuming an estimated mutation rate of 4·2x10⁵ nt¹ year¹ (Hannoun et al., 2000; Okamoto et al., 1987; Orito et al., 1989), the two strains would have taken 500 years to evolve from a common ancestor. The Nigerian strain NIE24072 shows evidence of a triple recombination of a non-A (E/D) sequence and an A3 sequence, separated by a G-specific insert (Fig. 4). Although genotype G has so far not been reported from Africa, the insertion in some of the quasispecies sequences suggests that the patient must have been in contact with a genotype G strain.

Considering the coexistence of genotypes A and E in West Africa, the frequency of coinfections is of interest. Mixed infections were identified by genotype-specific PCR in a number of patients, but sequences of both genotypes were confirmed only in 12 cases. Seven additional cases were identified by more than five ambiguous nucleotides in genotype-specific positions, paired with an unreadable reverse sequence (caused by the genotype A-specific insertion). Despite these additional criteria, false-positive mixed infections are unlikely and the 17·3 % mixed infections may still be a slight underestimation. Thirty-seven per cent of patients from Cameroon were coinfected with both genotypes and 79 % of these were children. As expected, a lower rate of coinfection was detected in Nigeria. All but one mixed-infected patient were HIV-positive, but this may be biased by the large number of HIV-positive donors from Cameroon. Thus, mixed infections seemed to be frequent when (these) two genotypes cocirculate. Together with earlier studies by us and others in Cameroon (Kurbanov et al., 2005; Mulders et al., 2004), this suggests that both genotypes are present in the population, independent of HIV status, and that mixed infections are not restricted to HIV-positive patients.

Considering the low genetic diversity of the most prevalent genotype, E, and its virtual absence in the Americas, we suggested a short evolutionary history and a recent introduction into humans (Mulders et al., 2004). This, however, is in contrast to the excessively high endemicity of acute and chronic hepatitis B infection throughout West Africa. Perinatal transmission is thought to be the most frequent cause of chronic infection in African children (Edmunds et al., 1996). The risk of becoming a chronic carrier is especially high for children born to HBeAg-positive mothers (Beasley et al., 1981a, b; Chu et al., 1985; Thomas et al., 1982). Eighty per cent of the donors tested were HBeAg-positive, with a clear genotype bias: 35 of 39 (89·8 %) genotype E carriers, but only two of six (33·3 %) genotype A carriers, were HBeAg-positive. If this can be confirmed in a larger study, it could partially explain the high(er) prevalence of genotype E. The HBeAg negativity associated with genotype A and E sequences corresponds to a number of mutations in preC/C sequences believed to affect HBeAg expression (1814, start-codon mutation; 1896 and 1897, encoding a new stop codon), but in more than half of cases, both wild-type and mutated nucleotides coexisted as quasispecies or mixed infections, explaining the expression of HBeAg. In the three HBeAg-positive cases, where only mutant strains are detectable, wild-type strains may have been missed. HBeAg-negative cases without the above mutations warrant the analysis of the core promoter region.

In all cases but one, the G1896A mutation was associated with the A1850T/C1858T double mutation. Only two sequences presented G1897A, encoding a stop codon, and for both, a C1857T mutation was detected instead of the double mutation. This might be related to the stemloop formation in this region and the pregenomic RNA-encapsidation signal, but the clinical significance of this is not clear.

This study confirms that genotype E is predominant throughout West Africa and presents a low diversity that is two to three times higher in the preC/C gene than in the preS/S gene, supporting our earlier speculation of a short evolutionary history and a recent introduction into humans of this genotype. The high prevalence of genotype E may be partially explained by the two- to threefold-higher rate of HBeAg positivity in comparison to genotype A. In contrast to genotype E, the diversity of genotype A suggests several new subtypes. In Cameroon, where genotypes A and E cocirculate, the frequency of mixed infections is high and A/E recombinations seem to have occurred. Recombinations with other genotypes or subtypes could partially explain the current genetic diversity.

We would like to thank Wim Ammerlaan for his excellent technical expertise and Dr Segolène Mahot for her valuable support during the generation and analysis of the sequence data. Without the important contributions of many health workers and physicians throughout West Africa, this comprehensive study would not have been possible. Support by the Ministry of Cooperation of the Grand-Duchy of Luxembourg and the Centre de Recherche Public Santé is gratefully acknowledged. C. M. O. was supported by a fellowship BFR of the Ministry of Research.