Transmission of hepatitis B virus infection in Gambian families revealed by phylogenetic analysis

Journal of Hepatology 35 (2001) 99±104

www.elsevier.com/locate/jhep

Transmission of hepatitis B virus infection in Gambian families revealed by phylogenetic analysis Uga Dumpis 1,2, Edward Charles Holmes 3, Maimuna Mendy 4, Adrian Hill 2, Mark Thursz 1, Andy Hall 5, Hilton Whittle 4, Peter Karayiannis 1,* 1

Department of Medicine A, Imperial College School of Medicine, St. Mary's Campus, South Wharf Road, London W2 1NY, UK 2 Wellcome Trust Centre for Human Genetics, Oxford University, Oxford, UK 3 Department of Zoology, Oxford University, Oxford, UK 4 Gambia Hepatitis Intervention Scheme, MRC Laboratories, Fajara, Gambia 5 London School of Hygiene and Tropical Medicine, London, UK

Background/Aims: Transmission of hepatitis B virus (HBV) in Africa occurs horizontally, with most people becoming infected between the ages of 1 and 5 years. The index cases in such events have been assumed to come from within the family unit or from sources outside the immediate family, such as other families or inhabitants of the same compound or village. Here, we de®ne these routes of transmission by phylogenetic tree analysis of sequences from the entire precore/core region of the virus, in Gambian chronic carriers. Methods: Ampli®cation by polymerase chain reaction of serum extracted HBV-DNA was followed by direct sequencing of the target region. Following editing and alignment of these sequences, phylogenetic tree analysis was performed using the neighbour-joining and maximum-likelihood methods. Results: Despite the overall conserved nature of the sequences of the pre-core/core region from 142 chronic carriers, distinct clusters were easily de®ned at the family and village level, but not on a wider geographical separation. Conclusions: Phylogenetic tree analysis of sequences obtained from family members provided strong evidence of intrafamilial transmission of HBV in at least two-thirds of the families studied from Gambia. q 2001 European Association for the Study of the Liver. Published by Elsevier Science B.V. All rights reserved. Keywords: Hepatitis B virus, Chronic carriers; Intrafamilial transmission; Phylogenetic tree analysis; Pre-core/core region; Gambia

1. Introduction Most hepatitis B virus (HBV) infections in sub-Saharan African infants and children are acquired through horizontal transmission of the virus in early childhood [1±3]. The exact mechanisms of spread, however, have never been documented. The virus has been found in exudates from tropical ulcers [4], and the possibility of transmission by blood-feeding arthropods has been considered [5], although a controlled trial of bedbug elimination in several villages failed to reduce HBV infection [6]. The risk behaviour Received 27 October 2000; received in revised form 15 February 2001; accepted 21 February 2001 * Corresponding author. Tel.: 144-207-886-6404; fax: 144-207-7249369. E-mail address: [email protected] (P. Karayiannis).

most strongly associated with HBV infection was shown to be the sharing of bath towels, sharing of chewing gum or partially eaten candies, sharing of dental cleaning materials and biting of ®ngernails in conjunction with scratching the backs of carriers [7]. Signi®cant clustering of HBsAg positive siblings suggestive of intrafamilial transmission was previously described in two Gambian villages [8]. The chance of the youngest child in a household being a carrier of HBsAg was strongly related to the number of antigen positive siblings. Thus, the risk of becoming a carrier if an older sibling was already a carrier was found to be signi®cantly higher than in the normal population. To date, only a few studies have tried to assess the likelihood of intrafamilial transmission. In order to determine the extent of intrafamilial versus intracompound or intravil-

0168-8278/01/$20.00 q 2001 European Association for the Study of the Liver. Published by Elsevier Science B.V. All rights reserved. PII: S 0168-827 8(01)00064-2

100

U. Dumpis et al. / Journal of Hepatology 35 (2001) 99±104

lage transmission, we have applied phylogenetic tree analysis to sequences of the pre-core/core and surface genes of HBV, obtained following ampli®cation of serum HBVDNA from a large number of Gambian carriers, including members of family clusters. 2. Materials and methods 2.1. Patients A total of 142 Gambian chronic HBV carriers from different areas and villages of the country who tested positive by PCR were included in the study. HBsAg positive index cases in households were identi®ed from data collected during previous studies [6,9,10] and family members were traced and bled. Seventy-®ve of the carriers were siblings from family clusters (32 families in total), whilst the remaining 67 patients were isolated cases of chronic HBV infection. Of these, 92 were hepatitis B e antigen (HBeAg) positive and 50 were negative for this marker.

2.2. Serological testing Following informed consent, a blood sample was drawn, and demographic data relevant to the study was entered on an appropriately designed form. Reverse passive agglutination (Hepato-test, Wellcome Diagnostics) was used to detect HBsAg. All negative samples were later retested with an immunometric assay (Sorin Biomedica, Italy). An immunoradiometric assay kit (Dia, Sorin) was used for the detection of IgM anti-core, HBeAg and anti-HBe.

2.3. PCR ampli®cation of the core promoter/pre-core/core region HBV-DNA was extracted from 50 ml of serum and resuspended in 20 ml of sterile distilled water as previously described [11]. Five microlitres of this were used to amplify the core promoter/pre-core/core region by PCR using the primers, M3 [12] and 5C (5 0 -CCCACCTTATGAGTCCAAGG, nt 2466±2486) [11]. Where necessary, second round ampli®cation was performed under the same reaction and cycling conditions with the primers, M3 and 8C (5 0 -GTCCCTGGATGCTGGATCTTGCT, nt 2130±2154), and the primers, 6C (5 0 -GGCAAGCCATTCTTTGCTGGGG, nt 2070±2092) and 5C, generating two partially overlapping fragments and covering the entire region under study.

2.4. PCR ampli®cation of the surface gene region Surface gene amplicons were obtained with the primers, S3 (5 0 -CAAGGTATGTTGCCCGTTTG, nt 457±476) and S4 (5 0 -GGGTTTAAATGTATACCCAGAGAC, nt 817±838), under the same reaction and thermocycling conditions described before [11].

2.5. Sequencing and phylogenetic analysis Amplicon DNA was puri®ed using the QIAquick Gel Extraction Kit (Qiagen, Crawley, West Sussex), following electrophoresis in 1% agarose gels. Puri®ed fragments were sequenced using the primers, 5C and M3, in an ABI 377 automatic sequencer. The sequences were edited manually (DNASIS, Hitachi, Japan) and subsequently aligned in FASTA format. Phylogenetic relationships among HBV sequences were reconstructed using the neighbour-joining (NJ) and maximum-likelihood (ML) methods, with bootstrap resampling in the latter case. All these analyses were conducted using the PAUP* package version 4.02b. Trees were drawn and rearranged using the TreeView programme.

3. Results 3.1. Surface gene sequence analysis Phylogenetic tree analysis of 35 S gene sequences revealed two major clusters, representing genotypes A and E (data not shown). However, this region was not ideal for demonstrating reliably clusters of common source infections, since it is relatively well conserved. 3.2. Phylogenetic tree analysis of pre-core/core sequences In order to determine the genotype and general patterns in the data, an initial NJ tree was constructed using all 142 sequences (not shown). Two major clusters representing genotypes A and E were again apparent, the latter being predominant (124/142). In addition, sequences from several family members were clustered together (22/32 families), indicating a close evolutionary relationship and therefore, likely transmission of the virus between them. In one half of these families, this was exclusively the case, whilst in the other half, there was evidence of transmission from within and from outside the family. Clustering according to village of birth was also evident in some cases, but no such clustering was observed in relation to geographical region. 3.3. Detailed analysis of family and village clustering To con®rm family or village clustering, viral sequences were re-analyzed after subdivision into four different data sets: (i), genotype A (n ˆ 14); (ii), genotype E (n ˆ 99); (iii), HBeAg positive (n ˆ 53); and (iv), HBeAg negative, genotype E sequences (n ˆ 30). Sequences which were the sole representative from any particular village were excluded. Genotype A sequences were not subdivided according to HBeAg status because of their small number. The general pattern of clustering according to family or village was maintained (Figs. 1 and 2). For genotype A, two separate clusters for families 125 and 214 with 95 and 68% bootstrap support, respectively were apparent, whilst sequences from family 222 did not cluster together (Fig. 1). Another cluster contained siblings from families 79 and 80, suggesting that the source of infection of these two families was the same or that they infected each other (82% bootstrap support). In genotype E, the number of samples was much larger, and clustering between families and villages was less well pronounced. Clusters from families 35, 89, 218, 238, 61 and 231 (coloured) were positioned separately in the tree and often supported by high bootstrap values (Fig. 2). Other family clusters grouped with one or two other sequences, such as families 31, 34 and 28 (family number in colour), or together, such as families 18 and 42 from different villages. Some families, therefore, had sequences very similar to those of families from different villages or even different areas. Identical viral sequences were observed in some families

U. Dumpis et al. / Journal of Hepatology 35 (2001) 99±104

101

clustering in villages and compounds. These are from the regional town V19, which has a much larger population. Thus, the risk of contracting the infection from an outside contact is higher. 3.4. Phylogenetic analysis according to HBeAg status Village and family clusters were de®nitely more distinct in the phylogenetic tree of sequences from HBeAg positive subjects (not shown). Some sequences (F33, 40 and 82), however, were still scattered throughout the tree and showed no relatedness. In contrast, family and village clusters were not so well pronounced in the tree of sequences from HBeAg negative subjects (not shown). The overall branch lengths were longer than those of HBeAg positive

Fig. 1. Unrooted ML phylogenetic tree of 14 genotype A pre-core/core sequences (Ts/Tv ˆ 1.09; a ˆ 0.25). Only bootstrap values higher than 50% are shown. Taxa information includes family number (F), if siblings are from families with multiple carriers, family member status (C, child; M, mother) and village number (e.g. V1). The bar indicates evolutionary distance.

between siblings (F18, 89, 217, 174), but also villages (V31 and V32). In the latter case, members of families 174 and 172 and one sibling from another family carried the same viral sequence. These two villages were small and geographically very close, so communication between them could not be excluded. Re-extraction of HBV-DNA followed by ampli®cation gave similar results, thus excluding contamination artefacts. Village clusters were also evident as in the case of six sequences from village V3 from non-relatives, two from V18, and one from V13 clustering with family 89 from the same village (Fig. 2). V9 in Fig. 1 is another example. Interestingly, families 40 and 41 do not show any signs of

Fig. 2. Unrooted ML phylogenetic tree analysis of 99 genotype E precore/core sequences (Ts/Tv ˆ 1.60; a ˆ 0.13). Only bootstrap values higher than 50% are shown. Full colour, intrafamilial transmission; coloured family number, transmission within and outside family; coloured village number, intravillage transmission.

102

U. Dumpis et al. / Journal of Hepatology 35 (2001) 99±104

sequences, possibly because of the increased number of nucleotide substitutions in HBeAg negative samples. 3.5. Sequences from mothers and their children Sequences from four mothers and their children were available for analysis. These were the only PCR positive mothers from those tested. Only one family (219) did not seem to share similar sequences (Fig. 3). If infection was acquired from the mother, it seems likely that this was acquired in early childhood, and therefore, the virus may have mutated, especially in those who were HBeAg negative. This could explain why the sequence from mother F89 was more distant than those of her children. The mother in family 36 and child shared both the core promoter and precore stop-codon mutations and the elevated rate of nucleotide substitution they experienced may explain the large phylogenetic distance of these viruses and why they did not cluster together in the main tree (Fig. 2). 3.6. Statistical analysis of clustering in families and villages A ML approach was also used to determine whether sequences were clustered by family and village more than might be expected by chance alone. First, for each of the four data sets, 200 random tree topologies were constructed. The log likelihood of each of these replicate trees was then estimated, using empirical transition/transversion (Ts/Tv) and a values. As these trees represent random assortments of the data, they effectively constitute a null distribution of log likelihoods. Using the TreeView program, three more trees were constructed: (i), the `family' tree, in which only sequences from the same family were grouped together; (ii), the `village' tree, in which sequences from the same village were grouped together, ignoring family origin; and (iii), the `family 1 village' tree, in which sequences from the same family and village were grouped together. In each case, all other branches were left unresolved. The log likelihoods of these model trees were then estimated under the data as before. If these likelihoods fell outside the null distribution, then we could conclude that the HBV sequence data showed more clustering by family and village than might be expected by chance. The ML randomization test provided strong evidence that

Fig. 3. Unrooted ML phylogenetic tree of sequences from families with HBsAg positive mothers and their children.

viruses tend to be related according to the family and/or village they were sampled from (Table 1). In the case of both the complete A and E genotype data sets, the log likelihoods for all three model trees fell outside the null distribution provided by the random topologies, strongly suggesting that there is more clustering by these variables than might be expected by chance. The combined family 1 village tree for genotype E also had a much higher likelihood than either component tree alone, suggesting that these two levels of population structure contained the most information. All three model trees also fell outside of the null distribution of log likelihoods in the analysis of the HBeAg positive genotype E sequences. The only case in which there was no support for village/family clustering was in the HBeAg negative cases. This may be due to common immunological selection pressures which increase the extent of convergent evolution.

Table 1 Comparison of log likelihoods for three different `model' phylogenetic relationships of HBV sequences a and a null distribution constructed using random tree topologies b Genotype

n

Null distribution

Family tree

Village tree

Village 1 family tree

A E E (HBeAg 1 ) E (HBeAg 2 )

14 99 53 30

2 1328.12 to 21389.84 2 4038.09 to 24197.20 2 2202.86 to 22394.43 2 2131.49 to 22213.21

2 1323.49 2 3851.96 2 2188.84 2 2162.7

2 1306.50 2 3816.84 2 2155.97 2 2139.66

2 1306.50 2 3731.27 2 2129.27 2 2139.66

a

Family, village and village 1 family. Log likelihoods that fall outside of the null distribution, and hence provide evidence for phylogenetic relationship at that level than might be expected by chance alone, are given in bold type. b

U. Dumpis et al. / Journal of Hepatology 35 (2001) 99±104

4. Discussion Nucleotide sequence comparisons have been used previously to demonstrate the transmission of HBV from mother to infant, but also in common source outbreaks of infection in different epidemiological settings. The presence of similar nucleotide sequences, or of well characterized variants of the virus, such as the pre-core stop-codon and surface variants, or viruses with deletions, have been reliable indicators of transmission from one person to another [13±17]. The exact mechanism of HBV transmission in Gambia has been the subject of speculation for a long time. In the present study, we have attempted to address this question using rigorous phylogenetic tree analysis of sequences obtained from the pre-core/core region of the virus, from family members and other chronic carriers. Crucially, this analysis provided evidence for signi®cant clustering of chronic HBV infection in families, as reported previously on the basis of serological ®ndings [1,8]. Sequencing of the targeted surface gene did not provide suf®cient information for reliable phylogenetic tree analysis, as it is well conserved and contains the `a' antigenic determinant which overlaps with the polymerase open reading frame. Twelve sequences from different families, villages and geographical areas were completely identical. The pre-core/core gene sequences were also very similar, but there was more phylogenetic signal in the data, and hence, an analysis of family and village clustering could be undertaken. However, both gene regions were able to clearly distinguish between genotypes A and E. At the geographical area level, it was not possible to see any signi®cant clustering of pre-core/core sequences. In some cases, sequences from completely different villages and geographical areas were identical. This suggests that extensive mixing of strains (`gene ¯ow') occurs at this level, most likely caused by the movement of infected individuals among villages. In contrast, strong evidence for phylogenetic clustering was seen at the family and village level. This was clear from both the ML randomization test and the fact that clustering of sequences in some families was supported by high bootstrap values (.70%). Sequences in the siblings remained identical in some families (F51, 81, 174, 217, 231), over the period from infection to their recruitment. This is consistent with a previous study of 42 HBsAg positive subjects from 11 Chinese families, where perinatal transmission was thought to be the predominant route of infection. A higher degree of sequence similarity was shown within, than between families [18]. A similar observation was made in Vietnamese and Turkish families [19], where sequences did not change, even after 20±30 years. Some sequences from non-relatives within villages were also identical over the pre-core/core region, a ®nding not previously described. In Gambia, infection occurs during early childhood [1]. All patients with identical sequences were HBeAg positive, which is consistent with previous

103

observations that before seroconversion to anti-HBe, sequence mutations are rare [18]. The fact that children from different families, but from the same village of birth, carry the same viral sequence is very strongly suggestive that these children were infected horizontally from a common source or from each other. For example, in the cluster from the village V9, families 79 and 80 were clustered together and these clusters were indistinguishable from each other. Both families were unrelated, but inhabited the same compound. There were families with no evidence of sequence clustering and spread right across the phylogenetic tree. Using genetic markers from host DNA, it is certain that these families were truly related and that the siblings had the same parents. Unfortunately, information on where individual siblings grew is not available. Children may be sent away during early childhood to stay with relatives living at some distance, particularly at the time of weaning. It is therefore possible that infection occurred in these families during such periods of separation. It is also very common that children from one, or even several compounds are kept together and handled by different women, thus increasing the risk of infection from other sources. Therefore, based on epidemiological observations and phylogenetic and statistical analyses, a model of transmission can be proposed. Children can get infected horizontally from their siblings, mothers, other compound members, and even children outside the compounds. However, this does not mean that all the children, even in the small villages, are infected from the same source. Children could become infected through close contact during play, sharing the same bed during sleep, or other activities. How this transmission actually occurs and what factors facilitate it, are unclear and merit further investigation. Acknowledgements The authors would like to thank to Dr Michael Wansbrough for his assistance in sample collection and ®eld assistants Lamin Giana, Joseph Bass and Adam Jeng. The authors also thank Professor Keith McAdam, Dr Abdoulie Jack and Ebrima Bah for assistance and help. The work was supported from grants from Roche and the Medical Research Council (UK). References [1] Whittle HC, Bradley AK, McLauchlan K, Ajdukiewicz AB, Howard CR, Zuckerman AJ, et al. Hepatitis B virus infection in two Gambian villages. Lancet 1983;1:1203±1206. [2] Tabor E, Bayley AC, Cairns J, Pelleu L, Gerety RJ. Horizontal transmission of hepatitis B virus among children and adults in ®ve rural villages in Zambia. J Med Virol 1985;15:113±120. [3] Davis LG, Weber DJ, Lemon SM. Horizontal transmission of hepatitis B virus. Lancet 1989;1:889±893. [4] Foster O, Ajdukiewicz A, Ryder R, Whittle H, Zuckerman AJ. Hepa-

104

[5] [6] [7]

[8] [9] [10] [11]

[12]

U. Dumpis et al. / Journal of Hepatology 35 (2001) 99±104 titis B virus transmission in West Africa: a role for tropical ulcer? Letter. Lancet 1984;1:576±577. Vall Mayans M, Hall AJ, Inskip HJ, Chotard J, Lindsay SW, Coromina E, et al. Risk factors for transmission of hepatitis B virus to Gambian children. Lancet 1990;336:1107±1109. Vall Mayans M, Hall AJ, Inskip HM. Do bedbugs transmit hepatitis B? Letter. Lancet 1994;344:962. Martinson FE, Weigle KA, Royce RA, Weber DJ, Suchindran CM, Lemon SM. Risk factors for horizontal transmission of hepatitis B virus in a rural district in Ghana. Am J Epidemiol 1998;147:478± 487. Whittle H, Inskip H, Bradley AK, McLaughlan K, Shenton F, Lamb W, et al. The pattern of childhood hepatitis B infection in two Gambian villages. J Infect Dis 1990;161:1112±1115. Inskip HM, Hall AJ, Chotard J, Loik F, Whittle H. Hepatitis B vaccine in the Gambian Expanded Programme on Immunization: factors in¯uencing antibody response. Int J Epidemiol 1991;20:764±769. Viviani S, Jack A, Hall AJ, Maine N, Mendy M, Montesano R, et al. Hepatitis B vaccination in infancy in The Gambia: protection against carriage at 9 years of age. Vaccine 1999;17:2946±2950. Karayiannis P, Alexopoulou A, Hadziyannis S, Thursz M, Watts R, Seito S, et al. Fulminant hepatitis associated with hepatitis B virus e antigen-negative infection: importance of host factors. Hepatology 1995;22:1628±1634. Carman W, Jacyna MR, Hadziyannis S, Karayiannis P, McGarvey M, Makris A, et al. Mutation preventing formation of hepatitis B e anti-

[13]

[14] [15]

[16]

[17] [18]

[19]

gen in patients with chronic hepatitis B infection. Lancet 1989;2:588± 591. Lin HJ, Lai CL, Lau JY, Chung HT, Lauder IJ, Fong MW. Evidence for intrafamilial transmission of hepatitis B virus from sequence analysis of mutant HBV DNAs in two Chinese families. Lancet 1990;336:208±212. Akarca US, Greene S, Lok AS. Detection of precore hepatitis B virus mutants in asymptomatic HBsAg-positive family members. Hepatology 1994;19:1366±1370. Barlet V, Zarski JP, Thelu MA, Bichard P, Seigneurin JM. Different prevalence of precore mutants in ®ve members of a hepatitis-B-virusinfected family: evidence for a precore variant type in an asymptomatic anti-HBs patient. J Hepatol 1994;21:797±805. Ho MS, Lu CF, Kuo J, Mau YC, Chao WH. A family cluster of an immune escape variant of hepatitis B virus infecting a mother and her two fully immunized children. Clin Diagn Lab Immunol 1995;2:760± 762. Santantonio T, Jung MC, Pastore G, Angarano G, Gunther S, Will H. Familial clustering of HBV pre-C and pre-S mutants. J Hepatol 1997;26:221±227. Bozkaya H, Akarca US, Ayola B, Lok AS. High degree of conservation in the hepatitis B virus core gene during the immune tolerant phase in perinatally acquired chronic hepatitis B virus infection. J Hepatol 1997;26:508±516. Hannoun C, Horal P, Lindh M. Long-term mutation rates in the hepatitis B virus genome. J Gen Virol 2000;81:75±83.