Recombination in the Genesis and Evolution of Hepatitis B Virus Genotypes

Hepatitis B virus (HBV) infection is widely distributed in bothhuman and ape populations throughout the world and is a majorcause of human morbidity and mortality. HBV variants are currentlyclassified into the human genotypes A to H and species-associatedchimpanzee and gibbon/orangutan groups. To examine the roleof recombination in the evolution of HBV, large-scale data retrievaland automated phylogenetic analysis (TreeOrder scanning) werecarried out on all available published complete genome sequencesof HBV. We detected a total of 24 phylogenetically independentpotential recombinants (different genotype combinations or distinctbreakpoints), eight of which were previously undescribed. Instancesof intergenotype recombination were observed in all human andape HBV variants, including evidence for a novel gibbon/genotypeC recombinant among HBV variants from Vietnam. By recordingsequence positions in trees generated from sequential fragmentsacross the genome, violations of phylogeny between trees alsoprovided evidence for frequent intragenotype recombination betweenmembers of genotypes A, D, F/H, and gibbon variants but notin B, C, or the Asian B/C recombinant group. In many cases,favored positions for both inter- and intragenotype recombinationmatched positions of phylogenetic reorganization between thehuman and ape genotypes, such as the end of the surface geneand the core gene, where sequence relationships between genotypeschanged in the TreeOrder scan. These findings provide evidencefor the occurrence of past, extensive recombination events inthe evolutionary history of the currently classified genotypesof HBV and potentially in changes in its global epidemiologyand associations with human disease.

INTRODUCTION

Top

Introduction

Infection with hepatitis B virus (HBV) is globally distributed,infecting approximately one-third of the world's human population.It is responsible for substantial proportion of liver diseasethat kills over one million people through uncompensated liverdisease or hepatocellular carcinoma each year (43). Populationsin South and East Asia, sub-Saharan Africa, and Central andSouth America show particularly high frequencies of HBV infectionthat may be maintained through mother-to-child perinatal transmission(13) or horizontal transmission in childhood (8, 41).

HBV variants infecting humans show genetic and antigenic heterogeneityand are currently classified into seven or eight genotypes thatdiffer from each other in nucleotide sequence by 10 to 13% (29-31,33). While genotypes A, D, and possibly G have global distributions,genotypes B and C are found predominantly in East and SoutheastAsia, genotype E is found in West Africa, and genotypes F andH are found among various population groups, including indigenouspeoples in Central and South America. Active and resolved HBVinfections are also found at high frequencies in chimpanzees(17, 17, 24, 42, 45) and Southeast Asian apes (12, 28, 46, 49)with species-associated HBV variants distinct from those foundin human populations. Finally, an HBV variant was recoveredfrom a captive woolly monkey (21), whose sequence is highlydivergent from all other human and ape sequences described todate.

It is increasingly accepted that, in addition to the currentlyclassified genotypes of HBV, recombination between genotypesgenerates novel variants that contribute to the genetic diversityof HBV. Recombination in human HBV was first detected in sequentialsamples from infected individuals and among integrated sequencesfrom cases of hepatocellular carcinoma (11, 44). However, recombinationbetween different genotypes was also suspected from incongruitiesbetween phylogenetic trees constructed from different regionsof the HBV genome (30) and was formally demonstrated throughthe identification of specific breakpoints in certain publishedsequences, such as HBVDNA (4), proposed to be a recombinantbetween genotypes A and D. Both the existence of well-definedbreakpoints and the exchange of large sections of sequence betweengenotypes indicate that recombination, rather than phenotypicallyselected accelerated evolution and/or convergence (e.g., suchas mediated by the immune system [22]), most plausibly accountsfor the existence of phylogenetic incongruities between HBVvariants in different parts of the genome.

Since then, using similar analytical methods, there have beenseveral published descriptions of intergenotype recombinantsamong newly obtained or existing variants of HBV. Among these,recombinant forms between A and D (5, 26, 34), B and C (5, 26,26, 27, 30, 40, 52), and C and D (7, 48) have been identified.However, in many cases, recombinants may be separate instancesof phylogenetically related viruses that have become widelydistributed geographically. For example, a recombinant generatedthrough the exchange of the core gene of genotype C into a genotypeB variant, often described as genotype Ba (with Bj designatingnonrecombinant genotype B), has become prevalent in mainlandEast Asia and independently recorded in several surveys (5,26, 40).

In this study, we have sought to identify independent intergenotyperecombinants of HBV by large-scale screening and phylogeneticanalysis of complete genome sequences available in public sequencedatabases. Using a series of novel phylogenetic and bioinformaticanalysis methods, we have mapped recombination hot spots topositions of phylogeny rearrangements within currently classifiedHBV genotypes. These findings, combined with the new evidencefor extensive recombination between members of the same genotype(intragenotype recombination), provide new data on the centralrole of recombination in the past evolution of HBV and in thegeneration of its current genetic diversity.

MATERIALS AND METHODS

Top

Materials and Methods

Nucleotide sequences. The study sequences comprised all available complete genomesequences of HBV available initially in October 2004 from GenBank,EMBL, and DDBJ. Sequences were screened to exclude patents,artificial mutants, and identical entries. The remaining sequenceswere aligned manually in the Simmonic sequence editor package(available from http://www.polio.vir.gla.ac.uk). From the initialdownload, identical sequences or sequences showing less than1% sequence divergence from each other were removed, leavinga total of 526. Using genotyping information in the originalsources, or through analysis of the hepatitis B surface antigengene, these sequences comprised 79 genotype A, 80 genotype B,161 genotype C, 56 genotype D, 6 genotype E, 23 genotype F,8 genotype G, and 7 genotype H sequences. A further 13 and 30complete HBV genome sequences from chimpanzees and gibbons/orangutans,respectively, were added to the study data set, along with thedivergent HBV variant (NC_001896) from the woolly monkey (usedas an outgroup in most subsequent phylogenetic analyses).

Sequence selection. To ensure as fair a weighting as possible in the analysis ofsequence groups and recombination, several complete genome sequencesof the overrepresented genotypes A to C were removed by similarityscanning. This ensured that the sequences used for analysiswere both representative and maximally informative phylogenetically.After recombinant sequences were removed, target numbers ofapproximately 40 to 60 sequences from each of the three genotypeswere obtained by excluding sequences with similarity thresholdsof below 1.5% over aligned complete genome sequences for genotypeA and B sequences (leaving 43 and 52, respectively) and 2.5%for genotype C (leaving 64). Genotype B sequences were furtherdivided by exclusion of B/C recombinants (Ba genotype [26])to leave 24 nonrecombinant variants (corresponding to Bj). Forother genotypes (including those from primates), all availablesequences were used, although genotypes unavoidably differedfrom each other in their intragenotype variability. These sequencescomprised set A and were used for group scanning analysis. Consensussequences from each group (50%) were used for SIMPLOT.

After the initial analysis, a second download of sequences publishedfrom October 2004 until June 2005 was carried out, and usingthe selection criteria described above to eliminate closelyrelated sequences, the following sequences were combined toform set B: genotype A, 45; genotype B (nonrecombinant), 25;genotype B/C recombinants (recB), 51; genotype C, 70; genotypeD, 59; genotype E, 13; genotype F, 25; genotype G, 9; and genotypeH, 9 sequences. No further nonhuman primate-derived sequenceswere available. A full listing of sequences in sets A and Band recombinant sequences identified in the study is availablefrom the authors.

Assignment of recombinant forms of HBV. HBV variants were assigned as confirmed intergenotype recombinantswhere there was published evidence for their sequences havingbeen obtained from two or more clones demonstrating identicalrecombination breakpoints or from variants with identical breakpointsdescribed in separate studies or from different HBV-infectedstudy subjects in the same study. The remaining sequences wereassigned as provisional recombinants. For these variants, therewas insufficient published or unpublished data to formally ruleout the possibility that the complete genome sequences werenot composites of fragments amplified from samples of individualswith mixed genotype infections.

RESULTS

Top

Results

Phylogenetic relationships between HBV genotypes. To visualize phylogenetic relationships in different parts ofthe HBV genome, a new method was developed to record positionsof each sequence in a series of phylogenetic trees generatedby sets of overlapping fragments across the genome. Sequencefragments of 250 bases, incrementing in 25-base steps were used(125 trees in total) (Fig. 1). Boot-strapped phylogenetic treeswere constructed using the programs SEQBOOT, DNADIST, NEIGHBOR,and CONSENSE in the PHYLIP version 3.62 package (10). The sequencedata set comprised a total of 348 human and ape HBV sequences(set B) from which all recombinant sequences had been excludedwith the exception of B/C recombinant sequences (recB) (Fig.1, bright green).

View larger version (58K):
[in this window]
[in a new window]
FIG. 1. TreeOrder Scan of nonrecombinant HBV and recB sequences, showing positions of individual sequences (y axis) in phylogenetic trees generated from sequential 250-base sequence fragments, incrementing by 25 bases (midpoints indicated in x axis). Changes in sequence orders resulting from changes in phylogeny at the 70% bootstrap level are shown. Sequences are color coded by genotype, as indicated by labels in left and right margins: genotype A, red; B, light green; recB (rB), green; C, yellow; D, blue; E, purple; F, pink; G, pale yellow; H, brown; chimpanzee (Ch), gray; gibbon (Gi), white; orangutan (Or, turquoise; woolly monkey (outgroup on line 1), black. The recombinant sequence, AY161141, is shown as a dotted line. The boundaries of sequence groupings with 70% or greater bootstrap support are indicated with black horizontal lines; for clarity, only clades with six or more members were demarcated. All sequence positions were numbered relative to the HBVADW2 reference sequence (length, 3,221 bp). For comparison, the TreeOrder Scan was aligned with a scale genome diagram of HBV (upper panel) and with positions of breakpoints in recombinant HBV variants (triangles in lower panel). Each independent breakpoint is indicated as a square, localized to one of 61 separate 250-base sequence fragments spanning the genome, with segment 62 representing the junction between the beginning and end of the alignment. Genotype transitions are shown in the upper-left and lower-right quadrants of the square and are color coded as for the TreeOrder Scan; unfilled triangles represent unknown genotypes; gibbon sequences are in black.

The vertical ordering of sequences in each tree was determinedby the following criteria. (i) Each tree was permuted to matchas closely as possible the sequence order of the tree generatedfrom the immediately preceding fragment. (ii) Allowable permutationscomprised branch rotations and positional reordering of sequencesor groups of sequences, provided this occurred within groupingsbelow the specified bootstrap threshold value (70% for the analysesshown in Fig. 1). (iii) Each tree was rooted using the woollymonkey HBV sequence (NC_001896), whose position was predefinedat position 1.

The ordering of sequences in sequential trees according to theabove rules was automated in the program TreeOrder Scan in theSimmonic 2005 version 1.4 Sequence Editor package (which containsfull information on required files and output formats). Theoutput of the program comprises a numeric array that recordsthe position of each sequence (columns) for each tree (rows).Each line in the graph (Fig. 1), representing the position ofeach sequence across the genome, was color coded by genotype(see left and right columns). Because the algorithm attemptsto match sequence order between fragments, alterations in thetree order of individual sequences and of whole genotypes betweenfragments indicate changes in the phylogenetic relationshipsof clades supported by bootstrap values, in this case, of 70%or greater. For sequences with constant phylogenetic relationshipsacross the genome, each sequence would follow a horizontal lineacross Fig. 1. In contrast, phylogenetic relationships betweenmany genotypes changed frequently in different parts of thegenome.

First, there were differences between sequence fragments inthe degree of phylogenetic differentiation between genotypes,with clear delineation of genotypes in some cases in bootstrap-supportedclades (delineated by black dividing lines). In other cases,such as between positions 1600 and 1850 (corresponding to theregion of the genome encoding the X gene), sequences of genotypesC, A, E, D, and B were phylogenetically undifferentiated atthe 70% bootstrap threshold level.

Second, phylogenetic relationships between genotypes changedfrequently, leading to several forced alterations in branchingorder. For example, genotype D became an outgroup for all genotypesexcept genotypes F and H between positions 780 (end of the Sgene) and 980. Genotype G moved to a similar outlier positionfrom 980 to 2750. Genotype B sequences split at position 1950(start of core gene), with the majority (recB) grouping withgenotype C sequences and rejoining the others at around position2500 (end of core gene). A similar change in phylogenetic relationshipswas observed between genotypes D and E. These genotypes werephylogenetically distinct over most of the genome, but genotypeE fell within the genotype D clade in the core gene (positions1950 to 2500). Genotypes F and H did not invariably form separateclades (350 to 500). Finally, sequences from orangutans areinterspersed with gibbon sequences throughout the genome exceptfor the pre-S1 and pre-S2 regions where they were distinct.

Overall, the major phylogenetic discontinuities detected inthe TreeOrder Scan were at positions 750, 900, 1600, 1950, 2500,2650, and 2750. Most of these occurred at, or close to, geneboundaries and frequently occurred independently in differentgenotypes. Particularly frequent changes in phylogenetic relationshipswere found at each end of the core and surface genes. Sincethe grouping of many genotype B sequences in the genotype Cclade in the core gene has been previously interpreted as evidencefor intergenotype recombination, this process plausibly accountsfor other changes in tree order evident in the scan.

Screening for HBV recombination. To further investigate the potential of recombination in thephylogenetic discontinuities observed between HBV genotypes,a complete catalogue of recombination breakpoints among previouslypublished HBV sequences was compiled. All sequences availablein the GenBank database in June 2005 were screened using theprogram TreeOrder Scan incorporating the previously shown dataset of nonrecombinant sequences as controls (set B; n = 347).The TreeOrder Scan provides a rapid method to detect intergenotyperecombination among individual sequences. To demonstrate this,a previously undescribed recombinant sequence AY161141 was includedin the data set (Fig. 1). This sequence fell within the genotypeA clade (red) throughout the genome except for the core gene,where it grouped with genotype D (blue). This method was appliedusing several sets of downloaded sequences for rapid, large-scaleinitial screening of all published HBV sequences for recombination.

Sequences identified as recombinant or potentially recombinantbecause of changes in branching order relative to the controlgroups were further analyzed using SIMPLOT, version 2.5 (23),and a newly developed program called Group Scan available inthe Simmonic Sequence Editor package. The latter program recordshow deeply embedded a test sequence lies within each of theclades formed by standard phylogenetic analysis of preassignedgroups (nonrecombinant HBV genotypes in set A, excluding recBsequences), from which it calculates a grouping score (Fig.2). The method for scoring is related to that used by the AssociationIndex, a method developed previously to quantify the degreeof segregation of different predefined groups (6, 47).

View larger version (45K):
[in this window]
[in a new window]
FIG. 2. (A) Comparison of Group Scanning and SIMPLOT analysis of the recombinant sequence AY161141. In each case, sequence fragments of 250 bases incrementing by 50 bases, 100 bootstrap replicates, were compared with sequence groups (Group Scan) or 50% consensus sequences of the eight human genotypes and four primate genotypes (color coded as described in the legend of Fig. 1). For the group scanning analysis, only values of >0.5 indicate grouping within a genotype group. (B) Group Scan analysis of examples of newly discovered recombinant HBV variants (listed in Table 1) and of novel recombinant sites in a previously analyzed sequence, HBVDNA. (C) Group Scan and SIMPLOT analysis of the sequence AB048704. Because the recombinant region in the S gene was relatively phylogenetically uninformative, longer sequence fragments of 400 bases incrementing by 15 bases (Group Scan) or 20 bases (SIMPLOT) were analyzed. Sequences are color coded as described in the legend of Fig. 1, except that the gibbon HBV variants are shown in black. The closely related variant, AB048705, produced comparable results (data not shown).

For the Group Scan, the grouping score of a test sequence ina phylogeny is computed by reference to the assigned groupsof its nearest neighbors in unrooted trees. Grouping scores(G) for a predefined group, a, can be represented by the followingformula:

1
where N is the number of nodesseparating the test sequence from each member of the group,and y is the total number of sequences in group a. By definition,the total of grouping scores of each preassigned group, G_a toG_n, adds up to 1. For robustness, each score is the mean ofa predefined number of bootstrap replicate trees (100 in theanalyses shown) generated from permuted data sets where eitherthe nucleotide sequence data (using SEQBOOT in the PHYLIP package)or the population (selection of a random two-thirds of sequencesfrom each preassigned group) is resampled. To localize sitesof potential recombination, consecutive fragments of predefinedlength and incremental steps between fragments were analyzed.

For each fragment (midpoint plotted on x axis), the depth ofgrouping in each genotype was plotted (y axis) with values rangingfrom close to 0 (no grouping) to close to 1 (deeply buried withina clade formed by sequences of a genotype). Changes in genotypegrouping, potentially arising from recombination, are characterizedby abrupt changes in grouping values of the sequence, as shownfor AY161141 (Fig. 2A). Concordant with its changing positionon the TreeOrder Scan (Fig. 1), this sequence fell within thegenotype A clade between positions 2400 to 1650 (values of >0.5;i.e., mean values for the 50 bootstrap replicated trees laywithin the outgroup value) and within the genotype D clade from1700 to 2400. The underlying methodology of the Group Scan isdifferent from SIMPLOT, which records the bootstrap value ofa test sequence with artificially generated consensus sequencesof each genotype. However, the output from both programs isfrequently comparable, with evidence for alternate groupingsof the AY161141 sequence with the genotype A (consensus) sequencebetween positions 2400 to 1600 and with genotype D between 1800to 2400 (Fig. 2A), as found in the Group Scan.

A detailed comparison of the two methodologies will be presentedelsewhere; for the purposes of the current study, both methodswere used to scan recombinant and potential recombinant sequencesin data set A of HBV complete genome sequences (Table 1). Onlybreakpoints identified by both methods were included in theanalysis, since inconsistencies between methods (such as theapparent recombination site between genotypes A and G at position700 in the SIMPLOT analysis of AY161141) (Fig. 2A) may representartifacts, in this case arising from a more general change inphylogenetic relationships between genotypes at the end of thesurface gene. However, in general, the methods produced concordantresults for almost all recombination sites.

View this table:
[in this window]
[in a new window]
TABLE 1. Compilation of recombinant HBV sequences

The screening carried out on GenBank sequences independentlyidentified all recombinant sequences described to date in theHBV literature (Table 1). However, the analysis in this studyidentified a series of previously undescribed recombinant forms,including several A/D recombinants, a C/B, a D/C, and an F/Crecombinant from Bolivia (Fig. 2B). For the sequence HBVDNA,reported as a D/A recombinant (4), there was evidence from bothGroup Scanning and SIMPLOT for additional recombination pointsbetween 1300 and 1450 (Fig. 2B). For the sequences, HPBADW1and HBVP4PCXX, which have previously been described as recombinant(4, 5, 26), neither group scanning nor SIMPLOT identified clearbreakpoints, and these sequences have therefore been omittedfrom subsequent analysis.

Finally, the Group Scan identified AB048704 and AB048705 fromVietnamese HBV patients as potential gibbon/genotype C recombinants(Fig. 2C and 3) (the only published analysis of them was carriedout before the gibbon-derived sequences were available). Betweenpositions 300 and 1000, both sequences either formed outgroupsof or fell within the gibbon clade (grouping values of $≥$ 0.5).The grouping of AB048704 and AB048705 with gibbon-derived variantswas confirmed by separate phylogenetic analysis, where bothsequences (along with those from orangutans) clearly fell withinthe gibbon sequence clade (Fig. 3). Because this region of theS gene is relatively phylogenetically uninformative, SIMPLOTfailed to record high bootstrap scores in the recombinant region,although the analysis confirmed a change in grouping from genotypeC to gibbon and back at the same breakpoints identified by theGroup Scan (Fig. 2C).

View larger version (18K):
[in this window]
[in a new window]
FIG. 3. Phylogenetic analysis of positions 251 to 650 (corresponding to the putative recombinant region in the S gene) (Fig. 2C) of sequences AB048704 and AB048705 (Rec), all available gibbon- and orangutan-derived sequences, and nonrecombinant human- and chimpanzee-derived HBV variants (set A). Since sequences from each genotype were monophyletic, only the most recent common ancestor (MRCA) is shown for non-gibbon genotypes. The tree was constructed by neighbor-joining using Jukes-Cantor corrected distances in the MEGA2 package (19), using 200 bootstrap replicates (values of 70% or greater shown).

For analysis of breakpoint positions, it was apparent that manyrecombinant HBV variants were phylogenetically related and containedidentical recombination sites (such as among the widespreadB/C recombinant form in East Asia and the newly discovered D/Crecombinants, such as AY057948 and AY817511 from Tibet and China[7, 48]). Therefore, only HBV variants that showed clearly differentbreakpoints and/or genotype composition were included in subsequentanalysis to ensure that the recombination events were phylogeneticallyindependent and not multiply scored. Breakpoints were localizedto the segment (from a total of 62) where grouping or bootstrapvalues crossed the threshold 0.5 value for the two sequences.The positions of the breakpoints were aligned with the HBV genomediagram and with the TreeOrder Scan of nonrecombinant sequences(Fig. 1).

Overall, the distribution of recombination breakpoints was nonrandom,with certain apparently favored sites (such as fragment 14 [positions650 to 900]), where several different recombination events betweendifferent genotypes occurred (between genotype A and D variants,as well as between D and C). Favored recombination sites frequentlylocalized to gene boundaries, such as the end of the S gene(fragment 14), the precore/C gene start, the end of C gene,and the pre-S1/pre-S2 junction. Elsewhere, however, breakpointswere more randomly distributed, such as between fragments 19and 32 (the 3' end of the P gene and the X gene).

Recombination within HBV genotypes. Apart from the changes in branching order of different genotypesand the multiple examples of phylogenetically independent recombinantevents (previous sections), further evidence for recombinationwas provided by changes in tree order between consecutive fragmentsamong variants conventionally regarded as nonrecombinant. Forexample, following the tree position of individual genotypeA (red) and D (blue) sequences across the genome (Fig. 1) revealsseveral sites where positions change in response to changesin phylogenetic relationships between members of the same genotype.

To investigate whether these violations of intragenotype phylogeniesbetween fragments arose more frequently than would be expectedby chance, frequencies of sequence order changes within genotypeswere compared with those of artificially generated, nonrecombinantcontrol data sets of sequences corresponding exactly in sequencedivergences between native sequences. A total of 50 independentlyreplicated data sets of the original 311 native sequences inset A were created using a maximum likelihood method implementedin the program EVOLVER in the PAML package (51), with likelihood-derivedparameters for the model, rate variation between sites, frequenciesof invariant sites, and sequence changes in different modelcategories. Separate analysis was also carried out on separatesubsets of certain HBV genotypes that contained more than 20representative sequences (A, B, recombinant B/C [recB], C, D,F/H combined, and Gibbon). Phylogeny violations in native andcontrol data sets were calculated as the mean number of treeorder changes observed on pairwise comparison of each sequencefragment, normalized by the number of sequences, using a rangeof bootstrap threshold values (Table 2).

View this table:
[in this window]
[in a new window]
TABLE 2. Phylogeny violations within HBV genotypes

For all meaningful bootstrap threshold values (i.e., equal toor greater than 50%), higher frequencies of phylogeny violationswere observed in native genotype A, D, F/H, and gibbon sequencescompared with mean values for control data sets (Table 2). Toestimate the statistical significance of these differences,the positions of native scores in the distribution of thosefrom control sequences was determined and expressed as a Z score.Values of greater than 2 represent frequencies of phylogenyviolations outside the 95th percentile for the control values(two-tailed test). Sequences within genotypes A, D, F/H, andgibbon-derived HBV variants showed significantly greater frequenciesof phylogeny violations than would be expected by chance (Zscores ranging from 3.4 to 7.7), while no significantly greatertree order changes were found in genotype B, recB, or C (Z scoresof –0.1 to 0.94).

The number of phylogeny violations attributable to recombinationwas calculated by subtracting the control from the native value(Table 2, column 9); this indicated substantial number of separaterecombination events in genotype A, D, and gibbon sequences,and none or close to none in genotypes B, recB, and C. Frequenciesof phylogeny violations were frequently, but not invariably,greater among the more divergent genotypes (measured as meanintragenotype pairwise distance) (Table 2, column 3). Althoughcarried out as a separate analysis, the number of phylogenyviolations in the whole data set of 312 sequences of all genotypes(n = 110) approximated to the total number of events from theindividual genotypes analyzed (n = 119). As expected, repeatingthe analysis using higher bootstrap threshold values recordedprogressively fewer recombination events (for the whole dataset: 62 at 70%, 52 at 75%, 44 at 80%, and 28 at 90%) but retaineddisparities in phylogeny violations between native and controldata sets for each genotype (Z scores for whole data set of4.91 at 60%, 3.44 at 70%, 3.15 at 75%, 2.6 at 80%, and 2.43at 90%).

The locations of phylogeny violations in gibbon-derived HBVsequences were determined by plotting frequencies for individualpairwise comparisons of fragments in the form of a half-diagonalphylogenetic compatibility matrix (Fig. 4). This analysis revealsseveral regions of low values (blue) where phylogenies werecompatible, interrupted by often sharp boundaries (arrows) splittingsequences into a series of noncompatible blocks. In two of thethree indicated positions, these corresponded to favored sitesof recombination among intergenotype recombinant HBV variants(Fig. 1, lower panel).

View larger version (56K):
[in this window]
[in a new window]
FIG. 4. Phylogenetic compatibility matrix of gibbon-derived HBV sequences, showing frequencies of phylogeny violations for each pairwise comparison of sequence fragments. Frequencies are color coded to indicate number of phylogeny violation per sequence (see key). Phylogenetically compatible regions are shown as deep blue. Arrows indicates the main sites where changes in phylogeny of the gibbon sequences lead to regions of incompatibility between fragments. For this analysis, sequence fragments of 400 bases, incrementing by 25 bases across the genome, were compared (117 fragments total), using a bootstrap value of 70% above which phylogeny violations were computed.

DISCUSSION

Top

Discussion

Intergenotype recombination. This study provides comprehensive information on the occurrenceand composition of HBV intergenotype recombinants. A total of24 phylogenetically independent recombinant forms of HBV werefound, involving all human genotypes as well as both chimpanzeeand gibbon variants. From the available data on their geographicaldistributions and the finding of the same recombinants fromdifferent published sources, it is apparent that recombinantforms are capable of spread in human populations and of developingtheir own specific distributions and epidemiologies. This isparticularly clearly demonstrated for the B/C recombinant, whichnow accounts for the majority of genotype B sequences in mainlandAsia (39). Another example is the D/C recombinant, found ina separate survey in Tibet (AY057948 [7]) and China (AY817511[48]). While the current study is as up to date as possible,clearly more extensive sampling and sequence analysis are likelyto reveal further examples of recombinant forms of HBV as awider range of human populations are surveyed in the future.

Apart from documenting individual examples of intergenotyperecombinants, this study provides evidence that the evolutionof what are conventionally regarded as "nonrecombinant" genotypesof HBV has been accompanied by a similar series of recombinationevents. This process is manifested most clearly in the frequentchanges in phylogenetic relationships between genotypes acrossthe genome in the TreeOrder Scan at positions that frequentlycoincide with breakpoints in recombinant viruses (end of S gene,start of pre-S2, and both ends of the C gene) (Fig. 1). Forexample, genotype G shows regions where it is highly divergentin sequence from genotypes A to E and primate viruses althoughin the S gene, genotype G is approximately equidistant fromgenotypes A to E; in a third region within the small S fragment,it becomes interspersed with genotype A (250 to 350). GenotypeE has a similarly variable relationship with genotype D (interspersedin the core gene and distinct elsewhere), as do genotypes Fand H (positions 350 to 500) and gibbon and orangutan variants,which are only phylogenetically distinct from each other inthe pre-S1/S2 region.

Part of the difficulty with conceptualizing the recombinationprocess is that, in some cases, one or other of the ancestralHBV variants involved in a recombination event may be missingfrom the data set. In the simple case, the widely distributedrecombinant virus, recB, has readily identifiable parents asthe genotype B and C nonrecombinant forms (5, 26), both of whichstill cocirculate in Asia with recB (40). However, a potentiallyanalogous recombination process may have occurred between genotypesD and E, although in this case, the putative ancestor genotypeE sequence with a phylogenetically distinct core gene sequencehas become rare or extinct and therefore not identifiable asa recombination parent. That both genotype E and recombinantB/C sequences are each monophyletic in the core region (althoughcontained within the genotype D and C clades) provides furtherevidence for their clonal origin from discrete recombinationevents in the past. It is similarly likely that genotype G isa recombinant virus between an ancestral virus with S gene sequencescomparable in divergence to those between genotypes A to E anda much more divergent HBV variant that contributed the restof the genome (18, 36). An abrupt change in phylogenetic relationshipsin indeed observed in the TreeOrder Scan, where genotype G sequences(pale yellow) migrated to an outlier position between positions950 and 2750 (Fig. 1). This observation is consistent with previouslyobserved differences in pairwise distances or phylogenetic treesof different genes of genotype G and other HBV variants (18,36). An analogous process of core gene replacement with a muchmore divergent HBV variant has been shown to have occurred inthe chimpanzee HBV variant, AF498266 (Table 2) (25, 45).

Intragenotype recombination. Using novel analysis methods based on recording frequenciesof phylogeny violations between members of the same genotype,we have been able to infer frequent recombination within genotypesA, D, and F/H as well as among gibbon variants recovered froma variety of species in Southeast Asia (Table 2; Fig. 3). Withoutany knowledge of what the recombination partners might be withinthe same genotype, it was nevertheless possible to show thattheir tree order changed more frequently in native sequencescompared to controls that were matched for their sequence diversitybut were nonrecombinant in phylogeny. The frequencies of recombinationevents varied by genotype, with frequent occurrences in genotypesA, D,F/H, and gibbon sequences but few if any in genotypes B,recB, and C (Table 2, column 9). These observations providefurther evidence for large-scale recombination in the evolutionaryhistory of several HBV genotypes.

Interestingly, the recB (B/C) recombinants (which were monophyleticin the core region; see above) showed no evidence for intragenotyperecombination, in common with the parental genotype C sequences(Table 2). The period over which their current sequence diversityarose (mean pairwise distances, 3% and 4%) indicates that thetimescale for intragenotype recombination can be extremely slow.For this calculation, the rate of sequence change of HBV of1.5 x 10^–5 to 5 x 10^–5 substitutions per site peryear observed in hepatitis B e antigen carriers (14, 32) ismost likely to approximate to the rate found among carriersinfected through perinatal transmission in Asia, where genotypesB, C, and recB predominate (2, 13). The current sequence diversitywithin recB (3%) therefore corresponds to a period of 300 to800 years of separate evolution without fixation of detectablerecombinant forms of this genotype. Mother-to-child transmissionand the establishment of a persistent carrier state resistantto adult reinfection may indeed be an epidemiological factorthat reduces opportunities for recombination among these Asiangenotypes (B, recB, and C). The higher frequencies observedin genotypes A and D may, in contrast, reflect greater epidemiologicalopportunities for recombination associated with horizontal transmissionpatterns later in life found in sub-Saharan Africa (41, 50)and in Western countries where these genotypes predominate.Recombination may also be favored over longer timescales overwhich some genotypes may have circulated, such as the more divergentgibbon HBV sequences (Table 2).

Recombination and the origin of HBV genotypes. The time scales of HBV recombination implied by this analysisare clearly compatible with the existence of geographicallywidespread individual recombinants (such as the B/C variantin Asia and genotype E in central Africa). Over this extendedtimescale, individual recombinant forms of HBV may spread andreplace their ancestors and therefore obscure the sequence ofrecombination events that led to their origin (such as hypothesizedfor genotype E). The evidence in this study for frequent intragenotyperecombination, combined with past multiple recombination eventswith often divergent HBV variants, as shown by the TreeScananalysis, indicates that many of the currently classified genotypesof HBV are themselves mosaics. The extraordinary difficultyin developing a plausible timescale in the evolutionary historyof HBV (3, 9) and understanding the relationship between humanand nonhuman HBV variants may be, at least in part, accountedfor by the complexity of the recombination events in their evolution(3).

While changes in the circulating pool of HBV variants are clearlysubject to stochastic events and complex epidemiological factors(such as transmission route), their spread may be enhanced bythe evolution of changes in frequencies of persistence and eantigen expression (37, 39, 40) generated though recombinationthat enhance transmissibility. In the future, observed differencesin treatment responsiveness (1), combined with the growing evidencefor frequent coinfection with different genotypes (15), mayprovide a further driving force for recombination to generateresistant viruses. Recombination potentially may therefore playa large, underestimated role in both the origins of currentlyknown genotypes of HBV found in human and nonhuman primate populationsand in the future evolution of their pathogenicity and transmissibility.

Finally, the evidence for recombination between human and apeHBV variants, such as the chimpanzee/genotype C recombinant(25, 45), and the evidence from group scanning for a gibbon/genotypeC recombinant among HBV variants in Vietnam (Fig. 2C and 3)have important implications for eradication and control measuresfor HBV. Their occurrence demonstrates that human and nonhuman-associatedHBV variants can indeed share hosts in nature. The existenceof widespread, endemic infection in nonhuman primates and theirpotential for cross-species transmission into humans will clearlyhamper efforts toward the ultimate goal of eradication of HBVinfection in geographical regions where humans and apes sharehabitats (35).

ACKNOWLEDGMENTS

The authors are grateful to Dan Haydon, Department of Zoology,University of Glasgow, for generating the control data set usedin the intragenotype recombination analysis (Table 2) and foruseful discussion of the results.

FOOTNOTES

* Corresponding author. Mailing address: Centre for Infectious Diseases, University of Edinburgh, Summerhall, Edinburgh EH9 1QH, United Kingdom. Phone: 44 131 650 7927. Fax: 44 131 650 6511. E-mail: Peter.Simmonds{at}ed.ac.uk.

Present address: Reference Laboratory, Department of Virology,Statens Serum Institut; Artillerivej 5, DK-2300, Copenhagen,Denmark.

REFERENCES

Top