Authors: Derek Tshiabuila, James E. San, Eduan Wilkinson, Graeme Dor, Houriiyah Tegally, Tongai G. Maponga, Marion Delphin, Philippa C. Matthews, Darren P. Martin, Cheryl Baxter, Tulio de Oliveira
Categories: Research, Hepatitis B virus (HBV), Genotypic diversity, Viral recombination, Recombination hotspots, Phylogenetics, Viral evolution
Source: Virology Journal
Authors: Derek Tshiabuila, James E. San, Eduan Wilkinson, Graeme Dor, Houriiyah Tegally, Tongai G. Maponga, Marion Delphin, Philippa C. Matthews, Darren P. Martin, Cheryl Baxter, Tulio de Oliveira
Hepatitis B virus (HBV) infection is a major public health concern, as chronic HBV infection can lead to liver cirrhosis and increase a person’s risk of developing hepatocellular carcinoma (HCC). HBV has been classified into ten genotypes (A to J). Here, we analysed the genotypic diversity and recombination patterns of HBV using 14486 publicly available HBV genome sequences. Partial sequences and sequences with no metadata were filtered out, resulting in a final dataset of 8823 HBV genomes. These sequences were then combined with 41 HBV reference genomes from NCBI GenBank, and a maximum-likelihood phylogenetic tree was constructed to generate ten HBV genotype datasets. Multiple sequence alignment was performed for each HBV dataset, and using RDP5.64, we identified 288 unique recombination events. Inter-genotype B/C recombination events were most common (found in 626/1194 identified recombinants), including 22/66 recombination events detected in viruses that are phylogenetically genotype B and 22/78 that are phylogenetically genotype C. The HBx (X) and pre-Core (pre-C) regions of the HBV genome were identified as recombination breakpoint hotspots, with the pre-C region also being the most frequently transferred genome region during recombination. As with many other viruses, the observed recombination breakpoint patterns in HBV genomes are significantly attributable to factors such as local sequence similarity, GC content, or selection against recombination-induced protein misfolding. This study highlights the complexity of the genetic diversity and recombination of HBV, with important implications for understanding its evolution and informing tailored public health interventions.
The online version contains supplementary material available at 10.1186/s12985-025-02829-0.
Chronic hepatitis B virus (HBV) infection (CHB) substantially increases a person’s risk of developing liver cirrhosis and/or hepatocellular carcinoma (HCC) [1]. HBV is transmitted by exposure to infected blood or bodily fluids [2, 3], and in 2022, CHB resulted in approximately 1.1 million deaths [4], the majority of which occurred in Southeast Asia and Africa, where HBV infections are most prevalent.
HBV is an enveloped DNA virus belonging to the Hepadnaviridae family. It has a partially double-stranded circular 3.2 kilobase (kb) genome comprising four overlapping P encoding a polymerase that mediates both reverse transcription (RT) and DNA-dependent DNA polymerization; pre-S1/pre-S2/S encoding the small (S), medium (M), and large (L) surface antigens (HBsAg); pre-C/C encoding the e antigen (HBeAg) and core protein (HBcAg); and X (HBx) encoding a transcriptional transactivator protein required for replication, gene expression, and immune evasion, respectively [5].
HBV genomes have been categorized into ten genotypes (designated A-J) [6] based on a genome-wide pairwise nucleotide sequence divergence from all other HBV genotypes of > 7.5%. In addition, over 40 HBV sub-genotypes have been identified based on 4–7.5% sequence divergence [7–9].
It is estimated that 96% of CHB cases are caused by five of the ten genotypes, with HBV-C being the most prevalent (26%), followed by HBV-D (22%), -E (18%), -A (17%), and -B (14%) [10]. In contrast, HBV-F, -G, -H, -I, and -J collectively account for < 2% of global chronic infections.
The clinical outcomes of CHB are variable and are associated with diverse factors, including the HBV genotype, specific viral polymorphisms, viral load, and quantitative HBsAg levels [11]. Compared with HBV-C, which is endemic in the same population, HBV-B tends to have a lower tendency for chronicity, earlier HBeAg seroconversion, higher HBsAg seroclearance, and better clinical outcomes (e.g., lower risk of liver cirrhosis and HCC) [12, 13]. Likewise, HBV-A shows earlier HBeAg seroconversion and better outcomes compared to HBV-D [14]. HBV-C and -D are associated with higher histological activity and worse clinical outcomes than HBV-A and -B [14]. Regarding response to treatment, HBV-A and -B have a greater response to interferon-alpha (INF-\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \alpha
The genetic diversity of HBV is partly explained by the absence of RT enzyme proofreading during viral replication, similar to that of HIV [16]. HBV has an evolution rate of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \approx $$\end{document}≈10-5 substitutions per genome site per year [17], which is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \approx $$\end{document}≈100–1000 times faster than that of many other small dsDNA viruses (such as *papillomaviruses* or *polyomaviruses*) [18]) and is only \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \approx $$\end{document}≈10–50 times slower than that of retroviruses such as HIV [19] or viruses with single-stranded RNA genomes such as hepatitis C virus [20]. In addition to having a relatively high mutation rate for a DNA virus, HBV also appears to have a very high recombination rate. Recombination in HBV was first detected in the 1980 s, and since then, over 30 inter-genotype recombinant strains have been described [3, 21, 22] and are denoted with reference to their parental genotypes, e.g., as A/D or B/C recombinants. There has also been evidence of recombination between human HBV genotypes and ape HBV genotypes, suggesting the possibility of cross-species transmission of some HBV variants [3, 23, 24]. In DNA viruses such as HBV, homologous recombination can occur through double-stranded break (DSB) repair pathways [25] or via a copy-choice mechanism during the formation of replicative intermediates, such as pregenomic RNA (pgRNA) or covalently closed circular DNA (cccDNA) [19, 20]. HBV replication involves the conversion of relaxed circular DNA (rcDNA) into cccDNA by host enzymes, followed by transcription of genomic and subgenomic RNAs by cellular RNA polymerase II, with pgRNA selectively packaged into progeny capsids and reverse transcribed into new rcDNA forms [25]. The DNA damage response (DDR) pathway plays a role in maintaining HBV genomic integrity by identifying and repairing lesions in cccDNA, which may arise from replication stress or host-induced DNA damage [26, 27]. During DSB repair, single-stranded DNA overhangs can invade an unbroken molecule, pairing with complementary sequences and enabling host DNA polymerase to extend the invading strand via the homologous template. Importantly, when co-infection involves genetically distinct HBV strains, this repair process can facilitate recombination between genomes, contributing to viral genetic diversity [26, 27]. As with other recombining viruses, factors such as HBV genetic diversity, viral loads within infected individuals, the geographical ranges of genetically distinct viral genotypes, periodic fluctuations in the prevalence of different genotypes, and rates of coinfection with genetically distinct viral genotypes are all expected to influence the likelihood of both recombination occurring between different HBV lineages and the detectability of recombination events that have occurred [28]. Here, we analyse patterns of recombination that are detectable in publicly available and newly determined human HBV whole-genome sequences and attempt to infer the factors underlying the observed recombination breakpoint patterns. ## Methods ### Data collection Publicly available consensus HBV whole-genome sequence data were accessed through the National Center for Biotechnology Information (NCBI) GenBank ([https://www.ncbi.nlm.nih.gov/genbank/](https://www.ncbi.nlm.nih.gov/genbank/)) in February 2025. The sequences were filtered to include samples from human hosts, genomic DNA/RNA, nucleotide sequence types, and sequences between 3000 and 3300 bp in length. Only sequences with available metadata (country, collection date) were used for further analysis. ### Phylogenetic analysis The HBV dataset was combined with 41 HBV reference genomes (Table 5 in the Appendix) [29] and aligned using Muscle v5.3 [30] to both root the alignment and enable differentiation between intra-genotype and inter-genotype recombination events. Multiple sequence alignments (MSAs) generated by Muscle were edited via visual inspection using AliView [31] and Geneious 2024.0.7 ([http://www.geneious.com/](http://www.geneious.com/)) to remove sequences with large gaps and multiple ambiguous bases. A maximum likelihood (ML) phylogenetic tree was constructed using IQ-TREE 2 [32], and 10 HBV genotype clusters (HBV A to J) were obtained. The HBV genotype clusters were converted to HBV datasets and combined with the 41 HBV reference genomes to result in 10 HBV genotype datasets. The nucleotide diversity and GC content of the HBV datasets were calculated via R with appropriate bioinformatics packages. The GC content per position was determined using a custom R function that iterates through each nucleotide position and calculates the proportions of guanine (G) and cytosine (C) nucleotides. Nucleotide diversity was calculated using the pegas package [27]. ### Recombination Recombination detection and analysis were performed using RDP5.64. The HBV dataset underwent a full exploratory automated scan with the RDP [33], GENECONV [34], and MaxChi [35] methods as “primary scanning methods” to detect recombination signals and the Bootscan [36], Chimaera [37], SiScan [38], and 3Seq [39] methods to verify the recombination signals (secondary scanning methods). Characterization of each unique recombination signal involved characterization (i) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 5' $$\end{document}5′ and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 3' $$\end{document}3′ paired breakpoint locations and their associated probability distributions, (ii) a list of one or more sequences carrying evidence of the recombination event, and (iii) a list of analysed sequences that are closely related enough to the parents of the recombinant that they could be used as proxies to detect the recombination events. Visualization of the overall recombination patterns observed in the datasets was performed using recombination region count matrices and recombination breakpoint distribution plots produced using RDP5.64. For each genotype, a recombination breakpoint distribution plot was constructed for each detected recombination event using RDP5.64. A 200-nt sliding window was used along the entire genome to sum the probabilities of all identified breakpoints within the window. This information was then plotted against the nucleotide coordinates. A permutation test implemented in RDP5.64 was then used to identify recombination breakpoint clustering patterns that significantly varied from breakpoint distributions expected under random recombination [27]. The permutation-based breakpoint clustering tests used for detecting recombination breakpoint hot- and cold-spots were also used to determine the associations between breakpoint sites and (i) GC content and (ii) sequence similarity between inferred parental genomes [40]. Here, the average GC proportion or pairwise sequence similarity was calculated across the HBV genomes using a sliding window (either 10 or 20 nucleotides long), and the mean GC content and pairwise similarity at observed breakpoint sites were compared with those at randomized breakpoint sites [40]. SCHEMA protein folding disruption tests [41] were performed using RDP5.64 as described previously [42, 43]. By comparing in silico reconstructions of real recombination events (M-recombinants) and simulated recombinant events (S-recombinants), we evaluated whether the predicted fold disruptions of proteins expressed by M-recombinants were significantly lower than those expressed by S-recombinants. Reference protein sequences for HBV-B, -C, and -D were downloaded from GenBank, and their expected structures were predicted using AlphaFold3 ([https://alphafoldserver.com/](https://alphafoldserver.com/)). Specifically, AlphaFold3 was used to infer the structures of the S, pre-S1, pre-S2, C, pre-C, X, and P genome regions (Fig. 6). The crystallographic information files (.cif) obtained for these structures were converted to protein data bank (.pdb) atomic coordinate format files using a Python script utilizing the MMCIFParser and PDBIO modules from Biopython ([https://github.com/CERI-KRISP/HBV-Recombination-Patterns/tree/main/python_scripts](https://github.com/CERI-KRISP/HBV-Recombination-Patterns/tree/main/python_scripts)). The.pdb files were then uploaded together with the appropriate RDP5.64 project file for a given HBV genotype-focused recombination analysis, and SCHEMA (protein folding disruption tests) was run for each protein for each of the HBV-B, -C and -D datasets. ## Results ### HBV-C and -B recombinants are the most common recombinant genotypes A total of 14486 HBV genomes were obtained from NCBI GenBank in February 2025. Of these, 9685 genomes had available sampling-location metadata. Muscle v5.3 was used to construct an MSA that was edited via visual inspection using AliView and Geneious to remove sequences with large gaps and multiple ambiguous bases, resulting in a final dataset of 8823 HBV genomes. A maximum-likelihood phylogenetic tree was constructed to categorize the different HBV genotypes (Fig. 1). HBV-C (*n* = 3981), -B (*n* = 2049), and -D (*n* = 1304) contained the highest numbers of sampled sequences. Recombination analyses of the various genotype-focused datasets revealed that the highest numbers of recombinants were found in the HBV-C (*n* = 671), -B (*n* = 421), and -D (*n* = 55) datasets (Table 1), thus implying that comparative recombination event distribution and breakpoint distribution analyses between different HBV genotypes would be most productive when focused on the HBV-B, -C, and -D whole-genome datasets. Notably, however, the large number of recombinants detected in these 3 genotypes does not imply that they have greater recombination than other genotypes these numbers merely reflect the fact that these were the best-sampled genotypes.Fig. 1Subtype distribution and categorization of HBV genomes. Maximum likelihood phylogenetic tree for HBV genomes coloured by genotype classification inferred using IQ-TREE 2 under the GTR+F+I+R10 model with 1000 bootstrap replicates. The 8823 completely sequenced genomes are divided into ten HBV genotypes (A-J). The different colours indicate the different HBV (A = Gold, B = purple, C = green, D = blue, E = maroon, F = dark tan, G = orange, H = pink, I = yellow, J = light green, recombinant genomes = red) Table 1Genotype-specific characteristics and detected numbers of recombination events in Hepatitis B Virus whole genome datasetsTypeNo. of sequencesNucleotide diversity (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \pi $$\end{document}π)GC content (%)Recomb. eventsNo. of recombinantsHBV-A6910.007042.31127HBV-B20490.005050.0246421HBV-C39810.01055.6198671HBV-D13040.006447.955055HBV-E3920.0112.161711HBV-F2590.01761.90168HBV-G260.03120.182213HBV-H180.019100134HBV-I1020.02334.59144HBV-J1NA48.5000 Based on this sequence dataset, the genotype distribution reflects what has previously been well described [14]. Recombinants were found across all continents (Fig. 2A and B). Fig. 2Global distribution of HBV genotypes and recombinants. **A** Bars represent the frequency of whole genomes sampled within each region assigned to each genotype, and numbers at the end of each bar indicate the absolute number of sequences in each dataset. **B** The global distribution of HBV genotypes (A-J) and recombinants, with each genotype represented by a different colour. Each pie chart indicates the prevalence of multiple genotypes within any given country, and the greyscale indicates the number of sequence counts per country. The colour code used in this figure associated with each genotype is displayed in panel A Geographical differences were also evident in the distributions of different sub-genotypes within the HBV-B, -C, and -D datasets (Fig. 3) (the other genotypes had too few sampled genomes to make a meaningful comparison) and in the distributions of classified recombinant genomes (Fig. 4). Sub-genotypes B2 (n = 1347), B3 (n = 178), C2 (n = 2338), C1 (n = 329), D1 (n = 737), and D3 (n = 255) had the highest number of genomes for the HBV-B, -C, and -D datasets. The other sub-genotypes included in the HBV-B dataset were B1, B4, and B6; in the HBV-C dataset; C4, C5, C8, C10, and C11; and D2, D4, D5, and D7 in the HBV-D dataset. Notably, however, sub-genotype B2 is a recombination of HBV-B and -C (B/C).Fig. 3Global distribution of HBV sub-genotypes and recombinants for major HBV genotypes (HBV-B, -C, and -D). Panel **A** shows the combined genotype distribution; panels **B**, **C**, and **D** present the major sub-genotypes mapped by sequence prevalence; and panel **E** shows the prevalence of recombinant forms for each genotype A total of 1194 recombinant genomes were identified in this study, and these were distributed across the different HBV genotypes. The most common recombinant forms were B/C (n = 623), followed by C/D (n = 45), B/C/D (n = 29), B/I (n = 12), and A/D (n = 15) (Fig. 4). Recombinant genotypes with fewer than 10 genomes were classified as “other” (Table 6 in the Appendix).Fig. 4Global distribution of major recombinant genotypes. Recombinants with fewer than 10 genomes were classified as “other”. Different colours represent the different recombinants, and the size of the circle represents the recombinant proportion within the region ### The X and pre-C genome regions are the sites of conserved recombination breakpoint hotspots We studied intra-genotype recombination, which is likely to be a major contributor to HBV diversification but has been far less studied than inter-genotype recombination [3, 8, 10, 11, 20, 28, 44–60]. This analysis revealed a total of 288 unique intra- and inter-genotype recombination events distributed across the 1194 recombinant HBV genomes. The most prevalent inter-genotype recombinants detected across the ten different genotype-focused datasets were B/C (50 unique recombination events) and C/D (33 unique recombination events), which are likely to be the most common since these genotypes are the most represented. Notably, each of the genotype-focused datasets contained a set of reference sequences from the other genotypes to enable the detection of and differentiation between both intra- and inter-genotypic recombinants. We were therefore able to determine that, among the sampled sequences, B/C recombinants were the most common of the inter-genotypic recombinants detected. These events accounted for 28/66 unique recombination events detected in the HBV-B dataset and 22/78 of the unique recombination events detected in the HBV-C dataset. Owing to the low numbers of recombinants detected in the HBV-A, E, F, G, H, I, and J datasets (Table 1), only HBV-B, C, and D were used for further comparative analyses of recombination patterns between the genotypes. Breakpoint distribution plots (Fig. 5A), which are useful for determining the positions of recombination breakpoint hot- and cold-spots, were generated for HBV-B, -C, and -D. Across all three genotype-specific datasets, statistically significant (local permutation *P* < 0.01) clusters of recombination breakpoint sites were identified in the X and pre-C genome regions (highlighted in red in Fig. 5A). Statistically significant clusters of recombination breakpoint sites were also identified at the beginning of the pre-S2 and pol genome regions for HBV-C and beginning of the pol region for HBV-D. Significant recombination cold spots were identified within the pol region for the three genotypes (highlighted in blue in Fig. 5A). Recombination region count matrices (Fig. 5B) were constructed to quantitatively represent the relative frequency with which different genome regions were transferred during the inferred recombination events. These results indicated that, for all three genotypes, the most commonly transferred genome region during detected recombination events was the pre-C gene (note that the red diagonal extends leftwards from the pre-C genome regions in Fig. 5B). The genome regions that were most consistently transferred during recombination as complete fragments (i.e., they were most rarely broken by internal recombination breakpoints) were those encoding small, medium, and large HBsAgs and the first half of the polymerase gene (indicated by blue triangles spanning these genes along the bottom axis of the recombination region count matrices). This was expected, as these genome regions are overlapping.Fig. 5Analysis of recombination breakpoint sites across HBV-B, -C, and -D. **A** Recombination breakpoint distributions across the HBV-B, -C, and -D genomes. All detectable breakpoint positions are represented by vertical lines above each graph, with the gene map shown above each panel. The gene regions are coloured as pre-S2, S, and pre-S1 (shades of green), pol (black), and pre-C and C (shades of brown). The plot also shows confidence intervals for breakpoint clustering under random the dark green, light green and lower and upper dashed lines represent the local 95% confidence intervals, local 99% confidence intervals, and 95% and 99% confidence limits, respectively. The black line represents the number of breakpoints within a 200-genome site window surrounding (100 nt upstream and 100 nt downstream) each genome location. Areas where the black line emerges above the global 99% confidence limit are considered locally significant recombination hotspots and are marked in red. The purple bar on the x-axis represents the origin of pgRNA synthesis and the direct repeat 1 (DR1) region. **B** Recombination region count matrices highlighting areas of the genome that are most and least commonly separated from one another during detectable HBV recombination events. Unique recombination events for HBV-B, -C, and -D were mapped onto recombination region count matrices based on pairs of determined breakpoint positions bounding transferred genome fragments. Each cell in the matrix represents a pair of genome sites, with the colours (heat) of the cells indicating the number of times the represented pairs of sites were separated from one another during recombination events. Whereas blue triangles at the base of the matrix represent individual genome regions that tend to be most frequently inherited as unbroken blocks of sequences from a single parent during recombination events (the larger the triangle is, the larger the region), red rectangles in the interior of the matrix represent pairs of genome regions that tend to be most frequently inherited from different parents during recombination events. Gene maps are displayed alongside each matrix to indicate protein-coding regions, color-coded similarly to Panel A. The purple bar on the genome axis shows the origin of pgRNA synthesis and the DR1 region ### Recombination breakpoint patterns are weakly impacted by local sequence similarity and local GC content To explore factors influencing the predisposition of different genome sites to recombination, we examined the associations of detected recombination breakpoint positions with local GC content and pairwise sequence similarity (Fig. 6). Although there was no observed association for HBV-C, it should be noted that for HBV-B and -D, breakpoint sites tend to occur in regions with high GC content (*p*-value < 0.019) and high pairwise sequence similarity (*p*-value < 0.04) when considering regions between 10 nucleotide sites upstream and downstream of detected breakpoint sites (Tables 2 and 3). This was, however, only the case for HBV-D when considering genome regions between 20 nucleotide sites upstream and downstream of the detected breakpoint sites (*p*-value < 0.024).Fig. 6Regional variations in average pairwise sequence similarity (purple) and GC content (green) across HBV genomes. The plotted values indicate the pairwise sequence similarity and GC proportions within a moving 50-nucleotide window. The detected recombination hotspots are indicated by red vertical stripes while cold-spots are indicated by blue vertical stripes in the graphs Table 2Association of GC content with recombination breakpoint sitesWithin 10ntWithin 20ntHBV-TypeLower/Higher*P*-valueLower/Higher*P*-valueHBV-BHigher0.019NonensHBV-CNonensNonensHBV-DHigher0.001Higher0.024 Table 3Association of sequence similarity with recombination breakpoint sitesWithin 10ntWithin 20ntHBV-TypeLower/Higher*P*-valueLower/Higher*P*-valueHBV-BHigher0.04NonensHBV-CNonensNonensHBV-DHigher<0.001Higher<0.001 ### Recombination hotspots within the HBV genome are at least partially attributable to selection against recombination-induced protein misfolding The SCHEMA method (implemented in RDP5.64 [61]) was used to determine whether the observed distribution of recombination breakpoints was at least partially attributable to natural selection disfavouring recombinants that produce chimeric proteins with disrupted structural folds. Given the paucity of available protein structure data for HBV, we relied on protein structures inferred using the Alphafold3 server ([https://alphafoldserver.com/](https://alphafoldserver.com/)) (Fig. 7). The chimeric polymerase proteins expressed by the detected HBV recombinants display degrees of predicted structural disruption that are not significantly different than would be expected under random recombination in the absence of selection (all multiple testing-corrected SCHEMA test *p*-values < 0.05 [62]; Table 4). This finding indicates that the locations of the detected recombination breakpoint sites impacting this protein are unlikely to have been strongly influenced by negative selection. In particular, they do not appear to reflect the removal of recombinants that would have been most likely to produce misfolded proteins. Conversely, for each of the other tested proteins, we detected evidence in at least one genotype of negative selection against the expression of misfolded proteins having impacted the observed recombination breakpoint distributions (Bonferroni corrected *p*-value < 0.05). Table 4SCHEMA protein misfolding test for evidence of negative selection against misfolded chimeric proteins impacting observed recombination breakpoint distributions within HBV protein-coding regions (tests where too few breakpoints were detected in a coding region to yield a meaningful result, are indicated as “NULL”). BPs = Breakpoint numberHBV-BHBV-CHBV-DGeneProteinBPs*P*-valueBPs*P*-valueBPs*P*-valuepre-S1Small HBsAg130.02130.01660.072pre-S2Medium HBsAg21<0.001180.001970.015SLarge HBsAg240.01420.0085140.098PolPolymerase70.68560.41280.21HBxHBx70.4212<0.00190.021pre-CHBeAg4NULL50.4060.017CHBcAg2NULL4NULL80.0036 ## Discussion We set out to better describe and understand the epidemiological, genetic, and evolutionary patterns of HBV recombination. Our analysis of all publicly available whole-genome HBV sequences validates existing knowledge of the distinct distributions of recombinant HBV genotypes across different continents and explores the characteristics of recombination patterns found in HBV genotypes that continue to make a substantial contribution to global HBV diversity [3, 11, 63]. Although we do not know how many of the HBV recombinants that arise during mixed infections of genetically distinct HBV lineages will be transmitted, we do know that recombination has already had a significant influence on phylogeny/diversity. An example is sub-genotype B2, which is a B/C recombinant that is circulating in Southeast Asia and is the most common inter-genotypic recombinant found in our dataset (which almost certainly has sampling biases in favour of the developed world). A previous study revealed that more than 60% of recombination breakpoint patterns within HBV genomes fall between nucleotides 1640 and 1900 (which encompasses the end of X and complete pre-C regions). Many breakpoints also occur within the S gene (nts 350–500, 3150-100, or 650–830) [3]. Recombination breakpoints have previously been reported between nt positions 1800-2000 almost five times more frequently than at the remaining sites in the HBV genome, which is consistent with our findings that the recombination hotspot is located in the X and pre-C regions (nt 1814-1901). A possible mechanistic predisposition of the pre-C region to recombination may be related to the crucial role that this region plays in HBV replication. Specifically, it contains the start site of pgRNA synthesis and the DR1 region, which is also a known hotspot for recombination with the host genome [64]. As described previously [57], this site could also plausibly be a hotspot for recombination between different cccDNA HBV forms when, instead of integrating into the host genome, pairs of HBV cccDNA molecules recombine with one another to form genome dimers. Multimeric HBV cccDNA forms (which such a mechanism might generate) are known to occur commonly [25] and, if used as templates for pgRNA synthesis, would be expected to yield pgRNA molecules with one breakpoint position at the site where the recombination event occurred and one breakpoint at the origin of pgRNA synthesis. Specifically, one of these breakpoint sites would be expected to always occur at the origin of pgRNA synthesis, whereas the other would simply display a tendency to occur near the DR1 region. Collectively, these findings might yield an elevated rate of detectable recombination within the region we identified as a conserved pre-C/X recombination breakpoint hotspot across the different HBV genotypes. Another non-exclusive explanation for this hotspot could be related to the relatively selective advantages of recombination events that transfer the pre-C/X regions between HBV genomes, as this region also encompasses the HBV encapsidation signal and is therefore expected to evolve under strong selection pressure. Apart from the P region, we detected consistent evidence that recombination patterns have been impacted by selection acting against recombinants that express misfolded chimeric proteins. This finding suggests that recombination breakpoints contributing to the pre-C/X recombination hotspot occur at positions in these genes that minimally affect protein structure and function. This observation is consistent with the hypothesis that recombination events in this genome region may have been favoured by natural selection simply because they have avoided disrupting protein structure. Recombination breakpoint patterns in HBV are also likely to be impacted by replication biology, explaining the impact of local GC content and degrees of sequence similarity between parental genomes. Across various virus families, the GC content has previously been shown to increase the frequency at which recombination breakpoints occur [65–68]. Accordingly, our analysis revealed that the breakpoint positions detected in HBV-B and HBV-D display a significant tendency to occur at genome sites with higher-than-average GC contents. It is expected that homologous recombination between parental HBV genomes likely requires a high degree of sequence similarity at sites where breakpoints occur, as observed in HIV [69–71]. Accordingly, we noted that for HBV-B and D, the degrees of sequence similarity were greater than the genome average at the detected breakpoint sites. It is noteworthy, however, that the pre-C/X recombination hotspot encompasses the genome region with both a lower-than-average GC content and a lower-than-average degree of parental sequence similarity. It is therefore very unlikely that either of these factors have mechanistically predisposed this region to increased recombination frequencies. It is most plausible, therefore, that the pre-C/X recombination hotspot is a consequence of both a replication-associated mechanistic predisposition to recombination, such as that described above, coupled with a tendency for the recombination breakpoints that occur in this region to have a minimal impact on the folding of encoded proteins. It also remains plausible that many of the detected recombination events in the pre-C/X region and in other genome regions have been impacted by selective processes that are more difficult to test for directly. Specifically, some recombination events might be favoured by natural selection if they transfer persistence and/or transmissibility-enhancing mutations between genomes. For example, we observed that pre-S/S-encoding genome segments tend to be transferred as an entire unit during recombination and that some pre-S/S mutations are related to vaccine failure, immune escape, and occult HBV infection. Similarly, mutations within the part of Pol that also tends to be recombinationally transferred as a complete unit are associated with resistance to nucleos(t)ide analogue-based antivirals; however, while drug resistance and HBsAg vaccine escape are theoretically subject to the influence of recombination, based on this dataset, recombination is unlikely to have a significant influence at the public health/population level because Pol is a recombination cold spot. Furthermore, it is also difficult to test for direct evidence of selective processes that impact virus pathogenicity. For example, there are a multitude of mutations within the C and pre-C regions that, by disrupting viral replication, result in pathogenicity-enhancing defective replicative intermediates [70]. These defective replicative intermediates are not encapsidated. This increases the chance of HBV DNA integrating into chromosomal DNA and leading to advanced liver disease, cirrhosis, and HCC [70]. Other pre-C mutations more directly result in liver function impairment. Our study was limited by the availability of sequence metadata and differences in HBV genome numbers between different genotypes. The analysis of global HBV genotype, recombinant, and sub-genotype distributions using available whole-genome sequence data is also likely strongly impacted by geographical and temporal sampling biases; specifically, it is expected that the dataset analysed is most representative of the Northern Hemisphere and regions in which resources and infrastructure have supported the generation of sequence data. Furthermore, the recombination analyses of the larger genotype-focused datasets (i.e., for HBV-B, -C, and -D) were likely underpowered because the multiple testing correction applied to the fully exploratory recombination analyses scaled cubically with the number of sequences analysed, reducing the statistical sensitivity in the context of a large dataset. Determining why, when, and where recombination events occurred between viruses belonging to different HBV lineages is important, as it reveals the discrete outcomes of an evolutionary process that is contingent on the circulation of different strains in the same geographical areas, the co-infection rates of individuals with genetically distinct HBV lineages, the degrees of genetic compatibility of co-infecting lineages, and the viral loads within coinfected people [26]. ## Conclusion The analyses performed in this study revealed conserved recombination patterns between divergent HBV lineages (Table 7 in the Appendix), with the most strikingly similar recombination patterns being those observed in HBV-B and D. The concordance of recombination patterns across HBV genotypes implies that both the mechanistic processes that determine where recombination breakpoints occur and the selective processes that determine which recombinants survive are likely broadly conserved across all HBV genotypes. Recombination is highly relevant to HBV diversity, may impact persistence and pathogenesis, and can aid in the development of insights into replication biology; however, although it has been understudied to date, gaps in sampling for certain regions mean that we still have a limited view of the full extent and implications of HBV recombination across global populations, highlighting the need for broader and more geographically diverse sampling methods in future studies. ## Supplementary Information Supplementary Material 1.