Selection, escape and the limits of inference in Middle Eastern hepatitis B virus genotype D: an analysis of 824 curated public sequences
Virus Evolution · 2026-11-12
DOI: 10.1093/ve/veag071 ↗Abstract
Abstract Background Hepatitis B virus genotype D (HBV/D) predominates across the Middle East, yet questions asked of its public sequence archive are rarely preceded by a test of whether that archive can answer them. Methods HBV records of 500–4,000 nt from fifteen Middle Eastern countries were retrieved from GenBank, genotype assigned against a 98-genome A–H reference panel, and records filtered on collection date, surface-gene coverage, ambiguity and within-country identity. Curated sequences were projected onto a genotype D reference (LC519816.1), giving a reference-anchored matrix across the surface and reverse-transcriptase frames. A maximum-likelihood tree was inferred under GTR+Γ, temporal signal tested by root-to-tip regression, and ancestral country states reconstructed under a discrete migration model and re-estimated on five country-balanced subsamples. Results Of 4,398 records retrieved, 824 sequences from ten countries passed curation (Iraq, 94). Canonical resistance substitutions (rtL180M, rtM204V/I, rtV173L) and surface escape substitutions (sP120S/T, sS143L, sD144A/E/G, sG145R/A) were recovered, with rtL180M and rtM204V/I most frequent (each >2%) the compensatory and escape substitutions each below 2%, as in clinical series of treated genotype D infection. Twelve codons showed diversifying selection by two methods, only one inside a determinant and none at classical escape positions; antigenic-loop selection did not differ from the rest of the protein (Mann-Whitney p = 0.50; permutation p = 0.61). Four planned analyses failed: root-to-tip regression gave R² = 0.002, so the dataset cannot be dated; the apparent Iranian source signal (+200 net exports) fell to +1.2 under balanced subsampling, where Syria (+27.0) displaced Iran; median pairwise distance was 0.0128 substitutions/site, with 97.0% of sequences in one component at 0.015; and subgenotype assignment left only 7.2% of sequences resolvable at a 0.005 margin. Conclusions Substantively, the HBsAg antigenic loop shows no excess of diversifying selection despite immune surveillance, and monitored escape positions carry no site-level signature, consistent with escape arising as rare individually selected events rather than a selective regime. Methodologically, the archive supports mutational and selection epidemiology but not dating, source attribution, transmission clustering or subgenotype assignment from surface-gene data; the constraint is sequence length and sampling design, not sample size.
EUVIMED