Abstract
<title>Abstract</title> <p>Synonymous sequence evolution can act through both codon composition and codon order, but these components are rarely separated in viral genomes. We analysed 16 exact-unique natural-host complete genomes of cetacean morbillivirus (CeMV), comprising 96 canonical coding sequences (N, P, M, F, H and L). For each CDS, 9,999 conditional randomizations preserved every amino-acid position and the exact inventory of synonymous codons while changing only their order. All 96 CDS formed fewer codon-boundary CpG sites than expected, and all remained significant after false-discovery-rate correction; gene medians ranged from 0.535 to 0.637. UpA depletion was less uniform and was concentrated in N and P. The CpG-depleted direction persisted after excluding P, thereby removing the principal P/C overlap and V-editing context, and in all three represented identity clusters, although two clusters contained single genomes. P and H were the most amino-acid-divergent genes, but sliding-window protein entropy was unrelated to local CpG or UpA ordering. Codon-correspondence rankings obtained from bottlenose dolphin and harbour porpoise references were nearly identical and are interpreted as robustness checks rather than evidence of host adaptation. These results identify a conserved CpG-avoiding arrangement of synonymous codons superimposed on gene-specific protein evolution and UpA organization. The pattern is consistent with an RNA-level evolutionary constraint, while its molecular mechanism and phenotypic consequences require experimental testing.</p>