Genome evolution of Kaposi sarcoma-associated herpesvirus (KSHV)

Abstract

Kaposi sarcoma (KS) is the most common cancer in people living with HIV (PLWH), particularly in sub-Saharan Africa (SSA), where Kaposi sarcoma herpesvirus (KSHV or human herpesvirus 8 [HHV-8]) is endemic. In KSHV endemic areas, the overall survival of KS patients has changed little over the past 20 years. A phylogenetic analysis of available full-length viral genomes (n = 164) identified two different virus lineages that co-circulate in KSHV endemic regions today. Their sequences differ from the GenBank reference sequence and those of common laboratory strains, which originated in the 1990s in the US and Europe. Targeted short-read sequencing accuracy was validated by PacBio-based long-read sequencing to resolve repeats. This analysis identified over 1,000 single nucleotide variants (SNV) in a new 14-member sequence collection from tumor biopsies and blood in Malawi with 127 ± 32 (median ± SD) SNV per genome. Most were private, i.e., specific to one individual's virus. Within each of the two lineages, KSHV continues to evolve over time and across national borders by genetic drift and recombination. Analyses of shared SNVs by AlphaFold2 predicted some changes in the conformation of key viral proteins. These findings may help our understanding of herpesvirus evolution.

Importance: To understand viruses, the field needs to know their genetic makeup. To develop mechanistic models, targeted therapies, and vaccines, we need comprehensive and up-to-date sequence information on the viral strains that circulate where the diseases appear today. Our knowledge of Kaposi sarcoma herpesvirus (KSHV) sequence distribution and evolution is behind that of other human herpesviruses and RNA viruses. Here, we add to community knowledge using new technologies and artificial intelligence.

Keywords: AlphaFold; KSHV; Kaposi sarcoma herpesvirus; Kaposi sarcoma-associated herpesvirus; Kaposi's sarcoma; PacBio; evolution; human herpesvirus 8; human herpesviruses.

Fig 1

Overview of KSHV sequences. (A) Comparison of BC1 isolates U75698 obtained by Sanger-sequencing (Sanger) of a cosmid library and MK7333609 obtained by genome-enriching NGS (NGS). MK7333609 is missing the first 20 nucleotides present in U75698. The ORFs are shown in blue and repeat sequences in red. Individual changes are indicated as white lines in the identity bar and annotated above. Regions of 100% identity are in blue. (B) Comparison of non-identical bacmid BAC16 whole genome sequences (GQ99435, KY246443, KY246444, MK208823) and the rK219 isolate (KF588566). Entry OK358814 was 100% identical to GQ994935. Individual changes are indicated in the identity bar and annotated above. (C) Comparison of non-identical bacmid BAC36 and BCBL-1 whole genome sequences HQ404500, JX228174, MN205539, MT936340, MZ712172, OR117738, OR333977. (D) Comparison of early PEL, KS, and bacmid-contained KSHV genomes: JSC-1^Bac16 (GQ994935), JSC-1^rKSHV219.1 (KF588566), BC-1^Russo (U75698), BC-1^unc (MK733607), BC3 (MK87631), BCBL-1^PacBio (OR117738), BCBL-1^Bac36 (HQ404500), BCBL-1^Bac52 (OR333977), KS^GK18 (NC_009333), KS^Sau3a (U93872), HBL6 (OR573937). Individual changes are indicated in the identity bar. Also shown is the LUR (from ORF16 to ORF58) between the two Ori-Lyt sites comprising 65,840 bp of continuous sequence. (E) Identity graph for just the PEL isolates of (D).

Fig 2

Alignment summary of untargeted, PacBio-derived KSHV and EBV genomes for BC-1 cells. (A) _BC1KSHV^PacBio de novo assembled from human-depleted reads. The lines diagrammed below indicate de novo assembled contigs in either the forward (green) or reverse (red) direction. (B) _BC1KSHV^PacBio de novo assembled from all reads. The lines diagrammed below indicate de novo assembled contigs in either the forward (green) or reverse (red) direction. (C) _BC1KSHV^PacBio consensus based on long-read sequencing mapped to GenBank reference NC_009333, a K15 type P strain, vs BC1, which is the M strain. (D) _BC1EBV^PacBio de novo assembled from human-depleted reads. The ptg00027 designation refers to de novo assembled contigs, which are diagrammed as lines below. (E) De novo assembled from all reads, with the black lines below indicating the longest contigs originating from the assembly. (F) Individual PacBio long-reads mapped to GenBank reference NC_007605. Shown in yellow and green are selected ORF and transcripts. Shown in blue are the coverage graphs (solid blue indicates a single, the longest, contig aligned). Individual reads are green (forward) or red (reverse). Reads in yellow are non-unique mappings.

Fig 3

Phylogenetic analysis of LUR, K1, and K15. The results of a Bayesian phylogenetic analysis of the African KSHV genomes using BEAST are shown. (A) A tree based on the LUR extracted from 164 genomes, the ML tree identifies two subtypes of KSHV genomes isolated from Africa. The country of origin is indicated by color: Malawi in green, Uganda in red, Zambia in blue, Cameroon in yellow, and USA/Europe/Japan in black. The four Japanese isolates are indicated by an asterisk. (B) A tree based on the K1 region (right side). The additional Malawian genomes described in this study are classified as the A5 subtype. (C) A tree based on the K15 gene (left side). The additional Malawian genomes described in this study are classified as the P-subtype.

Fig 4

SplitTree analysis. This analysis used n = 164 KSHV LUR sequences to construct the ML phylogenetic tree. Darker branch shadings represent branches with higher confidence (1,000 bootstrap replicates). The Phi test for recombination found statistical evidence for recombination (P < 0.01). A total of 1,393 informative sites were found using a window size of 100. Mean = 0.3132622742; variance = 4.70900; observed = 0.093065.

Fig 5

Model of KSHV ORF49. (A) The ORF49 crystal structure (5IPX.pdb) in gray is aligned to the AlphaFold2 model for the reference sequence, as shown in the pLDDT color scoring. (B) The highly confident model of the ORF49 sequence from the Malawian isolates is colored in the pLDDT color scoring. (C) The sequence coverage plot indicates that over 40 sequences were used to generate the MSA and their sequence similarity to the query ORF49. The sequences were 80%–100% similar to ORF49, except at the N- and C-terminals. (D) PAE indicates that AlphaFold2 is highly confident that the predicted structure has two domains and spatial arrangement. The axes represent the residue index, i.e., position of the amino acid.

Fig 6

AlphaFold2 model of KSHV ORF22/gH. (A) The ORF22 crystal structure (7czf.pdb) in gray is aligned with the reference model, as shown in the pLDDT color scoring. The domains of the structure are indicated. (B) The sequence coverage plot indicates that 140 sequences were used to generate the MSA and their sequence similarity to the query ORF22. The sequences were 40%–60% similar to ORF22, except at the N- and C-terminals, where similarity was 20%. (C) PAE indicates that AF2 is highly confident the predicted structure is multidomain and in its spatial arrangement. The axes represent the residue index, i.e., position of the amino acid. (D) The highly confident reference model of ORF22 aligned with the model of MZ712180 of subtype A and was colored in the pLDDT color scoring. The structures are almost identical except for the N- and C-terminal misalignment and the additional alpha helix at the N-terminal. (E) The two non-synonymous SNVs identified in the D-I domain are V68A and V104A. (F) The two non-synonymous SNVs are found in the D-III and D-IV domains, N406 and L583I, respectively (green line residue represents the reference sequence, and pink stick residue represents the Malawian sequence).