2.1 Characteristics of Study Population

Samples from 280 individuals from two separate studies conducted at Parkland Health's outpatient HIV clinic in Dallas, Texas were sequenced. The first study was a cross-sectional cohort investigating KSHV seroprevalence which included 206 individuals with no history of KAD and 14 participants with KS, as previously reported [17]. Participants were recruited between January 2020 and September 2021. Sera were collected for serological assays while whole blood and oral fluids were collected for molecular studies.

The second study enrolled 60 PWH with history of KS, from September 2022 until April 2023. Participants provided oral fluid samples for KSHV viral load (VL) quantitation and sequencing to determine viral subtypes. Both studies were approved by the University of Texas Southwestern Institutional Review Board (STU 2019-1204 and STU 2022-0355) and all participants provided informed consent at enrollment according to the Declaration of Helsinki. Figure 1.

2.2 DNA Extraction and KSHV qPCR Assays

Whole blood was collected from study participants utilizing Qiagen PAXgene vacutainer tubes (Qiagen, Hilden, Germany), while oral fluids were collected in mouthwash. Samples were processed and DNA extracted as previously reported [17]. KSHV VL was measured in the extracted DNA using qPCR assays targeting the human endogenous retrovirus 3 gene (ERV-3), used as a cell quantitation marker, and the KSHV K6 gene region [18, 19].

2.3 Library Preparation and NGS Sequencing

Genomic DNA was fragmented using a Covaris focused-ultrasonicator (Covaris, Woburn, MA, USA). Sequence libraries were generated using two different target enrichment kits: Agilent SureSelect XT (Agilent Santa Clara, CA) [10, 20, 21] and KAPA HyperCap (KAPA Biosystems Inc, Wilmington, MA). The genomic DNA input ranged from 200 ng to 3 µg for SureSelect and 100 ng for HyperCap. Both library preparation kits contain KSHV specific bait sets based on representative KSHV genomes of all K1 and K15 subtypes.

Samples were sequenced with Illumina MiSeq and NexSeq. 2000 instruments (Illumina, Hayward, CA) with either library preparation method generating 250 bp or 150 bp paired end reads. Some samples were sequenced with both methods; four, UTSW101, UTSW113, UTSW139, and UTSW141 to resolve poor KSHV coverage due to interference from high Epstein-Barr virus (EBV) load and one, UTSW107, to confirm KSHV multiple infections observed with SureSelect XT (Supporting information Table S1).

2.4 Viral Genome Assemblies

Near full-length KSHV genomes were generated combining reference-guided alignment against the KSHV reference genome NC_009333.1 (GK18) with a de novo assembly approach [9, 10]. The internal repetitive regions (NC_009333.1:g.24230-25045, 29927-30055, 118229-113914, 124784-126456, and 137169-137969), and the large terminal repeats between the KSHV K1 and K15 genes were masked in the final alignments as they cannot be computationally resolved. Samples with read depth coverage less than 30X or with high percentages of unresolved base calls were not assembled (Supporting information Table S1). All assembled genomes were manually curated to confirm sequence variants and to resolve areas of high variability including K1, vIRF-2, and K15 gene regions. Variable gene regions were manually examined for evidence of multiple infections observed as overlapping reads with distinctive polymorphisms. Additionally, de novo K1 and K15 gene-specific subassemblies were used to distinguish KSHV subtypes by mapping overlapping reads in these highly variable regions [10]. The curated sequence alignments were exported from Geneious (Geneious Prime 2022.0.2) as consensus FASTA files for submission to GenBank and further downstream analyzes [22].

2.5 Sanger Sequencing of the K1 Gene

For samples with KSHV VL estimates determined to be suboptimal for NGS, KSHV K1 subtype was identified by Sanger sequencing of nested-PCR products as previously reported. Briefly, the outer nested primers used were ATGTTCCTGTATGTTGTCTGC (outer forward) and AGTACCAATCCACTGGTTGCG (outer reverse) followed by inner nested primers GTCTGCAGTCTGGCGGTTTGC (inner forward) and CTGGTTGCGTATAGTCTTCCG (inner reverse). The PCR cycling conditions for both rounds of PCR were similar, consisting of 1 min 45 s at 95°C and 35 cycles of 1 min at 96°C, 45 s at 51°C (outer nest) and 58°C (inner nest), and 1 min at 72°C. The inner nested procedure used 5 µl of first round product and both rounds ended with a 5-min hold at 72°C. The final K1 sequence product size was 840 base pairs [23, 24]. All Sanger sequencing was performed using an Applied Biosystems 3130XL genetic sequencer (Thermo Fisher Scientific). The sequencing of each sample was performed multiple times and a minimum of 4 overlapping reads were used to assemble the K1 gene sequences [9].

2.6 Phylogenetic Analyzes

Three samples, UTSW107, UTSW595, and UTSW601 had unresolvable genomes indicating infections with multiple KSHV genomes and were consequently excluded from phylogenetic analysis. The remaining 22 new near full-length consensus sequences were aligned with 35 published KSHV genomes featuring all available K1 and K15 gene subtypes, using MAFFT v7.511 and default settings before importation into Geneious (Geneious Prime 2022.0.2) [22]. The internal repeat regions were masked in the alignment, and the remaining nucleotide bases were used for phylogenetic analysis. A SplitsTree v4.15.1 analysis was performed using the masked alignment [25]. The Neighbor-Net method with default settings of 1000 bootstrap replicates was used to construct a phylogenetic tree [26].

KSHV K1 gene subtypes were identified using a combination of BLAST similarity searches [27] followed with confirmation by phylogenetic tree analysis. The K1 gene sequences were translated to amino acid and an alignment was made in Geneious using the MAFFT module (v1.5.0) with default settings (Geneious Prime 2022.0.2) [22]. Individual K1 sequences in four samples with multiple KSHV infections were obtained using the K1 de novo subassembly for inclusion in the final alignment. The resulting amino acid alignments, including 63 publicly available KSHV subtype-specific references, were used to infer phylogenomic trees via the neighboring-joining method using IQTree version 2.2.0.5 [28] with default settings. The IQTree output file was visualized in FigTree version v1.4.4 (https://tree.bio.ed.ac.uk/software/figtree).

2.7 Analysis of Co-Infections by Multiple KSHV Genomes

Dot plots mapping the variant nucleotide positions across the KSHV genomes for the three multiple infection samples, UTSW595, UTSW601, and UTSW107 used previously published methods [10]. Visual examples of reads constituting mixed infections were illustrated for each genome using the K1, ORF25 and ORF47 gene regions respectively, generated by the reference guided assembly pipeline, and exported as screen shots from Geneious (Geneious Prime 2022.0.2) [22].

3.1 KSHV DNA Load Measurement

The VLs of samples collected during the 2020–2021 recruitment period were previously reported for the 206 individuals without history of KAD [17]. In the current study, 14 individuals from the earlier study and an additional 59 people with history of KAD recruited more recently were tested by qPCR. KSHV DNA was detected by qPCR in 30% (85 of 280) of study participants in either whole blood or oral fluids (Table 1).

3.2 Patient Characteristics

The characteristics of individuals with detectable KSHV DNA assessed for sequencing are shown in Table 1. The participants were men living with HIV with a median age of 43 (range 19–70 years), and 38% had histories of KAD. The median HIV load in the entire cohort was 24 copies/ml with an interquartile range of 0 to 66. All participants were prescribed ART; however, adherence was not assessed in this study. A total of 22 individuals from both studies had active KS at the time of enrollment. The population was predominantly Hispanic (41%) and Black (36%) with 22% White participants. KSHV K1 A and C subtypes predominate in the study (82%).

3.3 Newly Sequenced Near Full-Length KSHV Genomes and K1 Gene Subtyping

Fifty-nine of the 85 samples with detectable KSHV DNA by qPCR were Sanger sequenced to determine K1 gene subtype, which was successful in 41 participants as summarized in Table 1 and shown in Figure 2. A wide variety of KSHV K1 subtypes were observed including A1, A2, A3, A4, A5, B1, C1, and C3, as well as less frequently observed subtypes C7, E2, and F2. Data is available in GenBank using accession numbers PP789823-PP789863; PP952419-PP952426.

Using the KSHV load measured by qPCR, oral fluid DNA samples from twenty-six individuals were selected for NGS. Near full-length genomes were successfully obtained from 22 samples at greater than 30X (range 38X-5622X) read depth coverage (Supporting Information Table S1). One sample, UTSW124, with an estimated input of 17 KSHV copies/100 ng DNA, was sequenced by KAPA HyperPrep during optimization of the protocol to inform lower limit of KSHV load required for full genome coverage. UTSW124 was not included in full genome analysis, but the K1 gene sequence was included in the KSHV subtype analysis. The twenty-two new near full-length KSHV genomes were analyzed using SplitsTree v4.15.1 (Figure 3) and are available in GenBank as accessions PP768312-PP768333 [26].

The individual KSHV genomes for samples UTSW595, UTSW601, and UTSW107 were unresolvable and had clear evidence of more than one KSHV genome. This is shown by subtype specific variations within variable gene regions spanning across the viral genome (Figure 4). Infection with multiple lineages was confirmed for UTSW107 using two independent library preparations. To visualize the mixed infections across the KSHV genome, positions of nucleotide polymorphisms identifying the mixed infections are plotted based on frequency of occurrence (Figure 4A). Minority genomes constituting less than 5% of reads are more easily detectable in highly variable gene regions like K1, as demonstrated for UTSW595, while multiple genomes with coequal frequencies are distinguishable across the viral genome, as shown in UTSW601 (Figure 4A).

A wide range of KSHV K1 subtypes were identified in the new near full-length KSHV genomes, including the E2 and F2 for which limited information is available (Figure 3). K15 gene subtypes P and M were observed, while the N, more commonly seen in people born in Africa, was not present in the data set. Both individuals with KSHV K1 E2 subtypes were born in Mexico. Both E genomes were K15 M subtypes, in contrast the only other E genome currently available is from South America and has a K15 P allele (Figure 3). All participants with K1 F2 subtypes were born in the United States.

3.4 KSHV Gene Variations

Inspection of the genomes confirmed KSHV K1 subtype-specific indels and structural features recently described [10] including a sequence inversion between ORF8 and ORF9 observed in sequences with a F2 K1 subtype namely UTSW113, UTSW118, UTSW130 and UTSW624. The F2 subtype samples also have a two amino acid deletion in the ORF64 CDS shared with K1 A and C subtype samples UTSW535 and UTSW646. A K3 amino acid insertion was previously reported in samples of persons not born in Africa. This same K3 insertion was observed in K1 A subtype samples UTSW101, UTSW105, UTSW144, UTSW500, UTSW557, UTSW575, and UTSW607. A previously reported subtype-specific sequence inversion between ORF9-10 was observed in K1 B1 subtype sample UTSW547. Interestingly, the two K1 E2 genomes UTSW560 and UTSW657 share a distinctive 303 bp deletion between miR-K12-6 and miR-K12-5 with the only other published E2 sequence available, FNL0062 [10]. Additional observed structural features are summarized in Table 2.