INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Kaposi's sarcoma (KS) herpesvirus or human herpesvirus 8 (HHV8) DNA is present in virtually all tumor samples of both classical and AIDS-associated forms of KS (1a, 5), as well as in peripheral blood mononuclear cells in up to 50% of homosexual AIDS patients with KS (35). Infection without disease is also widespread in parts of Africa, the Middle East, and Mediterranean countries, with up to 60% seropositivity in Central Africa and 15% seropositivity in Italy, where classic and endemic KS rates are relatively high (11, 12, 16). Seropositivity is also very high among both KS positive and KS-negative male homosexual AIDS patients (85 and 50%, respectively) but not usually in human immunodeficiency virus (HIV)-positive intravenous drug user or hemophiliac AIDS patients. The seroprevalence in blood donors in the United States, the United Kingdom, and Japan may be no greater than 1% (11, 12), which also correlates with low KS incidences in renal transplant patients in those countries compared to reports of significant iatrogenic KS in patients of Middle East or Mediterranean Jewish ancestry in Toronto and in Saudi Arabia (13, 29).

HHV8 is a class gamma-2 herpesvirus that is distantly related to herpesvirus saimiri and Epstein-Barr virus (EBV) but contains several novel loci that include diverged viral homologues of exogenously acquired cellular genes encoding proteins such as interleukin-6, MIP-IA, MIP-IB, BCK, dihydrofolate reductase, TS, IRF, BCL-2, OX-2, FLIP, GCR, and CYC-D in place of the latent-state EBNA proteins and several other specific regulatory gene products of EBV (4, 22-27, 31). The nearly complete primary nucleotide sequences of the 190-kb double-stranded DNA molecules of two HHV8 genomes, one derived from an AIDS body cavity-based lymphoma cell line (31) and the other derived from a KS lesion (24), have been determined, and they differ overall by only 0.4% from each other.

Comparison of low-level sequence variability among different HHV8 samples in the open reading frame 26 (ORF26) and ORF75 gene regions originally led us to conclude that at least three distinct subtypes of HHV8 genomes occur (38). Subsequently, we found that the 289-amino-acid ORF-K1 membrane protein of HHV8 displays unusually high levels of genetic variability, resulting in four major subtypes, referred to as A, B, C, and D, that differ by 15 to 30% at the amino acid level (27, 37). Even within these ORF-K1 subtypes, amino acid differences of up to 6 to 8% among different clusters of subtype A ORF-K1 proteins and of 9 to 12% between two major branches of subtype C ORF-K1 proteins, together with small in-frame deletions, permitted the division of 46 genomes within the A-plus-C supergroup into 10 distinct variants plus several more narrowly diverged clades. Most U.S. AIDS KS samples are A1, A4, or C3 variants, whereas most classic KS cases from the Middle East, Asia, Europe, and United States are C2 variants. In contrast, samples from Africa are predominantly of the B subtype (plus occasional A5 variants) and the rare D subtype appears to be of Pacific Island origin. For example, among a total of 63 genomes examined at the ORF-K1 locus, 10 of 11 samples from central and southern Africa were subtype B; 6 of 7 from Saudi Arabia were subtype C2, C4, or C5; 7 of 9 from Taiwan represented a novel C3' clade; 6 of 7 from AIDS patients in Maryland were A1 variants, and all 3 samples from classic KS patients of native Pacific Island heritage were the only examples found of the novel D subtype.

We interpreted these results to imply that the major evolutionary branches of HHV8 correlate with the modern human population divisions that occurred via migrations from Africa first to southern Asia and Oceania 60,000 years ago and second to Europe and northern Asia some 35,000 years ago (37). The fact that cladal populations of the virus are still recognizable from paleolithic times presumably reflects both the relatively low-level distribution of the virus in many present-day populations and a relatively slow rate of horizontal transmission, which would together reduce recombination and scrambling of the differences to minimal levels. This scenario provides an appropriate evolutionary time scale for the application of some unknown but powerful biological selective pressure that generated ORF-K1 diversity. The ORF-K1 amino acid variability levels are similar to those found in immunoglobulin variable chains and in the HIV ENV protein, but we presume that unlike those situations, herpesviruses have no specialized mechanisms to rapidly generate this level of diversity. Nevertheless, it is plausible that some of the HHV8 cladal distribution patterns, especially within the hypervariable VR* loop, could reflect much more recent founder effects associated with higher horizontal infectivity rates within the AIDS epidemic. Further studies to continue to explore the patterns of HHV8 strain differences were undertaken because of the many intriguing questions raised about the molecular evolution of HHV8 genomes and the possibility that strain differences are associated with different disease states.

In the present study, we have evaluated variability at six other loci across the HHV8 genome, including three segments in the central conserved portions of the genome and three at the extreme right-hand end of the DNA molecule. Overall, the results have (i) confirmed that the divisions into three or four major subtypes can also be recognized within the more conserved portions of the genome, (ii) led to the identification of two highly diverged alternative allelic forms (referred to as predominant [P] and minor [M]) of the ORF-K15 region at the right-hand side (RHS) of the genome, and (iii) provided strong evidence that some genomes represent mosaics derived either by recombination between major subtypes or with some unknown but related exogenous virus source. In addition, we present evidence from reverse transcription (RT)-PCR analysis that the orphan extreme RHS regions of both the P and M subtypes of HHV8 encompass complex spliced latent-state genes encoding highly diverged integral membrane proteins that are distantly related to the LMP2 protein of EBV.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

HHV8 genomic phage library and subclones. A set of phage clones derived from two partial Sau3A BCBL-R tumor DNA genomic libraries in EMBL3 and DASHII backgrounds were described previously (25, 38). A EMBL3 clone (B3-2) containing a 17-kb insert from position 120.6 to position 137.4 proved to encompass the RHS terminus across the boundary of the unique region into the RHS terminal tandem repeats (TTRs). A 3.3-kb HindIII-SalI plasmid subclone, pDJA62, encompassing all of ORF-K15(P) and the RHS TTR from a subtype A genome was generated and sequenced by shotgun M13 procedures (part of the updated GenBank ORF75/K15 file [accession no. U85269]). Two additional EMBL3 clones, E-A2 (108.3 to 120.4) and E-C2 (115.7 to 129.0), encompassing the T0.7 gene region were also isolated, and a 5.0-kb EcoRI-to-EcoRI plasmid subclone was generated for initial PCR sequencing of the BCBL-R T0.7 region equivalent to coordinates 115470 to 120379 of BC1 (31).

HHV8-positive DNA samples. The sources and relevant characteristics of all of the KS and primary effusion lymphoma (PEL) DNA samples used here, as well as the procedures used for extraction of archival paraffin or frozen blocks, were described previously (37, 38).

PCR amplification and sequencing primers. Direct or nested PCR products from the following regions of the HHV8 genome were generated with Bethesda Research Laboratories (BRL) Taq or Stratagene extend polymerase in a Techne PHC3 thermocycler set at 94°C for 1 min, 50°C for 1 min, and 72°C for 2 min over 35 cycles using 20 to 100 ng of template DNA: ORF26 (330 bp), LGH1701 [5'-(GGAT)GGATCCCTCTGACAACC-3'] and LGH1702 [5'-(ACGT)GGATCCGTGTTGTCTACG-3']; ORF75-E (854 bp), LGH2087 (5'-CAGGTCGTCTACTATTCTG-3') and LGH1704 [5'-(GTAC)GGATCCACGGAGCATAC-3'], as well as LGH1984 [5'-(CTAG)AGATCTGTTTAGTCCGGAG-3'] and LGH2000 (5'-GGAAACAGGGTGCTGTG-3'); T0.7 (646 bp), LGH2076 (5'-GCTGCAATGTACTGCCATG-3') and LGH2075 (5'-CTCCAATCCCAATGCATGGA-3'); ORF72 (vCYC-D) (594 bp), LGH2045 (5'-GTAGAACGGAAACATCGCA-3') and LGH2046 (5'-GATTGGTATTGGGACCTTTC-3'); ORF-K1 (1,065 bp), LGH2089 (5'-GTTCTGCCAGGCATAGTC-3') and LGH2088 (5'-AATAAGTATCCGACCTCAT-3'); K14.1(P) (362 bp), LGH2079 (5'-GAGATCACTCTCCAACCAC-3') and LGH2033 (5'-GGAGTGCCTTCCGTATAG-3'); K14.1(M) (450 bp), LGH2079 (5'-GAGATCACTCTCCAACCAC-3') and LGH2506 (5'-CACAGTCACCTATGCTAG-3'); K15(P) (580 bp), LGH2476 (5'-GCAGTGTTTTATTAACGTC-3') and LGH2477 (5'-CAAACCCCATTTACTTC-3'); K15(M) (370 bp), LGH2473 (5'-CATGCAGCGAGCTTGAGA-3') and LGH2474 (5'-CTTTGAGTACTGTTTGTG-3'); U/TTR (850 bp), LGH2098, (5'-AAGATATAGACCCACCATAC-3') and LGH2097 (5'-CACGTAGCAAGCACTGAG-3'). The underlined sequences represent added restriction sites for cloning purposes.

Cloning and sequencing of RT-PCR products. Total RNA was isolated from uninduced HBL6 or BCBL-1 cells by using Trizol reagent (GIBCO BRL). For each cell line, RNA samples (3 µg each) were either DNase treated and subjected to RT-PCR or not DNase treated but subjected to RT-PCR or DNase treated and subjected to DNA PCR. DNase treatment involved incubation with 10 U of RNase-free DNase I (Boehringer Mannheim) for 4 h at 37°C. The samples were then extracted with phenol-chloroform, and the RNA was precipitated in 2.5 volumes of ethanol and 1/10 volume of 3 M sodium acetate (pH 5.2). Samples were resuspended in diethylpyrocarbonate-treated distilled H₂O, and reverse transcription was carried out at 50°C for 50 min by using Superscript II (GIBCO BRL) and an oligo(dT) primer (Boehringer Mannheim). Subsequent DNA amplification was performed by PCR for 45 cycles using Taq DNA polymerase (Promega) with the following 5' and 3' primers designed to encompass the complete ORF-K15 coding region for both the P and M forms: HBL6(M) N terminus (LGH3129), 5'-CTAGGAATTCATGAATTACAAAAAATACCTGTGGGG-3'; HBL6(M) C terminus (LGH3132), 5'-CTAGGAATTCGTCCGTGGGAAACAAAAC-3'; BCBL-1(P) N terminus (LGH3110), 5'-CTAGGGATCCATGAAGACACTCATATTCTTCTGG-3'; BCBL1(P) C terminus (LGH3111), 5'-CTAGAGATCTGTTCCTGGGAAATAAAACCTCC-3'.

Nucleotide sequence accession numbers. The GenBank accession numbers for these cDNA sequences are AF156886 [BCBL1 ORF-K15(P)] and AF156885 [HBL6 ORF-K15(M)].

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Linkage analysis at multiple genetic loci reveals evidence for recombinant genotypes. An overview map of the HHV8 genome which shows the positions of the six PCR loci involved in this work relative to other key features of the genome and to several characterized phage lambda clones from the RHS of the BCBL-R genome is presented in Fig. 1A. In our initial strain difference studies (38), the DNA sequence patterns found in the ORF26, ORF75, and UPS75 gene loci each fell into three distinct clusters represented by the prototypes of subtypes A (BCBL-R), B (431KAP), and C (ASM72), which all differed by 0.8 to 1.5% at the nucleotide level at each locus. Among the first 12 genomes tested, there were seven examples of subtype A, three of B, and two of C, and in 10 of these cases, all three genetic loci cosegregated but in 2 they did not (38). Addition of the more recent ORF-K1 data introduced additional complexities (37). For example, HBL6 was interpreted to belong to subtype C (identical to ASM72) at both the ORF75 and UPS75 loci, but was subtype A for ORF26 and ORF-K1. Similarly, ST1 was subtype A at both the ORF75 and UPS75 loci but gave a subtype C pattern for ORF26 and a subtype B ORF-K1 pattern. Finally, a third sample, EKS1, although giving a subtype A-like pattern at all three original loci, proved to be subtype C in ORF-K1.

View larger version (23K): [in this window] [in a new window]   FIG. 1.   Organization of the RHS of HHV8 genomes relative to the loci used for PCR analysis. (A) Relative genomic positions of the ORF-K1, ORF26, T0.7, ORF75, and ORF-K15 loci are indicated as solid bars with approximate genomic percentile coordinates included. Solid circles and overhead arrows refer to the duplicated predicted ORI-Lyt regions (DL and DR) (27). The G+C-rich TTRs are indicated. The names and map positions of four relevant overlapping phage lambda clones from the RHS of HHV8(BCBL-R) genomic DNA are illustrated by horizontal two-headed arrows. (B) Overall structure of the 3.3-kb sequenced region representing the RHS terminus of the unique region that encompasses the ORF-K15(P) allele in HHV8(BCBL-R). The intact region spanning the N terminus of the ORF75 gene across to the TTR (solid bar) on the RHS was subcloned into plasmid pDJA62 as a HindIII/SalI fragment of phage B3-2 derived from a PEL tumor genomic DNA library (27). The relative organization of ORF75 and the predicted short ORF-K14.1 gene, together with the multiple spliced exons of the extended ORF-K15 gene(s), are indicated below the plasmid map. The comparative structures of the P and M alleles are also shown. Parentheses (del) denote the unstable G+A-rich repeats near the 5' end of the M allele of ORF-K15 (31). TTR regions are shown as solid bars, and ORF-K15(M) regions are shown as dotted bars. (C) Map locations and sizes (in base pairs) of the PCR loci used to characterize P and M subtype genomes. Among the three ORF75 PCR products shown (hatched bars), only the new ORF75-E region was used here for extensive sequence analysis (see Fig. 6). K14.1(P) and K14.1(M) represent PCR products of the triple-primer reaction covering the divergent junction of the two subtypes of ORF-K15 genes (see Fig. 3). The K15(P) and U/TTR products are unique to the P subtypes (open bars), and K15(M) is unique to the M subtypes (dotted bars).

View larger version (37K): [in this window] [in a new window]   FIG. 2.   Representative examples of PCR DNA products involved in these studies. (A) Photograph of ethidium bromide-stained DNA bands in an agarose gel displaying a complete set of directly amplified PCR products from across the HHV8 genome present in the JSC1 cell line for use in DNA sequence analysis. Lanes: 1 and 9, 123-bp oligomer size marker ladder; 2, ORF26 primers (LGH1701 and LGH1702); 3, ORF75-E (LGH2087 and LGH1704); 4, U/TTR (LGH2097 and LGH2098); 5, T0.7 (LGH2076 and LGH2075); 6, CYC-D (LGH2044 and LGH2045); 7, ORF-K1 (LGH2089 and LGH2088); 8, ORF-K14.1 (triple-primer reaction of LGH2079, LGH2033; and LGH2506). (B) Photographs of ethidium bromide-stained RHS junction region PCR DNA products (K14.1) from various P (362 bp) and M (450 bp) allele genomes. Lanes 1 and 9 contained a 123-bp oligomer size marker ladder. The KS and PEL DNA samples used included the following: lanes 2 and 10, HBL6 (M); lane 3, BC2 (P); lane 4, BC3 (P); lane 5, AKS1 (P); lane 6, BCP1 (P); lane 7, BCBL-B (P); lane 8, ASM72 (M); lane 11, ASM76 (M); lane 12, BCBL1 (P); lane 13, EKS1 (P); lane 14, C282 (P); lane 15, a mixture of BCLB1 and HBL6 DNAs. The triple-primer combination of LGH2079 (both) plus LGH2033 (P) and LGH2506 (M) was used in standard direct PCRs.

Two distinct types of ORF-K15 region sequences at the extreme RHS of the genome. Our initial primary DNA sequence analysis of the extreme RHS of the HHV8 genome just upstream of ORF75 was carried out from a genomic phage library of our prototype subtype A KS genome, HHV8(BCBL-R). Curiously, several different PCR primer pairs derived from this region of the BCBL-R sequence all failed to yield any amplified products from either our prototype subtype C genome (ASM72) or the A/C hybrid genome (HBL6) (38). Direct PCR sequence analysis of the proximal upstream ORF75 region in all three prototype strains (BCBL-R, ASM72, and 431KAP) revealed that although subtypes A and B continued to be nearly identical upstream of the ORF75 coding region, both subtype C patterns proved to diverge toward a nearly totally different sequence beginning within 300 bp of the N terminus of the ORF75 coding region. To analyze the remainder of the RHS unique region from our subtype A genome, we first mapped and subcloned a 16-kb genomic region from the extreme RHS unique-segment terminus of BCBL-R DNA in a selected phage lambda clone (B3-2) that proved to encompass both the ORF75 region and the adjacent G-plus-C-rich TTR sequences (Fig. 1B). M13 sequencing on both strands of a 3.3-kb HindIII-to-SalI subclone containing the junction fragment (in plasmid pDJA62) revealed the presence of a 2,354-bp unique sequence between the ORF75 initiator codon and the beginning of the TTR region. The junction with the RHS unique region begins at position 653 of the published TTR sequence of HHV8(BC1) (GenBank accession no. U75699) (31) and then crosses into the next TTR unit and ends at position 551, representing the Sau3A site at the end of the RHS of B3-2 (Fig. 1B). As judged by the finding of identical-size PCR-amplified DNA products for a region spanning the unique-region-TTR junction from the original BCBL-R tumor DNA, as well as in both the B3-2 lambda clone and the M13-sequenced pDJA63 plasmid DNA (data not shown), we conclude that this sequence represents the complete undeleted RHS unique region from a prototype subtype A genome (GenBank accession no. U85269). Subsequent PCR sequencing of BCBL-1 DNA in this region also showed it to be almost identical to that of BCBL-R.

View larger version (53K): [in this window] [in a new window]   FIG. 3.   Comparison of DNA sequences across the divergent junctions between the leftward-oriented ORF75 and ORF-K15 gene regions in both the P and M subtypes of HHV8. The hypothetical rightward-oriented ORF-K14.1 genes encompassed by this region are also shown. Potential transcriptional control elements and the predicted amino acids encoded by the extreme N terminus of ORF75, the extreme C terminus of ORF-K15, and the intervening ORF-K14.1 region are shown. Nucleotide identities between the P allele in BCBL-R (top line) and the M allele in HBL6 (lower line) are indicated by asterisks, and missing bases are represented by hyphens. The positions of the PCR primers, LGH2079, LGH2033, and LGH2506, that were used to discriminate between the P and M allele forms of ORF-K14.1 in the triple-primer PCR (Fig. 2B) are indicated. The nucleotide sequences shown correspond to genomic positions 1268 to 1803 in the P allele (BCBL-R; GenBank accession no. U85269) and 134365 to 134881 in the M allele (BC1; GenBank accession no. U75698).

Many subtype A, B, and C HHV8 genomes contain the alternative M form of the RHS junction region. To evaluate the relative abundance of each of the two types of RHS among HHV8 genomes, we generated a three-primer PCR set containing a single primer within the LHS common region at position +42 in ORF75 and two alternative RHS primers in the divergent regions either at position 320 of BCBL-R (generating a 362-bp product) or at position 408 of HBL6/BC1 (giving a 450-bp product including genome positions 134399 to 134848). Representative examples of the results are shown in Fig. 2B, and the positions of the primers are indicated in Fig. 3. Analysis of all 63 different HHV8 DNA samples for ORF-K14.1 PCR product size variations revealed 45 of the BCBL-R/BCBL1 type and 18 of the HBL6/BC1/ASM72 type. The latter were associated with more than half (13 of 24) of the genomes with subtype C ORF-K1 genes plus 3 of 19 subtype A genes and 2 of the 3 American subtype B genomes but not with any of the 12 African subtype B genomes or the 3 Pacific subtype D genomes. Therefore, the two versions of the RHS ORF-K14.1 region appear to be only partially linked to the LHS ORF-K1 subtype patterns. To avoid confusion and to clearly distinguish them from the subtype C nomenclature used for ORF-K1, we have altered the previously introduced subtype C terminology for the ORF75 regions found in ASM72 and HBL6/BC1 (38) by referring to these alternative minor population forms of the RHS as M alleles compared to the predominant P alleles found in BCBL-R and the majority of the HHV8 genomes tested.

Evaluation of the coding capacity of the two ORF-K15 regions. In their original analysis of HHV8(BC1) coding regions, Russo et al. (31) identified only a single short, leftward-oriented coding region of 100 amino acids (termed ORF-K15) within the 3,500-bp region between ORF75 and the RHS TTR. Although our DNA sequence in the equivalent 2,500-bp block in HHV8(BCBL-R) revealed no significant nucleotide homology, it displayed a very similar overall distribution of GC-rich and AT-rich patterns, as well as a small ORF-K15-like region, but with no associated ATG codon. Furthermore, both sequences also contained numerous potential consensus splice signals located on one strand only (leftward orientation) together with a second potential larger ORF without ATG motifs just upstream of ORF75. Most strikingly, there were also several very small ORFs with potentially conserved amino acids, as well as numerous potential hydrophobic transmembrane (TM) domain-like domains present at apparently colinear loci in the leftward orientation (27).

View larger version (24K): [in this window] [in a new window]   FIG. 4.   The structures and intron-exon boundaries of the ORF-K15(P) and ORF-K15(M) proteins and mRNAs are remarkably similar and resemble those of the EBV LMP2 protein. Overall spliced structure and organization of coding exons of sequenced RT-PCR products from uninduced latent-state mRNAs from the BCBL-1 (P) and HBL6 (M) PEL cell lines are shown. Numbers and solid bars show the relative size and amino acid (aa) content of each coding exon (E1 to E8) from ORF-K15 (P) and ORF-K15 (M) mRNAs compared to the organization and size of exons (E1A to E9 or E1B or E9) from EBV LMP2A and LMP2B mRNAs. Seven additional amino acids are created across the splice junctions in both ORF-K15 mRNAs. Hatched and solid bars indicate coding regions, and hatched regions denote the 12 TM domains. Open bars indicate noncoding regions and exons. Circles represent conserved consensus YXXL-type SH2-binding tyrosine kinase interaction motifs. pA, polyadenylation signal.

View larger version (45K): [in this window] [in a new window]   FIG. 5.   Comparison of the complete predicted protein sequences of the ORF-K15(P) and ORF-K15(M) alleles of HHV8. Each protein consists of eight matching exons and 12 TM domains (TM1 to TM12) with an extended C-terminal cytoplasmic domain. The ORF-K15(P) allele coding region occupies nucleotide positions 1578 to 3652 and is oriented leftward in the HHV8(BCBL-R) and HHV8(BCBL1) DNA sequences (GenBank accession no. U85269). The ORF-K15(M) allele coding region occupies nucleotide positions 134671 to 136771 and is oriented leftward in the HHV8(BC1) DNA sequence (GenBank accession no. U75698) but has an additional A inserted after position 136610 in our HBL6 version. The 12 presumed TM domains are denoted by the broken overlines, and intron-exon junctions are indicated by carets. Amino acid identities are signified by asterisks, and similarities are shown by vertical lines. The potential protein tyrosine kinase binding motifs YASIL and YEEVL plus a conserved QSG(M/I)S motif in the C-terminal cytoplasmic domain in exon 8 are highlighted.

The unique-sequence-TTR junction in the P form of the RHS end differs between the B/D and A/C subtypes. To further evaluate the level of heterogeneity found within the two forms of the extreme RHS of HHV8 genomes, we also analyzed another PCR locus representing the 850-bp unique-sequence-TTR boundary immediately upstream and to the right of the ORF-K15(P) genes from several genomes representing each of the major ORF-K1 subtypes. The results revealed two principal alternative PCR products that differed in size by approximately 45 bp, with the shorter forms being detected only in genomes with subtype B and D ORF-K1 genes (data not shown). As expected, this PCR primer pair failed to give any products with M subtype genomes.

The M allele of ORF-K15 is linked to a specific M subtype of the adjacent ORF75 gene. To try to understand the differences noted previously between ASM72 and HBL6/BC1 compared to subtype A and B genomes within the ORF75 coding region, we PCR sequenced an expanded 854-bp region (ORF75-E) from 58 of the HHV8 genomes in our ORF-K1 set. The results revealed that even in the adjacent downstream ORF75 coding region (38), the influence of the M allele, versus that of the P allele, was still readily detectable. A chart showing the positions of all 20 of the nucleotide polymorphisms detected within ORF75-E and our interpretation of subgroup patterns is presented in Fig. 6. The data revealed that 25 of the 28 subtype A and C genomes having the P type of RHS were identical except for a single-nucleotide change in both subtype A3 genomes (BCBL1 and BKS14) and another in one of the subtype C2 samples (EKS1). Among the 14 subtype B genomes tested, two distinct patterns emerged. Six gave a characteristic B pattern, which differed from that of all of the A-plus-C versions at up to seven positions. The other eight samples (plus OKS3) were identical to the A/C pattern, except for 159-G in JKS20 and 563-G in the three American B samples. The three subtype D1 and D2 ORF75-E sequences were also unique and differed at six positions from the A/C pattern, as well as at seven positions from the B pattern. Remarkably, among genomes with the M type of RHS region, 14 of the 16 examples tested, representing a variety of C1, C2, C3, C5, A1, and A2 ORF-K1 genomes, were all identical to one another and to HBL6/BC1, except at one position. However, they also differed from the B/P subgroup ORF75-E pattern at 9 positions, from the D/P pattern at 13 positions, and from the A/P and C/P subtype RHS genomes at 14 positions. The two B/M genomes OKS7 and OKS8 from Florida were the only exceptions found in which M allele ORF-K15 genes were associated with P subtype ORF75-E patterns.

View larger version (29K): [in this window] [in a new window]   FIG. 6.   Comparison of polymorphic nucleotide patterns that identify four distinct subgroups of HHV8 genomes within the extended 854-bp ORF75-E block (map coordinate 0.96). Dashes indicate identities to the prototype A1 class sequence (BCBL-R) in the top line. The numbering system refers to positions (+ or ) relative to those used by Zong et al. (38) for ORF75 based on the ORF75 631-bp RDA fragment of Chang et al. (5). Positions 223, +1, and +631 are equivalent to positions 132875, 133098, and 133728 in the BC1 complete genome sequence (primers LGH2087 and LGH1704). ORF-K1 and ORF75-E subclass assignments are given in the far left-hand and far right-hand columns, respectively. Importantly, genomes with P allele ORF-K15 genes that have subtype A or C ORF-K1 genes (and some with subtype B ORF-K1 genes) cannot be distinguished here (=A/C class), whereas nearly all of the genomes tested with M allele ORF-K15 genes (denoted by asterisks) form a distinctive M subclass of ORF-75E patterns. Note that all of the genomes examined differed from the original 631-bp sequence of Chang et al. (5) by having 167-T, 442-C, and 496-T instead of 167-C, 442-T, and 496-C.

Further penetration of M allele-associated patterns into the T0.7/K12 gene region. A short and highly abundant viral RNA species referred to as T0.7, which potentially encodes the small ORF-K12 membrane protein, was the first HHV8 gene product identified that is unambiguously associated with latent-state expression in both PEL cell lines and KS tumor spindle cells (33, 34, 36). This gene lies within the DL-E divergent locus toward the RHS of the genome at map position 0.85. The latent-state ORF72 (vCYC-D) gene region at map coordinate 0.90 (data not shown) was also examined initially as a potential RHS locus that we expected might not be linked to the P and M genotypes for ORF75 and ORF-K15. Although both loci gave subtype-specific polymorphisms, the T0.7 gene proved to display a more useful level of subtype divergence (up to 5% nucleotide variation). Therefore, PCR sequence analysis of the 646-bp T0.7 segment equivalent to nucleotide positions 118114 to 117469 of the BC1 genome was carried out on a total of 58 different HHV8 samples representing almost all of those included in our previous ORF-K1 analysis (37). The results are summarized in the chart in Fig. 7 in the form of variations from our usual prototype subtype A1 genome BCBL-R. Among a total of 35 polymorphic positions within the T0.7 region, four major subgroup patterns were detected, with the A-plus-C group giving two very distinctive types of patterns and the B and D patterns also each being very different from the others. However, again there were some significant changes from the typical ORF-K1 groupings within the A-plus-C subtypes and several additional minor variant patterns were discernible among both the A and B subtypes.

View larger version (41K): [in this window] [in a new window]   FIG. 7.   Comparison of polymorphic nucleotide patterns that identify four distinct subgroups of HHV8 genomes within the 646-bp T0.7/K12 gene locus (map coordinates 0.85). Dashes indicate identities to the prototype A1 class sequence (BCBL-R) in the top line. The numbering system refers to the leftward noncoding strand direction of the gene, which is inverted relative to the genomic sequence. Therefore, position 118114 in the BC1 complete genome sequence represents position 1 of the PCR product and position 117469 in BC1 is position 646 (primers LGH2076 and LGH2075). ORF-K1 and T0.7 subclass assignments are given in the far left-hand and far right-hand columns, respectively. Importantly, genomes with P allele ORF-K15 genes that have subtype A or C ORF-K1 genes cannot be distinguished and many (but not all) of the genomes with M allele ORF-K15 regions fall into a distinctive M subclass. Note that all of the genomes examined, including HBL6, differed from BC1 of Russo et al. (31) by having 348-A in place of 348-T (genomic position 117769).

Re-evaluation and extension of the ORF26 region data. Although we originally defined the major A, B, and C subtypes in the ORF-K1 region based on the presumption of linkage to the original assignments for the ORF26 and ORF75-C regions in our prototype strains (i.e., BCBL-R as A; 431KAP as B, and ASM72 as C), it has become evident that ORF26 data alone are insufficient for accurate discrimination between subtype B and C patterns, especially within the smaller 233-bp ORF26 block used in most previous studies. Partly because of the presumed need to revise some tentative assignments in Table 2 of Zong et al. (38) and partly because there were no subtype C3 genomes in our original analysis, we have re-evaluated a somewhat larger 330-bp ORF26 block at map position 0.35 in the genome after sequencing 60 of the HHV8 genomes included in the ORF-K1 analysis. A summary chart of the results and interpretations is presented in Fig. 8. The data include our reanalysis of BC2, which differs at two positions from the original published sequence of Cesarman et al. (3) for BC2, as well as our data for HBL6, which agrees with that for BC1 of Russo et al. (31) but not that of Cesarman et al. (3). We have also omitted the KSHV sequence of Chang et al. (5) because we suspect that the unusual base 1033-C, rather than 1033-T, in ORF26 and three nonconsensus bases in ORF75-E, which have not been found in any other samples, may all represent PCR errors introduced during representational difference analysis (RDA).

View larger version (23K): [in this window] [in a new window]   FIG. 8.   Comparison of polymorphic nucleotide patterns that identify several distinct subgroups of HHV8 genomes within the 330-bp ORF26 gene locus (map coordinate 0.35). Dashes indicate identities to the prototype A1 class sequence (BCBL-R) in the top line. The numbering system conforms to that used by Zong et al. (38) and introduced by Chang et al. (5) for the ORF26 RDA fragment. Positions 893 (=1) and 1222 (=330) in these PCR products are equivalent to positions 47193 and 47522 in the complete BC1 genomic sequence (primers LGH1701 and LGH1702). Overall, the ORF26 sequences resolve into four distinct patterns roughly equivalent to the ORF-K1 A, C3, C2, and B subtypes. However, the patterns for several of the B and C variants are either very similar or identical and the D patterns are indistinguishable from those of subtype A patterns. Therefore, the use of ORF26 patterns alone for classification is not recommended. Note that all of the genomes examined here differ from the original data of Chang et al. (5) by having 1033-T instead of 1033-C (genomic position 47333). All of the genomes examined here, including HBL6 and BC2, also differed from the data given for BC1 and BC2 by Cesarman et al. (3) by having 926-A, 989-C, and 1199-T.

Complexities and possible overlaps among some ORF26 patterns within the B and C subgroups. In the ORF-K1 analysis (37), only 1 of 10 African KS samples tested did not belong to subtype B, and that was the single example found of subtype A5 (OKS3). The ORF26 pattern for OKS3 is also distinctive, being unlike any of the other A patterns and different from the prototype C3 (BC2) and B (431KAP) patterns. However, the ORF26 sequence pattern for OKS3 matches those of TKS11 (C2) and BKS13 (C1), as well as our unusual ST1 chimeric African sample, which proved to have a subtype B ORF-K1 gene, although it clearly had an A/C subtype ORF75-E pattern. Furthermore, four other African KS samples reported by Huang et al. (14), as well as four KS samples from Saudi Arabian renal transplant patients, have this type of ORF26 pattern (9). Subsequent sequencing of the ORF-K1 region of six Saudi Arabian samples with this type of ORF26 gene revealed four C2 patterns, as well as our prototype examples of the C4 and C5 patterns (37). Therefore, we now believe that the 981-C, 1086-T, and 1139-C pattern for ORF26, as found in TKS11, BKS13, SKS1, SKS2, SKS3, and SKS6, which all differ by only 1 bp (A-1094) from ASM72 (C1), probably represents the prototype C2 pattern. Since most C3 genomes have the distinctive 981-C, 1032-A, 1055-T, 1132-G, and 1139-C pattern for ORF26 (Fig. 8), we also have to conclude that the Taiwan clade TKS1 to TKS9 genomes, which are identical to C2 in ORF26 except for one additional distinctive change to 935-A, presumably all represent a novel C3'/C2'/M recombinant genome. Evidently, KS-F and EKS1 must both also represent chimeric C/A recombinant genomes based on discordant data for their ORF-K1 and ORF26 loci (see later).

Correlations and linkage between multiple loci. An overall summary of the available data and subgroup assignments determined at each of up to seven distinct central and RHS loci across the genomes of 63 KS and PEL patient samples are compared with the LHS ORF-K1 data from Zong et al. (37) in Fig. 9. Ignoring the complexities introduced by the ORF-K15(M) RHS recombinants, a reasonably coherent picture has emerged showing linkage and cosegregation between subgroup patterns across the whole genome. In most of the 49 subtype A, C, and D genomes examined, the subgroup patterns of the ORF-K1, ORF26, T0.7, ORF75-E, and UPS75 loci are fully consistent across the entire DNA molecule. Obvious exceptions that are best interpreted as intertypic recombinants between different A and C variants include EKS1 and SKS7 (C2/A2), SKS3 (C2/A3), KS-F (C3/A1), and BKS12 (C3/A2), as well as the seven TKS1 to TKS9 genomes (C3'/C2'). Note that many of these also have an M allele pattern at the RHS. In addition, BCBL1, which may be an A3/C3 recombinant, has ORF26, T0.7, and ORF75-E patterns slightly different from those of all of the other subtype A samples (although in each case differing at only one position) but has a novel RHS U/TTR junction pattern (4-bp differences from all of the other subtype A and C samples tested), including a C3-like variant of the TTR. Our current interpretation of the overall structures of 13 clearly chimeric HHV8 genome types are illustrated and summarized in Fig. 10. Most of the intertypic and intratypic chimerism observed was associated with the nine distinct types of M allele-containing genomes, and it apparently reflects multiple sequential recombination events that generated these genomes, but four types of intertypic chimerism within P allele genomes are also illustrated.

View larger version (34K): [in this window] [in a new window]   FIG. 9.   Summary of the overall subclass assignments and patterns of linkage obtained at eight different loci across the HHV8 genome. The comparison includes all 63 HHV8 samples for which complete ORF-K1 sequence data (K1 column) were generated previously (37) and also includes updated ORF26 and UPS75 locus data relative to the original 12 genomes evaluated (38). ORF72 (vCYC-D) data are not included, and because there were only limited amounts of DNA available for some samples, sequences were not obtained at all loci in all cases (blanks). Type information in the column alongside the sample designations summarizes sample characteristics where + is HIV or AIDS associated, B is BCBL or PEL, K is KS, C is classic or endemic, and R is renal transplant. The source column indicates geographic locality (or origin), where A is Africa, F is Florida, G is Germany, P is Pacific Islands, S is Saudi Arabia, T is Taiwan, Z is New Zealand, and U is United States (other than Florida). Designations in parentheses in the K14.1 and K15 columns indicate data derived from PCR product size measurements only. All other entries are derived from PCR sequence data. Only P subtype information can be obtained at the U/TTR locus, and dashes in this column indicate where no PCR products were detected from M subtype genomes. Note that 9 samples are considered to be intertypic recombinants and 18 others are chimeras containing M allele sequences at the RHS (see Fig. 10).

View larger version (32K): [in this window] [in a new window]   FIG. 10.   Summary of the deduced overall genome structures of 13 different mosaic HHV8 genotypes that are interpreted to represent either intertypic recombinants or chimeras that have acquired ORF-K15(M) alleles and associated M-linked ORF75 and T0.7 genes. The relative genomic locations of the seven PCR loci involved in these studies are shown at the top. Key for recombinant genomes: solid horizontal lines, subtype A and RHS P allele sequences; open bars, subtype C sequences; dotted bars, subtype B sequences; hatched bars, M allele-linked sequences; solid bars, presumed exotic nonhuman ORF-K15(M) sequences. (A) Four examples of exclusively intertypic P allele genomes. Note that the EKS1 pattern is also found in SKS7 and that variations of the RKS2/5 pattern are also found in OKS3, ST1, JKS15, and JKS20. (B) Nine examples of chimeric genome patterns containing M allele RHS sequences, including both ORF-K15(M) itself and adjacent M-linked genes. TKS1/9 and BKS16 are also considered to be intratypic recombinants independently of the M allele sequences, whereas BKS12, SKS3, and OKS7/8 are also intertypic recombinants. Note that the ASM72 pattern is also found in BKS13 and the TKS1/9 pattern is also found in TKS2, TKS3, TKS5, TKS6, and TKS7. (C) Model for the origin of M allele sequences from a hypothetical original intact M subtype parent genome. Note that this is itself interpreted to be a chimera containing a highly diverged ORF-K15(M) gene that was likely to have been derived initially by recombination between an ancient human M subtype HHV8 and an exotic nonhuman HHV8-like virus source.

Complex chimerism among some subtype B genomes. Approximately half (7 of 12) of the subtype B KS samples from Africa that have been analyzed appear to have simple uncomplicated subtype B characteristics throughout their genomes. However, the other five, as well as the three subtype B genomes from America, display a complex mosaicism that may reflect a history of both intertypic and intratypic recombination. Interestingly, all 10 of the samples obtained directly from Africa (as well as those from two recent African emigrants) have some features of subtype B genomes. This includes ST1, in which only the LHS is B, and OKS3, in which ORF-K1 is an A5 variant, but other parts of the genome (especially T0.7) have distinctive subtype B characteristics.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The ORF-K1 genes from more than 60 HHV8 genomes analyzed previously have evolved into four major subtypes with distinctive ethnic and geographic associations that we believe are likely to have arisen during the migrationary divergence of modern humans in Paleolithic times (37). Thus, the isolation of the precursor D subtype of HHV8 from a precursor B subtype was predicted to have occurred approximately 60,000 years ago as humans first spread from Africa to Southern Asia and the Pacific, whereas the derivation and subsequent divergence of the A and C branches probably correlated with later waves of migration into Europe and Northern Asia via the Middle East about 35,000 years ago. The nature of the biological selection pressure that has been driving the unusually high level of amino acid diversity displayed by the ORF-K1 membrane receptor signalling protein (18, 20, 28) is not understood, but obvious additional questions have arisen about whether or not the ORF-K1 protein subtypes also contribute to the different disease patterns observed in different regions of Africa compared to Mediterranean countries, etc. However, before this concept can be evaluated seriously, we considered it necessary to resolve whether other segments of the HHV8 genome display parallel strain heterogeneity patterns, as well as to determine whether previous hints that some genomes may be chimeric recombinants between subtypes represent additional complicating factors.

Therefore, we carried out a systematic analysis of sequence variability at three internal constant-region loci and at the extreme RHS of the genome in the same large set of samples that had been analyzed for ORF-K1 (37). The results of these studies at genomic map coordinates 0.35 (ORF26), 0.85 (T0.7/K12), and 0.96 (ORF75-E), as well as extensive evaluation of an apparent dichotomy in the DNA sequence patterns at the RHS of the genome (27), have led to three major conclusions. Firstly, consistent subtype A, B, C, and D patterns matching those in ORF-K1 are detectable throughout the more conserved internal regions of the genome, although the level of divergence is greatly diminished. Secondly, a very different pattern of strain subgrouping occurs at the extreme RHS of the genome, whereby two alternative highly diverged P and M subtypes or alleles of the ORF-K15 protein are essentially unlinked to the LHS subtype patterns. Thirdly, a combination of the presence of distinctive alleles of other adjacent genes linked to the P and M forms, together with evidence for simple recombination between some major subtypes, complicates the analysis of strain assignments in as many as 30% of all HHV8 genomes.

Origins of the ORF75 and T0.7 M-linked alleles. Our original interpretation (38) that the ORF75 sequence pattern present in both the ASM72 and HBL6/BC1 genomes represented the prototype subtype C pattern has had to be revised. Instead, this pattern clearly represents that of the M-linked allele of ORF75, which was present in 14 of the 16 genomes examined that have the M subtype of the ORF-K15 gene, and includes examples of C2, C3, C5, A1, and A2 variant ORF-K1 genomes. Furthermore, examination of a number of ORF-K15(P) allele genomes having subtype C ORF-K1 genes revealed that the presumed true C pattern for ORF75-E cannot be resolved from that of most subtype A genomes, although the B and D patterns are distinctive. The level of nucleotide divergence of the subtype P[D] and P[B] ORF75-E loci from P[A/C] is 0.7% and 0.8% respectively, whereas the M allele ORF75-E pattern differs from P[A/C] by 1.4% and from P[B] by 1.0%. Therefore, it is tempting to argue that the M allele of ORF75 represents an older form that diverged from the P prototype before modern HHV8 genomes split into the A, B, C, and D subgroups. Similarly, at the T0.7 locus, all remaining M patterns (10 of the 18) are linked to both ORF75(M) and ORF-K15(M) genes, and again the M form is at least as far diverged from the P[A/C] and P[D] patterns as is the P[B] pattern.

Specific features of M subtype recombinant genomes. The viruses found in all seven of the classic and AIDS-associated KS lesions examined from patients of Northern Chinese heritage in Taiwan (whose ancestors first emigrated no more than 400 years ago) represent a particularly interesting and complex clade. Although these seven genomes all differ from one another in the VR* loop and other variable loci within ORF-K1 (37), they all contain a distinctive subset of C3' ORF-K1 genes and a unique C2' variant of the ORF26 gene, together with both an apparently invariant M allele of the ORF-K15 gene and linked, almost invariant M subtypes of T0.7 and ORF75 (Fig. 9 and 10). Other than this clade, only 11 of the 56 genomes that we have examined from elsewhere around the world contained the ORF-K15(M) type of RHS. Note that in contrast, all five of the classic KS patients of Hwalien heritage that have been tested (whose ancestors are believed to have arrived in Taiwan from the Pacific Islands about 4,000 years ago) had D1/P subtype HHV8 genomes (33a, 37).

Lack of divergence between subtypes A and C at the RHS is paralleled within the constant regions of ORF-K1. Our interpretation that all subtype A and C genomes with P allele RHS patterns are almost invariant within their T0.7, ORF75-E, and ORF-K15 genes was initially rather surprising considering that the A and C subtypes of ORF-K1 at the LHS of the genome differ by as much as 14% at the amino acid level. However, a re-evaluation of the distribution of sequence variations across relatively well-conserved segments of the ORF-K1 membrane signalling protein confirms the notion that they are, in fact, much more closely related than the overall level of variation indicates. As we pointed out previously (37), there are two distinct types of variation in ORF-K1, firstly, the relatively consistent and characteristic intertypic variations and, secondly, the seemingly random intratypic hypervariability displayed both between and within the different variants and clades of each subtype. Both levels of divergence appear to result from powerful biological selection pressures, but the latter are concentrated mostly within the two 40-amino-acid blocks (VR1 and VR2) within the extracellular domains and especially within the 23-amino-acid Cys-bridged VR* loop (37). In fact, most of the variations between subtypes A and C of ORF-K1 occur within these two hypervariable domains. In contrast, when only the four most conserved regions (totaling nearly 60% of the protein) from between amino acids 1 to 36, 105 to 147, 173 to 199, and 229 to 289 are considered, the prototype A and C subtypes differ by only 2.5% (six amino acid changes), whereas A and D still differ by 19% (32 of 164) and A and B differ by 27% (47 of 164). Therefore, we conclude that the hypervariability displayed within and between the A and C subtypes and variants predominantly represents a much more recent ongoing intratypic process (that has occurred even within the Taiwan C3' clade, for example). In contrast, the specific variations common to all members of a subtype obviously represent a more accurate reflection of older evolutionary divergence. Therefore, in this sense, even on the LHS, the much greater similarity between the constant regions of the subtype A and C proteins compared to subtype B and D proteins matches quite closely to the patterns observed at the RHS. A similar situation applies to the D1 and D2 subtypes, which are virtually identical at the ORF26, T0.7, ORF75, and UPS75 gene loci, despite showing 21% overall amino acid differences between them in ORF-K1.

Nature and origin of the two types of ORF-K15 genes. The M and P subtypes (or alleles) of the HHV8 ORF-K15 integral membrane protein genes, although they display less than 30% amino acid identity, have practically identical intron-exon splicing patterns in latent-state mRNA, and they show considerable overall structural and exon pattern resemblances (but no homology) to the LMP2 latency protein of EBV. All three proteins have 12 similarly spaced hydrophobic TM domains, and in fact, our initial prediction of the complete ORF-K15 protein structure, which has been confirmed by RT-PCR analysis, depended partly upon the presumption of a linear splicing pattern closely parallel to that of LMP2 (17). However, the ORF-K15 proteins contain two conserved likely SH2-binding tyrosine kinase signalling motifs and other proline-rich SH3-like binding domains within a large C-terminal cytoplasmic domain, whereas LMP2 has functional ITAM, SH2, and SH3 binding domains in an N-terminal cytoplasmic domain originating from across the other side of the TTRs (Fig. 4). Nevertheless, there appears to be little doubt, because of their equivalent genomic positions and orientation, as well as the similar sizes and splicing patterns of exons 2 to 7, that they are evolutionarily related genes.

	ADDENDUM

We are aware that A. Davison (6a) has made similar independent predictions about the ORF-K15 structure, with supporting evidence from cDNA analysis carried out by Glenn et al. (12a).