INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The family Retroviridae consists of seven genera of avian and mammalian viruses that are grouped by common morphology, genomic structure, and host range (14). Two piscine retroviruses that are not congruous with the classification of these genera have been cloned and analyzed: walleye dermal sarcoma virus (WDSV) and snakehead fish retrovirus (SnRV) (20, 22), both of which are complex viruses that have type C morphology (18, 37). WDSV is etiologically associated with a nodular skin lesion, walleye dermal sarcoma (WDS) (36, 52), whereas the pathogenicity of SnRV is unknown (20). Based on pol sequences and genomic organization, WDSV and SnRV are only distantly related, making it difficult to judge if either is representative of fish retroviruses in general.

Recently, we identified two novel retroviruses associated with walleye epidermal hyperplasia (WEH) (52, 57), WEH virus types 1 and 2 (WEHV1 and WEHV2, respectively) (31). WEHV1 and WEHV2 are type C retroviruses that are closely related to one another (76% amino acid [aa] identity in pol) and to WDSV (ca. 65% aa identity in pol) (22, 31). Interestingly, like WDS, epidermal hyperplasia appears and regresses on a seasonal basis. Lesions develop in the fall, regress in the spring, and are absent in the summer (7, 8). Evidence suggesting that WEHV1 and WEHV2 are the causative agents of WEH includes the following: (i) WEHV1 and WEHV2 pol sequences were identified and cloned by reverse transcriptase PCR (RT-PCR) from virion preparations derived from hyperplastic tissue (31); (ii) WEHV1 and WEHV2 genomic DNA is found in diseased tissue, but not in surrounding normal tissue (31); and (iii) WEH has been experimentally transmitted to walleye fingerlings with cell-free inocula from hyperplasias, and WEHV1 and/or WEHV2 viral DNAs were detected in the lesions by PCR (6). While the two are commonly found together in lesions, WEHV2 has been detected in the absence of WEHV1, suggesting that WEHV2 alone can cause disease (6, 31). However, the relative capacities of WEHV1 and WEHV2 to cause disease have not been investigated.

Members of our group recently found that WEHV1, WEHV2, and WDSV encode cyclin D homologs (30). Since cyclin D1 is an oncogene that has been implicated in many types of human tumors, the retroviral cyclins (rv-cyclins) may induce the cell proliferation that ultimately leads to WEH and WDS (1, 30). As part of our efforts to develop this unique model of tumor induction and tumor regression, we have completed the DNA sequences of WEHV1 and WEHV2 and have identified their transcripts in developing and regressing tumors. The comparisons of the DNA sequences and transcriptional patterns of WEHV1, WEHV2, and WDSV emphasize the relatedness of these viruses and suggest that they use similar replication strategies that likely play a role in pathogenesis. In addition, the sequence analyses suggest that two of their accessory genes, orfA (rv-cyclin) and orfB, arose by gene duplication. This represents the first example where the origin of complex retrovirus accessory genes can be traced to a cellular gene.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Characterization of WEHV1 and WEHV2 genomic clones. All probes were labeled with ³²P by using a random priming kit (Boehringer).  DNAs were prepared, digested with restriction enzymes, electrophoresed on 1% agarose gels, and blotted onto nitrocellulose by standard procedures (46). All blots were hybridized with probes in 50% formamide buffer at 37°C for 24 h and washed in 0.2× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) at 55°C. Fragments for subcloning were gel purified using a Qiaex II gel extraction kit (Qiagen) and were cloned into Bluescript II SK() (Stratagene).

DNA sequencing. Plasmids for sequencing were purified with a Qiagen Maxi Prep kit. Both DNA strands of the viral clones were sequenced by walking using automated fluorescent sequencing with rhodamine dye terminator chemistry at the Cornell University Bioresource Center DNA Sequencing Facility. cDNA clones from transcriptional mapping were sequenced with the Sequenase sequencing kit, version 2.0 (U.S. Biochemicals).

Database searches and amino acid alignments. Database searches were done by the BlastX algorithm (National Center for Biotechnology Information). Amino acid alignments were done with Megalign (DNAstar) and Geneworks (IntelliGenetics, Inc.). The amino acid substitutions allowed for determination of sequence similarities were as follows: F = Y = W, M = L = V = I, S = A, S = T, D = E, N = Q, and R = K = H.

Northern blotting. Total RNA was isolated from WEH samples collected from individual fish in the spring (0.2 g) with RNAzol B (Tel-test, Inc.). Fall hyperplasias were difficult to obtain, and therefore lesions from 2 to 3 fish were pooled to produce enough material (1 g of tissue) for each of two independent poly(A)⁺ RNA isolations. poly(A)⁺ RNA was enriched from fall lesions with the PolyATract IV mRNA Isolation System (Promega). Spring and fall lesions were not derived from the same fish. Ten micrograms of total RNA or 1 µg of poly(A)⁺ RNA was electrophoresed in formaldehyde gels and blotted onto nitrocellulose (46). The blots were hybridized by using WEHV1 or WEHV2 virus-specific LTR probes to detect all viral transcripts. WEHV1 or WEHV2 LTR probes were made and hybridized for 48 h as described above. Blots were washed in 0.2× SSC at 55°C and exposed to film for 24 h (spring lesions) or 1 week (fall lesions).

Transcriptional mapping. Transcriptional mapping was done by RT-PCR; the oligonucleotides used are listed in Table 1 and shown schematically in Fig. 1, panels c. Total RNA was isolated from three or more spring WEH lesions derived from individual fish for the analysis of WEHV1 and WEHV2 transcripts. cDNA synthesis (50-µl reaction mixture) was done with 1 µg of total RNA, 10 pmol of antisense primer (U3-1 RT or U3-2 RT), and 4 U of murine leukemia virus (MLV) reverse transcriptase (New England Biolabs). The cDNA (2 µl) was amplified by PCR (50-µl reaction mixture) with Taq polymerase (Gibco-BRL). All samples were denatured at 96°C for 5 min, and 2.5 U of Taq polymerase was added to the reaction mixture. The amplification program consisted of 30 s at 94°C, 30 s at 50°C, and 2 min at 72°C for 35 cycles and 10 min at 72°C for 1 cycle. The PCR products were extracted with phenol-chloroform and precipitated with ethanol. WEHV1 products digested with XhoI and NotI and WEHV2 products digested with XhoI and SacI were cloned into Bluescript II SK(). Some products were gel purified prior to cloning.

Start of transcription. 5' rapid amplification of cDNA ends (RACE) (Life Technologies) was used to identify viral transcriptional start sites. cDNA was synthesized (50-µl reaction mixture) with 1 µg of total RNA, 2 pmol of gag-specific downstream primers for WEHV1 (5'-AGATCCAGGCCATCCTGCTAA-3') and WEHV2 (5'-ATGTTCTGCTATCTGATCTC-3'), and 4 U of MLV RT. The cDNA was purified as previously described (31). Two microliters of the cDNA was amplified by PCR with an upstream GI anchor primer (Gibco-BRL) and downstream primers in orfC for WEHV1 (5'-ATTAGCGGCCGCACCATGGCCTGGAACAAAAAGCAT-3') and WEHV2 (5'-CTCTCTAGATTACTAAAAGGTATTTCCGGTTTG-3') (engineered restriction sites are underlined). The samples were denatured for 5 min at 96°C prior to the addition of 2.5 U of Taq polymerase and amplified as follows: 94°C for 30 s, 50°C for 30 s, and 72°C for 2 min. The WEHV1 and WEHV2 PCR products were sufficiently abundant to be gel purified and cloned into Bluescript II SK() with SalI-NotI and SalI-XbaI, respectively.

Nucleotide sequence accession numbers. The DNA sequences of WEHV1 and WEHV2 were deposited in GenBank under accession no. AF133051 and AF133052, respectively.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Sequence analysis of WEHV1 and WEHV2. The complete DNA sequence for each virus was obtained from a single pair of 5' and 3' overlapping subclones (Fig. 1, panels b). The sequences of the 5' and 3' LTRs and overlapping pol regions showed little variation. Therefore, we assume that the sequence analyses reported here are representative of the complete WEHV1 and WEHV2 genomes. The sequence analyses showed that the genomic organizations of WEHV1 and WEHV2 are similar to that of WDSV; each virus contains three open reading frames in addition to the genes found in all retroviruses, gag, pol, and env (Fig. 1, panels a). The sizes of the WEHV1 and WEHV2 proviruses are 12,999 bp and 13,125 bp, respectively, and the analyses of their genomes are described below.

View larger version (25K): [in this window] [in a new window]   FIG. 1.   WEHV1 and WEHV2. For each, panels a to d are as follows. (a) Genomic organization. The numbers mark the boundaries of the open reading frames and of the LTRs relative to the start of transcription (+1). (b) Schematic of the subclones used for DNA sequence analysis. The numbers mark the ends of the overlapping subclones. (c) Primers used for RT-PCR. (d) Transcriptional profiles. The positions of the major splice donor and SAs relative to the start of transcription are indicated.

LTRs. The boundaries of the 5' LTRs were identified by comparison with their 3' LTR sequences. The WEHV1 and WEHV2 LTRs are shown in Fig. 2. The boundary of the U3 promoter region and R is defined as the start site for transcription. The WDSV transcriptional start site begins with the sequence GTCTCA (22). This sequence is conserved in WEHV1 and WEHV2, and by 5' RACE we confirmed that the G of this sequence is the transcriptional start site (position +1) for WEHV1 and WEHV2 (data not shown). TATA boxes are located at positions 31 and 30 in the WEHV1 and WEHV2 U3 regions, respectively. The boundaries of U3, R, and U5 for WEHV1 and WEHV2 were predicted as was done previously with the WDSV LTR (22). The WEHV1 LTR is 643 bp in length, with U3, R, and U5 regions of 503, 81, and 59 bp, respectively (Fig. 2A). The WEHV2 LTR is 550 bp in length, with U3, R, and U5 regions of 410, 79, and 61 bp, respectively (Fig. 2B). The consensus polyadenylation signal sequence, AATAAA, occurs at position 61 in both viruses, and the putative 3' ends of R for WEHV1 and WEHV2 are located at positions 81 and 79, respectively.

View larger version (38K): [in this window] [in a new window]   FIG. 2.   WEHV1 (A) and WEHV2 (B) LTRs. The predicted boundaries of U3, R, and U5, along with the start of transcription (+1), are indicated. Potential transcription factor binding sites are shown. A 7-bp repeat sequence, which is identical to the core sequence of the growth hormone factor 1 binding site, is shown. pA, polyadenylation signal sequence; PEA3, polyoma enhancer activator; GRE, glucocorticoid response element. The lowercase letters represent nucleotides in the PBS region that are conserved in WEHV1, WEHV2, and WDSV.

PBSs for synthesis of proviral DNA. Like WDSV, WEHV1 and WEHV2 are predicted to use a histidyl-tRNA (tRNA^His) primer for initiation of minus-strand DNA synthesis (22). The nucleic acid sequence flanking the PBS is conserved in WEHV1, WEHV2, and WDSV (Fig. 2). A polypurine tract, AAAAGGGG, presumably used for priming plus-strand DNA synthesis, is located upstream of the 3' LTR, and its sequence is conserved in WEHV1 (position 11842), WEHV2 (position 12154), and WDSV (22). Like spumaviruses and lentiviruses, WDSV, WEHV1, and WEHV2 have two additional polypurine tracts in the 3' portion of pol (13, 22, 28). The sequence CAAAGGGGG is found in WEHV1 pol at positions 4427 and 5028 and in WEHV2 pol at positions 4345 and 5081. One of the polypurine tracts in WDSV pol has this sequence, and the other varies in two positions (22).

Leader region and orfC. The leader sequence of retroviruses is usually defined as extending from position +1 to the initiation codon of gag. By this definition, the predicted lengths of the leader sequences for WEHV1 and WEHV2 are 910 and 861 bp, respectively. Like WDSV, WEHV1 (position 494) and WEHV2 (position 452) each have a large open reading frame (orfC) found in the leader region (22). The predicted OrfC proteins of WEHV1 and WEHV2 are 134 aa in length, share 46% aa identity, and have 25 and 30% aa identity with the WDSV OrfC, respectively (Table 2). All three OrfC proteins are very basic. Notably, several tryptophan (W) residues are conserved in the OrfC proteins of WEHV1, WEHV2, and WDSV (Fig. 3). Four of the Trp residues have a striking periodicity (WX₄₁WX₃₆WX₄₁W) that is more evident for the WEHV1 and WEHV2 OrfC proteins than for the WDSV OrfC, which has a tyrosine in place of the second conserved Trp (WX₄₂YX₃₆WX₃₃W). Additionally, many of the hydrophobic residues in the X regions are also conserved. Interestingly, the spacing of the Trp residues is reminiscent of the WW domain that is found in some signaling, regulatory, and cytoskeletal proteins (4, 5). The WW domain is believed to play a role in mediating protein-protein interactions with proteins containing proline-rich regions (4, 5). Recently, it was shown that the Rous sarcoma virus Gag protein late domain interacts at the plasma membrane with the WW protein, Yes-associated protein, in a process that may be important to viral budding (19). However, since we were unable to detect any obvious homology to proteins in the databases, we cannot assign a putative function to the OrfC proteins.

View larger version (23K): [in this window] [in a new window]   FIG. 3.   Alignment of the WEHV1, WEHV2, and WDSV predicted OrfC proteins. The conserved tryptophan (W) residues comprising a possible WW domain are marked with asterisks. The basic region near the C terminus of these proteins is bracketed.

gag. The amino acid sequences of the WEHV1 and WEHV2 gag open reading frames were aligned with that of WDSV Gag to identify putative cleavage products. The N termini of the WDSV virion proteins MA, CA, p10/p20, and NC have been determined (22). The MA proteins of WEHV1 (position 910), WEHV2 (position 861), and WDSV begin with the amino acid sequence MGN and are predicted to be myristylated, as shown schematically in Fig. 4. WEHV1 and WEHV2 share 58% aa identity in MA and they share 23 and 27% aa identity, respectively, with WDSV MA (Table 2).

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Schematic of the WEHV1, WEHV2, and WDSV Gag polyproteins. MA, matrix; CA, capsid; NC, nucleocapsid; MHR, major homology region. The Gag proteins are predicted to be myristylated at a glycine residue located at the penultimate amino acid residue. p10/p20 is the WDSV protein located between the MA and CA proteins, and p? represents the hypothetical proteins for WEHV1 and WEHV2 in the same region. Underlines denote signature residues. The GR-rich and proline-glutamine (PQ)-rich regions are shown with slashed boxes.

pol. The WEHV1 and WEHV2 genomic 5' and 3' subclones overlap in the pol region. The WEHV1 pol overlap is 3,105 bp and the WEHV2 pol overlap is 1,516 bp (Fig. 1, panels b). Sequence analysis showed that the pol overlap regions in the WEHV1 5' and 3' subclones are identical. The WEHV2 pol overlap regions in the 5' and 3' subclones differ by 18 nucleotides, resulting in three amino acid substitutions (31). Like WDSV and MLV, the gag and pro-pol genes of WEHV1 and WEHV2 are in the same translational frame, and suppression of the amber stop codon at the end of gag is likely responsible for the translation of pro-pol (22, 59). pro-pol is the most conserved gene among WEHV1 (position 2515), WEHV2 (position 2430), and WDSV, and the alignment of the three pol genes has been previously reported (31). The protease signature sequence (LVTGA) for WEHV1 and WEHV2 is located at positions 2599 and 2514, respectively. By analogy with other retroviruses, the N terminus for RT is expected to begin 10 to 20 aa downstream of the GRD (position 2809 for WEHV1 and position 2724 for WEHV2) sequence in the protease (45). As in other retroviruses, a putative zinc finger motif is present in the integrase proteins (24): HGVSH at positions 5008 and 5061, followed by CX₂C (underlines denote signature residues) at positions 5110 and 5163 for WEHV1 and WEHV2, respectively. The RT-RNase H and RNase H-integrase boundaries cannot be predicted from the amino acid sequence alignments.

env. The N terminus of the WDSV transmembrane protein (TM) and, therefore, the C terminus of the extracellular surface protein (SU) were determined previously by amino acid sequencing (22). The sequences of the WEHV and WDSV Env proteins are sufficiently similar to allow the WEHV1 and WEHV2 SU and TM components be identified. The SUs of WEHV1 (position 5910), WEHV2 (position 5957), and WDSV are not highly conserved; the WEHV1 and WEHV2 SUs share 36% aa identity, and each shares 19% aa identity with WDSV (Table 2). The SUs of WEHV1 (711 aa) and WEHV2 (739 aa) are substantially larger than that of WDSV (470 aa) (Fig. 5). The N termini of the WEHV1, WEHV2, and WDSV SUs contain a similar hydrophobic sequence of 14 aa that may be the hydrophobic core of the secretion signal peptide (Fig. 5). Analogous to the WDSV SU-TM boundary, the putative WEHV1 and WEHV2 SU-TM boundaries are marked by a consensus proteolytic cleavage site, RX(R/K)R.

View larger version (14K): [in this window] [in a new window]   FIG. 5.   Schematic of WEHV1 and WEHV2 (A) and WDSV (B) envelope proteins. SU, extracellular surface protein; TM, transmembrane protein; TMA, transmembrane anchor; CD, cytoplasmic domain of the TM. A possible hydrophobic leader peptide is shown at the N terminus of the SU. The cleavage site between the WEHV SU and TM proteins was predicted based on the N-terminal amino acid sequencing of the WDSV TM. The sequences of the TMAs as predicted by hydrophilicity plots are shown. The hydrophobic tail of the WDSV TM is shown by a hatched box.

3' orfs. By analogy with other complex retroviruses, we presume that the open reading frames located downstream of env encode proteins that may contribute to the regulation of viral gene expression and/or replication. The orfA genes of WEHV1 (position 10081), WEHV2 (position 10277), and WDSV were found to encode cyclin D homologs (233, 231, and 297 aa in length, respectively), and their relationships to each other and to cellular cyclins have been discussed (30). The similarity of the rv-cyclins and known cyclins is restricted to the conserved cyclin box motif (3). Analogous to cellular cyclins, the rv-cyclins may promote cell cycle progression (47), thereby stimulating the cell division needed for viral replication. It is also likely that the rv-cyclins induce the cell proliferation that ultimately leads to WEH and WDS (30).

View larger version (87K): [in this window] [in a new window]   FIG. 6.   Alignment of the WEHV1, WEHV2, and WDSV OrfA and OrfB proteins within the cyclin box region. Shaded regions 1 through 12 represent areas of greatest similarity between the OrfA and OrfB proteins. The amino acid substitutions allowed are listed in Materials and Methods. The amino acids shown in bold represent the following: (i) those that are similar in four or more of the OrfA or OrfB proteins (i.e., excluding human cyclin D1) or (ii) those that are similar in an OrfA or OrfB pair within a single virus (between solid lines). Both conditions may be shown in a single column. Human cyclin D1 is included to show the relationship of the OrfAs and OrfBs to cyclin proteins.

Analysis of viral gene expression by Northern blotting and transcriptional mapping. Previous Northern blot analyses have shown that abundant levels of WDSV viral RNA (full-length and subgenomic transcripts) are present in regressing dermal sarcomas (spring) and that low levels of WDSV viral RNA (predominantly orfA and orfB) are present in developing sarcomas (fall) (11, 44). Northern blot analysis with LTR-specific probes showed a similar expression pattern for WEHV1 and WEHV2 (Fig. 7). WEHV2 RNA was present at levels much lower than WEHV1 RNA in the fall samples and was not detected in one of the samples (Fig. 7, lane 3). A low level of a 2.6-kb transcript, which was shown earlier to hybridize with an orfA probe (30), was observed in WEHV fall lesions (Fig. 7, lanes 1, 2, and 4). In contrast, abundant levels of WEHV1 and WEHV2 genomic and subgenomic transcripts (~13, 7.0, 2.6, and 1.8 kb) were detected in spring hyperplasias collected in different years (Fig. 7, lanes 5 to 9). The sizes of the transcripts corresponded to those predicted for genomic RNA, env, orfA, and orfB transcripts, respectively. Notably, one sample collected in the spring contained abundant levels of the orfA and orfB transcripts but only low levels of full-length and genomic RNA (Fig. 7, lane 7). Qualitatively, this pattern is similar to that observed for developing hyperplasias and dermal sarcomas. This sample was prepared from a fish that was 80% covered with hyperplasias (data not shown). The significance of this observation is not known. Based on the intensity of hybridization, amount of RNA loaded, and exposure time, we estimate that there is 10- to 50-fold more of the 2.6-kb transcript (orfA) in the spring lesions than in the fall lesions (Fig. 7, compare lanes 1 and 2 with lanes 5 and 6). The differences in the patterns and levels of RNA expression between spring and fall lesions are similar to those observed for WDSV, for which there is at least 100-fold more of the 2.8-kb message (orfA) present in spring than is present in fall lesions (44).

View larger version (56K): [in this window] [in a new window]   FIG. 7.   Northern blot analysis of WEHV1 and WEHV2 viral transcripts in fall and spring lesions collected in different years, using LTR-specific probes. poly(A)+ RNA (1 µg) from fall lesions was hybridized with WEHV1 (lanes 1 and 2) or WEHV2 (lanes 3 and 4) LTR probes. Lesions from two to three fish were pooled to generate each of the two RNA samples represented in lanes 1 to 4. Total RNA (10 µg) from spring lesions isolated from individual fish was hybridized with WEHV1 (lanes 5 and 6) and WEHV2 (lanes 7 to 9) LTR probes; RNA shown in lane 7 was isolated from lesions obtained from a fish with a severe case of epidermal hyperplasia; i.e., 80% of its body was covered with the disease.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The sequence analyses of WEHV1, WEHV2, and WDSV have shown them to be closely related, with characteristics in common with several retrovirus genera: spumaviruses (large size, location and structure of 3' orfs, and multiple polypurine tracts); lentiviruses (long cytoplasmic domain in Env and multiple polypurine tracts); and mammalian type C simple retroviruses (type C particle morphology, a single Cys-His box in NC, and predicted amber suppression for translation of the Gag-Pol polyprotein). Additionally, these viruses are unique among the Retroviridae: they use a histidyl-tRNA as the primer for first-strand synthesis; they encode a large open reading frame in the leader region upstream of gag (orfC); they encode cyclin homologs (orfA) that may be involved in the induction of cell proliferation (30); and they encode accessory genes, orfA and orfB, that likely arose by gene duplication and whose origin can be traced to a cellular gene.

The leader sequence at the 5' end of genomic retroviral RNA is required for RNA dimerization, RNA encapsidation, and initiation of reverse transcription (34). In some cases, this region encodes peptides that may be required for viral replication. For example, the avian leukemia and sarcoma viruses encode three short peptides in their leader sequences that are important for packaging genomic RNA (16). By comparison, each of the leader sequences of the walleye retroviruses harbors an open reading frame, orfC, that encodes a protein of approximately 14 kDa. Although we cannot infer a function based on sequence analysis, the conservation of the OrfC proteins strongly suggests that they are biologically significant to the viruses. Alternatively, or possibly in parallel, we cannot rule out the possibility that an underlying RNA secondary structure in this region is responsible for conservation of orfC. Interestingly, the OrfC proteins contain tryptophans at regularly spaced intervals, similar to those found in the protein-binding module, the WW domain (48). WW domain-containing proteins are a diverse group of cellular proteins that bind ligands containing a PPPPY (PY motif) or PY-like motifs (42). The WW domain of the signaling protein, Yes-associated protein, binds to the PPPPY late assembly domain of Rous sarcoma virus in vitro, suggesting that cellular WW domain-containing proteins may be involved in the budding process (19). It is tempting to speculate that the tryptophan repeats found in the OrfCs form a WW domain that could mediate protein interactions with Gag at the plasma membrane necessary for viral budding, but we have no data suggesting that this is the case.

If orfC is translated, it is unclear how the Gag proteins are made by the walleye retroviruses. Simple stop codon suppression or frameshifting mechanisms are not likely to produce an OrfC-Gag polyprotein because different stop codons and spacing between orfC and gag exist in the three viruses and these mechanisms seem too inefficient to produce a sufficient amount of Gag for virion assembly. It is possible that Gag is translated from a spliced mRNA from which orfC has been removed, but we have failed to identify such a spliced transcript by RT-PCR (29). It has been shown that a cap-independent internal ribosomal entry mechanism can be used by MLV for Gag and Gag-Pol polyprotein translation (4, 27). It is possible that a similar mechanism could be used by the walleye retroviruses to produce Gag, but this remains to be tested.

The envelope proteins of the walleye retroviruses are quite different from those of other retroviruses. In the most general view, the envelope proteins are much larger, with the TM proteins of WEHV1 and WEHV2 (ca. 600 aa) and WDSV (755 aa) being two to four times larger than the TMs of prototypical retroviruses like MLV and human immunodeficiency virus (HIV) (ca. 170 aa and 345 aa, respectively). Previously, it was suggested that the long hydrophobic tail at the C terminus of the WDSV TM protein was the transmembrane anchor and that WDSV Env lacked a cytoplasmic domain (22). We have identified a small hydrophobic region (26 aa) in the WEHV1, WEHV2, and WDSV TM proteins that is similar in sequence and position in all three viruses. Since the WEHVs lack the hydrophobic tail at the C terminus of Env, we predict that this is the transmembrane anchor for the three TM proteins. The cytoplasmic domains for the WEHVs (ca. 200 aa) and WDSV (ca. 350 aa, including the 65-aa hydrophobic tail) are longer than those observed for any other retrovirus (40). Additionally, WDSV is the only retrovirus whose cytoplasmic domain ends in a long hydrophobic tail that may be buried within the protein or imbedded in the cell membrane (22). The biological significance of the unusual envelope proteins encoded by these viruses is not known, but it is speculated that they play some role in stabilizing the virus in an aquatic environment. It is likely that WDS and WEH are horizontally transmitted during the walleye spring spawning run when viral particles are shed from the lesions (11). Experimental transmission of WDS to walleye fingerlings has been achieved by injection of cell-free tumor filtrates, topical application to fingerlings, and oral lavage, suggesting that contact and/or the oral ingestion of tumors is a route of natural transmission (9, 37). Recently, efficient experimental transmission of WDS was observed when tumor-negative fish were housed in flowing stream water that harbored tumor-positive fish upstream, indicating that virions remain infectious in water (10).

By analogy with other complex retroviruses, we suggest that the proteins encoded by the orfA and orfB transcripts provide accessory or regulatory functions for the walleye viruses (15). The orfA genes of WEHV1, WEHV2, and WDSV encode cyclin D homologs that are likely to play a central role in the induction of cell proliferation observed in developing lesions and may be necessary for viral replication (30). While inducing cell proliferation may be the central function of the rv-cyclins, it is also possible that the rv-cyclins contribute to other aspects of viral replication, e.g., gene regulation. Independent of cell cycle functions, cyclins have been shown to be important in transcription (35, 60). A potentially relevant example of a cyclin regulating viral gene expression is the novel Cdk9-cyclin T complex that is believed to activate HIV transcription by acting as a cellular cofactor for binding of the HIV Tat protein to the TAR sequence (53, 58).

While the OrfB proteins have sequence similarities with the D cyclins and the rv-cyclins within the cyclin box, we can only speculate about their biochemical properties. It seems unlikely that they interact with cellular cyclin-dependent kinases because they lack the critical lysine and glutamate residues in -helices C and E that are necessary for cyclin-dependent kinase binding (12, 23, 31). However, because the amino acid sequences of the cyclin box can vary significantly between cyclin subfamilies (3), it is possible that all or some of the cyclin box tertiary structure has been retained in OrfB. Notably, two important cellular proteins have structural similarity with cyclins: the general transcription factor TFIIB contains a partial cyclin box and its crystal structure is remarkably similar to that of cyclin A (2), and the retinoblastoma tumor suppressor protein (Rb) has two cyclin box motifs that form the Rb pocket, which contains several potential binding sites for cellular and viral proteins (25). The N termini of the OrfB proteins contain polyproline tracts that may be involved in protein-protein interactions, and their C termini contain clusters of acidic residues analogous to those found in VP-16 of herpes simplex virus and Taf of simian foamy virus (38, 43, 54).

The similarity, albeit limited, of the OrfA and OrfB proteins and the adjacent positions of orfA and orfB in the genomes of the walleye viruses suggest that these genes arose by duplication (50). This is the second example of inferred gene duplication in a retrovirus, the first being that the vpx gene of the HIV type 2/simian immunodeficiency virus group arose by gene duplication of the vpr gene (50). The additional observations that orfA and orfB are distantly related to cellular cyclins and that the WDSV OrfA protein can rescue cyclin-deficient yeast from growth arrest suggest that they were derived from the host genome. It has been proposed that complex retroviruses evolved from simple retroviruses that acquired additional sequences from the host by transduction (39, 49), but sequence and functional data supporting capture and evolution of accessory genes from cellular genes in complex retroviruses have not been forthcoming (39). HIV/SIV Nef is the only example of an accessory protein that has sequence similarities to cellular proteins, including Ras and Src kinases, mammalian G proteins, and other cellular proteins, but the significance of this is unknown (41). In contrast, the finding that the WDSV OrfA protein complements cyclin deficiency in yeast clearly suggests that it was derived from a cellular gene with a similar function (30). Also, unlike the genes acquired by acutely transforming simple retroviruses, the sequences of orfA and orfB are very divergent from cellular cyclin sequences, suggesting that the capture of the cellular cyclin event occurred early in the emergence of the walleye retroviruses (30). Although convergent evolution cannot be excluded, we suggest that orfA and orfB are the first accessory genes of complex retroviruses whose origin is unambiguously cellular.

The pattern of gene expression observed for WEHV1, WEHV2, and WDSV in fall and spring tumors parallels the early and late phases of viral replication of complex retroviruses in cell culture; low levels of predominantly subgenomic transcripts (orfA and orfB) are observed in developing lesions collected in the fall, and high levels of genomic and subgenomic transcripts are observed in lesions collected in the spring. The transcriptional profiles of WEHV1 and WEHV2 are similar to that observed for WDSV, suggesting that they use similar strategies for replication and that the expression of specific viral genes correlates with different stages of their cognate diseases. The major difference between the profiles of the WEHVs and WDSV is that only one singly spliced transcript is predicted to encode each of the WEHV OrfA proteins, whereas several multiply spliced transcripts are predicted to encode the WDSV OrfA protein (44). Although a transcript of predicted size for WEHV2 orfB was identified by Northern blotting, we were unable to analyze this cDNA by RT-PCR and clone it. At present, we do not know why we were unable to isolate this cDNA; possibly the gene is unstable in Escherichia coli or the cDNA was underrepresented in the population of amplified products.

Previous phylogenetic analyses based on the conserved region of RT suggested that these retroviruses were most closely related to the mammalian type C (MLV-related) and spumavirus clades, but they are clearly divergent from both groups (31, 51). In addition, WEHV1, WEHV2, and WDSV are very different from SnRV, the only other fish retrovirus for which sequence is available (20). Based on the data presented herein, we suggest that WEHV1, WEHV2, and WDSV warrant consideration as a new genus in the family Retroviridae. Further characterization of these viruses will likely contribute to our understanding of retroviral biology, including viral replication, pathogenesis, and evolution.