Kaposi’s Sarcoma-Associated Herpesvirus Encodes a bZIP Protein with Homology to BZLF1 of Epstein-Barr Virus

Abstract

Kaposi’s sarcoma-associated herpesvirus (KSHV) is a recently discovered human gamma herpesvirus strongly implicated in AIDS-related neoplasms. We report here the identification in the KSHV genome of a gene for a protein designated K-bZIP and belonging to the basic-leucine zipper (bZIP) family of transcription factors. K-bZIP shows significant homology to BZLF1, which plays a key role in the replication and reactivation of Epstein-Barr virus. K-bZIP is a homodimerizing protein of 237 amino acids with a prototypic bZIP domain at the C terminus. The N-terminal portion of K-bZIP is derived from the K8 open reading frame which, through in-frame splicing, adjoins the ZIP domain. This structure was revealed by rapid analysis of cDNA ends, followed by cloning of the entire cDNA. A 1.35-kb transcript encoding K-bZIP was detected in BCBL-1 cells treated with 12-O-tetradecanoylphorbol-13-acetate. The synthesis of this transcript was blocked by the protein synthesis inhibitor cycloheximide but not by the viral DNA synthesis inhibitor phosphonoacetate, a result which classifies it as an early lytic gene. RNase protection analysis further mapped the major transcription start site for the 1.35-kb K-bZIP mRNA and identified two other splice variants which encode proteins with the N-terminal portion of K-bZIP but lacking the C-terminal ZIP domain. Full-length K-bZIP forms dimers with itself, and the C terminus encompassing the ZIP domain is required for this process. Our studies set the stage for understanding the role of K-bZIP in the replication and reactivation of the KSHV genome.

Kaposi’s sarcoma (KS)-associated herpesvirus (KSHV), the eighth type of human herpesvirus, was discovered in 1994 from a skin lesion of a Kaposi’s sarcoma patient (10). Phylogenic analysis of nucleic acid sequences placed KSHV in the lymphotropic gamma Herpesviridae family, showing significant homologies with herpesvirus saimiri and Epstein-Barr virus (EBV) (36). While HVS and EBV are considered oncogenic agents in primates (19, 32), definitive evidence for the tumorigenic potential of KSHV is lacking. However, a number of viral gene products, such as ORF K1, ORF K12 (kaposin), ORF K9 (vIRF), and ORF 72 (v-cyclin D), were shown to have mitogenic and transforming properties when overexpressed in certain cell types (11, 22, 29, 37). KSHV is also armed with several cellular homologues with immunomodulatory functions, including vIL6, vMIPs, and vGPCR (2, 6, 27, 35, 40). These gene products are likely to be involved in the progression of KS, a disease originating from uncontrolled paracrine signalings of vascular endothelium and spindle cells (15).

Although the presence of KSHV DNA has been repeatedly demonstrated in KS lesions, KS cell lines established in vitro usually do not harbor viral genomes (1, 18). However, various KSHV-infected human B-cell lines derived from primary effusion lymphomas are available for molecular studies (7, 8, 41). Complete sequences of the viral genomes from one such line and one KS biopsy specimen have been independently determined (38, 42). In the primary effusion lymphoma lines, most of the viral genes are not expressed, suggesting that the resident virus is predominantly in a latent state (33, 41, 43). The addition of phorbol esters or sodium butyrate to the culture medium activates the expression of viral lytic genes and results in the release of virus particles (28, 33). The identities of the KSHV target genes directly responding to stimulation by phorbol esters or sodium butyrate are not clear, nor is the gene expression cascade leading to the lytic phase. Nonetheless, for many other gamma herpesviruses, the viral immediate-early gene(s) responsible for the activation of lytic genes has been determined (13, 14, 39, 47–49). Among the notable examples is the BZLF1 (also known as ZEBRA, Zta, or EB1) product of EBV which, when overexpressed, can reactivate latent EBV, enabling it to enter the lytic cycle (14, 16, 30, 31). BZLF1 is also involved in the replication of EBV DNA in the lytic stage (17).

The genomic organizations of KSHV and EBV are similar in certain regions. By positional analogies (i.e., downstream of the BRRF2-BRRF1-BRLF1 complex), KSHV ORF K8 appears to be a homolog of BZLF1. Indeed, the N-terminal domain of ORF K8 shows some similarity to that of BZLF1. However, the leucine zipper (ZIP) motif, which is crucial to the function of BZLF1, is conspicuously missing from ORF K8. In addition, there is no canonical poly(A) signal within 1 kb downstream from ORF K8, and a potential splice donor site (44) can be identified immediately before the terminator UAG codon (nucleotide 75567). We therefore hypothesized that splicing may be involved in the generation of functional ORF K8. In this regard, it is noteworthy that the BZLF1 transcript also undergoes two splicing events, and the C-terminal domains are linked together (31).

Here, we report the successful cloning, by rapid analysis of cDNA ends (RACE) and reverse transcription (RT)-PCR, of multiply spliced transcripts encoding ORF K8 and the discovery of a prototypic ZIP domain encoded by one of the exons. Expression of these transcripts is absent in latent BCBL-1 cells but can be induced by phorbol esters. This induction is sensitive to cycloheximide but not to phosphonoacetic acid (PAA), a result which classifies these transcripts as early genes. The most abundant transcript yields a protein, designated K-bZIP, of 237 amino acids with a basic-ZIP (bZIP) motif. Functional analysis shows that K-bZIP forms homodimers. We have also mapped the transcriptional start site of the K-bZIP gene, which reveals the putative promoter sequence. Our studies provide a framework for studying the role of this protein in KSHV replication and the latency phase/lytic phase switch.

MATERIALS AND METHODS

Cell culture.

BCBL-1 cells (41) were grown at 37°C in RPMI 1640 supplemented with 10% fetal bovine serum in the presence of 5% CO₂. Virus replication was induced by the treatment of log-phase cells with TPA (12-O-tetradecanoylphorbol-13-acetate) (20 ng/ml) for various times. COS-1 cells (24) were maintained in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum in the presence of 5% CO₂. When transient expression was specified, BCBL-1 cells were electroporated with the desired plasmids by a standard protocol (3), while COS-1 cells were transfected with plasmids by use of Lipofectamine as described by the manufacturer (Gibco BRL, Gaithersburg, Md.). Forty-eight hours after transfection, cellular protein lysates or RNA was prepared and stored at −70°C until further use.

RACE.

The RACE assays were carried out essentially by the procedures reported by Frohman et al. (21) with the following modifications. Specifically, poly(A)⁺ RNA was purified by use of oligo(dT)-cellulose (type 7; Pharmacia, Piscataway, N.J.) spin columns. Poly(A)⁺ RNA (2.5 μg) then was primed with oligonucleotide SS-dT (CGTAGGTTACCGTATCGGATAGCGGCCGCATTTTTTTTTTTTTTTTTT) for 3′ RACE or dT₂₀ for 5′ RACE, and RT was carried out with avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim, Biochemicals, Indianapolis, Ind.) at 42°C for 1 h under the conditions specified by the manufacturer. The resulting first-strand cDNA was used directly as a DNA template in 3′ RACE. The DNA template for 5′ RACE was prepared by further treatment of the first-strand cDNA with 60 U of terminal deoxynucleotidyltransferase (Gibco BRL) in the presence of 0.2 mM dATP for 15 min at 37°C. Second-strand synthesis was then primed with primer SS-dT, and the mixture was incubated with 120 U of T4 DNA polymerase, 24 U of Escherichia coli DNA ligase, and 5 U of RNase H in a buffer containing 0.2 mM deoxynucleoside triphosphates, 100 mM KCl, 10 mM ammonium sulfate, 5 mM MgCl₂, 0.15 mM β-NAD, 20 mM Tris-HCl (pH 7.5), and 50 μg of bovine serum albumin per ml for 4 h at 15°C. The primers used for 5′ RACE were SS (CGTAGGTTACCGTATCGGATAG) and K8-AS (TTTTCCCCACCGTCAGTATTGTCC) (or K8-AS-N [CTTTCTCAGAATTGTCCGTTCCCG]). The primers used for 3′ RACE were SS and K8-S (AGAGGAACGCTTATGCACTAAGGC). Full-length cDNAs of K-bZIP were synthesized with K8FL-S1 (TTCCGAGACTGAAGTGTTCGCAAG) and K8FL-AS1 (GACAAGTCCCAGCAATAAACCCAC) or with K8FL-S2 (TGCCAAATGCCCAGAATGAAGGAC) and K8FL-AS1. PCR conditions for denaturation, annealing, and polymerization were 93°C for 50 s, 56°C for 1 min, and 68°C for 3 min for 5′ RACE and 93°C for 50 s, 62°C for 1 min, and 68°C for 3 min for 3′ RACE, respectively. High-fidelity Taq polymerase (Boehringer) was used in all the RACE assays, and 30 cycles of amplification were used for each reaction. PCR products were subsequently cloned into the pCR2.1-TOPO vector (Invitrogen, Carlsbad, Calif.). The DNA sequence of each insert was determined on both strands by the dideoxynucleotide chain termination method.

Plasmids.

Plasmid C1 is a pCR2.1-TOPO-based clone which contains the full-length cDNA of the K-bZIP coding region generated by RT-PCR. Inserts of C1 were transferred to pcDNA3.1 (Invitrogen) derivatives to introduce either the hemagglutinin (HA) tag (MGYPYDVPDYASGP) or the T7 tag (MASMTGGQQMGGP) to the N terminus. The resulting plasmids were denoted pHA-KBZIP and pT7-KBZIP, respectively. pBS-P2 was obtained by cloning a PCR fragment spanning nucleotides 74745 to 74947 of viral DNA (42) into pBluescript KS(+) (Stratagene, La Jolla, Calif.) with XbaI and SacII as cloning sites. pBS-IVS was constructed by cloning a PCR fragment spanning nucleotides 75376 to 75621 of viral DNA into pBluescript KS(+) with BamHI and SacI as cloning sites. All the plasmids were sequenced on both strands to confirm that no mutations were introduced during the cloning process.

RNase protection assay.

XhoI-linearized pBS-P2 or pBS-IVS was used as a template for the synthesis of [α-³²P]UTP-labeled antisense RNA probes by use of a MAXIscript kit (Ambion, Austin, Tex.). Fifty nanograms of probe (specific activity, 2 × 10⁶ cpm/μg) was hybridized to 15 μg of total RNA from BCBL-1 cells treated with TPA for 20 h. RNase protection reactions were carried out by use of a HybSpeed RPA kit (Ambion). Protected RNA fragments were resolved on a 6% polyacrylamide–8 M urea denaturing gel. A sequencing reaction containing ddA, ddC, and ddG of a DNA fragment with a known sequence (chicken CSK gene) was run in parallel to the sample, providing size markers.

Northern blot analysis.

Cells were lysed in TRIZOL reagent (Gibco BRL), and total cellular RNA was isolated as specified by the manufacturer. Twenty micrograms of total RNA from each sample was separated by electrophoresis through a 6% formaldehyde–1% agarose gel and blotted overnight onto a nylon membrane (Nytran; Schleicher & Schuell) by standard procedures (3). DNA probes used in Fig. 4 were either PCR amplified or obtained by digestion with restriction enzymes from genomic subclones of KSHV DNA. Detailed genomic locations of each probe were as follows: K8, nucleotides 74850 to 75104; and K8.1, nucleotides 75905 to 76207.

FIG. 4.

Expression kinetics of K-bZIP in BCBL-1 cells. BCBL-1 cells were treated with TPA for different times as indicated. For the 12- and 48-h time points, duplicate cell cultures were prepared and additionally treated with cycloheximide (CH, 100 μg/ml) or PAA (100 μM), respectively. Total RNAs were extracted at the end of the treatments. Twenty micrograms of RNA from each sample was loaded in each lane and transferred to a nylon membrane after electrophoresis. (A) The filter was hybridized with a K-bZIP-specific probe (nucleotides 74850 to 75104). (B) The filter was hybridized with an ORF K8.1-specific probe (nucleotides 75905 to 76207). RNA loading was assayed by hybridizing the same filter with DNA encoding H1 RNA of human RNase P (5) (bottom panel).

Immunoprecipitation and Western immunoblot analysis.

Immunoprecipitation was performed as described previously (25). Briefly, cells were rinsed in ice-cold phosphate-buffered saline and lysed in radioimmunoprecipitation assay (RIPA) buffer with protease inhibitors (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% sodium dodecyl sulfate [SDS], 1 mM phenylmethylsulfonyl fluoride [PMSF], 1 μg of pepstatin A per ml, 0.2 U of aprotinin per ml, 0.5 μg of leupeptin per ml). Two hundred micrograms of cell lysate was cleared by centrifugation and mixed with 1 to 2 μg of monoclonal antibodies against HA (BAbCO, Richmond, Calif.) or T7 (Novagen, Madison, Wis.) peptides for 2 h at 4°C with gentle rotation. The immunocomplex was then captured by the addition of a protein A-protein G-Sepharose mixture (Zymed, South San Francisco, Calif.) and rocking for an additional 2 h. Beads were washed three times in RIPA buffer and then boiled for 10 min in 60 μl of 2× SDS sample buffer (125 mM Tris-Cl [pH 6.8], 4% SDS, 2 mM EDTA, 20% glycerol, 0.6% bromphenol blue). Protein samples from total cell lysates or immunoprecipitations were resolved by SDS–10% polyacrylamide gel electrophoresis and then transferred to polyvinylidene difluoride membranes (Biotechnology System, Boston, Mass.) by use of a semidry apparatus (Pharmacia). After being blocked in TBST (20 mM Tris-HCl [pH 7.6], 137 mM NaCl, 1% Tween 20)–5% bovine serum albumin for 1 h at room temperature, the filters were incubated with primary antibodies (1:1,500 dilution) for 2 h at room temperature. The filters were subsequently washed with TBST three times for 10 min each time. The filters were then incubated with horseradish peroxidase-conjugated goat anti-mouse antibodies (1:2,000 dilution) for 1 h, washed three times with TBST, and developed with enhanced chemiluminescence (ECL) reagents (Amersham, Arlington Heights, Ill.).

DNA and protein sequence analyses.

DNA sequences of RACE clones were compiled, aligned, and analyzed with MacVector (version 6.0) software. Deduced amino acid secondary structures were analyzed by use of Chou-Fasman, Kyte-Doolittle, or Robson-Garnier programs. A comprehensive collection of bZIP transcription factor families was obtained from a database of protein families, Pfam (45). Alignment of consensus sequences among different bZIP proteins was performed by ClustalW analysis.

Nucleotide sequence accession number.

The cDNA sequence of K-bZIP described in this study has been deposited in GenBank under accession no. AF072866.

RESULTS

Major transcripts originating from the ORF K8 gene are spliced.

Previously, we noted that ORF K8 shows homology with EBV BZLF1 but lacks a canonical ZIP domain. We hypothesized that the ORF K8 gene encodes only part of the protein, with the remaining portion being connected by in-frame splicing(s). To investigate this possibility, the RACE approach was used to obtain cDNAs containing both ends of the transcript encompassing the ORF K8 gene. The use of antisense primer K8-AS in the 5′ RACE reaction generated a PCR product with an estimated size of 150 bp (Fig. 1B), while a nested primer, K8-AS-N, produced a smaller band of about 75 bp. The size difference of these two PCR products correlated well with the difference in the map locations of these two primers (compare Fig. 1A and Fig. 1B). Both bands were cloned, and five transformants each were randomly selected for sequencing.

FIG. 1.

RACE analyses of KSHV ORF K8 transcripts. (A) Schematic drawings of the genomic locations of ORF K8 and RACE primers used in this study. The nucleotide coordinates in parentheses are from reference 42. Primer orientation is depicted by arrows. The sketch is not drawn completely to scale. AATAAA, polyadenylation signal; ORF, open reading frame; SA, splice acceptor; SD, splice donor; TR_L: terminal repeat of left end; TR_R, terminal repeat of right end. (B to D) Agarose gel (1%) electrophoresis of RACE products. (B) 5′ RACE with K8-AS-N or K8-AS as 3′ primers. (C) 3′ RACE with K8-S as a 5′ primer. (D) Full-length cDNAs of K-bZIP amplified by K8FL-S1 and K8FL-AS1 (lane 1) or by K8FL-S2 and K8FL-AS1 (lane 2). Resulting PCR fragments were about 0.8 kb shorter than the sizes expected from the genomic sequence (1.2 kb rather than 2 kb in lane 1 and 1.0 kb rather than 1.8 kb in lane 2), suggesting that splicing events occurred.

Compilation of these sequences revealed a major transcriptional start site located 5 bp upstream of the first ATG of the ORF K8 gene (nucleotide 74850). The cDNA sequences in this region amplified by 5′ RACE were identical to the genomic sequence, indicating that no splicing was involved. However, a more complicated pattern was seen in the 3′ RACE reaction with sense primer K8-S (Fig. 1C). Three major bands of 1.4, 0.8, and 0.75 kb were seen, with the 0.75-kb band being the most intense. The presence of multiple bands suggested that splicing may have occurred in this region.

To confirm this notion, sequences from 15 individual 3′ RACE cDNAs were determined and compared to the genomic sequence. The results are summarized in Fig. 2A. Alignment of the cDNA sequences with the genomic sequence revealed the presence of three introns (IVS1, IVS2, and IVS3). Based on the presence or absence of these introns, three types of alternatively spliced transcripts could be discerned. Type I has all three intervening sequences (IVSs) spliced out, type II retains IVS2, and type III preserves both IVS1 and IVS2. All three transcripts used a canonical polyadenylation signal (AATAAA) located at nucleotide 76714. These results provide conclusive evidence that ORF K8 transcripts undergo splicing and that such splicing events result in the attachment of a potential ZIP domain to the ORF K8 protein (see below).

FIG. 2.

Sequence analysis of K-bZIP. (A) Summary of three types of cDNAs obtained by RACE cloning. Exons are represented as open boxes with roman numerals. IVSs between exons are represented as wavy lines. Peptides corresponding to each transcript are shown as solid lines. ∗, translation stop codon; +++, basic region. aa, amino acids. (B) Fine structure of the cDNA sequence of K-bZIP. Nucleotide sequences derived from the longest cDNA are boxed and are shown in uppercase. Introns are shown in lowercase, and consensus splice donor and splice acceptor sites are shown in bold. The 237 amino acids of K-bZIP deduced from three exons are depicted beneath the nucleotide sequences. The heptad repeat leucines and isoleucine are circled. Genomic coordinates (42) of the sequences are given at the left. Features discussed in the text are underlined: tga at 74627, translation stop codon of Rta/ORF 50; tataa at 74816, putative TATA box of K-bZIP; atg at 75915, translation initiation codon of ORF K8.1; AATAAA at 76714, polyadenylation recognition sequence. (C) Comparison of bZIP domain of K-bZIP to those of representative human bZIP transcription factors. Accession numbers for each sequence are as follows: CEBA (CCAAT/enhancer binding protein alpha, P49715); JUN (transcription factor AP-1, P05412); FOS (p55c-fos proto-oncogene protein, P01100); CREB (cyclic AMP response element binding protein, P16220). HHV8, eighth type of human herpesvirus. (D) Comparison of bZIP domain of K-bZIP to those of herpesvirus homologues: MEQ oncoprotein encoded by Marek’s disease virus (MDV) EcoQ fragment, A44083); BZLF1 (BZLF1 transactivator protein encoded by EBV, P03206). Amino acid sequences were aligned with the ClustalW program. Conserved leucine residues are marked by dots. Dark-gray shading indicates identical residues; light-gray shading indicates similar residues.

Molecular cloning of K-bZIP.

To ensure that the splicing schemes derived from the 5′ and 3′ RACE products were authentic, full-length cDNAs of the transcripts were synthesized by RT-PCR with primers derived, respectively, from the 5′- and 3′-terminal sequences (e.g., K8FL-S1 and K8FL-AS1). As shown in Fig. 1D, lane 1, a major band(s) of about 1.2 kb was amplified with primers K8FL-S1 and K8FL-AS1. When a nested primer, K8FL-S2, was used, a 1-kb PCR fragment was amplified (Fig. 1D, lane 2), as expected. These sizes are consistent with those of the spliced products (the unspliced transcripts would be 2 and 1.8 kb, respectively).

The entire sequence of type I cDNA superimposed on the genome sequence is shown in Fig. 2B. The protein coding sequences are contained within the first three exons. There are two basic regions, one in exon I and the other in exon II. A prototypic ZIP domain can be found in exon III. The basic region of exon II and the ZIP domain of exon III form a typical bZIP domain (discussed below). We have tentatively assigned the first methionine of the open reading frame as the putative initiation codon and, hence, amino acid 1. Amino acid 4 is also a potential initiator methionine, within a better Kozak context. There are two in-frame termination codons (Fig. 2A), one located in IVS2 (nucleotide 75567) and the other located within exon III (nucleotide 75789). Type I cDNA (Fig. 2A) encodes a protein of 237 amino acids. Because the transcript is fully spliced, it skips the termination codon in IVS2 and contains an intact bZIP domain. The type I cDNA product is therefore denoted K-bZIP. Type II cDNA retains IVS2, and the protein, 189 amino acids long, terminates within IVS2. Type III cDNA contains both IVS1 and IVS2 and encodes a protein of 239 amino acids, as was previously predicted from the genomic sequence of the ORF K8 gene (42). Both type II and type III cDNAs are predicted to encode proteins that carry the basic regions but not the ZIP domain.

bZIP domain of K-bZIP.

Perhaps the most interesting structural feature of K-bZIP is the presence of a heptad repeat of 4 leucines and 1 isoleucine, which constitutes the ZIP domain involved in dimerization. The ZIP domain is preceded by an arginine- and lysine-rich region which presumably functions to bind DNA. Compared to other known cellular bZIP proteins, such as Jun and Fos, K-bZIP shows similarity in the ZIP domain and, to a lesser extent, in the arginine- and lysine-rich region (Fig. 2C) (23, 26). The space between the basic region and the ZIP domain is invariable among the bZIP proteins, and K-bZIP is no exception. Figure 2D shows the alignment of the bZIP domain of K-bZIP with those of two other herpesvirus bZIP proteins, EBV BZLF1 and Marek’s disease virus Meq. The overall similarity between K-bZIP and BZLF1 is about 37%.

Mapping of the transcription start site of K-bZIP by an RNase protection assay.

The 5′ RACE results suggested a major transcription start site for all the transcripts. To precisely map the 5′ start site, plasmid pBS-P2 was constructed by the insertion of a DNA fragment extending from −105 to +98 (with the A of the first ATG set as 1) into a pBluescript vector. [α-³²P]UTP-labeled antisense RNA (relative to the K-bZIP gene) was transcribed from pBS-P2 in vitro and used as a probe in an RNase protection assay (Fig. 3A, lane 4). When the probe was incubated with RNAs from TPA-treated BCBL-1 cells and monitored by RNase digestion, protection of a distinct 103-nucleotide fragment was observed. An additional 203-nucleotide fragment, presumably derived from the readthrough transcript, was also detected. This protection was specific, since RNAs from uninduced BCBL-1 cells or with unrelated sequences (yeast RNA) did not give rise to such a signal (Fig. 3A, lane 3, and data not shown). As the schematic drawings in Fig. 3A show, the transcription start site of K-bZIP is located at nucleotide −5, a position which agrees well with the 5′ RACE data. In addition, a canonical TATA box found 29 nucleotides upstream of the transcription start site (Fig. 2B) could serve as the promoter for the K-bZIP transcripts.

FIG. 3.

RNase protection assays. (A) Mapping of the transcription start site of the K-bZIP transcript. Fifty nanograms of in vitro-transcribed [α-³²P]UTP-labeled RNAs derived from pBS-P2 (lane 4) was hybridized with 15 μg of RNAs from TPA-treated BCBL-1 cells (lane 1) or with an equal amount of yeast RNA (lane 3) at 68°C for 10 min. An RNase A-RNase T₁ mixture was then added to digest the unhybridized RNAs. The RNA hybrids were subsequently resolved on a 6% polyacrylamide–8 M urea denaturing gel. A sequencing reaction containing ddA, ddC, and ddG fragments was run in parallel as a DNA ladder marker (lane 2). Schematic drawings of the hybridization between the antisense probe (solid line) and the expected transcript (wavy line) are depicted to the left of each protected band. For simplicity, nucleotide 74850 (A of the first ATG in the K-bZIP gene) is arbitrarily defined as 1. (B) Mapping of the splice variants. Reactions were performed under the same conditions as those described for panel A, except that 50 ng of in vitro-transcribed [α-³²P]UTP-labeled RNAs derived from pBS-IVS was used as a probe. The numbers to the left of lanes 1 are in units of base pairs.

Splice variants of K-bZIP.

RNase protection assays were also performed to verify the splicing pattern of K-bZIP as well as to quantify the relative amounts of the three types of transcripts. As illustrated in the schematic drawings in Fig. 3B, the RNA probe used in these assays encompasses the 3′ half of IVS1, all of exon II, and the 5′ half of IVS2. Therefore, a type III or unspliced message which retains both IVS1 and IVS2 would protect a fragment of 246 nucleotides, type II transcripts would protect a 150-nucleotide fragment, and type I transcripts would protect a 93-nucleotide fragment. When total RNAs from TPA-induced BCBL-1 cells were used, all three predicted fragments were detected, while the intensities of the bands were distinct (Fig. 3B). Band intensity was further quantified by storage phosphor screen analysis and corrected against the length of the protected fragment. The molar ratio of type I, II, and III transcripts was determined to be 16:4:1. These results confirm the coexistence of differentially spliced variants of K-bZIP in chemically induced BCBL-1 cells and the notion that type I transcripts represent the predominant species.

Expression kinetics of K-bZIP.

To determine the stage of the viral life cycle in which KSHV K-bZIP was expressed, BCBL-1 cells were treated with TPA in combination with a protein synthesis inhibitor (cycloheximide) or a herpesvirus DNA polymerase inhibitor (PAA). Total RNAs from each sample were probed with a K-bZIP-specific DNA fragment. As shown in Fig. 4A, without treatment, there was little expression of this gene; the expression of the 1.35-kb K-bZIP transcript(s) began as early as 6 h, peaked at 24 h, and was maintained to 72 h after TPA treatment. The expression of the K-bZIP transcript(s) was blocked by cycloheximide but not by PAA, indicating that it is an early gene. As a control, a duplicate blot was hybridized with an ORF K8.1-specific probe (Fig. 4B). The ORF K8.1 gene resides in the IVS3 region and shares sequences with exon IV of the ORF K8 gene. It encodes two immunogenic glycoproteins via differential splicing (9). Despite the close proximity of the ORF K8 and ORF K8.1 genes, the expression kinetics of these two genes were quite different. The synthesis of ORF K8.1 began much later, at 24 h, and peaked at 48 h after TPA treatment. Furthermore, the expression of the ORF K8.1 transcript was sensitive to PAA treatment. Thus, the ORF K8.1 gene is classified as a late gene. These results suggest that within this 2-kb gene complex reside two independently regulated promoters. In a Northern blot analysis with the K-bZIP probe, we noted two transcripts of about 4 kb. We determined these to be readthrough transcripts initiated at the promoter for ORF 50, the homologue of EBV BRLF1 (31) (data not shown). Interestingly, one of these transcripts was not detected by the ORF K8.1 probe, indicating that this transcript lacks all or a portion of IVS3.

K-bZIP forms homodimers in vivo.

One characteristic of bZIP proteins is their ability to form homo- or heterodimers. As the first step in assessing the function of K-bZIP, we studied homodimer formation by K-bZIP. K-bZIP was differentially tagged with either the HA or the T7 epitope sequence at the N terminus. These two constructs were coexpressed in COS-1 cells. If homodimers were formed, coprecipitation of HA-K-bZIP and T7-K-bZIP would be expected. In the experiment shown in Fig. 5A, the cell extracts were first immunoprecipitated with T7 antibody, followed by Western blotting with HA antibody. HA antibody failed to detect K-bZIP in the T7 antibody immunoprecipitates of vector-transfected control cells, HA-K-bZIP-transfected cells, or T7-K-bZIP-transfected cells, attesting to the specificity of the antibody. An intense band corresponding to HA-K-bZIP was, however, readily detected in the T7 antibody immunoprecipitates of HA-K-bZIP- and T7-K-bZIP-cotransfected cells, suggesting that these two types of molecules form a complex. This conclusion was further reinforced by a reciprocal experiment in which HA antibody was used to immunoprecipitate the complex, followed by Western blotting to detect T7-K-bZIP (Fig. 5B). The above data provide strong evidence that K-bZIP forms homodimers.

FIG. 5.

Dimerization of K-bZIP. Total cellular protein from COS-1 cells transfected with pcDNA3.1 (lanes 1, 5, 9, and 13), pHA-KBZIP (lanes 2, 6, 10, and 14), pT7-KBZIP (lanes 3, 7, 11, and 15), or both pHA-KBZIP and pT7-KBZIP (lanes 4, 8, 12, and 16) was recovered and quantitated, and equal amounts were used in each reaction. Immunoprecipitation (IP) assays were performed by incubation of protein lysates with monoclonal antibodies against the T7 tag (lanes 5 to 8) or the HA tag (lanes 13 to 16), and the immunocomplex was captured with a protein A-protein G-Sepharose mixture. Protein samples from total cell lysates or from immunoprecipitations were subjected to Western blot (WB) analysis and probed with antibodies against the HA tag (A) or the T7 tag (B). Filters were developed by the ECL method. Kd, kilodaltons; Ig-H and Ig-L, heavy and light chains of immunoglobulin, respectively.

While previous studies (e.g., 23) strongly implicated the leucine heptad repeats or the ZIP domain in dimer formation, we wished to confirm that this is the case for K-bZIP. We noted that the type II transcript encodes a protein, designated K-bZIPΔLZ, whose truncation point coincides with the start of the ZIP domain. K-bZIPΔLZ was tagged with the T7 epitope at the N terminus and coexpressed with HA-K-bZIP in COS-1 cells, followed by immunoprecipitation and Western blot analysis. In contrast to the results obtained with full-length K-bZIP (Fig. 5), K-bZIPΔLZ (Fig. 6A, lane 5) failed to immunoprecipitate K-bZIP and vice versa (Fig. 6B, lane 3), suggesting the involvement of the ZIP domain in dimer formation.

FIG. 6.

The C terminus of K-bZIP is required for dimer formation. COS-1 cells were transiently transfected with HA-K-bZIP and T7-K-bZIPΔLZ for 48 h. Aliquots of total protein extract (lane 1) were immunoprecipitated (IP) either with anti-HA antibody (lane 3) or with anti-T7 antibody (lane 5) before Western blot analysis. Lanes 2 and 4 are blanks. Western blots (WB) were probed with antibodies against the HA tag (A) or the T7 tag (B). Filters were developed by the ECL method. Kd, kilodaltons.

DISCUSSION

In this report, we describe the identification and cloning of the gene for a novel KSHV protein with a prototypical bZIP structure, designated K-bZIP. The K-bZIP mRNA is generated through multiple in-frame splicings and thus was not recognized by direct examination of open reading frames in the genomic sequences. K-bZIP shows significant similarity to EBV BZLF1 (4, 31) with respect to genomic location, splicing pattern, and overall homology.

BZLF1 has been extensively characterized and is known to form homodimers in vivo and to recognize a group of sequences (BZLF1-responsive elements [ZREs]) that are related to the AP-1 consensus sequence and that are found in a number of EBV early promoters (16). The BZLF1 gene is also one of the master genes that, through binding to ZRE sequences, invokes the viral lytic cycle (12, 30, 31, 34). In addition, BZLF1 directly binds to the replication origin of the lytic cycle (OriL) and is essential for the synthesis of the concatemeric form of viral DNA for viral packaging (17).

We show here that K-bZIP also forms homodimers and, judging from its bZIP structure, likely will bind AP-1 or AP-1-like sequences. Experiments are in progress to determine the DNA binding specificity of K-bZIP dimers. While there are similarities between K-bZIP and BZLF1, there are also important differences. We note that the ZIP repeats of K-bZIP are much more extensive, including four leucines and one isoleucine, compared to those of BZLF1, which have only one leucine and one methionine. On the other hand, BZLF1 has a basic region with an arginine and lysine content much higher than that of K-bZIP. In addition, the DNA contact residues conserved in Jun, Fos, and BZLF1 are not well conserved in K-bZIP (Fig. 2) (23, 26, 46). K-bZIP homodimers therefore may recognize a DNA sequence motif distinct from conventional AP-1 or ZRE sequences.

Another interesting finding is the existence of differentially spliced variants of K-bZIP. Neither type II nor type III variants encode the ZIP domain, but they retain an intact N-terminal sequence, including the basic region. In the present study, it was found that the type I K-bZIP message is 4 times more abundant than the type II transcript and at least 15 times more abundant than the type III transcript. Whether proteins encoded by the type II and type III messages serve as modulators of the type I-encoded protein or themselves have independent functions remains to be ascertained. What we do know, based on the experiment shown in Fig. 6, is that type II proteins and, by the same token, type III proteins do not form dimers.

The induction kinetics, as well as the sensitivity toward inhibitors, suggest that the K-bZIP gene belongs to the class of early lytic genes. These results also parallel a report showing that the expression of BZLF1 was severely attenuated in the absence of de novo protein synthesis (20). In contrast, the ORF K8.1 gene, which is located within the third intron of K-bZIP and which shares the 3′ portion of exon IV, is regulated as a late gene. Thus, there is a complex regulatory mechanism of genes within this small transcriptional unit. In addition, a doublet (4.4- and 3.8-kb transcripts in Fig. 4A) and a single band (4.4-kb transcript in Fig. 4B) with a slower migration mobility were detected by K-bZIP- and ORF K8.1-specific probes, respectively. cDNA cloning of these transcripts demonstrated that they originate from Rta/ORF 50, the homologue of the EBV BRLF1 immediate-early protein (31) (data not shown). Indeed, similar bicistronic transcripts for both EBV BRLF1 and BZLF1 were identified and shown to be capable of translating both open reading frames in COS-1 cells (31). Furthermore, in a more synchronous induction system, the expression of monocistronic BZLF1 dramatically decreases at 8 h postinduction, while the bicistronic transcripts for both BRLF1 and BZLF1 are still detectable until 48 h postinduction (20). If a similar mechanism operates in KSHV, the translation of K-bZIP protein from the bicistronic or polycistronic RNAs, despite its low level, may bear significance. The lack of the 3.8-kb transcript in the Northern analysis when ORF K8.1 was used as a probe indicates that the ORF K8.1 gene is present only in the 4.4-kb message. The ORF K8.1 gene is located in the K-bZIP IVS3 (594 bp), an intron which is efficiently spliced in K-bZIP transcripts (RACE results and Fig. 3B). Since IVS3, where the ORF K8.1 coding sequence begins, is efficiently spliced out of K-bZIP transcripts, it is likely that the 3.8-kb transcript is derived from the 4.4-kb transcript by the removal of K-bZIP IVS3.

In summary, we have identified in the KSHV genome a gene for a bZIP protein and its splicing variants which lack the ZIP domain. We have also mapped the start and poly(A) sites of the transcriptional unit. The transcription of K-bZIP follows the kinetics of an early lytic gene. Functional analysis has revealed that K-bZIP is capable of forming a complex with itself. The transactivation or repression ability and the target genes of K-bZIP remain to be characterized.