Epstein-Barr virus encoded nuclear protein EBNA-3 binds a novel human uridine kinase/uracil phosphoribosyltransferase

Background Epstein-Barr virus (EBV) infects resting B-lymphocytes and transforms them into immortal proliferating lymphoblastoid cell lines (LCLs) in vitro. The transformed immunoblasts may grow up as immunoblastic lymphomas in immuno-suppressed hosts. Results In order to identify cellular protein targets that may be involved in Epstein-Barr virus mediated B-cell transformation, human LCL cDNA library was screened with one of the transformation associated nuclear antigens, EBNA-3 (also called EBNA-3A), using the yeast two-hybrid system. A clone encoding a fragment of a novel human protein was isolated (clone 538). The interaction was confirmed using in vitro binding assays. A full-length cDNA clone (F538) was isolated. Sequence alignment with known proteins and 3D structure predictions suggest that F538 is a novel human uridine kinase/uracil phosphoribosyltransferase. The GFP-F538 fluorescent fusion protein showed a preferentially cytoplasmic distribution but translocated to the nucleus upon co-expression of EBNA-3. A naturally occurring splice variant of F538, that lacks the C-terminal uracil phosphoribosyltransferase part but maintain uridine kinase domain, did not translocate to the nucleus in the presence of EBNA3. Antibody that was raised against the bacterially produced GST-538 protein showed cytoplasmic staining in EBV negative Burkitt lymphomas but gave a predominantly nuclear staining in EBV positive LCL-s and stable transfected cells expressing EBNA-3. Conclusion We suggest that EBNA-3 by direct protein-potein interaction induces the nuclear accumulation of a novel enzyme, that is part of the ribonucleotide salvage pathway. Increased intranuclear levels of UK/UPRT may contribute to the metabolic build-up that is needed for blast transformation and rapid proliferation.


Background
EBV is a gamma-herpes virus that is present in nearly all humans in form of lifelong infection. EBV infection in ad-olescence causes infectious mononucleosis. The virus associates with several human malignancies, particularly, with Burkitt lymphoma (BL), nasopharyngeal carcinoma (NPC) and post transplant lymphoproliferative disease [1].
The role of the EBNA-3 family in the viral strategy is not well understood [9,10]. We have previously identified two cellular proteins that can bind to EBNA-3 in yeast two-hybrid system. The ε-subunit of the TCP1 chaperonin complex may assist the initial folding of the nascent EBNA-3 [11]. The Xap-2 protein is a minor subunit of the aryl hydrocarbon receptor complex [12]. It is also a cellular target for the Hepatitis B virus encoded X antigen. HBX is believed to be involved in HBV associated carcinogenesis [13].
In this paper we have identified yet another human protein, designated F538, through its binding to EBNA-3. We have found that it is homologous to human and mouse uridine kinases, human uridine-cytidine kinase, and to uracil phosphoribosyltransferases of Toxoplasma gondii, C. elegans and Cryptosporidium parvum .

Results
RBP-Jk is one of the known interacting partners of EBNA-3. It binds to the N-terminal part of EBNA-3. In order to find additional targets of this large viral protein we used an N-terminus truncated EBNA-3 cDNA clone (encoding amino acids 127-945) for screening of a human lymphoblast cDNA library. We identified an interactive clone (designated clone 538) that could grow on His, Leu and Trp deficient medium and expressed β-galactosidase from an interaction dependent reporter locus. Plasmid rescue from the yeast to E. coli produced a clone with a ~800 bp long insert at its XhoI site.
Retransformation of clone 538 DNA together with the EBNA-3 construct into the yeast tester strain (SFY526) confirmed the interaction. Plasmids containing EBNA-1, ∆EBNA-4, EBNA-5, EBNA-6 and the empty pGBT9/BR vector were also retransformed together with the clone 538 DNA as negative controls. None of them showed any significant interaction in β-galactosidase test ( Table 1).
The strongly interacting p53/SV40LT protein pair was used as positive control.
XhoI cleaved insert of clone 538 was used as a probe on Northern blots. The 2,4 kb transcript was detected ubiquitously in all tissues with highest level expression in skeletal muscle, heart and kidney ( Figure 1). We have detected the same transcript in Burkitt lymphoma (BL) lines DG75, DG75-EBNA-3, DG75-EBNA-5 and freshly established LCL (data not shown).
Sequencing of the insert showed that clone 538 was identical to a part of a putative human gene with GenBank accession number of NM_017859.1, encoding the hypothetical gene product FLJ20517 (accession number NP_060329). The gene is consisted of fifteen exons that are distributed on a 19580 nucleotide long region on chromosome 20 (1325898-1306319).
To obtain the full-length cDNA (F538) we used a human heart cDNA library for PCR-amplification. The full-length cDNA was sequenced and cloned into green fluorescence protein fusion vector pEGFP-C1 (GFP) and AD vectors. The cDNA clone contained 1630 bp, encoding a 538 aa long protein. The first 10 amino acids were omitted from the cloning strategy because no selective primers could be designed for this region.
We have also identified an alternatively spliced variant of this gene (F538∆C) where exon 12 was spliced to the middle of exon 11. This splice created a frameshift that led to a stop-codon. The encoded protein product is truncated from residue 386 with 10 extra amino acids prior to the stop-codon ( Figure 2). The alternatively spliced variant was also recovered as EST from GenBank as mRNA expressed in different normal tissues such as optic nerve (BM702108), retina (BM693202) and testis (AA382318). The exon structure of the gene and the location of PCR primers are shown on Figure 3.
To analyse the protein-protein interaction in vitro, the insert from yeast 538 clone (the coding region corresponding to amino acids 216-473) was cloned into glutathione-S-transferase bacterial expression vector (GST-2TK). Upon induction, a 56 kD fusion protein was detected on silver or Coomassie blue stained SDS acryl amide gels.
This and several other control GST proteins were used to precipitate interacting proteins from lysates of CV-1 cells that were infected with recombinant vaccinia virus expressing full length EBNA-3. GST-538, but not GST or GST-EBNA-5 could precipitated EBNA-3 ( Figure 4). GST-Full436 containing the Xap-2 gene was used as positive interaction control [12]. Lysates of cells, infected with recombinant vaccinia virus that expressed EBNA-2 were included as non-specific precipitation controls.
To study the subcellular localization of the protein, GFP-F538 and GFP-F538∆C constructs were transfected into CV1 cells. Protein expression was already detectable after 4-6 hours by direct fluorescence. At this time the protein was homogeneously distributed in the cytoplasm. 24 hours after the transfection the protein started to form granular precipitates of varying size that were restricted to the cytoplasm. After 48 hours the protein formed large cytoplasmic inclusion bodies as the result of overexpression ( Figure 5).
We raised rabbit polyclonal antibodies against the bacterially produced GST-538 protein. Immunofluorescence staining detected an almost exclusively cytoplasmic distribution of F538 in the EBV negative BL cells DG75 ( Figure  6J) and BL21 but gave a predominantly nuclear staining in the EBV positive BL Raji ( Figure 6K) or EBV transformed lymphoblastoid cell lines Nadja, IARC171 and 940110 ( Figure 6L).
Importantly the stable transfected clone of DG75 that constitutively expressed EBNA-3 in more than 95% of the cells showed nuclear accumulation of the F538 protein ( Figure 6N) whereas F538 remained cytoplasmic in DG75 cells stably transfected with EBNA-5 ( Figure 6M). Optical sections taken at high magnification using fluorescence 3D microscope showed that F538 preferentially accumulated in the low DNA density areas of the nucleus of the EBNA-3 expressing cells (Fig 6O).

Figure 1
Northern blot for 2.4 kb F538 mRNA (top) shows an ubiquitous expression in different human tissues. β-actin probe is used as loading control on the same filter (bottom).

Figure 2
Sequence of F538 that is identical to the hypothetical protein FLJ20517. The sequence of the cDNA fragment isolated from the yeast two hybrid screen is underlined (top) Sequence of the C terminal part of the alternatively spliced variant with the additional ten amino acids preceding the newly created stop-codon (bottom).

Figure 3
Exon structure and cloning strategy for the full length F538 and the C terminal truncated splice variant  I, II and  IV) where the binding sites for ribose (R) and uracil (U) are located (Figure 8). Furthermore the predicted 3D structure shows an almost identical folding even for the region III despite the sequence difference. Structure and composition of the active centre in UPRT (T. gondii) and in F538 shows also very high similarity (Figure 9). The flexible loop that is involved in the binding to another monomeric unit is present as well. The 3D model of F538 suggests that the flexible loop is folded in a similar fashion as in UPRT (T. gondii) where it is involved in the formation of homo-dimer as well as homo-tetramer, that is the active form of the enzyme ( Figure 10).
The N-terminal part contains an ATP/GTP-binding site (Ploop) suggesting that F538 also has a uridine kinase activity. High (37-42) aligning scores to the known uridine kinases support this proposition ( Figure 11). F538 most probably consists of two domains with separate enzymatic activity in a similar fashion as the C. elegans UK/UPRT enzyme ( Figure 12).

Discussion
In the cell nucleotides are made either by the de novo pathway, or by the salvage pathway. The nucleosides can be salvaged by nucleoside kinases or cleaved by nucleoside phosphorylases to yield free bases and ribose-1-phosphate (desoxyribose-1-phosphate). The bases can be either salvaged by phosphoribosyltransferases or catabolized further.    uracil salvage might take place also in mammalian (rat) cells, but the responsible enzyme was still unidentified [18]. It was also shown that the activity of uridine kinases is increased 5-13 fold in human colon tumors [19] ovarian carcinomas, hepatocellular carcinomas [20] and damaged tissues [21]. The level of expression was however very low in normal tissues [22]. At present, two human uridine kinases are known: human uridine kinase 1 (HUCK1 -AAK49122) and human uridine kinase 2 (HUCK2 -AAK14053).
The N-terminal half of F538 shows high sequence similarity to both UCK1 and UCK2. UCK1 and UCK2 share 70% of identity at the protein level. F538 shows 38% identity to the UK/UPRT of Cryptosporidium parvum (accession number AAG53652), 50% identity to the UPRT of Toxoplasma gondii (accession number Q26998) and 51% identity to the UK/UPRT of C. elegans (accession number CAA93459).
Three of four conservative domains and binding sites are present in F538 and UK/UPRT as well as in UPRT.
On the basis of the high homology to UK and UK/UPRT and the similar 3D structure of the C-terminal part of F538 to the T. gondii UPRT, we propose that F538 is a novel uridine kinas/uracil phosphoribosyltransferase (UK/UPRT) -an enzyme with double catalytic activity. The protein coded by F538∆C lacks the UPRT domain but may function as uridine kinase.
Uridine kinases and UPRTs play an important role in the salvage pathway of RNA-synthesis. It was suggested [23] that de novo synthesized UTP is preferentially used for the production of UDP-sugars and phospholipids, while UTP, made by the salvage pathway, is used for RNA-synthesis. Uridine kinase activity is increased in the tumor cells. The increase of the nuclear UTP pool may be an important factor for maintenance of high proliferation rate. We suggest that the nuclear accumulation of this novel human UK/

Figure 8
Protein sequence alignment of F538 and UPRT of Toxoplasma gondii. The four conserved regions are underlined. R -ribose binding residue, U -uracil binding residues.

Figure 9
The structure of T. gondii UPRT active center with coordinated UMP (white balls and sticks) as compared to the predicted SWISSModel structure of F538. The UPRT conserved region III that shows the highest difference between the two proteins still exhibits the same folding pattern (black line on the backbone)

Figure 10
The structure of the predicted dimer derived from SWISSModel for F538. Reaction product -UMP is shown in red. N -and C-termini are indicated.
UPRT may be part of the viral strategy to convert a resting B cell to continuously proliferating lymphoblasts by securing a salvage pathway enzyme at the site of the RNA synthesis.
We have shown that F538 interacts specifically with EBNA-3 in the yeast two-hybrid system and in GST pull down assays. The predominantly cytoplasmic GFP-F538 translocates to the nucleus when EBNA-3 is expressed. GFP-F538∆C is not targeted to the nucleus by EBNA-3. This together with the sequence position of the cDNA fragment recovered from the yeast two hybrid screen indicates that binding site of F538 to EBNA-3 is located between residue 368 and 473, where the conserved domains of UPRT are located.

Conclusions
Our data suggest that EBNA-3 may bring an enzyme that participate in the uridine salvage pathway to the euchromatic part of the nucleus. Through this, EBNA-3 may contribute to the increase of UTP nuclear pool, that in turn is needed for increased cell proliferation.

Plasmids
The polylinker region of the GAL-4 DNA-binding domain containing pGBT9 vector (Clontech) was modified to make it compatible with the GST-2TK expression vector (designated as BD). For AD-F538 cloning we have used pGAD424 vector from Clontech.
N-terminus truncated mouse p53 in pGBT9 (pVA3) and SV40 Large T-antigen in pGAD10 (pTD1, both from Clontech) were used as positive interaction controls.
The GFP-F538 and AD-F538 were generated by inserting the 1642 bp PCR-product into the respective vectors (primers are described below) corresponding to amino acids 11 -548 of the NP_060329.