RNA binding proteins - Europe PMC

3 downloads 0 Views 2MB Size Report
Department of Virology, Royal Postgraduate Medical School, Du Cane. Road, London W12 ... State University of New York, Stony Brook, NY 11794-5222, USA.
The EMBO Journal vol.13 no.20 pp.4840-4847, 1994

EBNA-1, the major nuclear antigen of Epstein-Barr virus, resembles 'RGG' RNA binding proteins

Dee K.Snudden, Janet Hearing1, Paul R.Smith, Friedrich A.Grasser2 and Beverly E.Griffin3 Department of Virology, Royal Postgraduate Medical School, Du Cane Road, London W12 ONN, UK, 'Department of Microbiology, State University of New York, Stony Brook, NY 11794-5222, USA and 2Institut fur Medizinsche Mikrobiologie und Hygiene, Universitatskliniken des Saarlandes, D-66421 Homburg, Saar, Germany 3Corresponding author Communicated by B.E.Griffin

Nuclear antigen 1 (EBNA-1) is one of the key functions of the oncogenic DNA virus, Epstein-Barr virus (EBV), and is the only viral protein consistently expressed in EBV-associated malignancies. EBNA-1 binds in a site-specific manner to the viral DNA and is essential for viral replication, as well as for maintaining the genome as an extrachromosomal episome within infected cells. EBNA-1 is not recognized by the cellular immune system. Here we demonstrate that, in addition to its known DNA binding properties, EBNA-1 can also act as a strong RNA binding protein, interacting with diverse substrates in vitro, including the EBV-encoded RNA polymerase III transcript EBER1 and the HIV-encoded transactivation response (TAR) element. We also show that EBNA-1 can bind exon sequences derived from its own RNA expressed from the Fp promoter, as found in Burkitt's lymphomarelated cells and in nasopharyngeal carcinomas. EBNA1 has been identified as a component in an RNA complex; moreover, an anti-EBNA-1 antibody 1H4-1, that does not inhibit DNA binding, blocks binding to RNA. Arginine/glycine-containing (so-called 'RGG') motifs have been found in an increasing number of proteins that interact with RNA. The EBV antigen contains three potential 'RGG' motifs located around an internal glycine/alanine-rich repetitive sequence in the protein, and outside the region of EBNA-1 mapped previously as essential for viral DNA replication and other functionally defined properties. These motifs could be involved in the observed binding between EBNA-1 and RNA. Our data suggest that the mechanism of RNA binding may be complex and raise the possibility that EBNA-1 may play a role in EBVassociated tumours through transcriptional (or posttranscriptional) viral and/or cellular regulation. Key words: EBERS/EBV BamHI Q/gel retardation/RNA binding/TAR

Introduction The multi-functional nuclear antigen Epstein - Barr nuclear antigen 1 (EBNA-1) of Epstein-Barr virus (EBV) has

been shown to play a critical role in both viral replication and maintenance (Yates et al., 1984; Rawlins et al., 1985). In the principal tumours associated with EBV, endemic Burkitt's lymphoma (eBL) and nasopharyngeal carcinoma (NPC), it is the only one of the six latent viral nuclear antigens to be expressed (Rowe et al., 1986; Fahraeus et al., 1988). In transgenic mice, expression of EBNA-1 can induce tumours (Wilson and Levine, 1992). Thus, a full understanding of significant interactions between this important human virus and its host cell will require detailed knowledge of the role(s) played by EBNA-1. EBNA- 1 is a phosphoprotein (Hearing and Levine, 1985) with an unusual structure; it is separated into unique N- and C-terminal domains (amino acids 1-89 and 327641, respectively, in B95-8 cells) joined by internal glycine/alanine-rich short repeat sequences (Baer et al., 1984). Its size varies considerably in individual EBV-containing cell populations (from -69 to 94 kDa), depending on the number of internal glycine/alanine-containing (IR3) repeats within the protein. It has been characterized thoroughly as a DNA binding protein interacting as a homodimer with three discrete regions of the viral genome, two of them (I and II) lying in a region within the BamHI C restriction fragment of the genome, designated oriP, that consists of repetitive and dyad symmetry elements (Yates et al., 1984; Rawlins et al., 1985). The third region (III), composed of two consensus binding sites in a subregion of BamHI Q, binds EBNA- 1 with weaker affinity (Jones et al., 1989; Ambinder et al., 1990; Hsieh et al., 1993). In in vitro assays, EBNA-l also acts as a transcriptional transactivator (Polvino-Bodnar et al., 1985; Yates and Camiolo, 1988; Sugden and Warren, 1989). Numerous studies, carried out to map functional sites within EBNA- 1 for DNA binding, dimerization, phosphorylation, nuclear localization, transactivation and DNA looping, localize these activities mainly to the C-terminal region of the protein (Inoue et al., 1991; Polvino-Bodnar and Schaffer, 1992; Goldsmith et al., 1993, Frappier et al., 1994), with a short sequence in the N-terminal domain possibly contributing to some of the functions (Figure 1). Polypeptides of EBNA-1, in which most of the repetitive component has been deleted, have been found to function as well as, or better than, intact protein in transactivation, replication and plasmid maintenance assays (Yates and Camiolo, 1988). Recently, it has been shown that EBNA1 (amino acids 35-58) shares an epitope with SmD, a ribonucleoprotein frequently targeted by autoantibodies in systemic lupus erythematosus (SLE) patients (Sabbatini et al., 1993). Significantly, EBNA-1 is not seen by the host cellular immune system (Khanna et al., 1992; Murray et al., 1992), although antibodies to the antigen are produced. In tumour-related cells a novel promoter, Fp, is used

4 Oxford University Press 4840

EBNA-1 binding of RNA !!/cineialanine (IR3) repeat 40

amino acid

EBNAl ORF

57

90

r1

38C 405

327

sequence .

m

1

3

45?

641

1~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-

DNA binding

Dimerization

-

U

-

Phosphorylation

I

Transactivation Nuclear localization

.

Intranuclear localization

DNA looping

.

GEX clone

Fig. 1. Schematic diagram of the EBNA- I open reading frame. The 641 amino acid sequence of EBNA-1 (from B95-8 cells; Baer et al., 1984) divided into three domains: the unique N-terminus (amino acids 1-89), a glycine/alanine-rich (IR3) repetitive sequence (90-327) and the unique C-terminus. Also indicated (black bars) are location limits for known functions of EBNA-1 (Yates and Camiolo, 1988; Ambinder et al., 1991; Inoue et al., 1991; Polvino-Bodnar and Schaffer, 1992; Goldsmith et al., 1993; Chen et al., 1993; Frappier et al., 1994). The region of EBNA-1 cloned and expressed in the GEX expression system of Smith and Johnson (1988) is shown (cross-hatched).

for EBNA- 1 expression. Fp is active in some (tightly latent) BL-derived cell lines (Sample et al., 1991; Schaefer et al., 1991) and in NPCs (Smith and Griffin, 1992). EBNA- 1 itself is able to down-regulate Fp-directed gene expression. Thus, in addition to acting as a transactivator, EBNA- 1 can also repress gene expression (Sample et al., 1992). These findings, together with the tight control being exerted over viral gene expression in tumour cells, prompted us to investigate further the activities of this antigen. In particular, we explored the possibility that EBNA- 1 might also act as an RNA binding protein (the weak binding observed, for example, with the BamHI Q region of the DNA being a reflection of a stronger RNA interaction) and exert in this manner transcriptional or post-transcriptional influences on gene expression. Theoretically, this would afford a more cogent explanation for the key relevance of this particular antigen in EBVinfected cells and the down-regulation of other immunoreactive viral functions, than seemingly provided solely by its role in DNA replication and maintenance of EBV as an episome. Here we show in gel retardation and filter binding studies that EBNA-1 can indeed act as a (apparently) non-specific RNA binding protein. In vivo, EBNA- 1 binding to RNA may well be specific. The antigen probably uses 'RGG' motifs characterized originally in a subfamily of ribonuclear protein particles (Kiledjian and Dreyfuss, 1992). RNA recognition can be blocked by one anti-EBNA-1 monoclonal antibody, but not by another that recognizes a DNA binding epitope. Interactions between RNA and proteins, an area less well explored and less clearly understood than that of protein-DNA interactions, are clearly equally critical to many biological processes (reviewed in Frankel et al., 1991; Mattaj, 1993). Our findings suggest that functions for EBNA- 1 in human cells, other than those identified to date, merit further

examination.

Results Binding of EBNA-1 to RNA For our investigation of a potential RNA binding role for EBNA-1, we initially used isolates of full-length viral protein expressed in a baculovirus vector and purified to apparent homogeneity by immunoaffinity chromatography. The purity of the EBNA- 1 as obtained by this technique is described elsewhere (Hearing et al., 1992) and its ability to bind DNA in a sequence-specific manner is retained. Subsequently, a purified polypeptide fusion protein incorporating the C-terminus of the antigen was assayed for its ability to bind RNA. For these experiments, the binding mainly to four different radiolabelled RNAs containing variable amounts of potential secondary structure, and to a single-stranded homopolyribonucleotide, was assessed (see Materials and methods). Among the RNAs selected for study were recombinant constructions containing sequences corresponding to complementary strands of the BamHI Q region III DNA binding site (designated RIII sense and antisense), the known RNA binding sequence (TAR) for the HIV transactivator TAT protein (Cordingley et al., 1990; Dingwall et al., 1990) and a small polymerase III transcript of EBV (EBER1). The latter has been shown to bind a cellular auto-antigen, La (Lerner et al., 1981; Howe and Shu, 1988), a cellular protein designated EAP (Toczyski and Steitz, 1993) and an interferon-inducible protein kinase (Clarke et al., 1991). These RNAs all had considerable secondary structure, the calculated free energies of folding (Devereaux et al., 1984) varying from -57.9 (EBERI) to -27.3 (TAR) kcal/mol. The data obtained in our experiments, some of which are given in Figure 2A, show that EBNA- 1 bound to all the RNAs tested in a concentration-dependent manner. In the cases where the substrates were RIII antisense or TAR, the binding consistently produced a ladder of discrete bands on gels. With RIII sense and EBERI a 'smear' of

4841

D.K.Snudden et al.

A

,4

i.

i

.

; .,

CI1

0A8

.:

8

L/

.* 7:-

F:F f:'

.& a;

I

LUb _-

I I

'SEE. NS'E1F

-Sc:NSE

i-'II

r A 1-3

E B F..F)

B 1

7 8

234

2 3 4 5 6 7 8

ioA

W,

-RNA

ub

|

i., ._

_

+RNAse

1RNA,e

+RNAse

Rill SENSIL

+RNA1 s Ae - RNAs e-

I"A R

Fig. 2. (A) Complex formation between EBNA- I and various RNAs as assayed by gel retardation. RIII sense, RIII anti-(a) sense, EBERI and TAR RNAs, as noted, incubated with increasing amounts (0-1 ,ug) of purified EBNA- I (at 0.2 mg/ml) from baculovirus, as indicated. Unbound (ub) probe is identified (arrowhead). (B) EBNA- binding to RNase-resistant structures. Assays were performed using equal numbers of counts of radioactive RIII sense and TAR RNAs. Mobilities were assessed (as indicated): lane 1, in the absence of protein; lanes 2-4 respectively, after RNAs were digested with (i) 0.5 U RNase Tl + 1.25 ,ug RNase A, (ii) 10 U RNase TI + 10 ,ug RNase A or (iii) 100 U RNase Tl + 50 jg RNase A for 20 min at 30°C before incubation with 0.8 jg purified EBNA-1; lane 5, RNAs were incubated with 0.8 jig EBNA-1 in the absence of RNase; lanes 68, as in lane 5, except that after standing for 20 min at 300C, the EBNA- I -RNA complex was digested with RNase A and RNase TI, as described for assays in lanes 2-4, respectively. In RIII sense experiments, materials in lanes 2-4 and 6-8 have been exposed for longer than that in other tracks to show the weak RNase-resistant bands (0) bound by EBNA-1; the two bands which are faint here were none the less clearly visible on the original photograph. The high molecular weight doublet in these tracks can also be seen in lane 5 on longer exposure. Unbound TAR probe was run off the bottom of the gel.

retarded radiolabelled RNA was observed. In the case of EBERI, this could be resolved into a single strong band and a weak slower migrating band using a 4% native polyacrylamide gel (see below). These data are suggestive of multiple binding sites for EBNA- 1 with regard to either the RNAs or the protein, or both. Binding to singlestranded polyuridylic acid was also observed using a filter 4842

binding assay under standard conditions (Dingwall et al., 1990; data not shown), an aspect of binding that needs to be explored further. As a control for our studies, a protection experiment was performed in which the binding reactions of RIII sense and TAR radiolabelled RNAs (using equal amounts of radioactivity) were assessed after digestion with a

EBNA-1 binding of RNA

mixture of ribonucleases (RNAse A and TI) in either the presence or absence of EBNA- 1 (See Materials and methods). Our data, given in Figure 2B, suggest that EBNA-1 might at least in part be binding to structured components of the RNA since the same pattern was observed in the gel retardation assays irrespective of whether EBNA-1 was added to the substrate before or after digestion with the RNAses. Interestingly, the 'smear' binding observed with RIII sense RNA (Figure 2B, lane 5) was resolved into distinct, albeit weak, bands on digestion of the RNA. The TAR-RNA binding pattern was also changed, with disappearance of the faster migrating bands observed. Thus, the data given in Figure 2A may represent binding to single- and double-stranded regions of the RNA; with RNase treatment, the binding observed (Figure 2B) should reflect mainly that to residual secondary structure.

Mapping of RNA binding sites on EBNA-1 Having determined that EBNA- I appears to be an efficient RNA binding protein, the next question raised was whether the C-terminus, known to contain the DNA binding domain of EBNA- 1 inter alia, was responsible also for RNA binding. To this end, the C-terminal portion of EBNA-1 (Figure 1; amino acids 403-641) from the NPC tumour, C15 (Busson et al., 1988), was prepared as a 51 kDa fusion protein with the 26 kDa glutathione S-transferase using the GEX expression system (Smith and Johnson, 1988). After purification by affinity chromatography on immobilized glutathione, the identity of the fusion protein was confirmed by immunoblotting with polyclonal antibodies to EBNA-1 (Hearing et al., 1985; Figure 3A). In a mobility shift assay, this hybrid protein was also found to bind efficiently to oriP DNA (Figure 3B). When the RNA species used above (as substrates for EBNA-1 in gel retardation binding experiments) were assessed with the fusion protein, however, no binding was observed with any of them (Figure 3C; also data not shown), suggesting that EBNA- 1 binding to DNA and RNA is not part of a common mechanism.

Identification of EBNA-1 in an RNA complex To demonstrate that the RNA binding activity observed was genuinely due to EBNA-1 and not, for example, to a copurifying contaminating protein from baculovirusinfected insect cells, an anti-EBNA-1 polyclonal antibody was used to detect the presence of EBNA- 1 in an RNA-protein complex. Following fractionation on a polyacrylamide gel, the complex was transferred onto a nylon membrane and probed with the antibody. The EBNA-1-EBERI RNA complex was identified by this procedure (Figure 4A, arrow). The position of the complex resolved on the same gel, but detected by conventional autoradiography of the labelled RNA, is given in Figure 4B. Comigration of the components of the complex was revealed by this study. Blocking of RNA binding As further confirmation of the nature of the complex, it was shown that preincubation of baculovirus EBNA- 1 with a monoclonal antibody, 1 H4- 1, inhibited EBER 1-RNA binding (Figure 5A, lane 3). The epitope for this antibody has been mapped around amino acids 407-450 on EBNA-

C

B

A c

H

a9

1 2 3

1 2 3

Rill SENSE

TAR

Wf 68

--

-12

.'

ub _-

Z

::

-^8^

Fig. 3. (A) Identification of EBNA-I fusion protein. Purified EBNA-I recombinant protein (20 jtg) from the C1 5 tumour separated on SDS-PAGE (12.5%) and stained with Coomassie Blue dye (lane C). The location of the 51 kDa EBNA- I - glutathione S-transferase fusion protein is indicated (arrow) and positions of size markers (kDa) are given (at left). An immunoblot of the material used in lane C was probed with a polyclonal rabbit antibody against EBNA-1 (lane I). The antibody recognizes the 51 kDa fusion protein as well as a lower molecular weight band, just visible by Coomassie staining, probably a degradation product of the protein. The doublet seen in lane C at 26 kDa, not recognized by the antibody (lane I), corresponds to glutathione S-transferase products. (B) Binding of C15 EBNA-1 fusion protein to oriP DNA. DNA from oriP region II (RII DNA, position 8572-95 10) on the EBV genome of B95-8 (Baer et al., 1984; Rawlins et al., 1985) end-labelled with [32P]phosphate, then I X 104 c.p.m. incubated with 0.0, 2.6 and 5.2 jg of C15 EBNA-I -glutathionine S-transferase fusion protein, as noted. (C) EBNA- I polypeptides assayed for RNA complex formation. Mobility of RNAs (as designated): in the absence of EBNA- I (lane 1); after incubation with 0.8 jg holoEBNA- I from baculovirus (lane 2) or after incubation with 2 gg of the C-terminal EBNA-I fusion protein (lane 3). Unbound probe was run off the bottom of the gel.

I (D.Mackey, personal communication). Binding of the

1H4-1 antibody appears to mask the binding motifs and thereby prevent formation of a complex between EBNA1 and the RNA. On the other hand, a polyclonal antiEBNA- 1 antibody and another monoclonal antibody, Aza2E8, did not inhibit binding of EBNA- 1 to RNA (Figure SA, lanes 4 and 5). The latter antibody inhibits the DNA binding activity of EBNA-1 (Orlowski et al., 1990; Figure SB, lane 5) and is therefore believed to bind at or near the DNA binding domain. The binding epitopes of the polyclonal antibody have not been mapped, although this antibody also appears to have an effect on DNA binding (Figure SB, lane 4). Antibody lH4-1 does not inhibit DNA binding (Figure SB, lane 3). These data support the conclusion reached above that EBNA- 1 binding to RNA and DNA involves different sequences on the protein.

'RGG' binding motifs in EBNA-1 We compared the sequence of EBNA-1 with those of other RNA binding families (Mattaj, 1993) and noted the strong homology with the heterogeneous nuclear protein, hnRNP U (a member of the family of polypeptides that bind hnRNP particles, controlling the post-transcriptional pathways of expression of genetic information). This protein, as we have demonstrated with EBNA-1, acts as 4843

D.K.Snudden et al.

B

A

B

A

1

1

1

2 3 4 5

2 3 4 5

2

.s :3 I

.--Om-

lk

EBE:

ub_ m-

Rll DNA

EBER1

1/

EBERi

EBER1

Fig. 4. Complex formation between EBNA-1 and EBER1 RNA. (A) Detected by immunoblotting techniques. EBNA- 1 -EBERI complexes electrophoresed on a native polyacrylamide gel (4%), then transferred onto a nitrocellulose membrane and probed with a rabbit polyclonal antibody against EBNA-1. The tracks contain: EBERI RNA in the absence of protein (lane 1) or incubated with I tg of purified EBNA-1 from baculovirus (lane 2). EBNA-I alone (4 tg) purified from baculovirus (Hearing et al., 1992) was also run as a control on this gel, but not observed (data not shown). Either it had run off the gel or alternatively migrated with a net positive charge under these conditions (see Polvino-Bodnar and Schaffer, 1992). The position of the EBNA-1 complex is indicated by an arrow. Unbound probe migrated with the bromophenol blue dye front and was run off the bottom of the gel. (B) Detected by conventional autoradiography. EBNA- I - EBER 1 complexes electrophoresed on the same gel shown in (A). In this case the appropriate region of the gel was dried down before exposure to X-ray film which resulted in a slight distortion of the gel relative to that shown in (A). The tracks contain: EBERI RNA in the absence of protein (lane 1) or incubated with I gg of purified EBNA-1 from baculovirus (lane 2). A comparison of tracks 2 (A and B) indicates that the smear at the top of the gel in (A) recognized by the antibody is not bound to RNA. It appears to be a higher molecular complex including at least one copy of EBNA-1, but its composition is unknown.

Fig. 5. (A) Inhibition of complex formation between baculovirus EBNA-1 and EBERI RNA by antibodies. The effect of anti-EBNA-1 antibodies on the formation of protein-RNA complexes investigated by preincubation of baculovirus EBNA-1 with the respective antibody. Complexes were formed in the usual manner and materials separated by electrophoresis on a 4% native acrylamide gel. The tracks contain EBER1 RNA in the absence of protein (lane 1) or incubated with 1 ,ug of purified baculovirus EBNA-1 (lane 2). Lanes 3-5, as in lane 2, except that the EBNA-1 protein was preincubated with: lane 3, 8.5 ,tl l1H4-1 antibody (culture supernatant); lane 4, 8.5 gl rabbit polyclonal antibody (1.84 mg/ml); lane 5, 8.5 ,ul Aza2E8 (0.42 mg/ml). The locations of EBNA-1 complexes are indicated (arrow). Unbound probe was run off the bottom of the gel. (B) Effect of antibodies on the binding of EBNA-1 -glutathione S-transferase fusion protein to DNA. Corresponding amounts of the same antibodies used in (A) preincubated with the C15-derived EBNA-1 fusion protein before binding to the end-labelled region II DNA (RII DNA). The tracks contain RII DNA in the absence of protein (lane 1) or incubated with 1.8 gg EBNA- I -glutathionine S-transferase fusion protein; lane 3, as in lane 2 except protein was preincubated with the 1H4-1 monoclonal antibody before being added to RII DNA; lane 4, as in lane 2 except protein was preincubated with the rabbit polyclonal antibody before being added to RII DNA; lane 5, as in lane 2 except protein was preincubated with Aza2E8 before being added to RII DNA. Unbound DNA (ub) is indicated. Aa

PROTEIN

33 R EL

EBV (EBNA

C DENH O I X X

330

354 R G R X

HSV LRP (Human) 1

hnRNP U

Consensus

general RNA binding protein and can also bind 'singlestranded' RNA (Swanson and Dreyfuss, 1988; Kiledjian and Dreyfuss, 1992). Of notable interest in regard to this comparison is the fact that EBNA- 1 and hnRNP U contain 'RGG' consensus (R = arginine; G = glycine) RNA binding motifs (Figure 6). These motifs, found in numerous RNA binding proteins (Kiledjian and Dreyfuss, 1992), are generally 19-26 amino acids long and have a strong positive charge but contain no lysines (Dreyfuss et al., 1993). As shown, in comparison with hnRNP U and another herpes virus function HSV-1 LRP1 (or LAT), whereas the latter each have one 'RGG' motif, EBNA-1 contains three. One of these motifs maps between amino acids 33 and 56 in the unique N-terminal domain of the protein and the other two between amino acids 330 and 350, and 354 and 377. These sequences lie on either

1)

305

PROI

695 X

X

a xO

S

G

RG

I

S

O OS

- R

A

R G X Q

-

RE

R A

--

R O It a

N

XF-

R

-

P OI t

o

-

C

I G X

-

X-

R

A P

N

O

X

G

x x :

sG

PGA g

R

a S

56 350

-:

0 R 0 377

-

Is

YN

O

-

0

PG

RO326 N 716

RO aO a

a

4844

Fig. 6. Identification of 'RGG' motifs in EBNA-1. The proposed consensus RGG box sequence that interacts with RNA, with the shaded area representing the most highly conserved RGG triplet sequence (the single letter amino acid abbreviation is used). The residues in EBNA-1, compared with HSV-1 LRP1 (LAT) and hnRNP U (human) (Kiledjian and Dreyfuss, 1992), conforming to the consensus motif (bottom) are noted in bold, with the first and last amino acid numbers for each sequence indicated. Dashes are gaps placed in the sequence to optimize alignment; X represents any, or no, amino acids (Dreyfuss et al., 1993).

side of the glycine/alanine repetitive regions (Figure 1). Notably, there is no 'RGG' consensus sequence in the unique C-terminal region of EBNA-1 where the DNA binding function has been mapped. In our experiments this region does not bind RNA. Thus, if EBNA- 1 is acting

EBNA-1 binding of RNA

as an 'RGG' type RNA binding protein, as seems possible, it is not surprising to find that smear patterns can be observed in gel retardation assays since the protein potentially possesses the capacity to bind the same RNA molecule using several different amino acid sites either singly or in conjunction with another site. Alternatively, it may bind to more than one region on the RNA. Possibly also involved in the observed RNA binding by EBNA-1 is a near perfect R/G-rich dodecamer repetitive sequence (RGRGRGRGXXRP) found between amino acids 43-54 and 370-381, which is similar but not identical to 'RGG' motifs (Figure 6). This has little homology with other known arginine-rich motifs, but could also be responsible for some of the observed binding to RNA.

Discussion In a tumour setting, tight control over EBV gene expression is exerted. In NPCs, for example, apart from EBNA- I, only the anti-sense (or complementary) strand BamHI I/ A transcripts (Hitt et al., 1989; Gilligan et al., 1990; Chen et al., 1992; Karran et al., 1992), and the small polymerase III RNAs, EBERs (Howe and Shu, 1988), are expressed consistently. In data drawn from immunoprecipitation experiments EBNA-1 is detected, but none of the other immunoreactive EBNAs (designated EBNA-2-6, or alternatively the EBNA-2, -3 family and LP) are found in either BL cells or NPCs (Rowe et al., 1986; F?ahraeus et al., 1988). Our findings suggest that such tight control over viral gene expression, postulated to allow the virus to escape immune surveillance (Khanna et al., 1992; Murray et al., 1992), might be regulated by EBNA- 1 itself, possibly by binding to its own message (as shown in Figure 2) and autoregulating gene expression. The EBNA1 transcript expressed from the Fp promoter contains the BamHI Q region III sense sequence (Smith and Griffin, 1992) which we show can bind to EBNA-1. This element of the sequence was shown to be necessary for the repressor activity of EBNA-1 identified by Sample et al. (1992). From earlier studies, we (Smith and Griffin, 1992) identified two different splicing patterns for mature EBNA1 mRNA involving sequences in BamHI Q, splicing giving rise in one case to a single consensus binding site (Jones et al., 1989; Ambinder et al., 1990) and in the other case to two such sites. Whether this finding can be attributed to protein binding to the pre-mRNA and controlling splicing remains to be determined. RNA binding through the Q exon or alternative sites could also control expression of other EBNAs by either regulating transcription, transport or processing, or negatively affecting mRNA stability (Hitt et al., 1989; Sample et al., 1992). This is a distinct possibility, at least in the case of the immunoreactive EBNA-3 family, since their mature transcripts in BL-related B cells and NPC might be spliced from the same primary pre-mRNA that gives rise to the EBNA-1 message (Hitt et al., 1989; Sample et al., 1991; Schaefer et al., 1991; Smith and Griffin, 1992). Neither mature EBNA-3 family messages nor their translation products have been detected in EBV-associated tumours (reviewed in Klein, 1989). In EBV-containing lymphoblastoid cell lines, on the other hand, where 'upstream' promoters are used for the EBNA-3 family, they are expressed and recognized by cytotoxic T cells

(Khanna et al., 1992; Murray et al., 1992). Our data thus allow the suggestion to be made that EBNA-1 might be exerting viral control over the cell, not only by mechanisms involving gene transactivation and viral replication, but also by binding through sites on the protein to either viral or host RNA molecules, thereby altering expression of other viral functions. The binding observed in vitro between EBNA-1 and EBER 1 RNA is worth noting, since both species are expressed together in tumours or in latently infected cells (although the levels of EBERs vastly exceed that of the protein). EBERs are known to bind, via a 3' stretch of uridines to the La protein, an auto-antigen (Howe and Shu, 1988), and more recently have been identified in association with a cellular protein designated EAP (Toczyski and Steitz, 1993) and to an interferon-responsive protein kinase (Clarke et al., 1991). Their structural similarity to the small polymerase III transcripts of adenoviruses, the VA RNAs, has led to the hypothesis that in EBV-infected cells EBERs, like VA RNAs (Mathews and Shenk, 1991), may play a role in translational control. Since both EBNA- 1 and EBER I co-reside within the nucleus, a role for the RNA-protein interaction observed in our studies might be sought in vivo. The binding of EBNA-1 to HIV TAR (Delling et al., 1992; Tao and Frankel, 1992, and references therein) detected here is an intriguing finding, although it is unlikely to be of relevance under normal circumstances in vivo. However, in the rare event where the two viruses infected a single cell, the effect on the host could be of considerable significance. We suggest from the data presented here that EBNA- 1 may belong to the increasingly important class of proteins that bind RNA through 'RGG' motifs, the minimal binding domain of which has been defined (Ghisolfi et al., 1992; Kiledjian and Dreyfuss, 1992; Crozat etal., 1993; Dreyfuss et al., 1993; Mattaj, 1993). EBNA-l has at least three such motifs. Our evidence for suggesting that these are used in RNA binding consists of showing that a large Cterminal polypeptide lacking such motifs fails to bind RNA. It retains its ability to bind DNA, however. Further, a monoclonal antibody to EBNA- 1 with an epitope just 'downstream' of the arginine-rich elements in the protein can be used to block RNA binding, at least to EBER 1. RNA binding proteins recognize a more structurally diverse set of both paired (double-helical regions) and unpaired (single-stranded loops, etc.) RNA residues than generally found with DNA binding proteins, utilizing both Watson - Crick and non-Watson - Crick base pairs to create structure binding elements (as reviewed in Ellington, 1993). The notable absence of lysine residues in the basic 'RGG' motifs has led to the suggestion that arginines provide the specific binding contacts (Dreyfuss et al., 1993). Whereas with EBNA- 1 many more studies are needed to define the RNA substrate(s) used by this antigen in vivo and the nature of the contacts made between the antigen and RNAs, studies on mutants of EBNA- 1 are also consistent with our suggestion that 'RGG' motifs in the antigen may be responsible for the observed RNA binding. That is, in 'transactivation' experiments when CAT assays were performed with in-frame deletion mutants where the RGG boxes (Figure 6) were either partially or totally removed, the percentage of transcriptional activation fell dramatically, approaching nil activity

4845

D.K.Snudden et al.

(Yates and Camiolo, 1988). As noted in these experiments, removal of the internal repetitive (Gly/Ala) sequence actually increased activity. The authors attribute their observed loss of activity to the deletion of a possible enhancer activity within EBNA- 1. Their mutant data could also be accounted for by diminished RNA binding, with a concomitant effect on transcription. Two recent papers from the tumour literature may be relevant to our EBNA-1 findings. In these, chromosomal translocations in human tumours [Ewing's sarcoma (Delattre et al., 1992) and myxoid liposarcoma (Crozat et al., 1993)] have resulted in a gene (ETS and CHOP, respectively) for a transcription factor found in normal tissues being fused to an RNA binding sequence. In both cases, the abnormal protein carried by the tumour has acquired 'RGG' binding motifs and, as with EBNA-1, there are multiple copies of them. Crozat et al. (1993) comment that juxtaposition of an effector domain from this new subclass of RNA binding proteins with the DNA binding domain of transcription factors may represent a novel molecular mechanism for tumour formation common to several solid tumours. We note that EBNA- 1 per se appears to carry both of these functions in its normal state and thus fulfils, without the need for a specific chromosomal translocation, the 'novel mechanism' for tumour formation mentioned; in transgenic animals this antigen can even induce tumours (Wilson and Levine, 1992). In searching for a mechanism for EBNA-l in tumorigenesis, the ability of this antigen (unlike the other EBV nuclear antigens) to go unrecognized by the cellular immune system is noteworthy. Whether RNA binding per se contributes to the viral escape mechanism is a subject that can be pursued.

Materials and methods Antibodies These include polyclonal antibodies purified from the serum of rabbit (:f125) anti-EBNA-1 monoclonal antibody Aza2E8 (Hearing et al., 1985) and anti-EBNA-1 monoclonal antibody IH4-1, which map in the region around amino acids 407-450 (D.Mackey and F.A.Grasser et al., manuscript in preparation).

Expression and purification of EBNA- 1 HoloEBNA- I from B95-8 cells was expressed in a recombinant baculovirus vector and purified to near homogeneity, as described previously (Hearing et al., 1992), with purity assessed by gel electrophoresis (immunoblots and silver staining) and biological properties.

RNA substrates The BamHI Q/EBNA-I low-affinity binding site (region 111) was amplified from C15 DNA (Busson et al., 1988; Hitt et al., 1989) by PCR using Taq polymerase (Perkin Elmer Cetus) as recommended by the manufacturer. Oligonucleotides used in the PCR were: 5'-GGGTCGACTTGAAAAGGCGCG (position 62419 in the B95-8 EBV sequence; Baer et al., 1984) and 5'-GGCAAGCTTGAAGCACCCCCAT (position 62524). Each primer included a 5' extension containing a Sall and HindIll restriction site, respectively. After purification by elution, amplified DNA was checked by sequence analysis, digested with the appropriate enzymes, then cloned into Bluescript (KS) vector (Stratagene). The 'sense' RNA transcript (RIII sense) was synthesized using T3 RNA polymerase (Boehringer Mannheim) following cleavage of the plasmid with HindIll; 'antisense' region III (RIII antisense) was prepared by cloning the same PCR fragment between the HindIll and Sall sites of the Bluescribe vector (Stratagene). This plasmid was then linearized with Sall and transcribed using T3 RNA polymerase. RNA tran.scripts corresponding to EBERI were synthesized by transcription of pRA386 plasmid (MacMahon et al., 1991) with T3 RNA polymerase. TAR RNA was prepared

4846

by transcription of HIV residues + I to +57 cloned between the HindIll and Pstl sites of the Bluescribe vector (Stratagene) (op. cit.) with T3 RNA polymerase. Calculated energies (kcal/mol) using the computer program of Devereaux et al. (1984) for RNAs used in EBNA- I experiments are: EBER1, -57.9; RIII sense, -44.5; RIII antisense, -28.2; and TAR, -27.3. Transcription reactions contained 0.1 ,ug linearized plasmid, 40 mM Tris-HCI (pH 7.5), 6 mM MgCI2, 1 mM spermidine - HCI, 10 mM DTT, 500,M each ribonucleotide triphosphate, 40 jCi [a-32P]CTP (800 Ci/mmol), U T3 RNA polymerase (Boehringer Mannheim) and 40 U RNasin (Promega) in a final volume of 30 ,ul. After incubation for I h at 37°C, 1 U of RNase-free DNase (Promega) was added and the reaction incubated for a further 15 min. The RNA was purified from a 9% polyacrylamide denaturing gel and, after phenol extraction and ethanol precipitation, resuspended in 50 jil water. The c.p.m. of the RNA probe was determined and appropriate amounts used in the binding assay.

Gel retardation assay for RNA binding Protein concentrations were determined by the bicinchonin assay using bovine serum albumin as standard, as described Smith et al. (1985). Binding reaction mixtures (20 ,ul) contained Xl104 c.p.m. of RNA probe, 1 ,tg poly(dl-dC) (Pharmacia), 0.4 jig yeast tRNA, 40 U RNasin (Promega), 25 mM Tris-HCI (pH 8.0), 50 mM NaCI, 1 mM MgCl,, 5 mM spermidine, 0.5 mM DTT and 5% (v/v) glycerol. After incubation at 30°C for 20 min in the presence of 0.1-1.0 pg baculovirus EBNA-1 protein or 2.0 ,ug fusion protein, the reaction mixtures were loaded onto a 6 (Figures 2 and 3) or 4% non-denaturing polyacrylamide gel (Figures 4 and 5) using an acrylamide:bisacrylamide ratio of 19:1 in 50 mM Tris-glycine (pH 8.8). After electrophoresis at 180 V for -2 h at room temperature, gels were dried and exposed to X-ray film at -80°C using intensifying screens. Where the effect of anti-EBNA-1 antibodies on binding was investigated, I ,ug of baculovirus EBNA-1 was incubated with 8.5 gl of antibody for 30 min at room temperature prior to the addition of RNA. Protection assays were performed as above, except that before and after the addition of protein, reaction mixtures were digested with 0.50-100 U RNase TI (Boehringer Mannheim) and 1.2550.0 jig RNase A (Boehringer Mannheim) for 20 min at 30°C. Following electrophoresis, the products of binding reactions were also electrotransferred (8 V/cm) onto a nitrocellulose membrane filter in blot buffer [25 mM Tris, 192 mM glycine (pH 8.3), 20% (v/v) methanol] in a cooled tank overnight. The filter was incubated for 30 min with PBS containing 0.05% Tween 20 and 5% fat-free milk powder (PTM) with gentle rocking. Protein blots were then incubated with a I in 100 dilution (in PTM) of rabbit polyclonal antibody against EBNA-1 for 90 min at room temperature, washed in PBS containing 0.05% Tween 20 (PT) three times for 20 min and then reacted with a I in 500 dilution of swine anti-rabbit immunoglobulin (Ig)-peroxidase conjugate (DAKOPATTS). After being washed in PT as described above, the filter was developed using the ECL detection system (Amersham).

Expression and purification of the C-terminus of EBNA-1 The C-terminal portion of EBNA-1 was prepared as a fusion protein with the C-terminus of Sj26, a 26 kDa glutathione S-transferase using the GEX expression plasmid (Smith and Johnson, 1988). The Cterminal coding region was amplified from C 15 DNA by PCR using oligonucleotides: 5'-TCATCCGGATCCCCACCGCGCAGG (position 109117 in the B95-8 EBV sequence; Baer et al., 1984) and 5'CGGGAATTCACGGCTTTTAATAC (position 109925). For cloning purposes the first oligonucleotide primer includes a G -* A and T -> C base change at positions 109125 and 109128 which generated a BamHI restriction site, and the second oligonucleotide includes an A -o T base change at position 109910 generating an EcoRI site. The amplified DNA was purified by elution from a 0.8% agarose gel. After verification by sequencing and digestion with BamHl and EcoRI enzymes, the fragment was inserted in-frame behind the C-terminus of Sj26. Screening of transformants and large-scale purification of the EBNA- I fusion protein was carried out as described by Smith and Johnson (1988). The protein was resolved by SDS-PAGE (12.5%). Material in one track was stained with Coomassie Blue, as described by Laemmli et al. (1970), and in another electrotransferred onto a nitrocellulose membrane. The latter was probed with an EBNA- I rabbit polyclonal antibody (Hearing et al., 1992) as described above, except that after the final wash in PT, the filter was developed with 3,3'-diaminobenzidine tetrahydrochloride (DAB) 200,ul of 20 mg/ml solution, 9.7 ml PBS, 100 j. 1% CoCk, 1% Ni(NH4)2(SO4)2 and 10 jd H2O, (30% v/v).

EBNA-1 binding of RNA

132Plphosphate incorporation into DNA binding substrate

PCR was used to clone the region of EBV DNA dyad symmetry (region II) localized within oriP (Rawlins et al., 1985). Amplification of region II from C15 tumour DNA was performed using the following oligonucleotide primers: 5'-CCATGAATTCGTGTGAGATG [position 8572 in the B95-8 genome; Baer et al., 1984)] and 5'-ATAAGGATCCCTTGTTAACC (position 9150). For cloning purposes. the first primer has an A -> G base change at position 8576 which generates an EcoRI restriction enzyme site, and the second primer C -* A and C -* T base changes at positions 9156 and 9157, respectively, which generated a BamHI site. The PCR product was verified by sequence analysis, then cloned into the Bluescribe vector (Stratagene). Plasmid DNA, digested with EcoRI and BamHI, was electrophoresed on a 0.8% agarose gel. The region II binding site was eluted from the gel, its ends 'filled in' with [32P]dCTP using the Klenow fragment of DNA polymerase, and the reaction product passed through a G-75 Sephadex column and collected. The labelled probe was ethanol-precipitated before use in mobility retardation assays.

Gel retardation assay for DNA binding Protein DNA complexes were formed by mixing 0.0-5.2 jg fusion protein with 2x 104 c.p.m. of 32P-labelled probe in 20 ,ul binding buffer [25 mM Tris-HCI, pH 8.0, 50 mM NaCl, I mM MgCl2, 5 mM spermidine, 0.5 mM DTT. 1 ,ug poly(dl-dC) (Pharmacia) and 5% (v/v) glycerol]. After incubation at 30°C for 20 min, the samples were loaded onto a 4% non-denaturing polyacrylamide gel and electrophoresed as described previously. Where the effect of anti-EBNA-I antibodies on DNA binding was being investigated, the fusion protein and antibody were incubated together for 30 min at room temperature, prior to the addition of labelled DNA, as above.

Acknowledgements We thank Dr C.Dingwall for the recombinant TAR construction and Drs M.Gait, J.Karn and N.Krauzewicz for helpful discussions. We (D.S., P.R.S. and B.E.G.) thank the Cancer Research Campaign (UK) for generous support of this work. J.H. wishes to acknowledge the American Cancer Society (grant VM-13A) for financial support.

References Ambinder,R.F., Shah,W.A., Rawlins,D.R., Hayward,G.S. and Hayward S.D. (1990) J. Viral., 64, 2369-2379. Ambinder,R.F., Mullen,M., Chang,Y.-N., Hayward,G.S. and Hayward, S.D. (1991) J. Virol., 65, 1466-1478. Baer,R. et al. (1984) Nature, 310, 207-211. Busson,P. et al. (1988) Itit. J. Cancer, 42, 599-606. Chen,H.-L., Lung,M.M.L., Sham,J.S.T., Choy,D.T.K., Griffin,B. and Ng,M.H. (1992) Virology. 191. 193-201. Chen,M.-R.. Middledorp,J.M. and Hayward,S.D. (1993) J. Virol., 67, 4875-4885. Clarke,P.A., Schwemmle,M., Schickinger,J., Hilse,K. and Clemens,M.J. (1991) Nucleic Ac-ids Res., 19, 243-248. Cordingley,M.G. et al. (1990) Proc. Natl Acad. Sci. USA, 87, 8985-8989. Crozat,A., Aman,P., Mandahl,N. and Ron,D. (1993) Natuire, 363, 640644. Delattre,O., Zucman,J., Plougastel,B., Desmaze,H., Joubert,I., de Jong,P., Rouleau,G., Aurias,A. and Thomas,G. (1992) Natuire, 359, 162-165. Delling,U., Reid.L.S., Barnett,R.W., Ma,M.Y.-X., Climie,S., SumnerSmith,M. and Sonenberg,N. (1992) J. Virol., 66, 3018-3025. Devereux,J., Haeberli,P. and Smithies,O. (1984) Nucleic Ac ids Res., 12, 387-395. Dingwall,C., Ernberg,l., Gait,M.J., Green,S.M., Heapy,S., Karn,J., Lowe,A.D., Singh,M. and Skinner,M.A. (1990) EMBO J., 9. 41454153. Dreyfuss,G., Mantunis,M.J., Pinol-Rome,S. and Burd.C.G. (1993) Annu. Rev: Biochem., 62, 289-321.

Gilligan,K., Sato,H., Rajadurai,P., Busson,P., Young,L., Rickinson,A., Tursz,T. and Raab-Traub,N. (1990) J. Virol., 64, 4948-4956. Goldsmith,K., Bendell,L. and Frappier,L. (1993) J. Virol., 67, 3418-3426. Hearing,J.C. and Levine,A.J. (1985) Virology, 145, 105-116. Hearing,J.C., Lewis,A. and Levine,A.J.(1985) Virology, 142, 215-220. Hearing,J., Mulhaupt,Y. and Harper,S. (1992) J. Virol., 66, 694-705. Hitt,M.M., Allday,M.A., Hara,T., Karran,L., Jones,M.D., Busson,P., Tursz,T., Emberg,I. and Griffin,B.E. (1989) EMBO J., 8, 2639-2651. Howe,J.G. and Shu,M.-D. (1988) J. Virol., 62, 2790-2798. Hsieh,D.-J., Camiolo,S.M. and Yates,J.L. (1993) EMBO J., 12, 49334944. Inoue,N., Harada,S., Honma,T., Kitamura,T. and Yanagi,K. (1991) Virology, 182, 84-93. Jones,C.H., Hayward,S.D. and Rawlins,D.R. (1989) J. Virol., 63, 101110. Karran,L., Gao,Y., Smith,P.R. and Griffin,B.E. (1992) Proc. Natl Acad. Sci. USA, 89, 8058-8062. Khanna,R., Burrows,S.R., Kurrila,M.G., Jacob,C.A., Misko,I.S., Sculley,T.B., Kieff,E. and Moss,D.J. (1992) J. Exp. Med., 176, 169-176. Kiledjian,M. and Dreyfuss,G. (1992) EMBO J., 11, 2655-1664. Klein,G. (1989) Cell, 58, 5-8. Laemmli,U.K. (1970) Nature, 277, 680-685. Lerner,M.R., Andrews,N.C., Miller,G. and Steitz,J.A. (1981) Proc. Natl Acad. Sci. USA, 78, 805-809. MacMahon,E.M., Glass,J.D., Hayward,S.D., Mann,R.B., Becker,P.S., Charache,P., McArthur,J.C. and Ambinder,R.F. (1991) Lancet, 338, 969-973. Mathews,M. and Shenk,T. (1991) J. Virol., 65, 5657-5662. Mattaj,I.W. (1993) Cell, 73, 837-840. Murray,R.J., Kurilla,M.G., Brooks,J.M., Thomas,W.A., Rowe,M., Kieff,E. and Rickinson,A.B. (1992) J. Exp. Med., 176, 157-168. Orlowski,R., Polvino-Bodnar,M., Hearing,J. and Miller,G. (1990) Virology, 176, 638-642. Polvino-Bodnar,M. and Schaffer,P.A. (1992) Virology, 187, 591-603. Polvino-Bodnar,M., Kiso,J. and Schaffer,P.A. (1985) Nucleic Acids Res., 16, 3415-3435. Rawlins,D.R., Milman,G., Hayward,S.D. and Hayward,G.S. (1985) Cell, 42, 859-868. Rowe,D.T., Evan,G.I., Wallace,L.E.. Farrell,P.J. and Rickinson,A.B. (1986) EMBO J., 5, 2599-2607. Sabbatini,A., Bombardieri,S. and Migliorini,P. (1993) Eur J. Immunol., 23, 1146-1152. Sample,J., Brooks,L., Sample,C., Young,L., Rowe,M., Gregory,C., Rickinson,A. and Kieff,E. (1991) Proc. Natl Acad. Sci. USA, 88,

6343-6347. Sample,J., Daniel Henson,E.B. and Sample,C. (1992) J. Virol., 66, 4654-4661. Schaefer,B., Woisetschlaeger,M., Strominger,J. and Speck,S. (1991) Proc. Natl Acad. Sci. USA, 88, 6550-6561. Smith,P.R. and Griffin,B.E. (1992) J. Virol., 66, 706-714. Smith,D.B. and Johnson,K.S. (1988) Gene, 67, 31-40. Smith,P.K. et al. (1985) Anal. Biochem., 150, 76-85. Sugden,B. and Warren,N. (1989) J. Virol., 63, 2644-2649. Swanson,M.S. and Dreyfuss,G. (1988) Mol. Cell. Biol., 8, 2237-2241. Tao,J. and Frankel,A.D. (1992) Proc. Natl Acad. Sci. USA, 89, 27232726. Toczyski,D.P. and Steitz,J.A. (1993) Mol. Cell. Biol.. 13, 703-7 10. Wilson,J.B. and Levine,A.J. (1992) Curr Topics Microbiol. Immunol., 182, 375-384. Yates,J.L. and Camiolo,S.M. (1988) Canicer Cells, 6, 197-205. Yates,J., Warren,N., Reisman,P. and Sugden,B. (1984) Proc. Natl Acad. Sci. USA, 81, 3806-3810. Received on April 18, 1994; revised on JulY 26. 1994

Ellington,A.D. (1993) Curr Biol., 3, 375-377. Fahraeus,R. et al. (1988) Int. J. Caancer, 42, 329-338. Frankel,A.D., Mattaj,I.W. and Rio,D.C. (1991) Cell. 67, 1041-1046. Frappier,L., Goldsmith,K. and Bendell,L. (1994) J. Biol. Chem., 269, 1057-1062. Ghisolfi,L., Joseph,G., Amalric.F. and Erard,M. (1992) J. Biol. Chem., 267, 2955-2959.

4847