Analysis of DNA binding by the adenovirus type 5 E1A oncoprotein

The human adenovirus type 5 (Ad5) early region 1A (E1A) gene encodes two major proteins of 289 and 243 residues; these proteins differ only by the presence of an internal sequence of 46 amino acids unique to the larger protein. Comparison of the E1A sequence from a number of adenovirus serotypes has identified three regions of sequence conservation (Kimelman et al., 1985 ; van Ormondt et al., 1986 ), designated conserved regions (CR) 1, 2 and 3 (Fig. 1A). CR 3 coincides almost exactly with the region unique to the 289 residue protein (Shenk & Flint, 1991 ).

(41K): Fig. 1. (A) Map of the major E1A proteins and CRs. The two major products of E1A are 289 and 243 residues in length and differ only by the presence of an additional 46 amino acids unique to the larger. Regions of sequence conservation between various adenovirus serotypes are indicated as CRs 1, 2 and 3. The regions of E1A expressed as GST fusions in E. coli are depicted and the indicated amino acid numbers are inclusive. (B) dsDNA cellulose chromatography of the GSTE1A protein fusions. The recombinant fusion proteins described in (A) were allowed to bind to dsDNA cellulose and bound proteins were eluted with a stepwise gradient of NaCl as indicated at the top of the gel. FT, flow-through fraction; W1 and W2, wash fractions. Equal portions of each fraction were analysed by 12% SDSpolyacrylamide gel electrophoresis and visualized by staining with Coomassie brilliant blue.

The multifunctional E1A proteins influence a variety of transcriptional and cell cycle events (Shenk & Flint, 1991 ; Dyson & Harlow, 1992 ; Peeper & Zantema, 1993 ; Bayley & Mymryk, 1994 ; Moran, 1994 ). E1A interacts with a variety of cellular proteins, including transcriptional co-activators, such as pCAF (Yang et al., 1996 ), the CREB-binding protein and p300 (Eckner et al., 1994 ; Arany et al., 1995 ; Lundblad et al., 1995 ), and cell cycle regulatory proteins, such as the retinoblastoma tumour suppressor gene product and related family members (Whyte et al., 1988 ; Egan et al., 1989 ; Ewen et al., 1991 ; Hannon et al., 1993 ). E1A also interacts with various components of the general and specific transcriptional machinery, including the TATA-binding protein (Boyer & Berk, 1993 ; Hateboer et al., 1993 ; Geisberg et al., 1994 ; Song et al., 1995 ), several of the TATA-binding protein-associated factors (Geisberg et al., 1995 ; Mazzarelli et al., 1995 ) and transcription factors, such as ATF-2 (Chatton et al., 1993 ; Liu & Green, 1994 ) and c-Jun (Maguire et al., 1991 ; Liu & Green, 1994 ).

In addition to its many interactions with cellular proteins, the 289 residue E1A protein binds non-specifically to double-stranded (ds) DNA in vitro (Chatterjee et al., 1988 ) and a C-terminal fragment of E1A spanning amino acids 201289 is sufficient to bind dsDNA in vitro (Zu et al., 1992 ). Substitution of aspartic acid for arginine at amino acids 205 and 206 or deletion of residues 201216 blocks binding of this C-terminal fragment of E1A to DNA (Zu et al., 1992 ). In vivo, both of these mutants function as well as the wild-type protein to activate and repress transcription, suggesting that the interaction of E1A with DNA is not required for either of these activities (Zu et al., 1992 ).

In this study, we tested a number of fragments of E1A for DNA binding and determined that only the C-terminal region of E1A bound DNA in vitro. This region of E1A bound both single-stranded (ss) and dsDNA and binding required a basic region spanning amino acids 201220. In addition, a minimal fragment of E1A encompassing amino acids 201218 was sufficient to bind DNA and several cation-exchange resins. Although fusion of the C-terminal region of E1A to a strong transcriptional activation domain (AD) inhibited growth when expressed in Saccharomyces cerevisiae, this did not require the region we identified as necessary for DNA binding in vitro. Our results suggest that E1A may not function as a DNA-binding protein in vivo.

Plasmid construction.
To examine their toxicity in yeast, E1A fragments were expressed from a modified pJG4-5 yeast vector (Origene Technologies), containing an expanded multiple cloning site (pJG4-5+). Proteins expressed from this vector are fused to an N-terminal tag consisting of a nuclear localization sequence (NLS), a strong transcriptional AD and an haemagglutinin antigen (HA) epitope tag. Large portions of E1A-encoding fragments, corresponding to amino acids 182, 93138, 139204 and 187289, were generated by PCR with flanking EcoRI, BamHI or XhoI restriction sites and cloned into the corresponding sites within pJG4-5+. Similarly, E1A-encoding fragments spanning residues 201289, 221289 and 201218 were generated and cloned into pJG4-5+. Mutants with deletions within residues 187289 of E1A were amplified by PCR using the templates described previously (Jelsma et al., 1988 ; Boyd et al., 1993 ) and cloned into pJG4-5+. To generate pJG4-5AD-, the AD was excised from pJG4-5 by removing the EcoRVHpaI fragment. The portion of E1A encoding amino acids 187289 was then subcloned into pJG4-5AD- using EcoRI and XhoI.

To express E1A fused to glutathione S-transferase (GST), DNA sequences encoding amino acids 182, 93138, 139204, 201289, 201218 and 221289 of E1A were excised from the corresponding yeast expression vectors described above with EcoRI and XhoI and ligated into the same sites of pGEX-4T1 (Amersham Pharmacia). pGEX-4T1 vectors encoding amino acids 201289 of E1A with deletions of residues 227239, 239254, 255270, 271284 and 285289 were constructed by replacing the wild-type XbaIXhoI fragment with the corresponding mutant fragments derived from a series of vectors provided by G. Chinnadurai, St Louis University Health Sciences Center, St Louis, MO, USA (Schaeper et al., 1995 ).

Expression and purification of GSTE1A fusion proteins.
GSTE1A fusion proteins were expressed and purified as described by the pGEX-4T1 supplier (Amersham Pharmacia), with minor modifications. An overnight culture of Escherichia coli BL21 cells transformed with either pGEX-4T1 or one of the pGEX-4T1 recombinants described above was diluted 1:10 in LB medium containing 50 µg/ml ampicillin and grown at 37 °C. After 1 h, expression of the fusion protein was induced by the addition of IPTG to a final concentration of 0·2 mM and cultures were grown for a further 3 h at 37 °C. Cells were harvested, resuspended in 1/200 vol. PBS and lysed by mild sonication. DNaseI was added to 20 µg/ml and lysates were incubated for 30 min at room temperature. Subsequently, 10% Triton X-100 was added to a final concentration of 1% and the lysates were centrifuged at 21000 g for 5 min at room temperature. Then, 1/10 vol. 50% glutathione bead slurry was added to the supernatant and mixed gently at room temperature for 5 min. Beads were washed twice with 100 vols PBS, once with TEK buffer (10 mM TrisHCl pH 7·4, 1 mM EDTA and 50 mM KCL) containing 1·0 M NaCl and twice with TEK buffer alone. The GSTE1A fusion proteins were then eluted from the beads with 50 mM TrisHCl pH 8·0 containing 5 mM glutathione and dialysed against TEK buffer.

Assay for binding to nucleic acid cellulose, cellulose phosphate and carboxymethyl Sepharose.
GSTE1A fusion proteins were assayed for binding to ds and ss calf thymus DNA cellulose (Sigma), cellulose phosphate (Whatman Nuclepore) and carboxymethyl Sepharose (Amersham Pharmacia). dsDNA cellulose, ssDNA cellulose and carboxymethyl Sepharose were allowed to swell in TEK buffer overnight at 4 °C, washed with TEK buffer twice and then resuspended in an equal volume of buffer. Weighed cellulose phosphate was stirred gently into 25 vols 0·5 M NaOH, left for 5 min at room temperature and washed with TEK buffer until the pH of the flow-through dropped to below 11·0. The resin was then mixed with 25 vols 0·5 M HCl, left for 5 min at room temperature and washed with TEK buffer until the pH of the flow-through reached 3·0. The cellulose phosphate resin was then resuspended in an equal volume of TEK buffer. Prior to binding assays, 1 ml of the slurry was washed twice with 25 ml TEK buffer and resuspended in 500 µl TEK buffer.

Binding assays of GSTE1A fusion proteins were performed as follows: 100 µl DNA cellulose, cellulose phosphate or carboxymethyl Sepharose resin slurry was centrifuged briefly and the supernatant was removed. Then, 30 µg recombinant protein was diluted to a final volume of 100 µl in TEK buffer, mixed gently with the resin in a microcentrifuge tube and allowed to bind for 30 min on ice. After centrifuging briefly, the supernatant (flow-through) was withdrawn into a fresh tube and the resin was washed twice with 100 µl TEK buffer. The wash fractions were also collected. Fusion protein was eluted sequentially with 100 µl TEK buffer containing 0·1 M NaCl, 0·2 M NaCl, 0·3 M NaCl, 0·4 M NaCl, 0·6 M NaCl and 1·0 M NaCl. Aliquots of 24 µl from each fraction were mixed with 8 µl 4x sample buffer (0·0625 M TrisHCl pH 6·8, 10% glycerol, 2% SDS, 0·05% b-mercaptoethanol and 0·0005% bromophenol blue), boiled for 4 min and separated on 12% SDSpolyacrylamide gels. Proteins were visualized by staining with Coomassie brilliant blue.

Analysis of growth inhibition in yeast.
To determine the effect of each fragment of E1A on yeast growth, yeast strain w3031a (obtained from M. Smith, University of Virginia, Charlottesville, VA, USA) was transformed with pJG4-5+ or corresponding derivatives expressing portions of E1A using the lithium acetate method (Adams et al., 1998 ). Cells were plated on synthetic complete (SC) selective medium containing galactose to induce recombinant protein expression. After 4872 h of growth at 30 °C, pictures of plates were taken using a Pharmacia Biotech ImageMaster VDS image capture system. Toxicity of E1A fragments was assessed by visually examining the size of yeast colonies and comparing them to those that received the empty pJG4-5+ plasmid. Similar tests were carried out in yeast strains FY86 (GCN5) and FY1370 (Δgcn5) (Roberts & Winston, 1997 ) (obtained from C. Brandl, University of Western Ontario, London, Ontario, Canada). Yeast doubling time analysis was performed by diluting overnight cultures of transformed yeast to 0·1 A₆₀₀ in SC selective medium containing galactose. Triplicate cultures were grown with agitation at 30 °C and A₆₀₀ was measured at various times post-inoculation.

Western blot analysis of recombinant protein expression in yeast.
Expression levels of the C-terminal fragments of E1A in yeast were determined by Western blot analysis. Yeast were transformed with pJG4-5+ or pJG4-5+ derivatives expressing the C-terminal portion of E1A and plated on SC selective medium containing glucose. Colonies of transformed yeast were picked from each plate and used to inoculate 5 ml liquid SC selective medium containing glucose. These cultures were then grown in a cell culture rotator at 30 °C for 24 h. Cells from 1·5 ml aliquots of each culture were harvested by centrifugation, washed with sterile distilled H₂O to remove residual glucose and grown for a further 24 h in 5 ml SC selective medium containing galactose to induce E1A expression. Cells were harvested from 1·5 ml aliquots of each culture, washed twice in 1 ml HSE buffer (20 mM HEPES pH 7·5, 150 mM NaCl and 1 mM EDTA), resuspended in 200 µl HSE buffer containing Complete Protease Inhibitor Cocktail (Roche) and transferred into fresh tubes containing 0·3 g acid-washed glass beads (425600 µm; Sigma). Cells were lysed by 15 cycles of vortexing for 30 s followed by 30 s of incubation on ice. After lysis, samples were centrifuged at 21000 g for 10 min. A sample of 200 µl of each supernatant was transferred to fresh microcentrifuge tubes, recentrifuged as above and the protein concentration of the supernatant was measured using the Bio-Rad DC Protein assay (Bio-Rad Laboratories). From each extract, 30 µg of protein was separated on 12% SDSpolyacrylamide gels and transferred to a PVDF membrane. Western blot analyses were performed with the ECL Plus system (Amersham Pharmacia) using the 3F10 anti-HA rat monoclonal antibody (Roche) and goat anti-rat IgG horseradish peroxidase-conjugated antibody (Pierce). Blots were visualized using a Molecular Dynamics Storm 860 system.

Identification of the dsDNA-binding domain in E1A
Previous reports have demonstrated that recombinant E1A made in E. coli could bind dsDNA in vitro, although no evidence for sequence-specific binding has been presented (Chatterjee et al., 1988 ; Zu et al., 1992 ). These experiments also showed that the C-terminal portion of E1A could bind DNA independently of the remainder of the protein (Zu et al., 1992 ). Currently, only full-length E1A or N-terminal truncations of E1A have been tested for DNA binding, hence it remained possible that other regions of E1A, in addition to the C-terminal domain, could bind DNA. To test this, we constructed vectors expressing various fragments of E1A fused to GST and purified them from E. coli (Fig. 1A). Fragments of E1A, spanning amino acids 182, 93138, 139204 and 201289, were tested for the ability to be retained on dsDNA cellulose resin, an assay used previously to demonstrate the interaction of E1A with dsDNA (Zu et al., 1992 ). Purified recombinant E1A fragments were mixed with dsDNA cellulose and eluted with increasing concentrations of NaCl. No retention of the fragments spanning amino acids 182, 93138 or 139204 of E1A was detected. However, the fragment spanning amino acids 201289 of E1A was retained on the resin and was eluted by 0·10·2 M NaCl (Fig. 1B), which is in agreement with previous observations (Zu et al., 1992 ). These results suggest that the C-terminal portion of E1A is the only region of E1A capable of interacting with dsDNA.

To precisely identify the region(s) of E1A required for interaction with DNA, we constructed a series of mutants with deletions spanning the entire C-terminal region of E1A and tested the ability of these recombinant proteins to bind to dsDNA in vitro (Fig. 2). Only the mutant lacking amino acids 201220 failed to interact with dsDNA cellulose, suggesting that the dsDNA-binding domain of E1A resides entirely within this region (Fig. 2B). To determine if this portion of E1A was sufficient for DNA binding, we expressed amino acids 201218 of E1A as a GST fusion protein and tested it for binding to dsDNA cellulose. This fragment alone was sufficient to bind dsDNA cellulose at least as efficiently as the fragment spanning amino acids 201289 and, thus, represents the minimal portion of E1A necessary and sufficient to bind dsDNA (Fig. 2B).

(45K): Fig. 2. Identification of the minimal region of E1A required for DNA binding. (A) Recombinant GST fusion proteins containing the region spanning amino acids 201289 of E1A or the indicated deletion mutants were expressed as GST fusions in E. coli. Fusion proteins are depicted as closed bars and amino acid deletions are indicated as open bars. Numbers indicating the deleted amino acids are inclusive. (B) The recombinant GST fusion proteins depicted in (A) were allowed to bind to dsDNA cellulose and bound proteins were eluted with a stepwise gradient of NaCl as described in Fig. 1(B).

E1A binds to ssDNA, cellulose phosphate and carboxymethyl Sepharose
Chatterjee et al. (1988) reported that full-length E1A could bind dsDNA but not ssDNA or RNA. This had not been confirmed by other methodology, so we tested the ability of the C-terminal region of E1A spanning amino acids 201289 or the minimal DNA-binding domain spanning amino acids 201218 to bind ssDNA cellulose resin (Fig. 3A). In contrast to previous results, both of these fragments of E1A bound ssDNA cellulose. Recombinant GST did not bind this matrix.

(52K): Fig. 3. ssDNA cellulose, cellulose phosphate and carboxymethyl Sepharose chromatography of GSTE1A protein fusions. Recombinant GST or GST fusion proteins containing the indicated regions of E1A were allowed to bind to (A) ssDNA cellulose, (B) cellulose phosphate or (C) carboxymethyl Sepharose and bound proteins were eluted with a stepwise gradient of NaCl as described in Fig. 1(B).

Based on the observation that E1A bound both ds and ssDNA, it appeared possible that binding occurred via ionic interactions between basic residues within E1A and the phosphate groups in the DNA backbone. For this reason, we tested the minimal DNA-binding domain of E1A, spanning amino acids 201218, and the C-terminal portion of E1A lacking this region for their ability to interact with cellulose phosphate (Fig. 3B). The portion of E1A spanning amino acids 201218 bound this resin. However, no interaction with cellulose phosphate was observed for the E1A fragment spanning amino acids 221289, which lacks the DNA-binding domain. Similarly, the portion of E1A spanning amino acids 201218 bound the cation-exchange resin carboxymethyl Sepharose, whereas the GST protein did not (Fig. 3C). These results suggest that in vitro DNA binding by E1A results from an ionic interaction between the basic residues in E1A and the phosphodiester backbone of ds or ssDNA.

Growth inhibition is mediated by the C-terminal portion of E1A
Although E1A binds dsDNA in vitro, no evidence has been presented to suggest that E1A binds dsDNA in vivo. Studies in the simple eukaryote S. cerevisiae have demonstrated that expression of a chimeric protein composed of a strong transcriptional AD fused to a DNA-binding domain can inhibit yeast growth, probably by trapping general transcription factors at genomic sites (Berger et al., 1992 ; Pina et al., 1993 ; Marcus et al., 1996 ). We reasoned that fusion of a strong AD to the DNA-binding region of E1A might inhibit growth in a similar manner, providing evidence for an interaction of E1A with DNA in vivo. We constructed vectors expressing portions of E1A fused to a strong AD and tested their ability to inhibit yeast growth, as measured by colony size (Fig. 4A). Yeast expressing fragments spanning amino acids 182, 93138 or 139204 of E1A fused to an AD did not strongly inhibit yeast growth, consistent with their inability to interact with dsDNA in vitro. However, the fragment encompassing amino acids 187289 of E1A had a strong negative effect on yeast growth, as demonstrated by the overall decrease in colony size, consistent with the presence of a DNA-binding domain within this fragment. In addition, growth inhibition by the C-terminal portion of E1A was only observed when this fragment was fused to an AD (Fig. 4A). The inability of the C-terminal fragment to inhibit growth without the AD was not simply a result of a decreased level of protein expression, as determined by Western blot analysis (Fig. 4B).

(64K): Fig. 4. Growth inhibition by the C terminus of E1A in S. cerevisiae. Yeast strain w3031a was transformed with vectors expressing the indicated portions of E1A fused to an N-terminal tag consisting of an NLS, a strong transcriptional AD and an HA epitope tag. Yeast were also transformed with a similar vector expressing the C-terminal fragment of E1A that was fused to an NLS and an HA epitope tag, but not to a strong AD (w/o AD). Transformed yeast were plated on selective media, allowed to grow for 4872 h at 30 °C and photographed. (B) Western blot analysis of the level of expression of the C-terminal fragments of E1A. Protein extracts were prepared from yeast transformed with the control expression vector, with a vector expressing the C-terminal portion of E1A fused to an AD or an otherwise identical vector expressing the C-terminal portion of E1A not fused to an AD (w/o AD). Extracts were subjected to Western blot analysis with a monoclonal antibody specific for the HA epitope tag.

The effect of these same portions of E1A on yeast growth in liquid culture was measured and yeast doubling times were calculated (Table 1). As observed for growth on solid media, only expression of the C-terminal portion of E1A had an appreciable effect on yeast growth rate, increasing the doubling time by approximately 20%. This lengthening of the doubling time in liquid culture was also only observed when this portion of E1A was fused to an AD.

Table 1. Effect of Ad5 E1A expression on yeast growth rate

It was shown previously that the Gal4p DNA-binding domain fused to the herpes simplex virus VP16 AD does not inhibit growth in a yeast strain with a deletion of GCN5, which encodes a component of the ADA transcriptional regulatory complex (Marcus et al., 1996 ). In contrast, the C-terminal portion of E1A fused to an AD inhibited growth, regardless of the absence or presence of GCN5 (Fig. 5). This result is not fully consistent with the hypothesis that the C-terminal region of E1A functions as a DNA-binding domain in vivo.

(66K): Fig. 5. Growth inhibition by E1A in S. cerevisiae lacking GCN5. Yeast strains FY86 (GCN5) and FY1370 (Δgcn5) were transformed with a yeast expression vector or otherwise identical vectors expressing the C-terminal portion of E1A and analysed for growth as described in Fig. 4(A).

Growth inhibition does not require the DNA-binding portion of E1A
Our studies in yeast do not fully support the premise that the C-terminal domain of E1A functions to inhibit growth in yeast by acting as a DNA-binding domain. Therefore, we determined whether the minimal dsDNA-binding region identified in vitro was necessary and sufficient for growth inhibition. We constructed yeast vectors expressing only the minimal DNA-binding domain, corresponding to residues 201218, and a deletion mutant lacking residues 205221 (Δ205221) of E1A (Fig. 6A) and tested them for the ability to inhibit yeast growth when fused to an AD (Fig. 6B; Table 1). Mutant Δ205221, which lacks the DNA-binding region, inhibited yeast growth on solid or liquid media as well as the wild-type E1A C-terminal fragment. In addition, the minimal DNA-binding region of E1A spanning residues 201218 was not sufficient to inhibit growth in either circumstance. These results suggest that although the C-terminal region of E1A can inhibit yeast growth when fused to a strong transcriptional AD, this is not mediated by the DNA-binding region identified in vitro.

(61K): Fig. 6. Growth inhibition in S. cerevisiae is independent of the DNA-binding region of E1A. (A) Schematic representation of the E1A mutants tested for growth inhibition in yeast. All fragments were expressed as fusions to an N-terminal tag consisting of an NLS, a strong transcriptional AD and an HA epitope tag. The amino acid deletions are indicated as open bars and numbers are inclusive. (B) Effect of the E1A mutants on yeast growth. Yeast were transformed with vectors expressing the fragments of E1A described in (A) and analysed for growth as described in Fig. 4(A).

To determine which regions within the C-terminal portion of E1A were required for growth inhibition, we transformed yeast with vectors expressing a strong AD fused to the same series of deletion mutants described in Fig. 2(A). All of the mutants tested inhibited yeast growth to some extent. However, with the exception of mutant Δ255270, none of the mutants was as effective as wild-type E1A (Fig. 7A). Western blot analysis demonstrated that only fragments 201289 and 221289 were expressed at a substantially reduced level with respect to the wild-type 187289 fragment (Fig. 7B).

(59K): Fig. 7. Growth inhibition in S. cerevisiae by mutants in the C-terminal portion of E1A. (A) Yeast were transformed with vectors expressing the C-terminal portion of E1A containing the indicated deletions fused to an N-terminal tag consisting of an NLS, a strong transcriptional AD and an HA epitope tag and analysed for growth as described in Fig. 4(A). All deletions are inclusive. (B) Western blot analysis of the level of expression of the mutants containing deletions within the C-terminal portion of E1A. Protein extracts were prepared and analysed as described in Fig. 4(B).

Previous reports have shown that E1A can bind dsDNA in a sequence-independent fashion (Chatterjee et al., 1988 ; Zu et al., 1992 ). We tested the ability of various portions of E1A that collectively span the coding region (Fig. 1A) to bind dsDNA in vitro. Of the four regions of E1A tested, only the C-terminal portion of E1A was able to interact with dsDNA (Fig. 1B). This is consistent with previous reports demonstrating that the C terminus of E1A was sufficient to bind dsDNA (Zu et al., 1992 ). Interestingly, we also determined that this region of E1A can interact with ssDNA as efficiently as it can with dsDNA (Figs 2 and 3). This has not been observed previously, perhaps because of differences in methodology (Chatterjee et al., 1988 ). Using a series of recombinant E1A proteins with small deletions that collectively span the C-terminal portion of E1A, we determined that a mutant protein lacking amino acids 201220 failed to bind DNA (Fig. 2B). Although previous work similarly showed that a mutant protein with a deletion in this region failed to bind dsDNA (Zu et al., 1992 ), it did not rule out a requirement for additional portions of E1A. In contrast, our results demonstrate conclusively that only this small region of E1A is essential for binding. Furthermore, we have shown that a recombinant protein containing amino acids 201218 of E1A was not only necessary but sufficient to interact with both ds and ssDNA (Figs 2B and 3A). This demonstrates that this 18 residue peptide comprises an isolated nucleic acid-binding module that can function independently of the remainder of the E1A protein. This portion of E1A contains six basic amino acids and four prolines and represents the most basic region of E1A. Taken together, these results suggest that the binding to ds or ssDNA is mediated by an ionic interaction between E1A and the phosphates in the DNA backbone. This is supported by our observation that the portion of E1A spanning amino acids 201218 bound the negatively charged cellulose phosphate and carboxylmethyl Sepharose resins (Fig. 3, B and C), whereas the 221289 mutant, which is unable to interact with dsDNA, also fails to interact with cellulose phosphate resin (Fig. 3B).

Although it is well established that E1A can bind dsDNA in vitro, we wished to determine whether this interaction occurred or was relevant in vivo. As E1A does not bind dsDNA in a sequence-specific fashion (Chatterjee et al., 1988 ; Zu et al., 1992 ), we took advantage of previous studies in the simple eukaryote S. cerevisiae, which demonstrated that fusion of a strong transcriptional AD to a DNA-binding domain can inhibit yeast growth. This probably occurs by sequestering limiting general transcription components at genomic sites, as growth inhibition can be overcome by mutation of transcriptional co-activator proteins, such as Gcn5p (Berger et al., 1992 ; Pina et al., 1993 ; Marcus et al., 1996 ). We reasoned that fusion of a strong AD to the DNA-binding region of E1A might inhibit growth in a similar manner, providing evidence for an interaction of E1A with DNA in vivo. We tested various fragments of E1A fused to a strong AD (Fig. 4A) for their ability to inhibit yeast growth and observed that only a fragment encompassing the C terminus of E1A had an effect on yeast growth (Fig. 4B; Table 1). In addition, growth inhibition by this portion of E1A was only observed when this fragment was fused to an AD (Fig. 4B; Table 1). These results are fully consistent with the presence of a DNA-binding domain that functions in combination with a strong AD to inhibit growth by trapping limiting general transcription factors.

However, further tests rule out this interpretation of the data. Firstly, growth inhibition by the C terminus of E1A was not dependent on the presence of Gcn5p, which differs from previous results using the Gal4p DNA-binding domain fused to a strong AD (Fig. 5). Secondly, the minimal DNA-binding domain was not sufficient to inhibit yeast growth when fused to a strong transcriptional AD (Fig. 6B; Table 1). Thirdly, mutant Δ205221, which lacks the DNA-binding region, inhibited yeast growth as well as the wild-type fragment of E1A (Fig. 6B; Table 1). These results indicate that the C-terminal region of E1A does not bind DNA in vivo and suggest that growth inhibition by this portion of E1A occurs via an alternative mechanism. Interestingly, mutational analysis demonstrated that growth inhibition requires most of the C-terminal region of E1A, with the exception of residues 205221 and 255270 (Figs 6 and 7A; Table 1). This pattern of activity does not coincide with the binding to CtBP or Yak1p, the two known cellular proteins targeted by the C-terminal portion of E1A (Schaeper et al., 1995 ; Zhang et al., 2001 ), suggesting that interaction with additional, as yet unknown, cellular factors may be responsible for growth inhibition. However, the lower level of expression of mutants with deletions within residues 187221 makes it difficult to unequivocally determine if this region is required for growth inhibition (Fig. 7).

Taken together, our results raise the distinct possibility that the observed binding of E1A to DNA is an artefact of in vitro analyses, resulting from a simple ionic interaction of a basic portion of E1A with the phosphodiester backbone of DNA and that additional cellular proteins are targeted by the C-terminal portion of E1A.

We thank Dr J. Torchia for advice and Jay Loftus for technical assistance. We thank Drs G. Chinnadurai, C. Brandl, M. M. Smith and F. Winston for generous gifts of plasmids and yeast strains. This work was supported by grants from the Canadian Institutes of Health Research, The London Health Sciences Centre and The University of Western Ontario Academic Development Fund awarded to J.S.M., who is a Scholar of the Canadian Institutes of Health Research. N.A. holds a McLauchlin Foundation studentship.