Cell, Vol. 67, 365-376, October18, 1991,Copyright© 1991 by Cell Press

Adenovirus E IA Activation Domain Binds the Basic Repeat in the TATA Box Transcription Factor Wes S. Lee, C. Cheng Kao,* Gene O. Bryant, Xuan Liu, and Arnold J. Berk Department of Microbiology and Molecular Genetics Molecular Biology institute University of California, Los Angeles Los Angeles, California 90024-1570

Summary The adenovirus large E I A protein is a potent activator of transcription. We use several different experimental approaches to demonstrate that the large E I A protein binds specifically and stably to the TATA box-binding factor (TFIID), the general polymerase II transcription factor that initiates assembly of transcription complexes. Sedimentation velocity centrifugation revealed that TFIID and E I A form a heterodimer in vitro. We demonstrate that the activation domain of E I A (conserved region 3) binds to TFIID. EIA interacts with a 51 residue region from the conserved C-terminal domain of TFIID that includes a repeat of basic residues between the homologous direct repeats of TFIID. Analysis of TFIID binding by various EIA mutants indicates that TFIID binding is necessary, although not sufficient, for E I A transactivation. Introduction The adenovirus large E1A protein is a prototype promiscuous transcriptional transactivator. It stimulates transcription from the five early adenovirus promoters, even though they have no common cis-acting DNA sequence required for activation (reviewed in Berk, 1986; Flint and Shenk, 1989). Moreover, cellular promoters such as the 13-globin, E-globin, and preproinsulin I promoters are also transactivated by E1A. Large EIA protein also stimulates transcription from simple synthetic promoters containing only a single TATA box, a single cAMP response element, or two binding sites for the E2F transcription factor (Pei and Berk, 1989). Consistent with this lack of sequence specificity in function, the isolated large E1A protein has only very low affinity for DNA and no apparent sequence specificity (Chatterjee et al., 1988). How can the large E1A protein stimulate transcription from so many different promoters? Experimental results support several models that are not mutually exclusive. In one model, E1A stimulates the activity of a number of different transcription factors. Consistent with this, expression of E1A is correlated with increased phosphorylation and activity of transcription factors E2F (Bagchi et al., 1989), E4F (Raychaudhuri et al., 1989), and TFilIC (Hoeftier et al., 1988). E1A also works synergisticallywith cAM P in the $49 lymphoma cell line to induce genes encoding * Present address: Departmentof Plant Pathology, Universityof Wisconsin, Madison,Wisconsin53706.

transcription factor polypeptides junB and c-fos (Muller et al., 1989). In a second model, for which there is considerable genetic evidence, large E1A protein binds to transcription factor ATF-2, which in turn binds to upstream promoter elements found in several early adenovirus promoters (Liu and Green, 1990). This is proposedto tether a strong activating domain of the large EIA protein to the promoter region, where it can interact with the basic transcriptional machinery (Lillie and Green, 1989). In a third model, E1A activates a general transcription factor, TFIID, required for transcription from most promoters. TFliD binds to the TATA box promoter element and initiates a cascade of assembly of general transcription factors and RNA polymerase II (reviewed in Sawadogo and Sentenac, 1990). In support of this model, certain TATA box sequences can mediate activation by E1A (Wu et al., 1987; Simon et al., 1988), and adenovirus infection increases the transcriptional activity of a partially purified TFIID fraction from HeLa cells (Leong et al., 1988). Any rigorous test of the third model, in which TFliD is a target of E1A transactivation, requires direct characterization of TFIID and any interactions that E1A may have with it. To generate the reagents for such analyses, we cloned .cDNAs encoding TFliD, first from yeast (Schmidt et al., 1989) and then from humans (Kao et al., 1990). Other groups have also cloned cDNAs for this important transcription factor from yeast (reviewed in Sawadogo and Sentenac, 1990) and humans (Hoffman et ai., 1990; Peterson et al., 1990). in this article, we report that the large E1A protein binds specifically to a region in the conserved domain of human TFIID. We show that the small E1A protein expressed from the viral 12S mRNA, which is a much poorer transactivator (Flint and Shenk, 1989), binds human TFIID much more weakly than large E1A protein. Moreover, we demonstrate that the activation domain of large E1A protein binds to TFIID. Furthermore, mutations in the activation domain that diminish TFIID binding in vitro have previously been shown to severely impair E1A activation in vivo. We conclude that the large E1A protein stimulates transcription in part through this direct interaction with TFIID. Results Coimmunoprecipitation of E I A and TFIID We tested the possibility that E1A might regulate TFIID activity by binding to it directly. To simplify the analysis, TFIID and the large E1A protein were expressed at a high level in HeLa cells using the vaccinia virus-T7 RNA polymerase vector system developed by Eiroy-Stein et al. (1989). In addition to expressing the proteins at a high level, this system has the advantage of supplying mammalian cell-specific posttranslational modifications that might be required for such an interaction. The vector system depends on transcription of a cDNA inserted into one vaccinia virus recombinant by T7 RNA poiymerase

Cell 366

aTF]ID

n

aEIA a19K d---I

+

~ t~ ¢.,J i=1

+

÷

÷

C~{o-

97-

.....

-9797`-

68-

~tjBi b ,al ,=llJ

_ 6868 -EtA

-TF~D

-29

29-

29

-18 18-

fSt 2 ~4 t

23

567

t

23

4

Figure 1. Coimmunoprecipitation of Large E1A and Human TFIID (A) Autoradiogram of an SDS-polyacrylamide gel of total protein labeled from 20 to 23 hr postinfection with vaccinia virus recombinants VE1A plus vTF7-3 (lane 1), VE1A plus VliD plus vTF7-3 (lane 2), or VIID plus vTF7-3 (lane 3). (B) Immunoprecipitates of the extracts shown in (A). Lanes 1-3 were immunoprecipitated with anti-TFIID rabbit serum. Lanes 4 - 6 were immunoprecipitated with anti-E1A rabbit serum. Lane 7 was immunoprecipitated with control anti-Ad2 E1B 19K protein rabbit serum. (C) Lane 1: total labeled protein of an extract prepared as in lane 2 of (A). Lane 2 : 5 p.g of purified Abl protein was added to this extract, which was then immunoprecipitated with anti-Abl rabbit serum. Lane 3: extract shown in lane I was immunoprecipitated with anti-human TFIID rabbit serum. Lane 4: extract shown in lane I was immunoprecipitated with control antiserum to the E1B 19 kd protein.

expressed from a second, coinfected vaccinia virus recombinant. Figure 1 (lane 1) shows ~S-labeled proteins extracted from HeLa cells coinfected with the T7 RNA polymerase vector (vTF7-3) and the vector expressing large E1A protein (VE1A). Most of the labeled proteins are late vaccinia virus proteins (Elroy-Stein et al., 1989), including virion structural proteins and their unprocessed precursors (Pennington, 1974). The large E1A protein is also one of the major labeled proteins. It separates into more than one band because of different extents of phosphorylation and migrates more slowly in a sodium dodecyl sulfate (SDS)poiyacrylamide gel than expected for its actual molecular mass of ,'~32 kd (Harlow et al., 1985). Lane 3 shows the profile of proteins expressed from the vaccinia virus vector producing TFIID (VIID). The indicated TFIID protein accumulates to a higher level than the E1A protein produced by VE1A. This is probably due to the short half-lite of E1A protein in HeLa cells (Spindler and Berk, 1984). Minor forms of TFIID are observed with slightly slower mobility than the main TFIID band (better visualized in lane 2 at this exposure). The main TFIID band comigrates with TFIID expressed in Escherichia coli (data not shown), suggesting that the minor, slower mobility forms may result from posttranslational modifications. Lane 2 shows the profile of proteins expressed in HeLa cells infected with an equal multiplicity of infection of VE1A and VIID. E1A and TFIID proteins are again apparent, but they accumulate to

lower levels than in cells infected with the individual vectors, probably because of competition for DNA replication, transcription, and translation between the two vectors. To search for a possible interaction between the large E1A protein and TFIID, the samples shown in Figure 1A were subjected to immunoprecipitation with specific antisere raised against either the large E1A protein or human TFIID (Figure 1 B). Immunoprecipitation of the extract from cells infected with both VIID and VE1A ~ith anti-TFIID serum precipitated E1A protein as well as TFIID (Figure 1B, lane 2). This was not due to cross-reaction of the antiTFIID serum with E1A, since E1A was not precipitated by this serum from the extract of cells infected with VE1A alone (Figure 1B, lane 1). As a control, the extract from cells infected with both VE1A and VIID was also subjected to immunoprecipitation with an antiserum raised against the E1B 19K protein, which is not expressed in these cells (lane 7). This control immunoprecipitation brought down only a trace amount of TFIID and no E1A. These results indicate that E1A protein was associated with TFIID protein in the extract from cells expressing both proteins. Consistent with this, anti-E1A serum immunoprecipitated both E1A and TFIID from the extract of cells expressing both proteins (lane 5). This was not due to cross-reaction of the anti-E1A serum with TFIID, since immunoprecipitation of the extract from cells expressing TFIID alone with this antiserum brought down only trace amounts of TFIID (lane 6) at a level comparable to that seen with the control anti19K serum (lane 7). No TFIID is apparent in lane 4 of Figure 1B, even though endogenous TFIID was present in these cells, because endogenous proteins are not labeled under these conditions (see Figure 1A). Vaccinia virus late proteins of "~95 and ',,65 kd were also brought down in each of the immunoprecipitations in which an immunoprecipitate was formed (Figure 1B, lanes 2-5). The extent of nonspecific precipitation of these polypeptides was variable from experiment to experiment. To determine whether the coprecipitation of E1A with TFIID could be due to a similar type of nonspecific interaction, we. performed the control experiment shown in Figure 1C. Five micrograms of purified, unlabeled P185 Abelson protein (Abl) was added to an extract from cells coinfected with VIID and VEIA (shown in lane 1). This extract was then immunoprecipitated with anti-Abl (lane 2), anti-TFIID (lane 3), or control anti-19K (lane 4) sera. Vaccinia virus proteins were not detected in the immunoprecipitate with control anti-19K serum. Equal amounts of the ~,65 kd vaccinia virus protein were detected in the immunoprecipitates with both anti-Ab! and anti-TFIID sera. This small fraction of the ,~65 kd vacoinia protein was presumably trapped nonspecifically in these immunoprecipitates. However, E1A protein was only detected in the immunoprecipitation with anti-TFIID serum. We conclude that E1A was immunoprecipitated because it was specifically bound to TFIID. The ratio of E1A to TFIID appears less in Figure 1C, lane 3, than in Figure 1B, lane 2, because the ratio of total EIA to total TFIID was less in the experiment shown in Figure 1C than in the experiment shown in Figures 1A and lB. The E1A-TFIID complexes were relatively stable. In the

AdenovirusE1A BindsTFIID 367

silver- stoin I

p.)l

i1~

J~l

~

western

F--

n"

F.-

i

i

m

i

for western

r

P--

i

i

.~

i

>

"

kd - 97.4

~ ~

RB-

"~

-68.0

-- 43.0

TF~D-

~

- =i ~

~

t 2 3

-- 29.0

456

7

89

Figure 2. LargeEIA Bindsto HumanTFIIDand RB in a Far-Western Blot Lanes 1-3: silver-stainedlanes of an SDS gel of (lane 1) a partially purified preparationof the glutathioneS-transferase-RBfusion protein expressedfrom pGT-RB(379-928)(Kaelinet aL, 1991);(lane2) an extract of HeLa cells infectedwith VIID and vTF7-3; or (lane 3) an extract of HeLacells infectedwith vTF7.-3alone. Lanes4-6: Westernimmunoblotof lanesfromthe samegel equivalent to lanes 1-3, analyzedwith anti-humanTFIID rabbit serum and goat anti-rabbitimmunoglobulincoupledto alkalinephosphatase. Lanes7-9: far Western of the samenitrocelluloseblot shownin lanes 4-6. The nitrocelluloseblot was incubatedwith ~S-labeledin vitro translated large EIA as described in ExperimentalProcedures.An autoradiogramof the washedblot is shown. The autoradiogramwas preparedbeforesubjectingthe blotto the Westernprocedureas shown in lanes4-6.

experiment shown in Figure 1B, the immunoprecipitates were washed with a buffer containing 0.05% SDS, a concentration of detergent that disrupts many nonspecific interactions. The E1A-TFIID coprecipitates were also stable to washing with 1 M LiCl. However, the E1A-TFIID interaction was not covalent. Washing with 0.1% SDS partially disrupted the complexes, and washing with 0.2% SDS at room temperature completely disrupted the complexes, so that the anti-E1A serum pellets contained only E1A and the anti-TFlID serum pellets contained only TFIID when prepared from cells coinfected with VE1A and VIID (data not shown). The E1A-TFIID interaction required native protein structure, since it was disrupted by heat denaturation. E1A crid not coprecipitate with TFIID, and vice versa, when the extracts containing both proteins were incubated at 51°C for 15 min (data not shown). T F I I D - E I A Interaction D e t e c t e d by F a r - W e s t e r n Protein Blotting

To provide additional evidence for the specific interaction of large E1A protein with TFIID inferred from the coimmunoprecipitation experiments, we performed protein blotting experiments (Figure 2). Extracts from HeLa cells coinfected with the VIID vector plus the vector expressing T7

RNA polymerase (vTF7-3) (lanes 2, 5, and 8) or from Hela cells infected with the vTF7-3 vector alone (lanes 3, 6, and 9) were resolved on an SDS-polyacrylamide gel. As a positive control for this experiment, we also ran on the same gel a partially purified preparation of an RB fusion protein expressed in E. coil (Kaelin et al., 1991) (lanes 1, 4, and 7). Others have shown that E I A protein forms a stable, specific complex with the RB protein (Whyte et al., 1988) and binds to this bacterially expressed fusion protein (Kaelin et al., 1991). Proteins detected by silver staining are shown in lanes 1-3, with TFIID and the RB fusion protein indicated. Neighboring lanes from the same gel were transferred to nitrocellulose paper, subjected to a denaturation-renaturation protocol, and incubated with in vitro translated ~S-labeled large E1A protein. TFIID protein was visualized on the same blot using anti-TFIID serum followed by colorometric detection of bound antibody. This Western blot is shown in lanes 4-6, and an autoradiogram of the blot showing bound ~S-labeled large E1A protein ("far Western") is shown in lanes 7-9. ~S-labeled large E1A protein bound to the RB fusion protein and a proteolytic fragment of the fusion protein (lane 7). E1A protein also bound to TFIID (lane 8). The position of the TFIID band in lane 8 was readily confirmed by overlaying the Western blot with the autoradiogram. It should be noted that the Western blot shown in lanes 4-6 was developed beyond the linear range of the procedure, so that a minor, _slower migrating form of TFIID in lane 5 appeared disproportionately dark compared with the major band. This can be seen by comparison with the silver-stained TFIID band in lane 2. This was not observed in other Western blots of TFIID with the same antiserum. The far-Western blot demonstrates the specificity of the interaction of E1A with TFIID. Multiple other proteins were present on the blot in higher concentration than TFIID, yet only TFIID and the RB fusion protein bound E1A to a significant extent. Furthermore, at least as much E1A bound to TFIID as bound to RB (Figure 2C) when roughly equal amounts of RB and TFIID were applied to the blot (Figure 2A). E I A and T F I I D Form a H e t e r o d i m e r

The size of the E 1A-TFIID complex was estimated by sedimentation velocity centrifugation. Figure 3 shows an experiment in which in vitro translated E1A protein was incubated with TFIID expressed in and partially purified from E. coll. The amount of TFIID used in the experiment was titrated so as to use the minimal amount of TFIID required to form a complex with all the E 1A added. In vitro translated E1A sedimented as expected for its molecular mass, 32 kd. The peak of E1A was in fraction 12, just ahead of the 29 kd carbonic anhydrase marker (Figure 3). TFIID sedimented more slowly than expected for a globular protein of its molecular mass, 38 kd. It peaked in fractions 13 and 14, just behind the carbonic anhydrase marker (Figure 3). We also observed that TFIID eluted from a Superose 12 gel filtration column more rapidly than expected for its molecular mass, at a position corresponding to ,~50 kd relative to bovine serum albumin (BSA) (67 kd) and carbonic anhydrase (data not shown). The slow sedimenta-

Cell 368

6 7 8

9 10 11 12 13 14 15 16

E1A

-- E1A

hemoglobin (65 kd), corresponding to a molecular mass of ,~,60 kd for a globular protein. The position of TFIID in the same gradient was determined by Western blotting (Figure 3, bottom). TFIID protein was also shifted to a higher S value after incubation with EIA and, importantly, cosedimented with EIA protein in the same gradient. Incubation of TFIID with an equivalent amount of reticulocyte extract without translated E1A did not alter the sedimentation of TFIID. These results establish several important points. First, the E1A-TFIID complex formed a peak during velocity sedimentation centrifugation as sharp as that observed for uncomplexed EIA, uncomplexed TFIID, and the marker proteins. This indicates that the EIA-TFIID complex is a discrete molecular species. Second, the S value of the E1A-TFIID complex is most consistent with a one-to-one complex of the two proteins, i.e., a heterodimer. A higher order complex would be expected to sediment more rapidly than hemoglobin. Third, no other proteins besides E1A and TFIID appear to be part of the complex. If additional unlabeled proteins present in the reticulocyte lysate were incorporated into the complex, we would have expected the complex to sediment faster than hemoglobin. However, we cannot rule out the possibility that a very small polypeptide is included in the complex in addition to EIA and TFIID. Finally, the results show that under these conditions, the interaction between E1A and TFIID is quantitative; virtually all the E1A and TFIID proteins interact with each other. "

E1A + TFIID

-- E1A

6 7 8

9 10 11 12 13 14 15 16

TFIID

-- TFIID

TFIID + E1A

-- TFIID

I 65 kd

I 29 kd

Figure 3. Sedimentation Velocity Centrifugation of the EIA-TFIID Complex In vitro translated large E1A protein (5 Id of reticulocyte extract) was incubated with TFIID purified from E. coli in 100 ~.1 of 0.5 M KCI, 20 mM Tris (pH 7.9), 5 mM MgCI2, 0.1 mM ZnCI2, 1 mM dithiothreitol (binding buffer) and incubated at 25°C for 30 rain. The solution was then layered on a 5%-20% sucrose gradient made in binding buffer and centrifuged in an SW50.1 rotor at 46 krpm for 24 hr at 4°C. The gradient was dripped into 20 fractions, which were TCA precipitated, subjected to SDS-PAGE, and blotted onto nitrocellulose. An autoradiogram of the blot was prepared to detect ~S-labeled E1A (second panel from the top). The Western procedure was performed with antiTFIID antiserum to detect TFIID (bottom panel). In vitro translated large EIA protein (5 ~1) was incubated for 30 rain at 25°C in binding buffer without added TFIID and sedimented as above during the same centrifugation. Fractions were precipitated with TCA, subjected to SDSPAGE, the gel was dried, and an autoradiograph was prepared to detect ~S-lebeled E1A (top panel). TFIID was incubated for 30 rain at 25°C in binding buffer and sedimented as above during the same centrifugation (third panel). TFIID was detected by Western blotting as for the bottom panel. Hemoglobin (65 kd) from the reticulocyte lysate peaked in fractions 9 and 10. Marker proteins were run in a parallel gradient. Carbonic anhydrase (29 kd) peaked in fractions 12 and 13. B-Amylase (200 kd) peaked in fraction 1. Fractions 6-16 are shown. Neither E1A nor TFIID was detected in the fractions not shown.

tion coefficient and rapid elution during gel filtration indicate that TFIID has an asymmetric rather than globular conformation. When E1A was mixed with TFIID, incubated, and then sedimented, it shifted to a higher S value with a peak in fraction 10 (Figure 3). This was just behind the peak of

Mapping the Region of Interaction in TFIID To facilitate mapping the region of TFIID that interacts with E1A, we tested the ability of in vitro translated TFIID to bind to large E1A protein in solution. In vitro translated 3sS-labeled TFIID (Figure 4, lane 1) was incubated with an unlabeled extract of HeLa cells infected with VE1A and vTF7-3. Immunoprecipitation with anti-E1A monoclonal antibody M73 coprecipitated labeled TFIID (Figure 4, lane 2). TFIID was not precipitated by M73 following incubation ~i!h a control extract from cells infected with vTF7.3 alone (lane 3). Furthermore, control monoclonal antibody to E. coil ~-galactosidase did not immunoprecipitate labeled TFIID after incubation in the VEIA plus vTF7-3 extract, even after addition of unlabeled ~-galactosidase protein in an attempt to trap TFIID nonspecifically in an immunoprecipitate (lane 4). A control in vitro translated protein, the Zta transcription factor encoded by Epstein-Barr virus (Lieberman and Berk, 1990, and references therein; shown in lane 5), was not coprecipitated by the anti-E1A monoclonal antibody (lane 6). These results provide additional evidence that E1A can bind to TFIID in vitro. To map the region of TFIID that interacts with EIA, a number of TFIID mutants were translated in vitro and assayed by the immunoprecipitation protocol described above. The primary sequence of human TFIID is diagrammed at the bottom of Figure 4. The conserved 180 residue C-terminal sequence, which is 80% identical to Saccharomyces cerevisiae TFIID (Kao et al., 1990), is represented by a rectangle. The nonconserved N-terminal region of TFIID is represented by a line. Arrows below the

Adenovirus E1A Binds TFIID 369

extract : -

EIA vT ~.~1

-

EIA

-

EIA vT

-

EIA vT

-

Aba : -

EIA EtA B-~l

-

EtA

-

EtA EIA

-

EIA EIA

-- EIA EtA

labeled protein :r

TFIID

II zlo

II

ANED

~1 StuI ]ZD

EtA vT

II SspIT[D

I

kd - 43

Ill m

¸

-d_~

O 9

~