Induction of AP-2 a Expression by Adenoviral Infection Involves Inactivation of the AP-2rep Transcriptional Corepressor CtBP1*

AP-2 transcription factors execute important functions during embryonic development and malignant transformation. Recently, we have isolated a transcriptional repressor of AP-2 a expression, the novel Kru¨ppel-related zinc finger protein AP-2rep (Klf12). Here, we show that repression of AP-2 a transcription by AP-2rep is dependent on an N-terminal PVDLS motif that interacts specifically with the corepressor CtBP1 both in vivo and in vitro . This interaction motif was previously identified in the C-terminal region of the adenoviral oncoprotein E1A. Infection of both HeLa and PA-1 cells with adenovirus type 5 strongly induced AP-2 a mRNA. Con-sistently, E1A was necessary and sufficient to mediate up-regulation of AP-2 a . Transiently transfected wild-type E1A protein activated an AP-2rep sensitive cis-reg-ulatory element of the AP-2 a promoter, but E1A protein harboring a mutation in the PVDLS motif failed to activate. In summary, we conclude that the adenoviral oncoprotein promoter-driven was amplified primers GCG TTC ATG GGC AGC TCC CAC GAA was amplified

The basic helix-loop-helix transcription factor AP-2␣ represents the prototype of a small family of closely related and evolutionarily conserved DNA-binding proteins harboring helix-loop-helix dimerization motifs, AP-2␣, AP-2␤, and AP-2␥ (1)(2)(3). Highly regulated expression patterns of AP-2␣ have been described during embryonic development of the neural tube, neural crest derivatives, eye and face, and limb bud, as well as urogential and ectodermal tissues (4 -6). Recently, germ line mutations in the human AP-2␤ gene have been shown to cause Char syndrome, a familial form of patent ductus artiosus involving a specific defect in thoracic neural crest cells (7). Studies of AP-2-deficient mice reveal specific defects in the neural tube, craniofacial structures, and kidney, suggesting that AP-2 genes execute essential functions in regulating specific gene expression programs and cell survival during embryonic development (8 -11). Consistently, a number of target genes that are transcriptionally activated or silenced in a cell type-specific fashion have been identified (12)(13)(14).
Specific expression patterns of AP-2 genes have further been implicated in malignant transformation and stress response of mammalian cells. AP-2␣ overexpression was found to result from and to be functionally involved in N-ras-mediated malignant transformation of PA-1 human teratocarcinoma cells (15). AP-2␣ and AP-2␥ overexpression in breast cancer cells direct transcriptional activation of the transmembrane tyrosinasecoupled receptor c-erbB2 (16) and correlate with regulation of multiple growth factor signaling pathways (17). A role of AP-2 in the mammalian stress response following ultraviolet A radiation has been identified (18). Specific interaction of AP-2 with the c-Myc-Max heterodimer negatively regulate c-Myc target genes and c-Myc-induced apoptotic cell death (10,19). Therefore, it has been proposed that AP-2 genes are involved in programming cell survival, particularly in fast-proliferating cells under stress conditions or under conditions of limited external growth factor supply that occur particularly during embryonic development and in neoplastic tissues. We have previously studied transcriptional mechanisms controlling the activity of the AP-2␣ promoter and identified a network of activating and silencing factors (20,21). The novel Krü ppel-related zinc-finger repressor factor AP-2rep (Klf12) mediated down-regulation of AP-2␣ expression (22). Interestingly, AP-2rep harbors a putative interaction domain for the transcriptional corepressor CtBP1, which also binds to the C terminus of the adenoviral oncoprotein E1A. Viral oncoproteins frequently reinstruct endogenous gene regulation and provide insights into molecular mechanisms of malignant transformation. In this study, we therefore investigated whether adenoviral infection elicits alterations in AP-2␣ expression patterns and explored the putative role of AP-2rep and CtBP1.

EXPERIMENTAL PROCEDURES
Cell Culture and Transient Transfections-HeLa and PA-1 clone 9117 (15) cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Sigma). Transient transfections were performed using a standard calcium coprecipitation protocol as described previously (16). Luciferase activity was assayed as recommended by the manufacturer (Promega, Mannheim, Germany) in a Luminometer Lumat LB 9501 (Berthold, Wildbad, Germany). Relative light units were normalized to ␤-galactosidase activity and protein concentration. All experiments were repeated at least three times, with standard deviations less than 10%.
Reporter and Expression Plasmids-AP-2 promoter constructs have been described previously, as well as the cytomegalovirus promoterdriven AP-2rep expression plasmid (21). The same cytomegalovirus promoter-driven vector (pCMX) vector was used to construct CtBP1 and E1A 13 S expression plasmids. The entire coding sequence of CtBP1 was amplified from 10 ng of plasmid DNA using the sense and reverse primers GCG GAA TTC ATG GGC AGC TCC CAC TTG C and GCG GAA TTC CTA CAA CTG GTC ACT CG, respectively. The E1A 13 S sequence was amplified using the primers GGC AAG CTT ATG AGA CAT ATT ATC TGC and GGC AAG CTT TTA TGG CCT GGG GCG TTT. The CtBP1 PCR 1 fragment was digested with EcoRI, E1A was digested with HindIII, and both were ligated into pCMX-PL1 and verified by sequencing the entire open reading frame.
Site-specific Mutagenesis-The CtBP-interacting motifs of AP-2rep and E1A (PVDLS and PLDLS) were mutated to PVASS and PLASS, respectively, with the transformer mutagenesis kit following precisely the manufacturer's protocol (CLONTECH, Palo Alto, CA). The following primers were used: AP-2rep mutation primer, GAG CCA GTG GCT AGC TCA ATC AAC; AP-2rep selection primer, GAA TTC GAT ATC AAA CTT CTG GAG; E1A mutation primer, CAA CCT TTG GCT AGC AGC TGT AAA CGC; E1A selection primer, GTC TGC AGG ACG ACT CTA. All mutations were verified by sequencing.
In Vitro Binding Assays-Murine CtBP1 was cloned into the vector pCMX-PL1, and 35 S-labeled in vitro translated protein was generated using T7 polymerase and the Promega TNT system. Glutathione Stransferase (GST) fusion proteins were purified as previously described (21). Protein concentrations were estimated on a Coomassiestained SDS-polyacrylamide gel. Approximately equal amounts of GST fusion protein were mixed with 35 S-labeled in vitro translated protein in HBBNP buffer (20 mM HEPES/HCl, pH 7.8, 50 mM KCl, 5 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Nonidet P-40) and incubated at 4°C for 2 h. Beads were washed four times in HBBNP buffer, and bound proteins were separated on a polyacrylamide gel and visualized autoradiographically.
Two-hybrid Screening and Two-hybrid Assays-The CLONTECH two-hybrid system was used according to the manufacturer's instructions. A murine embryonal (day 17.5) brain cDNA library in the gal4AD fusion vector, pGAD10, was transfected into the yeast strain Y190 harboring the gal4DBD-AP-2rep NT (N-terminal) fusion protein expressed from pAS2-1. 65 mCtBP1 clones differing in length were isolated. Interaction assays between mCtBP1 and AP-2rep NT were performed by co-transfecting pAS2-1/AP-2rep NT together with pGAD10/ mCtBP1 into Y190. Colonies were then transferred in liquid cultures, and ␤-galactosidase activity was quantified.
Adenoviral Infections-Adenoviral stock solutions were prepared from 293 cells using a standard protocol (24). HeLa and PA-1 clone 9117 cells were infected with adenoviruses by growing cells to 50% confluency and incubating them with adenoviral suspension in a minimal volume of Dulbecco's modified Eagle's medium without fetal calf serum for 1 h and adding of a larger volume of fetal calf serum-containing medium for 12 to 72 h. Cells were harvested after two phosphatebuffered saline wash steps in a minimal volume of phosphate-buffered saline and were resuspended in Laemmli buffer for Western blot analysis. Wild-type adenovirus-5 and the following mutated viruses were used: H5 dl312 (lacking E1A) and H5 dl1135 (lacking amino acids 225-238 of E1A exon 2, including the PLDLS sequence at 233-237 (25)).
Coimmunoprecipitation-For coimmunoprecipitation in vitro, the AP-2rep coding sequence was ligated into the EcoRI site of pCMC-PL2 containing a Kozak ATG translation start sequence followed by a FLAG-tag motif. The construct was in vitro transcribed by T7-RNApolymerase (Stratagene, Heidelberg, Germany) and then in vitro translated. The FLAG-tagged protein was coincubated with [ 35 S]methionine-labeled AP-2␣, CtBP1, AP-2rep-FLAG, or unmodified AP-2rep and precipitated by an anti-FLAG-antibody (Sigma) and protein A-agarose. Finally, the [ 35 S]methionine-labeled proteins were visualized on SDS-polyacrylamide gel electrophoresis gels and subsequent autoradiography.
For coimmunoprecipitation from cells, the FLAG-tagged AP-2rep construct was transiently transfected into 1 ϫ 10 7 HeLa cells. The cells were lysed in 1 ml of radioimmune precipitation buffer (Roche Molecular Biochemicals) and precleared by adding 200 l of protein G-Sepharose (Amersham Pharmacia Biotech) and shaking at 4°C for 6 h. The Sepharose was pelleted, and then 10 g of anti-FLAG-antibody was added to the lysate and incubated overnight at 4°C. Finally, 20 l of fresh protein G-Sepharose was added for 1 h, recovered by centrifugation, and washed three times with phosphate-buffered saline. The final pellet was resuspended in 20 l of Laemmli buffer, incubated at 95°C for 5 min, separated by SDS-polyacrylamide gel electrophoresis, and Western blotted. Western blots were immunoprobed with anti-CtBP antiserum to detect copurified CtBP as described previously (26).
Real Time PCR-First strand cDNA was synthesized using 2 g of total RNA template, 1 g of random primer (Amersham Pharmacia Biotech), 4 l of 5ϫ first strand buffer (Life Technologies, Inc.), 2 l of 10 mM dithiothreitol, 1 l of 10 mM dNTPs, and 1 l of Superscript Plus (Life Technologies, Inc.) in a total volume of 20 l. The amount of AP-2␣ cDNA was quantified using the real-time PCR LightCycler system (Roche Molecular Biochemicals). For PCR, 3 l of cDNA preparation, 2.4 l of 25 mM MgCl 2 , 0.5 M forward and reverse primer, and 2 l of SybrGreen LightCycler mix were combined in a total volume of 20 l (hAP-2␣ forward primer, AAT TTC TCA ACC GAC AAC ATT; hAP-2␣ reverse primer, ATC TGT TTT GTA GCC AGG AGC; ␤-actin forward primer, CTA CGT CGC CCT GGA CTT CGA GC; ␤-actin reverse primer, GAT GGA GCC GCC GAT CCA CAC GG). The following PCR program was performed: 20 s at 95°C (initial denaturation); 20°C/s temperature transition rate up to 95°C for 15 s, 10 s at 58°C, 22 s at 72°C, and 10 s at 82°C acquisition mode single, repeated for 40 times (amplification). The PCR was evaluated by melting curve analysis following the manufacturer's instructions and checking the PCR products on 1.8% agarose gels. Each quantitative PCR was performed at least in duplicate.

The PVDLS Motif of AP-2rep Functions as a Transposable
Repressor Domain-Conserved structural motifs in the transcriptional repressor AP-2rep (Klf12) include three Krü ppelrelated zinc fingers and an H/C knuckle TGE(K/R)P(Y/F)X in the C terminus and a serine/threonine-rich domain next to an PVDLS motif in the N-terminal region (Fig. 1A) (22). The consensus sequence P(V/L)DLS, present in the adenoviral oncoprotein E1A (27) and many mammalian transcriptional repressors, has been shown previously to recruit CtBP1 and CtBP2 corepressors (28,29). To identify transposable protein domains eliciting transcriptional regulation, we constructed fusion proteins (schematically shown in Fig. 1A) of the AP-2rep N-terminal region with the Gal4 DNA binding domain (Gal4/ repNT), of the AP-2rep C-terminal region with the Gal4 DNA binding domain (Gal4/repCT) and the AP-2rep C-terminal region with the viral activator VP16 (VP16/repCT), respectively. The effect of transiently transfected fusion proteins on the activity of appropriate luciferase reporters was assayed in HeLa cells and the human teratocarcinoma cell line PA-1 clone 9117 (Fig. 1B). As reporters we used TK-LUC plasmids containing three Gal4 binding sites (UAS) or two AP-2rep binding sites (A32 (21)).
Transfection of Gal4/repNT, but not of the unmodified Gal4 protein, resulted in significant repression of the luciferase reporter (ϳ10-fold). Interestingly, transcriptional repression of the Gal4/repNT fusion protein was entirely abrogated by a double point mutation in the putative CtBP interaction motif (PVDLS to PVASS). Transfection of a Gal4 protein fused with the AP-2rep C-terminal region did not change activity of the Gal4 reporter, indicating that the zinc finger domain does not harbor any intrinsic transcription regulatory activity. As expected, fusion of the viral VP16 activator to the zinc finger domain resulted in strong activation of the AP-2rep-dependent luciferase reporter. From these data, we concluded that the AP-2rep N-terminal region harbors a transposable repressor activity, which is entirely dependent on the PVDLS motif. Furthermore, the Krü ppel-related zinc finger domain of AP-2rep functions as a sequence-specific DNA binding domain without significant intrinsic transcriptional regulatory activity.
To address further whether the PVDLS motif conferred transcriptional repression we fused two copies of the wild-type and the mutated sequences (RPVDLSR and RPVASSR, respectively) with Gal4. Results from transient cotransfection with the Gal4-dependent luciferase reporter indicated that the pu-tative CtBP interaction domain was both necessary and sufficient for transcriptional repression (Fig. 1C).
Recruitment of CtBP1 by AP-2rep Is Dependent on the PVDLS Motif-To demonstrate that the AP-2rep transcriptional repressor motif interacts with CtBP corepressors, we assayed protein-protein interaction. Screening 2 ϫ 10 6 clones of a yeast two-hybrid library prepared from murine embryonal brain at gestational stage day 17.5 with the AP-2rep N terminus as a bait resulted in isolation of 71 independent clones. Determination of the respective cDNA inserts revealed that 65 of these clones contained partial or complete fragments of the murine CtBP1 cDNA and that the remaining six clones represented obvious artifacts (data not shown). A functional yeast two-hybrid assay indicated robust interaction that resulted in approximately one-third of ␤-gal reporter expression as compared with SV40 large T antigen interaction with p53, which was assayed for a positive control ( Fig. 2A, pVA3/pTD1). Importantly, the mutated AP-2rep protein, harboring the same two amino acid exchanges in the PVDLS motif as used for transient transfections in Fig. 1C, failed to interact with CtBP1.
To further assay in vitro interaction between AP-2rep and CtBP1, we performed a GST pull down experiment. Therefore, GST fusion proteins containing either wild-type AP-2rep or PVASS-mutated AP-2rep were immobilized to glutathione-coupled Sepharose and incubated with in vitro translated 35 Slabeled protein CtBP1 protein. As shown in Fig. 2B, CtBP1 was specifically retained by wild-type, but not by mutated AP-2rep fusion protein. In control experiments, 35 S-labeled CtBP1 also failed to interact with unmodified GST.
To verify specific protein-protein interaction of AP-2rep with CtBP1 by a second independent method, we also performed coimmunoprecipitations both with in vitro translated AP-2rep and CtBP1 proteins and with tagged AP-2rep transiently expressed in HeLa cells. Therefore, we used a flag-modified AP-2rep protein as a bait. As shown in Fig. 2C, we were able to coimmunoprecipitate specifically CtBP1 and also unmodified AP-2rep protein.
When we transiently transfected FLAG-tagged AP-2rep into HeLa cells, we were able to coprecipitate CtBP1 with an anti-FLAG antibody (Fig. 2D, lanes 5 and 6) from cell extracts. In comparison, we Western probed HeLa cell lysate transiently transfected with CtBP1 alone or with both CtBP1 and tagged AP-2rep. Our results shown in Fig. 2D (lanes 1-3) clearly show that a significant portion of endogenous CtBP-1 present in HeLa cells can be coimmunoprecipitated with the tagged AP-2rep protein, indicating tight interaction between the two proteins. In summary, we concluded from these data that the transcriptional repressor AP-2rep recruits the corepressor CtBP1 via physical interaction through the PVDLS domain.

Activation of Endogenous AP-2␣ Expression Is Dependent on Adenoviral E1A and the CtBP1 Interaction Domain PVDLS-
Based on previous reports that the C-terminal region of adenovirus E1A elicits gene regulatory activity through interaction with CtBP proteins (27,30,31), we determined the effect of adenovirus infection on endogenous AP-2 expression levels. Our results shown in Fig. 3A provide evidence that adenoviral infection of both HeLa and PA-1 (clone 9117) cells leads to strong induction of AP-2␣ protein with a maximum ϳ24 h after infection and 12 h after maximal expression of adenoviral E1A protein. In contrast to infection with wild-type virus, mutants lacking the entire E1A region (Ad5 312 in Fig. 3B) or harboring a specific microdeletion within the E1A-CtBP interaction motif, including the residues PLDLS at 233-237 (Ad5 1135 in Fig.  3B), did not induce AP-2␣ expression. In these experiments, we used identical virus titers (50 PFU/cell) in the case of all three strains and controlled equal protein loading by reprobing the Western blots with a ␤-actin antiserum.

FIG. 2. Interaction of AP-2rep and CtBP-1 dependent on the PVDLS domain.
A, yeast two-hybrid assay. ␤-Galactosidase activity mediated by p53/SV40-large T interaction (pVA3/pTD1) was assayed as a positive control. As bait, wild-type AP-2rep NT (as shown in Fig. 1A) or AP-2rep NT mutated in the PVDLS motif was coexpressed with CtBP-1. B, GST in vitro binding assay. Immobilized GST-AP-2rep fusion protein retained 35 S-labeled CtBP1, but fusion protein mutated in the PVDLS motif failed to interact. C, immunoprecipitation in vitro. Unlabeled AP-2rep/FLAG protein was coincubated with 35 S-labeled AP-2␣, AP-2rep/FLAG, CtBP1, or AP-2rep and immunoprecipitated. To analyze whether induction of AP-2 protein by adenoviral infection involved up-regulation of AP-2␣ mRNA, we performed quantitative real time reverse transcription-PCR on RNA templates extracted from HeLa cell cultures infected for 24 h (Fig.   3C). The ratio of AP-2␣ versus ␤-actin mRNA was increased ϳ8-fold by wild-type adenovirus but was not significantly altered after infection with identical titers of mutant adenovirus strains (50 PFU/cell) or in mock-infected cells. Further controls showed that the induction of AP-2 was dependent on the multiplicity of infection. Infection with a 1 ⁄10 titer of wild-type virus (5 PFU/cell) resulted in ϳ5-fold AP-2␣ induction, and infection with a 1 ⁄100 titer (0.5 PFU/cell) resulted only in 1.5-fold induction (data not shown). From these experiments, we inferred that induction of AP-2␣ by adenoviral infection is critically dependent on the intact E1A-CtBP1 interaction motif and involves both AP-2␣ mRNA and protein.
Therefore, we investigated the effect of transiently expressed E1A protein and explored the role of an intact CtBP interaction motif in regulation of AP-2␣ expression. In parallel, we measured the effect on a luciferase reporter under control of the dimerized A32 cis-regulatory element of the AP-2␣ promoter (Fig. 4A) and on endogenous AP-2␣ protein levels quantified by Western blots (Fig. 4B). Transient transfection of E1A resulted in ϳ2-fold activation of the A32-dependent luciferase reporter. Importantly, an E1A construct harboring the double point mutation in the CtBP interaction motif did not activate the luciferase reporter. In contrast, we observed moderate but reproducible repression to ϳ0.6fold activity (Fig. 4A, E1A mut). Consistent with these results, transcriptional repression of the A32-dependent reporter by AP-2rep was entirely abrogated by the double point mutation in the PVDLS interaction domain. Again, we observed reproducible activation of the reporter, possibly due to interference with endogenous wild-type AP-2rep repressor protein and to mutually exclusive binding with activating transcription factors such as BTEB-1 (21). Importantly, transiently expressed CtBP1 protein also repressed the A32 luciferase reporter, indicating that CtBP1 elicits intrinsic transcriptional repressor activity when appropriately recruited. All results obtained from measuring luciferase reporter activities were closely paralleled by changes of endogenous AP-2␣ protein visualized by Western blots of HeLa cell extracts (Fig. 4B). In general, the effects of transient transfections were significantly smaller than changes in AP-2 expression by adenoviral infections, because we observed much higher E1A expression levels after infections (data not shown), and the activity of an AP-2-dependent promoter depends also on AP-2 isoforms other than AP-2␣.
Taken together, our data provide evidence that both adenoviral infection and transfection of the adenoviral oncoprotein E1A activate expression of the endogenous AP-2␣ gene. Furthermore, activation of AP-2␣ transcription is dependent on the presence of a functional E1A-CtBP interaction motif and involves inactivation of the transcriptional silencer AP-2rep by direct interaction with the corepressor CtBP1. DISCUSSION Recently, we identified the A32 element, an important cisregulatory element in the AP-2␣ promoter mediating both activation and repression of AP-2␣ mRNA transcription. Promoter activity was silenced through sequence-specific binding of AP-2rep, a novel Krü ppel-related zinc finger repressor, which interacted mutually exclusive with activating factors (21). In the present study, we provide further evidence for a role of AP-2rep in transcriptional repression that involves physical association with the corepressor CtBP1. Data presented in Figs. 1 and 2 show that the transcriptional repressor activity of AP-2rep is tightly associated with a functional CtBP interaction motif and demonstrate strong interaction of the two proteins by coimmunoprecipitation both from endogenous cell extracts and in vitro. Therefore, specific interaction with CtBP1 through the PVDLS domain is functionally important and identifies AP-2␣ as a novel target gene of CtBP-mediated repression.
The function of CtBP1 and its close homolog CtBP2 as transcriptional corepressors for the zinc finger-transcription factor ␦EF1 has been previously detected by Furusawa et al. (29). Interestingly, these authors have described specific embryonic expression patterns of CtBP proteins, particularly in structures that are subject to regulated AP-2 expression, including the face, cephalic ganglia, dorsal root ganglia, and limb. Therefore, it is possible that CtBP corepressors may contribute to specify expression patterns of different AP-2 genes during development. The precise functional role of CtBP proteins in these developmental processes, however, will require analyses of their respective knockout mice.
Furthermore, data presented in Figs. 3 and 4 show that activation of AP-2␣ mRNA and protein expression by adenoviral infection closely parallels the ability of E1A protein to interact with CtBP1 through the PLDLS motif. Thus, our study for the first time identifies that regulation of an endogenous gene through E1A requires binding of CtBP1 and that sequestration of the transcriptional corepressor changes significantly gene expression patterns in infected host cells.
Exon 1 of E1A has been shown previously to confer transforming activity in cooperation with activated ras oncogenes (32). Functions mediated through the conserved regions CR1 and CR2 of E1A include interaction with pRb, p107, and p130, which leads to release of E2F transcription factor, and interaction with p300 (reviewed in Ref. 25). Additional functions of ras are required for cell cycle progression in order to overcome p53-induced apoptosis elicited by E1A expression (33). In contrast, the C-terminal region of exon 2 has been shown to negatively modulate transformation by E1A and activated ras. The transformation restraining activity of the E1A C terminus re-quires interaction with CtBP1 (25). Our data presented here raise the possibility that AP-2 activation may be causally involved in executing this activity.
Dependent on the cellular context, AP-2 may exert different effects. We have shown previously that AP-2 is a strong negative regulator of c-Myc transcriptional target genes and c-Myc-induced apoptosis (Ref. 19 and reviewed in Ref. 34). In nontransformed cells, activation of the cell cycle inhibitor p21 and negative regulation of c-Myc transcriptional target genes may restrain cell cycle progression and delay entry into G 1 /S-transition. However, in fully transformed cells harboring mutated ras oncogenes, transcriptional activity of AP-2 is defective (15), and therefore, the tumor restraining activities of the E1A C terminus may be overcome.