INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Oct-4 (also termed Oct-3) encodes the only known transcription factor that is specifically expressed in the mammalian germ line and stem cell lines derived therefrom (8, 34; for further references, see reference 36). Oct-4 is essential to maintain cellular totipotency (31, 36). All cells of preimplantation embryos that lack Oct-4 acquire a trophoblastic cell fate instead of developing into a normal embryo with both an inner cell mass and a trophoblast, strongly suggesting that its function is linked to the totipotent germ line cycle (36).

Members of the POU transcription factor family share a conserved DNA binding domain, the POU domain, originally identified in the transcription factors Pit-1, Oct-1, Oct-2, and Unc-86 (18). Like other Oct factors, Oct-4 contains a POU domain that binds as a monomer to the octamer sequence motif, ATGCAAAT (37). In addition, Oct-4 and other Oct factors bind as homo- and heterodimers to the novel palindromic Oct factor recognition element (PORE), which is composed of an inverted pair of homeodomain binding sites separated by exactly 5 base pairs (ATTTG +5 CAAAT) (6).

The N and C termini of Oct-4 function as transactivation domains. Whereas the N terminus is active in various cultured cell types, the activity of the C-terminal domain depends on the cell type and has been correlated with specific phosphorylation states (7). Cell type specificity of the C-terminal domain is maintained when tethered to a promoter by the Oct-2 POU domain but is lost when linked to the Pit-1 POU domain or to the GAL4 DNA binding domain. These results have been taken as an indication that the C-terminal activation domain of Oct-4 acquires its specificity through a precise functional interaction with its native DNA binding domain (7, 25).

Oct-4 can activate transcription from various positions within a gene. FGF-4 and osteopontin are two target genes of Oct-4 during early mouse development and contain binding sites that are remote from the site of transcriptional initiation. FGF-4 contains an octamer-containing enhancer in the 3' untranslated region (references 1 and 48 and references therein); osteopontin has a PORE in the first intron (6). In embryonal stem cells, the octamer sequence motif is active irrespective of the distance from the site of transcriptional initiation (33, 38). However, in differentiated cells, Oct-4 can transactivate only from an octamer motif at a proximal position (7, 39, 40). Transactivation from a distance appears to depend on stem cell-specific coactivators that bridge a remotely bound Oct-4 protein and the basal transcription machinery (39). Such bridging factors still have to be identified. However, the adenovirus (Ad) E1A oncoprotein mimics the function of a stem cell-specific coactivator(s) (E1A-like activities, hereafter referred to as ELA) (39). When coexpressed with Oct-4 in differentiated cells, E1A allows Oct-4-mediated transactivation from distal binding sites. E1A specifically stimulates the activity of Oct-4 and not that of other Oct factors, such as the ubiquitously expressed Oct-1. The nature of the Oct-4-E1A interaction is poorly defined and it is unclear if viral oncoproteins other than E1A also possess ELA function.

Here we identify two independent Oct-4 binding sites on E1A and show that the POU domain of Oct-4 is sufficient for binding to both E1A sites. The human papillomavirus (HPV) E7 oncoprotein, which is implicated in anogenital cancer, shares sequence and functional homology with E1A. We demonstrate that E7, like E1A, can bind to the POU domain of Oct-4. Oct-4 and E7 can form a complex in vivo and E7 expression stimulates Oct-4-mediated transactivation from remote binding sites in differentiated cells. Our data suggest that different DNA tumor viruses have evolved oncoproteins that share the ability to mimic a stem cell-specific activity.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Constructs. Oct-4 (pCMV-Oct-4), POU domain (pCMV-POU4), and Oct-2A (pCMV-Oct2A) expression vectors, as well as the reporters p6W-109tkCAT, p5Fd-109tkCAT, and p-actin-lacZ, have been described previously (7, 40). pRSV-E1A carries the Rous sarcoma virus enhancer-driven Ad5 E1A gene, and pSG5-vErbA carries the cDNA of vErbA under simian virus 40 (SV40) enhancer control. pX-E7 was created by inserting the HPV16 E7 cDNA contained in pLexA-16E7 (amino acids 1 to 98 of LexA) (50) as an EcoRI-XhoI fragment into the cytomegalovirus-driven expression vector pX (44).

Cell lines and transient transfections. Cell lines were maintained and transfected essentially as described previously (30, 44). Cells in 6-cm dishes were transfected with 2 µg of p-actin-LacZ internal standard, 0.25 to 2.0 µg of chloramphenicol acetyltransferase (CAT) reporter, 0.01 to 3 µg of expression vector, and pBS-KS carrier (12 µg of total DNA) and harvested 40 h later. Extracts of 50 µl were prepared in 250 mM Tris (pH 7.8) by freeze-thawing. Extracts were used for determination of -galactosidase and CAT activity and for electrophoretic mobility shift assays (EMSAs).

EMSA. EMSAs were performed as described previously, using 1 µl of cell extract and the 1W probe (38, 41).

Immunoprecipitations. Oct-4 reactive antibodies were purified from immunized rabbit serum by affinity chromatography. For precipitation of E7, a protein G agarose-conjugated monoclonal antibody (HPV16 E7 D6) was used. Proteins of interest were expressed in COS cells transiently transfected with 5 to 10 µg of expression vector. For precipitation with Oct-4, antibody cells were resuspended in 300 µl of lysis buffer (250 mM NaCl, 0.1% Nonidet P-40, 50 mM HEPES [pH 7.6], 0.5 mM -mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride, 10 µg of aprotinin/ml) (30 min, 4°C). Cleared lysate was incubated (2 h) with Oct-4 antibody and immunoglobulin G, respectively. Antibody complexes were precipitated with protein A Sepharose beads. For precipitation with E7 antibody 107 cells were extracted in 1 ml of lysis buffer (50 mM HEPES [pH 7.0], 150 mM NaCl, 0.1% Nonidet P-40, 0.2 mM phenylmethylsulfonyl fluoride, 0.1 mM Na₃VO₄, 10 mM -glycerophosphate, 1 mM NaF, 10 µg of aprotinin/ml) (30 min, 4°C). Cleared lysate was incubated (1 h, 4°C) with protein G-agarose (Santa Cruz Biotechnology, Inc.) followed by precipitation with HPV16 E7 D6 and protein G beads (3 h, 4°C). Oct-4 and E7 precipitates were extensively washed with lysis buffer, suspended in sodium dodecyl sulfate (SDS) sample buffer and subjected to SDS-polyacrylamide gel electrophoresis (PAGE) and Western blotting (17).

Western analysis. Immunoprecipitates were separated on SDS-10% (Oct-4-E1A) or -12.5% (Oct-4-E7 and vErbA-E7) polyacrylamide gels. Proteins were transferred onto nitrocellulose (Schleicher & Schuell) or polyvinylpyrrolidone (DuPont de Nemours) membranes, respectively. Immunoblot analysis using monoclonal HPV16 E7 (CIBA Corning, Inc., Alameda, Calif.) and vErbA (1g10) antibodies was performed as described elsewhere (50). Immunoblot analysis using Oct-4 antiserum was performed with blocking and incubation buffer containing 1 M NaCl, 1% Triton X-100, and 2.5% milk powder in phosphate-buffered saline.

In vitro binding studies. Proteins were in vitro translated according to standard procedures (TNT reticulocyte lysate system; Promega). The production of glutathione S-transferase (GST) fusion proteins and the GST pull-down assay were described previously (5).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

A functional interaction between Oct-4 and E1A has been demonstrated previously, both in vitro and in vivo (39). The E1A 289R gene product can form a trimeric complex with Oct-4 and an octamer-containing DNA fragment and allows Oct-4-mediated transactivation from remote binding sites in differentiated cells (39). In contrast, the E1A 243R alternative splice variant that lacks the conserved region 3 (CR3) found in E1A 289R fails to interact with Oct-4 in a mobility shift assay and does not stimulate Oct-4-mediated transactivation. Consequently, it has been suggested that E1A binds Oct-4 via the CR3 domain. However, overexpression of E1A 243R represses the activity of an Oct-4-driven enhancer, indicating that regions outside CR3 could also functionally interact with Oct-4. We assessed the binding of Oct-4 deletion mutants to GST-E1A fusion proteins in order to define the domain requirements for formation of the Oct-4-E1A complex.

E1A contains two independent Oct-4 binding sites. As expected, in vitro translated Oct-4 protein bound efficiently to GST-E1A but not to the GST control (Fig. 1a and b; compare lanes 1 and 3). In contrast, Oct-1 failed to bind to both GST and GST-E1A (lanes 2 and 4). Thus, the in vitro binding assay accurately reflects the specificity of the Oct factor-E1A interactions observed in vivo.

View larger version (40K): [in this window] [in a new window]   FIG. 1.   Oct-4 binds GST-E1A and GST-E7 in vitro. (a) Schematic representation of Oct-4 and E1A constructs used. (b) Equivalent amounts of GST (lanes 1 and 2) and GST-E1A (lanes 3 and 4) were used to bind 35S-labelled in vitro translated (IVT) Oct-4 (lanes 1 and 3) and Oct-1 (lanes 2 and 4) proteins, respectively, as indicated. Bound proteins were subjected to SDS-PAGE and autoradiography. Lane 5, 25% Oct-4 input; lane 6: 25% Oct-1 input. (c) Equivalent amounts of GST (lanes 1, 5, 9, 13, and 17), GST-E1A 1-90 (lanes 2, 6, 10, 14, and 18), GST E1A 127-204 (lanes 3, 7, 11, 15, and 19) and GST-E1A 140-204 (lanes 4, 8, 12, and 20) were used to bind IVT Oct-4 (lanes 1 to 4), 4N-POU4 (lanes 5 to 8), POU4-4C (lanes 9 to 12), POU4 (lanes 13 to 16) and Oct-1 (lanes 17 to 20). (d) Equivalent amounts of GST (lanes 1 to 5) and GST-E7 (lanes 6 to 10) were used to bind IVT Oct-4 (lanes 1 and 6), 4N-POU4 (lanes 2 and 7), POU4-4C (lanes 3 and 8), POU4 (lanes 4 and 9), and Oct-1 (lanes 5 and 10).

The POU domain is sufficient for Oct-4-E1A complex formation. Next we sought to determine which of the three characterized Oct-4 domains (the N-terminal activation domain [N domain], the C-terminal activation domain [C domain], and the POU domain [POU4] [Fig. 1a] [7]) was involved in Oct-4-E1A complex formation. Oct-4 deletion mutants lacking either the C domain (4N-POU4), the N domain (POU4-4C) or both (POU4) retained the ability to specifically bind the GST-E1A fusion proteins (Fig. 1c, lanes 5 to 16). In contrast, an interaction of the N or C domain alone with E1A could not be detected (data not shown). This indicates that the activation domains of Oct-4 are dispensable for complex formation. The POU domain of Oct-4 is sufficient to interact with either of the two Oct-4 interfaces on E1A.

Oct-4 binds HPV16 E7 in vitro. DNA tumor viruses, such as Ad, SV40, and HPV, target overlapping sets of regulatory proteins in order to drive the host cell into S phase and to create conditions favorable for viral replication (21). For example, E1A (Ad), T antigen (T-Ag) (SV40), and E7 (HPV) all bind to and inactivate the retinoblastoma (Rb) tumor suppressor protein by using a conserved LXCXE sequence motif (reference 24 and references therein). However, certain cellular factors appear to be targeted by only a subset of these oncoproteins. For example, E1A and T-Ag bind the CBP/p300 coactivator, but no interaction between E7 and CBP/p300 has been demonstrated to date (reference 13 and references therein; 14).

Oct-4 binds HPV16 E7 in vivo. We used a coimmunoprecipitation approach to determine if the E7-Oct-4 complex could form in vivo. HPV16 E7 and Oct-4 were coexpressed in transiently transfected COS cells. Cell extracts were then immunoprecipitated with a monoclonal antibody directed against E7, and the presence of E7 and Oct-4 in the immunoprecipitates was detected by Western analysis. The E7 antibody precipitated large amounts of E7 from extracts of cells transfected with E7 expression vector (Fig. 2B and C, lower panels). Neither E7 nor Oct-4 was precipitated from extracts of cells transiently transfected with Oct-4 expression vector alone (Fig. 2A, lower and upper panel, respectively), verifying that COS cells do not express E7 and that the E7 antibody did not cross-react with Oct-4. The E7 antibody coprecipitated Oct-4 from extracts of cells transiently transfected with both E7 and Oct-4 expression vectors (Fig. 2B, upper panel). The ability of the E7 antibody to precipitate the unrelated nuclear protein vErbA from extracts of cells cotransfected with E7 and vErbA expression vectors was also tested. Unlike Oct-4, vErbA was not detectable in the precipitate (Fig. 2C, upper panel). Thus, the E7 antibody specifically coprecipitated Oct-4 from cells coexpressing Oct-4 and E7. Furthermore, recombinant GST-E7 fusion protein was found to efficiently bind to Oct-4 when added to cell extracts (data not shown). Taken together, these data suggest that Oct-4 and E7 physically interact in vivo.

View larger version (36K): [in this window] [in a new window]   FIG. 2.   Oct-4 binds E7 in vivo. Whole-cell extracts (2 µg) were prepared from COS cells transiently transfected with 10 µg of pCMV-Oct-4 (A), cotransfected with 10 µg of pCMV-Oct-4 and 10 µg of pX-E7 (B), and cotransfected with 10 µg of pSG5-vErbA and 10 µg of pX-E7 (C). Extracts were immunoprecipitated with HPV16 E7 antibody. Immunoprecipitates [IP(-E7)] and 20 µg of cell lysate were separated by SDS-PAGE, and proteins were transferred onto polyvinylpyrrolidone membrane. Membranes were cut into two parts to allow the simultaneous detection of E7 (lower panels), Oct-4 (panels A and B, upper portion), and vErbA (panel C, upper portion) by Western analysis. The three proteins with molecular weights of approximately 43,000 recognized by the Oct-4 antibody probably represent different phosphorylated forms of the protein (7, 18a). Ig, immunoglobulin; MW, molecular weight (103).

HPV16 E7 stimulates Oct-4 transactivation from remote binding sites. Given that E7 and E1A both can form complexes with Oct-4 in vitro and in vivo, we asked if E7, like E1A, could stimulate Oct-4-mediated transactivation from promoter distal binding sites. To address this question, we employed the HPV16 E7 expressing cell line E7/2 and the control cell line M/1 (12). The E7/2 line was derived from 3T3 cells by stable transfection with an HPV16 E7 expression vector. M/1 cells are stably transformed with empty expression vector. Both cell lines were cotransfected with an Oct-4 expression vector and a CAT reporter gene bearing remote octamer sites. Oct-4 stimulated the reporter in E7/2 cells only, indicating that HPV16 E7 can cooperate with Oct-4 in transactivation from a distance (Fig. 3A). E7 not only shares with E1A the ability to bind Oct-4 but, like E1A, it also modulates Oct-4-mediated transactivation. Thus, like E1A, E7 appears to mimic a cellular activity that is restricted to embryonal cells.

View larger version (66K): [in this window] [in a new window]   FIG. 3.   Oct-4 and E7 cooperate in transactivation from remote binding sites. (A) The 3T3-derived cell lines M/1 and E7/2 were cotransfected with 2.0 µg of p6W-109tkCAT and 3.0 µg of pCMV-Oct-4. Fold activation refers to the quotient of CAT activity in the presence and absence of expression vector. The mean values of two independent transfection experiments are plotted. (B) HeLa cells were cotransfected with 0.5 µg of p5Fd-109tkCAT or p6W-109tkCAT and increasing amounts of pCMV-Oct-4 as indicated. (C) EMSA of extracts of HeLa cells transiently transfected with increasing amounts of pCMV-Oct-4 as indicated. Arrows indicate the positions of complexes containing Oct-1 and Oct-4. mock, extract from mock transfected HeLa cells; probe, octamer motif containing fragment from immunoglobulin heavy chain gene enhancer (1W) (41).

Distance-independent transactivation in differentiated cell lines. To extend our analysis of Oct-4-mediated transactivation in differentiated cells, we cotransfected several cell lines with the p6W-109tkCAT plasmid and varying amounts of Oct-4 and Oct-2 expression vectors, respectively (Fig. 4; only maximal activation shown). Expression of Oct-4 and Oct-2 in the cell lines tested was verified by EMSA (data not shown). As expected, Oct-4 was active in 293 (E1A) and HeLa (E7) cells. Interestingly, Oct-4-mediated transactivation (10-fold) was also observed in COS cells. Like 293 and HeLa cells, COS cells are transformed by a DNA tumor virus (SV40) and express T-Ag, a viral protein with analogous functions to E1A and E7. Low-level or no Oct-4 activity was detected in the 3T3 and CV1 cell lines. These cell lines do not express viral proteins. Oct-2, which has been shown to transactivate from remote positions in B cells, failed to significantly stimulate the reporter in any of the cell lines tested. This finding suggests that Oct-2 is dependent on B-cell-specific cofactors (reference 29 and references therein) that cannot be substituted for by E1A and E7. In contrast, Oct-4 requires stem cell-specific cofactors whose function is mimicked by the viral oncoproteins.

View larger version (11K): [in this window] [in a new window]   FIG. 4.   Cell lines were cotransfected with 0.5 µg of p6W-109tkCAT and increasing amounts of pCMV-Oct-4 (black bars) and pCMV-Oct-2A (white bars) as indicated. The value of the titration series representing maximum reporter activation is shown (see text).

Role of the Oct-4 activation domains. We have previously identified two independent activation domains of Oct-4 (N domain and C domain, respectively [7]). We tested Oct-4 deletion mutants lacking one or both activation domains for their ability to transactivate from distal octamer sites in tumor virus-transformed cell lines in order to assess the individual contribution of the activation domains. Although the POU domain was sufficient for E1A-Oct-4 and E7-Oct-4 complex formation (Fig. 1C and D) it failed to activate the reporter in both HeLa (Fig. 5A) and 293 (Fig. 5B) cells. Presence of the N domain allowed reporter activation in both cell lines tested. The C domain was active in HeLa cells only. These data correlate with the previously observed cell line specificity of the C domain and indicate that recruitment of E1A or E7 to the promoter in absence of an Oct-4 activation domain is not sufficient for transactivation.

View larger version (21K): [in this window] [in a new window]   FIG. 5.   HeLa (A) and 293 (B) cells were cotransfected with 0.5 µg of p6W-109tkCAT and increasing amounts of pCMV-POU4, pCMV-4N-POU4, and pCMV-POU4-4C as indicated. Fold activation refers to the quotient of CAT activity in the presence and absence of expression vector. The mean values of two independent transfection experiments are plotted.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The embryonal transcription factor Oct-4 has previously been shown to require a stem cell-specific cofactor to activate transcription from binding sites located in a promoter-distal position (39). The adenoviral E1A oncoprotein can replace the stem cell-specific cofactor and allow transactivation by Oct-4 from remote binding sites in differentiated cells. The E1A 289R protein can bind the Oct-4-octamer complex in a mobility shift assay (39). The E1A 243R protein that lacks the CR3 domain present in E1A 289R fails to interact with the Oct-4-DNA complex. These results implicated the CR3 domain in Oct-4 binding. Using recombinant proteins in a GST pull-down assay, we now show that E1A and Oct-4 can interact in solution in the absence of octamer motif containing DNA. Two nonoverlapping fragments of E1A CR1 (amino acids 1 to 90) and CR3 (amino acids 140 to 204) bind Oct-4 in vitro. Thus, the GST pull-down assay has identified a second potential Oct-4 interaction domain in addition to CR3. In contrast to the results obtained with the mobility shift assay, CR3 is not absolutely required for Oct-4 binding in the GST pull-down assay. This most likely reflects a difference in the stringency of the two assays or a difference in conformation between Oct-4 bound to DNA and Oct-4 in solution. An interaction between non-DNA-bound Oct-4 and E1A has been suggested to mediate the squelching observed when E1A is present in large excess over Oct-4 (39). Under these conditions, E1A does not mediate Oct-4 transactivation but rather represses Oct-4-driven enhancers. Indeed, E1A 243R that lacks the CR3 domain represses Oct-4-mediated transactivation in F9 cells when expressed at high levels. This strongly indicates that regions outside the CR3 domain can functionally interact with Oct-4 in vivo.

The involvement of multiple E1A domains in the binding of cellular factors is not without precedent: E1A harbors two binding sites for the Rb protein (43, 45). A high-affinity binding site is located in CR2. A second low-affinity binding site has been identified in CR1. Both binding sites play a role in the dissociation of the E2F-Rb repressor complex (19).

The POU domain of Oct-4 is sufficient for E1A binding. Members of the POU family are characterized by a high degree of sequence conservation in the POU domain but show little conservation in other domains. Nevertheless, E1A binds to Oct-4 but fails to interact with Oct-1 even though the POU domains of these two Oct factors are highly homologous. In accordance with the interaction data, previous studies and the results of the transactivation experiments reported here show that E1A can stimulate Oct-4-mediated transcription but fails to activate Oct-1. Thus, E1A can recognize subtle sequence differences in related POU domains, and this specificity is reproduced in the in vitro interaction assay.

The ability of E1A to specifically target the POU domain of one Oct factor but not the POU domain of its close relatives is shared by viral and cellular coactivators (reference 4 and references therein; 18). Indeed, the POU domain emerges as the main regulatory domain of Oct factors. The herpes simplex virus protein VP16 interacts with the POU domain of Oct-1 bound at viral promoters to direct high level transcription of viral genes using its powerful activation domain. In contrast, VP16 fails to bind to Oct-2, Oct-4, and Oct-6. The B-cell-specific coactivator OCA-B (Bob1, OBF1), which recognizes the POU domains of Oct-1 and Oct-2 but does not bind the POU domains of Oct-4 and Oct-6, provides another example. In this case, transcriptional activation functions are provided by both the Oct factor activation domain and by an activation domain of the cofactor OCA-B. The Oct-4-E1A complex appears to differ from the Oct-1-VP16 and the Oct-1,2-OCA-B complex in that transactivation depends solely on the presence of activation domains supplied by the Oct-4 protein. The POU domain of Oct-4 is unable to activate transcription in E1A-expressing cells even though it is capable of binding the E1A protein.

The HPV16 E7 oncoprotein shares the ability of E1A to stimulate Oct-4-mediated transactivation from remote binding sites. Oct-4 is not active in 3T3 cells, which represent differentiated fibroblasts (Fig. 3). However, Oct-4 can transactivate from distal binding sites if the same cell line constitutively expresses E7. This result indicates that E7 expression in differentiated cells is sufficient to stimulate Oct-4 activity. Furthermore, Oct-4 can transactivate from distal binding sites in HeLa cells that express E7 but not E1A. As has been previously observed for Oct-4-E1A cooperation (39), stimulation of Oct-4 by E7 appears to be very sensitive to the expression levels of both proteins. Overexpression of Oct-4 in HeLa cells abolishes the Oct-4-mediated transactivation most likely due to squelching of limiting E7 protein. Like E1A, E7 can interact with Oct-4 in vitro and in vivo and binds to the POU domain of Oct-4. These observations suggest that E7 and E1A use a similar mechanism to activate Oct-4 functioning. This mechanism appears to involve a direct interaction between the viral oncoprotein and Oct-4. How could a direct interaction between Oct-4 and E1A or E7 lead to transactivation in differentiated cells? Oct-4 transactivates from binding sites close to the TATA box in all cell lines tested irrespective of the presence of viral oncoproteins (7, 46). The failure of Oct-4 to activate from distal binding sites could be due to factors bound between Oct-4 and the TATA box obstructing access to the transcription complex. In addition, the increased distance between binding sites and the TATA box might decrease the rate of productive interactions with the transcription complex. E1A, E7, and T-Ag have all been shown to interact with components of the basal transcription machinery (10, 16, 22, 28). It is conceivable that they function as "bridging factors" by mediating a stable interaction between Oct-4 and the transcription apparatus. Viral oncoproteins have recently been demonstrated to bind coactivators and corepressors of transcription, such as histone acetyltransferases and histone deacetylases, which regulate transcription at the chromatin level (3, 5, 14, 26, 32). Thus, it is also possible that stimulation of Oct-4 activity by viral oncoproteins involves the recruitment of acetyltransferases or the displacement of deacetylases.

While our interaction data suggest that Oct-4 stimulation involves the direct binding of viral oncoproteins, we cannot exclude the possibility that Oct-4 activity is also affected indirectly. It is noteworthy that we observed only a weak stimulation of Oct-4 activity when E7 was transiently expressed in transfected cells (data not shown). This is in contrast to the strong stimulation of Oct-4 activity in cell lines which express E7 constitutively, such as 3T3 E7/2 (Fig. 3) and HeLa (Fig. 4). It is possible that Oct-4 stimulation involves an additional factor the expression or activity of which is altered by prolonged (but not by transient) expression of E7. E1A, E7, and T-Ag all target the pRb family of tumor suppressors. In fact, pRb is expressed only at low levels in cells in which Oct-4 efficiently transactivates from distal binding sites, such as early mouse embryo and P19 EC cells (42). It is therefore conceivable that viral oncoproteins activate Oct-4 function by counteracting a repressive effect of pRb. However, overexpression of pRb and p107 in HeLa and EC cells does not result in repression of Oct-4 activity, indicating that pRb and p107 are not involved in the regulation of Oct-4 (data not shown).

E7 has previously been shown to functionally interact with transcription factors such as c-Jun and ATF2 (2). We demonstrate for the first time a functional interaction between E7 and a member of the POU family of transcription factors. Our finding that Oct-4 can activate transcription from remote binding sites in T-Ag-transformed COS cells but not in the untransformed parental CV1 cell line suggests that T-Ag, like E1A and E7, can also bind Oct-4. However, we still have to demonstrate that T-Ag and Oct-4 can interact in vivo.

E1A and E7 stimulate proliferation and counteract differentiation of their respective host cells and share the ability to overcome normal cell cycle regulation (11, 15, 21, 23, 27, 47, 49). They both can stimulate Oct-4 function in differentiated cells where Oct-4 is not normally active. Thus, different tumor viruses have evolved proteins that share the ability to mimic an activity that is normally restricted to toti- and pluripotential embryonal stem cells. We propose that this leads to changes in cellular gene expression that are incompatible with differentiation and that favor proliferation. The identification of cellular genes affected in this way would greatly contribute to a better understanding of the differentiation process.

Embryonal stem cells are unique because they can form a tumor or contribute to a new organism, depending on the time and place of injection (9, 35). They can act as tumor stem cells when injected into an adult mouse but also as embryonal stem cells when injected into a blastocyst, where they resume embryonic development and participate in the formation of a normal chimeric mouse. An important goal is to understand the molecular mechanisms that govern the decision of an embryonal stem cell to either generate a tumor or develop into an organism.