The HMG-Box Transcription Factor HBP1 Is Targeted by the Pocket Proteins and E1A

The HMG-box transcription factor HBP1 is targeted by the pocket proteins and E1A

Paul Lavender, Laurence Vandel, Andrew J Bannister and Tony Kouzarides

Wellcome/CRC Institute and Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge, UK

A yeast two-hybrid screen has identi®ed HBP1 as a activation functions, which correlates with RB's ability transcription factor capable of interacting with the to induce growth arrest. Only a subset of E2F family pocket protein family. We show that HBP1 can interact members (E2F1, 2, 3) possess the ability to stimulate with one of these, RB, both in vitro and in mammalian S-phase. The fact that this subset of E2F proteins is cells. Two distinct RB binding sites are present within precisely the subset regulated by RB in vivo (rather HBP1 ± a high anity binding site, mediated by an than p107 or p130) may explain why RB is the only LXCXE motif and a separate low anity binding site pocket protein mutated in tumours (Lukas et al., present within an activation domain. GAL4-fusion 1996). experiments indicate that HBP1 contains a masked Another function of RB which also correlates with activation domain. Deletion of two independent N- and its ability to induce growth arrest, is the repression of C-terminal inhibitor domains unmasks an activation genes involved in protein biosynthesis. RB can repress domain which is 100-fold more active than the full most RNA polymerase III (Pol III) transcribed genes length protein. The released activation capacity is in vivo (White et al., 1996) and can regulate RNA repressed by RB, p130 and p107. In addition, E1A can polymerase I (pol I) transcription by repressing the repress the activity of HBP1 via conserved region 1 UBF transcription factor (Cavanaugh et al., 1995). sequences in a manner independent of the CBP co- Repression of pol I, pol III and E2F regulated pol II activator. We show by stable expression in NIH3T3 cells genes requires the pocket domain of RB (which is that HBP1 has the capacity to induce morphological found mutated in tumours) and is alleviated by viral transformation of cells in culture. transforming proteins such as E1A binding to RB. The ability of RB to recognize and repress a diverse Keywords: HMG-box; retinoblastoma; E1A array of factors which regulate pol I, pol II and pol III genes may be related to the sequence similarity displayed between the RB pocket domain and the two general factors TBP and TFIIB Introduction (Hagemeier et al., 1993a). Although the mechanism of this repression is not yet understood, masking of The pocket proteins, RB, p107, p130, represent a TBP binding capacity may be involved for both E2F family of transcriptional regulators which have a highly (Hagemeier et al., 1993b) and UBF (Cavanaugh et al., conserved transcription factor-binding `pocket' do- 1995). However, in addition, RB must contain main. The prototype of this family, RB, has been independent repressive capacity, which acts directly found mutated in all Retinoblastomas as well as in to silence the transcriptional machinery (Weintraub et many other tumours and this has led to its al., 1992, 1995). characterization as a tumour suppressor protein A growing body of evidence suggests that RB not (Weinberg, 1995). Although p107 and p130 have only functions to regulate cell proliferation but is also functions which overlap those of RB, these two genes intimately involved in regulating cell dierentiation. have not been found to be mutated in tumours. Indirect evidence has come from the generation of RB Perhaps, as a consequence, RB still remains the most de®cient mice which survive until 14.5 days post- studied member of this family. coitum, a time when dierentiation of a number of The pocket proteins (most commonly RB) have dierent cell types is known to be initiated (Jacks et al., been shown to bind and regulate a number of 1992; Lee et al., 1992; Clarke et al., 1992). More direct dierent transcription factors (Kouzarides, 1995). evidence comes from the analysis of RB-interacting Their most characterized target is the E2F family of proteins (reviewed in Kouzarides, 1995). A number of transcriptional activators, E2F1-5 (Weinberg, 1995). these (myoD, C/EBP, Elf1) are transcription factors The three pocket proteins bind and regulate a distinct which have a role in regulating the dierentiation of subset of E2F family members but the reason or speci®c cell lineages (Gu et al., 1993; Chen et al., 1996; mechanism behind this speci®city is not yet clear. The Wang et al., 1993). Thus RB may not only regulate the E2F1 protein has the ability to promote G1 to S cell cycle, but may also be involved in regulating cell phase transition by stimulating the activity of S-phase dierentiation. speci®c genes (Nevins, 1992; Johnson et al., 1993). The Here we report that HBP1, an HMG-box transcrip- binding of RB to E2F1 leads to a suppression of these tion factor, is bound and regulated by RB and the other pocket proteins. We show that HBP1 has the capacity to induce morphological transformation of Correspondence: T Kouzarides cells in culture and that its transcriptional activation Received 6 February 1997; accepted 1 May 1997 capacity is regulated by the viral E1A protein. RB targets HBP1 PLavenderet al 2722 1

Results P P

To identify novel targets for pocket proteins, we Transfected DNA HA-IDP HA-HB HA-HB employed a yeast two-hybrid screen using the pocket RB RB RB domain of p130 (aa 322 ± 1070) as a bait. A library of IP Ab RB RB Cyc B2 cDNA's derived from mouse embryos was screened Western Ab HA HA HA using this bait. This screen identi®ed multiple isolates of an open reading frame from an uncharacterized gene. During our isolation of the mouse full length gene, the full length sequence of the rat homologue was reported by Lesage et al. (1994). This rat cDNA, HBP designated HBP1 (HMG-Box containing Protein 1) had the features of a transcriptional regulator since it contained a region highly homologous to the HMG- Figure 1 HBP1 interacts with RB in vivo. U2OS osteosarcoma box DNA binding domain. HBP1 is a ubiquitously cells were transiently transfected with 5 mg of pCMV RB 379 ± expressed gene whose expression is stimulated during 928 and either 5 mg of pCMV 2N3T IDP1 (lane 1) or 5 mg pCMV 2N3T HBP1 (lanes 2 and 3). Cells were harvested after 36 h, the dierentiation of adipocytes and muscle cells. whole cell extracts were prepared and were subjected to We are con®dent that the mouse HBP1 protein immunoprecipitation with either an anti RB monoclonal anti- isolated by the two-hybrid screen represented a true body (G3-245, Pharmingen) or an anti cyclin B2 antibody (lane p130 interacting target in yeast, for several reasons; (i) 3). Samples were separated by SDS ± PAGE. Immunoprecipitated it was isolated multiple times from the library; 23 out proteins were subjected to Western blotting, probed with an anti HA antibody (12CA5, Boehringer Mannheim) and visualized of 110 clones examined were HBP1. (ii) each of the 23 using an ECL kit (Amersham), followed by exposure to X-ray isolates were overlaps of the same region of HBP1. (iii) ®lm all the HBP1 isolates contained the classic LXCXE motif known to mediate interactions with the pocket domain. (iv) the interactions were speci®c since the HBP1 isolates did not interact with a lamin control bait. (v) the E2F1 protein, known to interact with the pocket domain, was also isolated with an identical frequency to HBP1 from the same yeast screen, indicating that our assay was able to recognize true pocket-interacting proteins. We next tested whether HBP1 speci®cally interacts with p130 or whether HBP1 can recognize other pocket proteins such as RB and p107. Using several assays, it became clear that HBP1 was able to recognize the GST-fusions pocket domain of p130, RB and p107: (1) in the two- hybrid assay p130 and RB interact equally well with HBP1 (data not shown), (2) HBP1 can bind RB in vitro and (3) p130, RB and p107 are all capable of repressing the activation functions of HBP1 (see

below). Since we could not demonstrate speci®city for INPUT HBP 11-63 HBP 11-63 mut HBP 37-120 GST any pocket protein, we present here the characterization of the interaction between HBP1 and RB, since this represents the most characterized member of the RB pocket family. The HBP1 protein can interact with RB not only in yeast cells but also in mammalian cells. When U20S cells are co-transfected with RB and an HA-tagged version of the rat HBP1, immunoprecipitation of RB 1 2 3 4 5 co-immunopuri®es HBP1, as recognized by a Western Figure 2 Identi®cation of HBP1 sequences interacting with RB. blot with an HA-speci®c antibody (Figure 1, lane 2). GST, GST HBP 11 ± 63, GST HBP 11 ± 63 mut or GST HBP 37 ± The RB-speci®c antibody does not immunopurify a 120 were incubated with in vitro translated RB 379 ± 928 and subjected to GST pull down. Precipitated proteins were resolved control protein (IDP1) tagged with HA (lane 1) and by SDS ± PAGE, gels were ®xed, ¯uorographed, dried then immunoprecipitation with a control antibody (against autoradiographed cyclin B2) does not bring down HA-HBP1 under these conditions (lane 3). Thus RB and HBP1 can form a speci®c complex in mammalian cells. The HBP1 protein can also interact with RB in vitro. LXCXE motif of HBP1 represents a high anity A domain of HBP1 (11 ± 63) which contains the binding site for RB. HBP1 also contains a second LXCXE motif (LQCNE) linked to GST can interact binding site for RB which resides in a region directly very eciently with in vitro translated and radiola- C-terminal to the LXCXE motif. A GST fusion linked belled RB in a GST pull-down assay (Figure 2). to residues 37 ± 120 of HBP1 is able to interact with Mutagenesis of the C and E residues of the LXCXE RB, independently of the LXCXE motif, although this motif to A residues severely impairs the ability of interaction is of low anity when compared to the HBP1 11 ± 63 to bind RB. This result con®rms that the anity of the LXCXE containing HBP 11 ± 63. Thus, RB targets HBP1 P Lavender et al 2723 like the E1A protein, HBP1 contains a high and a low RB. Figure 5 shows that the activation capacity anity binding site for RB. residing within HBP 37 ± 120 is still repressed by RB The HBP1 protein has the characteristic features of but less eciently than the domain which contains the a transcription factor since it contains a domain highly high anity LXCXE motif. This repressive capacity is homologous to the HMG-box DNA binding domain. consistent with the presence of a second RB binding We therefore wanted to know whether HBP1 has the site in this domain (Figure 2). This repressive eect of capacity to regulate transcription. Since we did not RB is speci®c to HBP1 since the activation capacity of know the precise DNA binding site for the HBP1 HMG-box, we replaced this DNA binding domain with that of GAL4 and asked whether HBP1 can stimulate transcription from a GAL4-site bearing promoter. Figure 3 shows that a GAL4 fusion with almost the entire HBP1 sequence (GAL4-HBP 11 ± 394) has very weak, but detectable activation capacity relative to GAL4 alone. However, deletion analysis of HBP1 revealed that it contains a very powerful activation domain whose activity is masked by specific inhibitor domains. One of these domains is at the C- terminus of the protein since HBP 11 ± 120 has more than tenfold higher activity than HBP 11 ± 394. A second inhibitor domain resides within the N-terminus of the protein since HBP 37 ± 120 has almost tenfold higher activity than HBP 11 ± 120. Altogether, removal of both the N- and C-terminal inhibitor domains generates an 83 residue activation domain (HBP 37 ± 120) which has 100-fold higher activity than the almost full length HBP 11 ± 394. Further deletion of the activation domain reduces activity but does not abolish it (see HBP 93 ± 120), indicating that the Figure 4 Pocket proteins repress the transcriptional activation HBP1 activation domain has, like other activation capacity of HBP1. U2OS cells were transfected with 1 mgof domains, a modular structure composed of several 56Gal4 E1bCAT reporter and 1 mg of pHKGT HBP 11 ± 120 independently activating regions. together with either 500 ng of pCMV neo Bam (7); pCMV RB 379 ± 928 (RB); pCMV p130 (p130) or pCMV p107 (p107). The We next tested whether the transcriptional activation CAT activity is de®ned relative to HBP 11 ± 120, which was given capacity of HBP1 is regulated by pocket proteins. a value of 100 Figure 4 shows that all three pocket proteins, RB, p130 and p107, can repress the activity of HBP 11 ± 120, which contains both the high and low anity binding sites for RB. Further analysis of RB-mediated repression of HBP1 indicates that this function relies on an intact pocket domain, since a RB pocket mutant, RBD22, is unable to repress the activity of HBP 11 ± 120 (Figure 5). As the LXCXE motif lies outside the HBP1 domain which shows the highest activation capacity (Figure 3), we asked whether a domain lacking this motif, HBP 37 ± 120, can be repressed by

Figure 3 HBP1 contains a masked activation domain. Gal 4 Figure 5 HBP1 repression by RB is pocket-dependent. U2OS fusions of various fragments of HBP1 were generated. 1 mgof cells were transfected with pHKGT HBP 11 ± 120, pHKGT HBP Gal4 fusion was transiently transfected into U2OS cells along 37 ± 120 or pHKGT p53 1 ± 73 (1 mg) and 56Gal4 E1b CAT with 1 mgof56Gal4 E1bCAT reporter. Thirty-six hours after reporter (1 mg) along with either 500 ng pCMV neo Bam (7), transfection, cells were harvested, whole cell extracts prepared and pCMV RB 379 ± 928 (RB) or pCMV RBD22 (RBD22). The CAT CAT assays performed. Activation capacity of HBP1 is de®ned as activity is shown relative to the GAL4 activator control (7) fold over Gal4 DNA binding domain alone which was given a value of 100 RB targets HBP1 PLavenderet al 2724

Figure 6 HBP1 transcriptional activity is repressed by E1A 12S. U2OS cells were transfected with 1 mg of either pHKGT HPB 11 ± 120 or pHKGT HBP 93 ± 120 along with 1 mgof56Gal4 E1bCAT reporter and 50 ng of the indicated E1A 12S expression vectors. CAT activity is expressed relative to the Gal4 DNA binding domain alone

the p53 activation domain is not aected under the residues 93 ± 120. The activation of HBP1 93 ± 120 is same conditions (Figure 5). repressed by E1A in a CR1 dependent manner and The E1A protein is capable of regulating the activity once again, this repression does not depend on the of E2F, another target for the pocket proteins. We binding of CBP and RB (Figure 6). These results therefore asked whether E1A could regulate the demonstrate that a small 27 residue region of HBP1 activation capacity of HBP1. Figure 6 shows that has transcriptional activation capacity which can be expression of the E1A 12S product severely inhibits the repressed by E1A. activation capacity of HBP 11 ± 120. However, E1A To assess the biological functions of HBP1 we was unable to repress the activation capacity of cloned an HA tagged version of HBP1 cDNA in a another HMG-box protein, GAL4-mSry (data not retroviral vector, pBabe-puro, and generated an shown). Deletion mutants of E1A were then used to NIH3T3 cell line (NIHHBP) which expresses the map the region of E1A responsible for inhibiting HBP1 protein, as recognized by an HA-speci®c HBP1 functions. Removal of conserved region 2, CR2, antibody (data not shown). In addition, a control (which contains the LXCXE binding site for pocket cell line was generated which carried the pBabe-puro protein binding) does not aect E1A repression, vector (NIHpBabe). Microscopic analysis of the two whereas removal of conserved region 1 (CR1) almost cell lines indicates that the NIHHBP cells are completely abolishes E1A mediated repression. The morphologically transformed (Figure 7b) compared E1A CR1 region contains binding sites for both RB to the control NIHpBabe cells (Figure 7a). The and CBP. To assess whether binding of these factors NIHHBP cells are rounded and elongated compared was necessary for E1A's ability to repress HBP, we to the ¯attened morphology of the NIHpBabe cells, a used E1A mutants RB mut and CBP mut which feature which is most obvious when the cells are under abolish RB and CBP binding respectively (Trouche con¯uent. As the cells become more con¯uent, it and Kouzarides, 1996). These E1A mutants were becomes apparent that they have lost contact perfectly capable of repressing the activity of HBP1 inhibition, forming discrete areas of overgrown cells 11 ± 120 (Figure 6). These results indicate that E1A (Figure 7c). When the cells become fully con¯uent, the does not repress HBP1 activity by competing for the loss of contact inhibition in manifest as discrete foci of CBP co-activator, as it does for c-Fos (Bannister and morphologically transformed cell (Figure 7d). Thus, Kouzarides, 1995), and that RB binding is also not like other targets of RB and E1A, the HBP1 protein involved in this process. The domain of HBP1 which is has the capacity to induce morphological transforma- responsive to E1A mediated repression resides within tion when overexpressed. RB targets HBP1 P Lavender et al 2725 a c

b d

Figure 7 Morphology and focus forming ability of the NIH HBP1 cell line. NIH3T3 cells were infected with pBabe(puro)HBP1 or pBabe(puro) retroviruses and pools of puromycin resistant cells were generated. NIH HBP1, and NIH pBabe established cell lines were examined by light microscopy. Shown are representative phase-contrast micrographs of (a) NIH pBabe control cells, (b) NIH HBP1 cells at low density which exhibit altered morphology compared to control cells, (c) NIH HBP1 cells at subcon¯uence showing signs of loss of contact inhibition, and (d) NIH HBP1 cells at con¯uence exhibiting morphologically transformed foci

Discussion Zi, 1994; Dubendor et al., 1992). However, certain features exhibited by HBP1 (its masked activation Our analysis of HBP1 indicates that this transcription domain and RB binding characteristics) are very factor has the capacity to recognize the pocket reminiscent of the viral transcription factor E1A domains of RB, p130 and p107. This result is (Figure 8). Firstly, like HBP1, E1A is unable to consistent with the presence of an LXCXE motif activate transcription as a full length GAL4 fusion. within HBP1, a sequence known to interact with the This is due to an inhibitor domain at the C-terminus of pocket domain. The interaction between HBP1 and E1A which speci®cally represses the activity of the CR1 RB can be demonstrated using a number of dierent activation domain (Sollerbrant et al., 1996). Secondly, in vivo and in vitro assays, indicating that this like HBP1, the CR1 activation domain of E1A has a interaction is not due to the fortuitous presence of binding site for RB and is repressed by the binding of an LXCXE motif. Further evidence for the biological RB (Trouche and Kouzarides, 1996). Thirdly, like relevance of these interactions comes from the fact HBP1, E1A has a second high anity RB-binding that all three pocket proteins can repress the LXCXE motif which falls outside its activation domain transcriptional activation capacity of HBP1. and within CR2 (Moran, 1993). Thus HBP1 is the ®rst Although we cannot exclude the possibility that some characterized transcription factor which has transcrip- selectivity for a certain member of the pocket family is tional activation and RB binding characteristics which displayed when these proteins are not overexpressed, resemble those of the viral E1A protein. these experiments at least con®rm that the binding of The HBP1 protein shares some (but not all) of the a pocket protein to HBP1 leads to regulation of its properties of the E2F transcription factor. HBP1 has a functions. transcriptional activation domain which is repressed by The HBP1 protein clearly has the potential to pocket proteins and has the ability to mediate activate transcription, but this activity is masked by transformation of cells in culture. However, the inhibitor domains. Such inhibitor activity has been proliferative functions of HBP1 must be distinct in described for a number of transcription factors nature to those of E2F1 since HBP1 is unable to (Gashler et al., 1993; Brown et al., 1995; Nerlov and overtly regulate the cell cycle (data not shown). In RB targets HBP1 PLavenderet al 2726

Figure 8 HBP1 and E1A 12S exhibit similar functional domains. HBP1 has an activation domain whose activity is masked by inhibitor domains at the N- and C-terminus (only C-terminal domains shown). E1A has a C-terminal inhibitor domain which speci®cally represses CR1-dependent activation. The activation domains of both E1A and HBP1 have binding sites for RB and are transcriptionally repressed by the pocket protein. Both E1A and HBP1 contain a second, high-anity binding site for RB, the LXCXE motif, which lies outside the activation domain

addition, our preliminary experiments indicate that biological functions of HBP1 is regulated by the pocket HBP1 is unable to relieve RB-induced growth arrest in proteins. the ¯at cell formation assay. These results suggest that We have shown that E1A can repress the the biological functions of HBP1 are distinct from transactivation functions of HBP1. Given that E1A is those of the E2F family. a known suppressor of dierentiation pathways, these The biological functions of HBP1 may also be results also point to a dierentiation speci®c function related to the process of dierentiation given that for HBP1. Transcriptional repression could be the HBP1 expression is stimulated during myogenic and result of E1A binding a co-activator protein whose adipocyte dierentiation (Lesage et al., 1994). It seems function is required to stimulate the HBP1 activation unlikely that HBP1 is an inducer of a speci®c domain. We have shown that the sequestered co- dierentiating pathway, given its expression in two activator is not CBP, which is a protein capable of distinct lineages. One possibility we favour is that mediating the activity of numerous other activation HBP1 acts more broadly at an early stage of domains (Janknecht and Hunter, 1996). This raises the dierentiation, perhaps during commitment. Repres- possibility that the protein sequestered by E1A has a sion of its activation functions by RB and other pocket more speci®c role, such as in dierentiation. Recently, proteins may therefore be required in order to suppress E1A has been shown to possess residues within CR1 the commitment to dierentiation when cells are (distinct from the RB and CBP binding sites) which are quiescent or arrested in G1. In other words, repression required for the repression of skeletal muscle speci®c of HBP1 activity may be necessary to prevent the cell enhancers (SandmoÈ ller et al., 1996). A protein which from embarking on a dierentiating pathway, which binds such residues may have the characteristics may otherwise be triggered by the cessation of cell expected for a transcriptional co-activator of HBP1. proliferation. Although super®cially a dierentiation The HBP1 transcription factor represents the ®rst function for HBP1 seems contradictory to its characterized target of the pocket family, other than transformation characteristics, HBP1 may be analo- E2F, whose activation functions are repressed by this gous to a transcription factor such as c-jun which has family. Biologically HBP1 may be involved in the roles in both proliferation and dierentiation (Angel process of cell proliferation as well as dierentiation. and Karin, 1991). We therefore cannot rule out a role Such apparently diverse functions may explain the need for HBP1 in either process. Detailed characterization for the tight regulation of HBP1 activity by inhibitor of the transformation and possible dierentiation domains. Thus, signals which unmask the activation features of HBP1 (including the isolation of HBP1 capacity of HBP1 may regulate the dierent functions responsive genes) is required to establish which of the associated with it. RB targets HBP1 P Lavender et al 2727 Materials and methods Helin (p107); and P Whyte (p130). The Ad5 E1A 12S and E1A DCR1 and DCR2 were expressed from pBJ9O, gifts Yeast two-hybrid system from H Land. E1A p300 and Rb mutants and the p300.Rb double mutant were reported previously (Trouche and Components of the yeast two-hybrid system were provided Kouzarides, 1996). An HA-tagged expression vector, by S Hollenberg. The bait was constructed by insertion of pCMV 5'2N3T was constructed by replacing the SV40 the pocket domain of p130 (amino acids 322 ± 1070, gift promoter of pHK (Hagemeier and Kouzarides, unpub- from P Whyte) into the Lex A fusion vector pBTM116 lished) with the cytomegalovirus promoter, then inserting (Bartel and Fields, unpublished). The library screened was two nuclear localization signals and three HA epitope tags mouse embryonic cDNA prepared from 9.5/10.5 d.p.c. at the 5' end of a multiple cloning site. For transformation CD1 mouse embryos ligated into pVP16. Interaction assays an HA-tagged version of HBP-1 was cloned into assays were carried out in yeast strain L40 as described pBabe (puro), a gift from H Land, and expressed under the (Vojtek et al., 1993; S Hollenberg, personal communica- control of the murine Moloney Leukaemia Virus promoter. tion). 66108 primary transformants were screened and grownon±THULLplatesinthepresenceof3mM 3- aminotriazole. After 3 days, 110 colonies had grown, these GST fusion proteins and pull-down assay yeast were further cultured and library plasmids isolated by Various domains of HBP-1 were cloned into the relevant standard procedures. Sequence tags were obtained using pGex vector (Pharmacia), using PCR or inherent restric- United States Biochemicals Sequenase V 2.0 kits. tion sites. Oligonucleotide directed mutagenesis was used to generate mutants. Recombinant proteins were expressed Cell cultures, transfections, CAT assays and retroviral infection in, and puri®ed from, E. coli as reported previously (Bannister et al., 1991). Pull-down assays were performed U2OS human osteosarcoma cells, BOSC23 and NIH3T3 as described previously (Hagemeier et al., 1994). were maintained in DMEM supplemented with 10% fetal calf serum and grown at 378C(5%CO2) as monolayers and split 1 : 4 every 3 days. For transfection of U2OS, In vitro translation of proteins 8 *2.5610 cells were seeded 6 h prior to the procedure. The pocket domain of pRb (amino acids 379 ± 928) was Transfection was performed by the calcium phosphate co- expressed from pING 14, an SP6 transcription vector (S precipitation technique. DNA was left on the cells for 14 h Inglis, unpublished results). In vitro translation products before washing, cells were harvested 36 h after transfec- were generated using the TNT coupled transcription- tion. Extracts from transfected cells were then used for translation system (Promega) following manufacturers CAT assays and for Western blots, using a GAL4 DBD instructions. speci®c antibody to check for expression of transfected proteins. The retroviral packaging cell line BOSC 23 was transfected with 30 mg of pBabe (puro) or pBabe (puro) Coimmunoprecipitation of Rb and HBP-1 proteins HBP-1 plasmids by calcium phosphate co-precipitation as Approximately 16108 U2OS human osteosarcoma cells described by Pear et al., 1993). Medium containing the were transfected with pCMV Rb (379 ± 928) and pCMV retrovirus was harvested 24 h following removal of the 5'2N3T HBP-1(11 ± 394) as described above. Thirty-six precipitate and used to infect NIH3T3 cells at 30% hours after transfection cells were harvested by scraping con¯uency. Uncloned pools of puromycin resistant cells into PBS, pelleted then lysed by rotation in EBC buer were generated and maintained as described by Xu et al. containing 200 mM NaCl as described (Hagemeier et al., (1995), except that cells growing in the supernatant were 1994). Immunoprecipitations were carried out using an anti kept each time and reseeded when NIHpBabe or NIHHBP RB monoclonal antibody (G3-245, Pharmingen) or anti- cells were split. cyclin B2 antibody (gift from J Pines) in the presence of protein G-Sepharose and protein A-agarose (Sigma). In vivo expression plasmids Precipitates were washed ®ve times in NETN buer as described (Hagemeier et al., 1994), separated by SDS ± For the Ga14 fusion experiments, various domains of PAGE and blotted onto nitrocellulose. Western blotting HBP-1 were cloned into pHKGT (Hagemeier and was performed with anti-HA monoclonal antibody Kouzarides, unpublished) using PCR or inherent restric- (12CA5, Boehringer). Immunoreactive bands were re- tion sites. pHKGT has the GAL4(1 ± 147) DBD under the vealed using an ECL kit (Amersham) according to the control of an SV40 promoter. The reporter construct for manufacturers instructions. Ga14-fusions was G5E1b.CAT a gift from M Green. Control plasmids were Ga14.p53(1 ± 73) (Martin and Kouzarides, unpublished) and Ga14.mSry(138 ± 395), a gift from R Dubin. Pocket proteins were expressed under Acknowledgement the control of the cytomegalovirus promoter and were gifts We thank Dennis McCance for comments on this manu- from W Kaelin (Rb: amino acids 379 ± 928 and RbD22); K script.

References

Angel P and Karin M. (1991). Biochim. et Biophys. Acta, Cavanaugh AH, Hempel WM, Taylor LJ, Rogalsky V, 1072, 129 ± 157. Todorov G and Rothblum LI. (1995). Nature, 374, 177 ± Bannister AJ, Cook A and Kouzarides T. (1991). Oncogene, 180. 6, 1243 ± 1250. Chen P-L, Riley DJ, Chen Y and Lee W-H. (1996). Genes & Bannister AJ and Kouzarides T. (1995). EMBO J., 14, 4758 ± Dev., 10, 2794 ± 2804. 4762. Clarke AR, Maandag ER, van Roon M, van der Lugt NMT, Brown HJ, Sutherland JA, Cook A, Bannister AJ and van der Valk M, Hooper ML, Berns A and te Riele H. Kouzarides T. (1995). EMBO J., 14, 124 ± 131. (1992). Nature, 359, 328 ± 330. RB targets HBP1 PLavenderet al 2728

Dubendor JW, Whittaker LJ, Eltman JT and Lipsick JS. Moran E. (1993). Curr. Opin. Gen. Dev., 3, 63 ± 70. (1992). Genes & Dev., 6, 2524 ± 2535. Nevins JR. (1992). Science, 258, 424 ± 429. Gashler AL, Swaminathan S and Sukhatme VP. (1993). Mol. Nerlov C and Zi EB. (1994). Genes and Dev., 8, 350 ± 362. and Cell. Biol., 13, 4556 ± 4571. Pear WS, Nolan GP, Scott ML and Baltimore D. (1993). Gu W, Schneider JW, Condorelli G, Kaushal S, Mahdavi V Proc.Natl.Acad.Sci.USA,90, 8392 ± 8396. and Nadal-Ginard B. (1993). Cell, 72, 309 ± 324. SandmoÈ ller A, Meents H and Arnold HH. (1996). Mol. and Hagemeier C, Bannister AJ, Cook A and Kouzarides T. Cell. Biol., 16, 5846 ± 5856. (1993a). Proc. Natl. Acad. Sci. USA, 90, 1580 ± 1584. Sollerbrant K, Chinnadurai G and Svensson C. (1996). NAR, Hagemeier C, Caswell R, Hayhurst G, Sinclair J and 24, 2578 ± 2584. Kouzarides T. (1994). EMBO J., 13, 2897 ± 2903. Trouche D and Kouzarides T. (1996). Proc. Natl. Acad. Sci. Hagemeier C, Cook A and Kouzarides T. (1993b). Nuc. USA, 93, 1439 ± 1442. Acids Res., 21, 4998 ± 5004. Vojtek AB, Hollenberg SM and Cooper JA. (1993). Cell, 74, Jacks T, Fazeli A, Schmitt EM, Bronson RT, Goodell MA 205 ± 214. and Weinberg RA. (1992). Nature, 359, 295 ± 300. Wang C-Y, Petryniak B, Thompson CB, Kaelin WC and Janknecht R and Hunter T. Nature, 383, 22 ± 23. Leiden JM. (1993). Science, 260, 1330 ± 1335. Johnson DG, Schwarz JK, Cress WD and Nevins JR. (1993). Weinberg RA. (1995). Cell, 81, 323 ± 330. Nature, 365, 349 ± 352. Weintraub SJ, Chow KNB, Luo RX, Zhang SH, He S and Kouzarides T. (1995). Cancer Biol., 6, 91 ± 98. Dean DC. (1995). Nature, 375, 812 ± 815. Lee EY-HP, Chang C-Y, Hu N, Wang Y-CJ, Lai C-C, Weintraub SJ, Prater CA and Dean DC. (1992). Nature, 358, Herrup K, Lee W-H and Bradley A. (1992). Nature, 359, 259 ± 261. 288 ± 294. White RJ, Trouche D, Martin K, Jackson SP and Lesage F, Hugnot J-P, Amri EZ, Grimaldi P, Barhanin J and Kouzarides T. (1996). Nature, 382, 88 ± 90. Lazdunski M. (1994). Nuc. Acids Res., 22, 3685 ± 3688. Xu G, Livingston D and Krek W. (1995). Proc. Natl. Acad. Lukas J, Petersen BO, Holm K, Bartek J and Helin K. (1996). Sci. USA, 92, 1357 ± 1361. Mol. and Cell. Biol., 16, 1047 ± 1057.