Identification of discrete structural domains in the retinoblastoma protein. Amino-terminal domain is required for its oligomerization.

To characterize the protein product of the retinoblastoma tumor suppressor gene biochemically, a recombinant human protein was produced in an Escherichia coli expression system. The full-length protein, p110RB, and an amino-terminal truncated form, p56RB, were expressed and purified to near homogeneity by conventional chromatographic procedures. To probe the structural organization of the retinoblastoma protein the purified proteins were subjected to partial proteolysis by trypsin, chymotrypsin, and subtilisin. Four discrete structural domains were revealed in p110RB by this method. Two of these structural domains, found in both p56RB and p110RB, were mapped to the carboxyl-terminal half of the protein and corresponded to the SV40 large T binding domains defined previously by genetic methods. In addition two distinct domains in the amino-terminal half of the protein were also defined. A potential role for these newly defined amino-terminal domains was uncovered upon analysis of the purified proteins by nondenaturing polyacrylamide gel electrophoresis. p110RB revealed multiple bands by this method, suggesting the formation of oligomeric structures by the protein, while this property was not observed for p56RB. Electron microscopy of p110RB revealed linearly extended, macromolecular structures, further supporting the formation of homologous higher order structures by the full-length retinoblastoma protein. Analysis of the interactions between retinoblastoma protein molecules using the yeast two-hybrid system confirmed that the retinoblastoma protein could self-associate and that this association was mediated by interactions between the amino- and carboxyl-terminal ends of the protein. These observations suggest that the retinoblastoma protein contains multiple structural domains with the amino-terminal domains being required for oligomerization of the full-length protein.

TO characterize the protein product of the retinoblastoma tumor suppressor gene biochemically, a recombinant human protein was produced in an Escherichia coli expression system. The full-length protein, p l loRa, and an amino-terminal truncated form, p66=, were expressed and purified to near homogeneity by conventional chromato~aphic procedures. To probe the structural o r g a n i~t i o n of the r e t i n o b l~o m a protein the purified proteins were subjected to partial proteolysis by trypsin, chymotrypsin, and subtilisin. Four discrete structural domains were revealed in pllORB by this method. Two of these structural domains, found in both pWRB and p l loRB, were mapped to the carboxyl-terminal half of the protein and corresponded to the SV40 large T binding domains defined previously by genetic methods. In addition two distinct domains in the amino-terminal half of the protein were also defined. A potential role for these newly defined amino-terminal domains was uncovered upon analysis of the purified proteins by nondenaturing polyacrylamide gel electrophoresis. pllORB revealed multiple bands by this method, suggesting the formation of oligomeric structures by the protein, while this property was not observed for ~5 6~~. Electron microscopy of p l loRa revealed linearly extended, macromolecular structures, further supporting the formation of homologous higher order structures by the full-length retinoblastoma protein. Analysis of the interactions between retinoblastoma protein molecules using the yeast two-hybrid system confirmed that the retinoblastoma protein could self-associate and that this association was mediated by interactions between the amino-and carboxyl-terminal ends of the protein. These observations suggest that the retinoblastoma protein contains multiple structural domains with the amino-terminal domains being required for oligomer~ation of the fulllength protein.
Retinoblastoma protein (RB)' is the subject of intense * The work performed in the authors' laboratory was supported by the Council for Tobacco Research Grant 2992R2, National Institutes of Health Grant EY05758, and National Cancer Institute Grant CA58318. The first two authors have made equal contributions to this work. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section investigation in an effort to unravel its role in the regulation of cell growth (Goodrich and Lee, 1993b;Hollingsworth, et al., 1993). The protein was identified upon cloning of the gene whose defect predisposes to the pediatric cancer, retinoblastoma (Friend et al., 1986;Lee et aL, 1987b), and was subsequently found to be mutated in many cancers including osteosarcoma and small cell lung carcinoma (Harbor et d, 1988, Horowitz et at., 1990Shew et al., 1989). Reintroduction of the wild-type RB gene into these tumor cells, by retrovirusmediated gene transfer% consistently suppressed their tumorigenic ability, suggesting that RB is a general tumor suppressor (Huang et ai., 1988;. RB is a nuclear phosphoprotein that exhibits cell cycledependent changes in its phosphorylation state (Buchkovich et al., 1989;Chen et al., 1989;DeCaprio et al., 1989), this being the first indication that RB may play a role in the cell division cycle. Microinjection of the full-length or carboxyl-terminal half of the protein into cells inhibits G1 phase progression, suggesting that RB may function during early G1 (Goodrich et al., 1991). The RB protein has been shown to form specific complexes with the transforming proteins of several DNA tumor viruses, including SV40 large T, adenovirus ElA, and human papilloma virus E7 (DeCaprio et al., 1988;Dyson et al., 1989;Whyte et at., 1988). Deletion analysis indicates the existence of two noncontiguous regions, from amino acids 394 to 571 (domain A) and from 649 to 773 (domain B), both of which are necessary for the binding of these transforming proteins (Huang et ab, 1990;Kaelin et al., 1990). Recently, this region of the RB protein was found to bind many cellular proteins including the transcription factor, E2F-1 (Helin et al., 1992;Kaelin et al., 1992;Shan et al., 1992), p48 (Qian et al., 1993), RBP-1, RBP-2 (Defeo-Jones et al., 1991), RbAP46 , and a protein phosph~tase type 1 catalytic subunit (Durfee et al., 1993). This is also the region that suffers the most mutations found in human tumors (Bookstein and Lee, 1991). The carboxyl-terminal one-third of the RB protein contains both nuclear localization sequences (Shew et al., 1990) and a DNA binding activity, which to date has not been shown to be specific for any particular DNA sequence (Lee et al., 1987a;Wang et al., 1990b).
Availability of pure proteins in sufficient quantities is an important prerequisite for detailed structural and functional analyses of proteins. Here we report the production and purification of both pllORB and ~5 6~ from an Escherichia coil expression system in quantities sufficient for biochemical and structural analysis. It is widely recognized that globular proteins are composed of discrete globular regions, generally referred to as domains (Rose, 1979;Wetlaufer, 1973;Wilson, 1992). Limited proteolysis has frequently been found to result in cleavage between structural domains, as the relatively Structural Domains of the p l 1 ORB 1381 unstructured segments of polypeptide chain connect~g domains are usually quite susceptible to proteolytic digestion (Fontana et al., 1986). In the present study we have analyzed the praducts of limited proteolysis of purified pllORB and p56=, using the proteolytic enzymes, trypsin, chymotrypsin, and subtilisin. Four structural domains were identified in the protein by this method, the location of which was determined by microsequence analysis.
The structure and function of the amino-terminal half of the RB protein are largely uncharacterized. No naturally occurring mutant protein has been found to contain mutations in this region. However, two of the structural domains identified mapped to this portion of the protein, and a cellular protein that specifically binds to this region has been identified? In addition, while characterizing the full-length RB protein a novel biochemical property was observed. Specifically, it concerns the ability of the purified RB protein to form oligomers in uitro, a property that requires the presence of its amino-terminal region. This additional level of structural organization between Rl3 molecules suggests complex structure/funct~on relationships for this protein and a possible functionai significance for the amino-terminal domain. These results provided further insights into the structure of the RB protein thus helping to establish s t~c t u r e /~n c t i o n relationships within the protein.

Cloning and Expression of R e~~~~t o~
Proteins-Retinoblastoma proteins were expressed in E. coli using the pETRbc construct previously described (Huang et QL, 1991). The entire coding sequence of Rb cDNA (2.8 kilobase pairs) was placed under the control of the T7 palymerase promoter using the PET& vector (Rosenberg et ai., 1987). Two species of RE3 were expressed from this construct, a fulllength pllORB protein and p56=, an amino-terminal truncated form of the protein.
Cell Growth-Transformed E. coli cells were grown to log phase ( A m = 0.5) at 28 "C in the presence of ampicillin (50 pg/ml) and chloramphenicol (50 gg/ml). Cultures were then induced with 0.2 mM isopropylt~o~lactoside and growth allowed to continue for 5 h after which cells were harvested by centrifugation. Cells from 20 liters of culture were resuspended in 200 ml of lysis buffer (50 mM Tris-HCI, pH 8.5, 100 mM NaCI, 1 mM EDTA, 10 mM 2-mercaptoethanol, 1 mM PMSF) and stored at -80 "C.
Purification-All procedures were carried out at 4 'C. The cell lysate from a 20-liter culture was thawed, diluted in 2.5 volumes of DEAE buffer (10 mM Tris-HC1, pH 8.5, 1 mM EDTA, 10 mM 2-mercapt~thanol, 1 mM PMSF), and further d i s~p t e d by sonication. Cell debris was removed by centrifugation at 40,000 rpm for 1 h. The high speed supernatant (~1 liter) was loaded onto a DEAE-Sepharose CL-GB column (20 X 5 cm), ~uilibrated in DEAE buffer, at a flow rate of 1 ml/min. The flow through was directly loaded onto a P-11 cellulose phosphate column (20 X 2.5 cm). Following sample loading the columns were washed with 400 ml of DEAE buffer. The P-11 column was then disconnected from the DEAE column and washed with a further 500 ml of DEAR buffer. pllORB and p56= were eluted from the P-11 column with a 0-1 M NaCl linear gradient in DEAE buffer. ~1 1 0~ a n d~5 6~ were eluted together at 0.5 M NaC1. Fractions containing the RB proteins were pooled and concentrated by 90% saturation with ammonium sulfate, Protein was pelleted by centrifugation at 14,000 rpm for 15 min, and the pellets were resuspended in s-200 buffer (20 mM sodium phosphate, pH 7.5, 200 mM NaCI, 1 mM EDTA, 10 mM 2-mercaptoethanol). The sample was then loaded onto a Sephacryl S-200 column (115 X 2.5 cm) and protein eluted in S-200 buffer at a flow rate of 0.2 mlfmin. pllORB eluted first closely followed by ~5 6~. Purified protein was stored in 10% glycerol at -80 "e. Samples from each stage of the purification were analyzed by SDS-PAGE (Laemmli, 1970), and protein was quantitated by the Bradford dye binding assay using bovine serum albumin as a standard (Bradford, 1976). were incubated with TPCK-treated trypsin at a protease to substrate ratio of 1:lOOO (w/w) at 37 "C. At various time points the reactions were terminated by incubating an aliquot of the reaction mixtures with PMSF (final concentration, 40 mM). Digestions with chymotrypsin and subtilisin were carried out under similar conditions and at protease to substrate ratios of 1500 and 1:lOOO (w/w), respectively. The proteolytic fragments obtained were analyzed by 12% SDS-PAGE and visuaiized by Coomassie Blue staining. Molecular mass was determined from the electrophoretic mobility of standards. Following separation by SDS-PAGE proteins were electroblotted to nitrocellulose paper (Immobilon P), and the immunoreactive bands were detected using an alkaline phosphatase detection system. ~o~~t u r i~ PAGE 5nd I m~u~l o t A~l y s~-F o r nondenaturing gel electrophoresis, a gel containing 6% acrylamide, 0.16% bisacrylamide, and 123 mM Tris-glycine, pH 8.7, was set at 4 "C for 24 h.
Typically, Tris-HC1, pH 8.67, and glycerol were added to each protein sample to a final concentration of 100 mM and lo%, respectively, kept for 20 min on ice, loaded, and electrophoresed for 16 h at 4 "C under a constant 100 V setting. To perform immunoblot analysis, the electrophoresed gel was incubated in a 1% SDS solution for 30 min, after which samples were transferred onto a nitrocellulose paper by electroblotting for 20 h at 4 "C under a constant 200 mA setting in a 49 mM Tris glycine buffer, pH 8.9.
E~c~r o n Microscopy-To perform the i~~u n o g o l d labeling experiment, purified RB protein (0.1 pg/gl) in (20 mM NaP04, pH 7.0, 1 mM 8-mercaptoethanol, 1 mM EDTA) was placed on a carbon/ Formvar-coated grid. The grid was then incubated for 10 min with bovine serum albumin (0.1 gg/gl) in HN buffer (10 mM HEPES pH 7.4, 100 mM NaCl), followed by incubation with the first antibody (1:20 dilution) in HN containing bovine serum albumin (0.1 pg/pl) for 10 min, washed 5 times in HN, incubated with the second antibody (k20 dilution) in HN with bovine serum albumin (0.1 ~g/gl) for 10 min, washed 5 times in HN, incubated in a staining solution (1 part saturated uranyl acetate solution, 9 parts of double-distilled water), and viewed with an electron microscope.
Yeast Two-hybrid System-A cDNA encoding the RE3 carboxylterminal region (301-928) was cloned in frame with sequences for the Gal4 transactivating domain present on the expression vector pJBTB, to create pJD-1 (Fig. &I). On a second expression plasmid, pAS1, a cDNA encoding amino acids 1-300 of RB was joined to the Gal4 DNA binding domain, creating pAS/N-RB (Fig. SA). The yeast strain, Y 153, was cotransformed with the above plasmids, cotransformanta were assayed €or their ability to activate transcription of the lac2 gene, and the resulting @-galactosidase activity was measured. @-galactosidase activity was determined by the colony lift method and quantitated using an ONPG assay, as previously described (Durfee et ai., 1993).

~~~-~~r m~~
Sequence A~Lys~-Amino-terminal sequence of a number of proteolytic fragments was obtained. Protein fragments were separated by SDS-PAGE, electroblotted to Immobilon P, and sequenced using a model 477 Applied Biosystems protein sequencer. S e c o n d a~ Strttcture Prediction-Protein analysis and secondary Structure prediction were performed using DNASTAR software (DNASTAR, Inc., Madison, WE).

RESULTS
Purification of RB Proteins-Two species of RB were expressed from the pETRbc construct, the full-length protein, pllO=, and ~5 6~~ an amino-terminal truncated form of the protein. The amino terminus of p5SRE was determined by microsequence analysis and found to correspond to Met-379 in the full-length protein. Expression of p56= from Met-379 occurred due to the presence of an internal initiation site in the full-length RB cDNA. ~5 6~ was the major product expressed, with the ~l -l e n~h protein only constituting approximately 10% of the total RB produced. The reason for the low yield of pllORB is not understood. pllORB was resolved from p56= during the final pu~fication step, gel fttration on a Sephacryl S-200 column (Fig. IA). RB proteins were purified to near homogeneity as judged by SDS-PAGE (Fig. 1B).
Fohving this purification protocol, the yields from a 20-liter culture were typically 2 mg of pllORB and 15-20 mg of ~5 6~. Both proteins were capable of binding to SV40 large T antigen and caused cell growth arrest in early G1 when microinjected Structural Domains of the p l l ORE into the osteosarcoma cell line, Saos-2, indicating that the purified proteins were biologically active (Goodrich and Lee, 1993a).
Partial Tryptic Digestion of p56RB-To explore the structural domain(s) of the RB protein biochemically, ~5 6~~ was incubated with trypsin as described under "Materials and Methods," and the progress of proteolysis was followed by analyzing samples from various time points on SDS-PAGE (Fig. 2 4 ) . ~5 6~~ was first cleaved to a transient 48-kDa fragment (Fig. 2, lane 2 ) , which was further cleaved to two protease-resistant fragments, labeled A and B, of molecular mass 24 and 19 kDa, respectively (Fig. 2, lanes 3-7). Both proteolytic products A and B were resistant to further digestion for up to 2 h under the conditions described, even at a 10-fold higher ratio of trypsin to ~5 6~~ (data not shown). It is evident from these results that the sites of cleavage of p56RB, which contains a total of 76 potential trypsin cleavage sites scattered throughout the molecule, are restricted to a small segment of the molecule indicating a structural hierarchy in the protein causing many potential sites to be inaccessible.  Initial identification of these proteolytic products was performed by immunoblot analysis with specific antibodies against known regions of the RB protein. As shown in Fig.  2B, the monoclonal antibody against the A domain, a-1E5, recognized the 24-kDa protease-resistant fragment (Fig. 2B,  lanes 3-7), and an antibody against the B domain, a-B, recognized the 19-kDa fragment (data not shown). Thus, the genetically defined A and B domains of the T-antigen binding region correspond to these two protease-resistant fragments, A and B, of the protein.
Amino-terminal Sequence Analysis Defines the Boundaries of the Proteolytic Fragments-To define further the precise amino-terminal boundaries of these domains, the proteolytic products, resolved from each other by SDS-PAGE, were subjected to automated Edman degradation. The analysis revealed that the 24-kDa fragment had an amino-terminal sequence of MNTIQQLM, and the 19-kDa fragment had an amino-terminal sequence of VNSTANAE, which correspond to pllORB residues 379-386 and 622-629, respectively (Fig.  5A). The first methionine in domain A is the first amino acid in ~5 6~~ indicating that ~5 6~~ is not digested at all from the amino terminus. Initial digestion is occurring from the carboxyl terminus, which is highly susceptible to proteolytic degradation as evidenced by the fact that the carboxyl-terminal antibody a-11D7 did not recognize any of the proteolytic fragments generated (data not shown). The resulting transient 48-kDa fragment is then cleaved in two, yielding two protease-resistant fragments.
These domains contain a total of 8 cysteine residues. A naturally occurring mutation affecting one of these cysteines is known to eliminate RB's T antigen and DNA binding activity. Cysteines commonly contribute to structural interactions via hydrogen bonds, disulfide bonds, or coordinate metal ion binding. To check whether these structural domains were linked by disulfide bond formation the proteolytic digest was run on a nonreducing gel. The migration pattern of the bands was found to be the same as under reducing conditions (data not shown) indicating that the domains were not linked by disulfides. This observation is supported by a recent report, which showed that 4 cysteine residues within the binding pocket were necessary for binding to RB-associated proteins, but these residues do not seem to be involved in the formation of disulfide bonds (Stirdivant et al., 1992).
Trypsin-resistant Fragments of RB Are Resistant to Other Proteases-To confirm that the discrete structural domains

1383
were not unique to trypsin digestion, p56= was also subjected to cleavage by chymotrypsin and subtilisin, proteases that differ in their substrate specificities compared with trypsin. Digestion of ~5 6~~ with both these proteases produced a similar digestion pattern and yielded two major proteolytic fragments similar to those obtained by tryptic digestion (Fig.  3). In the presence of chymotrypsin ~5 6~~ was first cleaved to a transient 45-kDa fragment (Fig. 3A, lanes 1 and 2 ) ) which was in turn cleaved to yield two fragments of 24 and 17.5-kDa (Fig. 3A, lanes 3,4, and 5 ) . These correspond to fragments A and B from the tryptic digestion (as determined by immunoblot analysis) except that the B domain is slightly smaller and hence is labeled B'. Digestion by subtilisin proceeded in a similar fashion with a 46-kDa intermediate (Fig 3B, lane 1 ) being cleaved in two to give two fragments of 24 and 19-kDa) which are labeled A and B, respectively (Fig. 3B, lanes 3, 4,  and 5 ) . As with the tryptic digestion the fragments produced by chymotryptic digestion were very stable to further digestion even at a &fold increase in the ratio of protease to RB. Subtilisin, however, could totally digest the protein after longer incubation times and at higher amounts of proteolytic enzyme. This is to be expected because subtilisin cleaves peptide bonds nonspecifically. The fact that three proteases of differing specificity cleave ~5 6 '~ in close proximity to each other indicates that something other than the specificity of the protease is governing proteolysis.
Partial Tryptic Digestion of pllPB-To test whether the full-length protein, pllORB, contained the same and/or additional structural domains, pllORB was also subjected to tryptic digestion. Digestion of pllORB also occurred in a stepwise fashion with transient fragments giving way to more stable proteolysis-resistant domains (Fig. 4A). Tryptic digestion of pllORB yielded four protease-resistant fragments (Fig. 4A,   lane 4 ) labeled A, B, N, and R of molecular mass 24,19.5,30, and 10 kDa, respectively. Due to its small size, the R fragment can be seen clearly in the original gel but poorly after photography. The amino terminus of each peptide was sequenced to identify the exact location of these domains in the full-length protein. The amino-terminal sequence of A was found to be TVMNTIQQLM, which corresponds to amino acids 377-386 in pllORB. Therefore, this is the same A domain as found in ~5 6~~ with cleavage occurring two amino acids before the first methionine of ~5 6~~. The amino-terminal sequence of B was found to be identical to that of domain B obtained from the ~5 6~~ digestion. The 30-kDa fragment, labeled N, was first identified by the amino-terminal anti-peptide antibody, a-D (Fig. 4B), which recognizes RB amino acids 62-91, indicating that this fragment originates from the amino terminus of RB. Aminoterminal sequence analysis of the 30-kDa domain revealed the residues LTAATAAAAA, which correspond to amino acids 8-17 in pllORB. Fragment R was found to have the aminoterminal sequence IALQLENDTN, corresponding to amino acids 263-270 in pllORB (Fig. 5). A similar pattern was also obtained upon digestion of pllORB with subtilisin (data not shown) indicating that some structural organization in the protein is influencing proteolytic digestion. Whether R is a distinct domain or a sub-domain of N cannot be completely determined. However, the production of R corresponds to the formation of N from its higher molecular weight precursor, suggesting that there is a susceptible site between these two regions.
Oligomerization of p1lPB Revealed by Nondenuturing Gel Electrophoresis-Analysis of the purified proteins revealed an interesting biochemical property specific to the full-length RB protein. Following nondenaturing polyacrylamide gel electrophoresis multiple bands were revealed upon immunoblot analysis using an anti-RB antibody (Fig. 6A, lane 2). At least seven bands corresponding to various oligomers of the protein were detected. Because the distance migrated by a protein on a native gel is a complicated function of both protein size and charge it is difficult to establish the exact molecular weight of the oligomers. The ladder-like appearance of the bands observed suggested that each species may differ from the adjacent one by unit mass. Plotting distance migrated by each band in Fig. 6A, lane 2, against log10 (n), where n equals the ladder position from the bottom, we found a nearly linear relationship (Fig. 6B). This supports the notion that each band differs from the next one by the addition of another RB molecule. Recombinant RB from a baculovirus expression system, purified in a different manner, also produced a similar pattern (data not shown), eliminating the possibility that this may be an artifact of a purification protocol. When an aminoterminal truncated form of the protein, ~5 6~~ (amino acids Structural Domains of the p l 1 PB 379-928) (Fig. 6A, lune 3), was analyzed by a similar method, uranyl acetate, and viewed with an electron microscope. Imonly a single band was detected ( Fig. 6A; lune 4 ) . Pretreatment munogold labeling using an anti-RB antibody, Ab 245, shows of full-length RB protein with dithiothreitol (30 mM) failed numerous gold particles bound often in an apparently unduto abolish the formation of multiple bands, suggesting that lating or spiral manner along the structures indicating that oligomerization of the RB protein is mediated through an these structures contain the RB protein (Fig. 7, a-c). When interaction other than disulfide bond formation (data not the experiment was repeated without adding the RB protein, shown). The multiple band pattern was mhanced in the the labeled structures did not appear, indicating that the Presence of a small amount of SDS (0.002%), suggesting a structures do not originate from any of the reagents used in slight conformational change such as that induced by Postthe immunogold-labeling experiment. The linearly extended translational modification or binding to another protein might structures were not detected when purified p5GRB was exambe necessary for this process to occur in vivo.
ined (data not shown). Neither an anti-p53 antibody, Ab122 Electron M~~SCOPY Of RB shows the O1&mric Struc- (Fig. 7d), nor an antibody directed against SV40 large T ture-when the retinoblastoma Protein was analYzedbY elecantigen, Ab419 (data not shown), labeled the structures elimtron microscopy linearly extended macromolecular structures inating the possibility that the structures seen were due to were observed (Fig. 7). The Protein Was placed directly on nonspecific antibody binding. These data suggest that the carbon/Formvar-coated copper grids, negatively stained with observed structures consisted mainly of the RB protein, providing another indication that the retinoblastoma protein Intermolecular Interactions of the RB Protein-To identify regions of the retinoblastoma protein that could interact with Quantitation of @-galactosidase activity then serves as a re- promoter. It has been demonstrated that the physical association detected in this system reflects the known in vitro (Durfee et al., 1993). Using this assay we have studied RB's   expressed in E. coli, was electrophoresed through a 7.5% SDS-PAGE and stained with Coomassie Blue ( l a n e 1). Purified full-length RB protein (0.8 pg) (in 50 mM NaPO,, pH 7.5, 250 mM NaCI, 1 mM EDTA, 1 mM 8-mercaptoethanol, and 10% glycerol) was electrophoresed through a 6% polyacrylamide nondenaturing gel and analyzed by immunoblotting using Ab245, an anti-RB antibody ( l a n e 2). Arrowhead denotes the position where the RB protein was loaded. Purified ~5 6~ protein (1.5 pg), expressed in E. coli, was analyzed as described for lane 1 ( l a n e 3). Purified ~5 6~~ protein (0.5 pg) (in 50 mM NaPO,, pH 7.5, 250 mM NaCl, 1 mM EDTA, 1 mM 8-mercaptoethanol, and 10% glycerol) was electrophoresed as described in lane 2 and silver stained ( l a n e 4). Molecular size markers are in kilodaltons. B, distance migrated by each band in lane 2 of A was plotted against log10 (n), where n equals the band position from the bottom. the resulting 8-galactosidase activity was measured. Analysis shows that the amino and carboxyl termini of the RB protein can interact as evidenced by a 34-fold increase in transcription (Fig. 8B). To determine the effect of mutations in the aminoterminal region on this interaction a number of fusion proteins containing deletions or insertions in this region were also constructed and tested in this assay. All mutations, except for an insertion at amino acid 181, abolished binding, indicating the importance of the amino-terminal region in this interaction. The region examined contains the two structural domains mapped to the amino terminus of RB, and all deletions within this region have deleterious effects on interactions between the amino and carboxyl termini. This indicates the importance of the structural integrity of this region for the N-C interaction to occur. The carboxyl-terminal end of RB did not have the ability to self-associate as no interaction was detected upon co-transfection of two fusion proteins, which each contained the carboxyl-terminal end of RB (Fig.  8B).

DISCUSSION
To date there are a number of reports on the purification of recombinant RB from E. coli and baculovirus systems (Edwards et al., 1992;Huang et al., 1991;Wang et al., 1990a). While truncated forms of the RB protein have been produced, the successful production of the full-length retinoblastoma protein has proved difficult. We have previously produced pllORB from a baculovirus system, but yields were poor, and the protein tended to be denatured. The above purification scheme allows the isolation of significant amounts of stable and functionally active pllORB suitable for biochemical studies.
Analysis of the purified proteins by partial proteolytic digestion revealed a structural organization in the protein previously unidentified. ~5 6~~ is initially digested from the carboxyl terminus, which is likely an extended region of polypeptide highly susceptible to proteolytic cleavage. The resulting transient 45-48-kDa fragment is then cleaved in two, yielding the highly ordered structural domains A and B that correspond to the two regions previously defined as being required for binding to SV40 large T antigen. The domains mapped by deletion analysis fall within the two structural domains defined by partial proteolytic digestion. The exact carboxyl termini of these structural domains have not been determined, but they at least extend to the boundaries defined by deletion analysis, as judged by the molecular mass of the proteolytic fragments. Interestingly, these discrete domains do overlap with the positions of naturally occurring mutations in RB. The extreme susceptibility of RB to proteolytic cleavage in a region linking A and B strongly suggests that the continuity of the two domains is maintained by an accessible hinge region.
Using the Chou-Fasman method (Chou and Fasman, 1974) for predicting a-helices, @-sheets, and turns and the Karplus-Schultz method (Karplus and Schultz, 1985) for predicting the flexibility of the polypeptide chain the secondary structure of RB was analyzed. Domains A and B were found to have a much more defined secondary structure, while the region 1386 A.

B.
formed with various plasmids as indicated, and @-galactosidase activity was determined by the colony lift method and quantitated using an ONPG assay. "hinge" between the two domains was predicted to be a highly flexible turn region. Interestingly, this hinge region is also proline-rich compared with domains A and B. The carboxylterminal "tail" was the region with the least defined secondary structure and was predicted to be highly flexible with many turns. Nonetheless, this region is still important functionally; for example it contains the DNA binding domain (Wang et al., 1990a). It seems that p56= is digested by various proteases in the flexible regions of least defined secondary structure. This is consistant with other reports, which have indicated that cleavage sites in proteins can be associated with flexible hinge regions, protruding surface loops, or similarly exposed segments lacking a defined secondary structure (Fontana et al., 1986).
A structural analysis of p60m, an amino-terminal truncated RB that can be considered equivalent to p56RB, has been reported (Edwards et ai., 1992). The ultraviolet circular dichroism spectrum of purified p60m suggested that the protein contained 30% unordered conformations. Analysis of ~5 6~ by partial proteolytic digestion also suggests that approximately 30% of the protein contains unordered conformations that are presumably very susceptible to proteolytic attack with the remaining 70% forming two highly ordered structural domains within the protein.
pllOm also contained the two structural domains, A and B, which seem to be distinct structural entities within both the full-length protein and the amino-terminal truncated form. In addition, this analysis led to the identification of two structural domains in the amino-terminal portion of the retinoblastoma protein, a large 30-kDa domain and a small 10-kDa domain. Given that the carboxyl-terminal half of RB is biologically active in growth suppression, the question remains as to the function of the amino-terminal half of the protein. The identification of two structural domains in the poorly studied amino-terminal end of RB may be indicative but is by no means proof that this region of the protein may have some biological function.
In this regard, the oligomerization of purified pllORB, which requires the amino-terminal domains, provides a new basis for conceptualizing its role inside the nucleus. The gel migration pattern obtained is indicative of proteins with oligomerization properties such as SV40 large T antigen (Prives et al., 1991) and several DNA helicases (Patel and Wingorani, 1993). These results suggest that the oligomerization is an intrinsic property of the full-length RB protein, requiring the presence of the amino-terminal region. Results from the yeast twohybrid system suggest that there is a direct interaction between the amino and carboxyl termini of the protein, which would allow the formation of an oligomeric structure. Of course, the possibility of intramolecular interactions also exists with the amino and carboxyl terminus of the same molecule interacting. Interestingly, the other tumor suppressor gene, p53, can also oligomerize (Kraiss et at., 1988, O'Reilly andMiller, 1988). It has been reported that the unphosphorylated RB protein binds to several cellular proteins, including a number of transcription factors, and that such interactions may be disrupted when the RB protein is phospho~lated in late GI, S and M phases ( C h e~a p p~ et aL, 1991; Shirodkar et ai., 1992). Though speculative, perhaps the RB protein may serve to regulate the activities of these other proteins by forming a "corral" that sequesters these factors . Whether oligomerization of the RB protein is a cell cycle-regulated phenomenon, occurring in uiuo, is currently being investigated.
For the first time it has become possible to examine at the structural level the retinoblastoma gene product. These studies provided new insights into the structural organization of the retinoblastoma protein revealing the presence of four distinct structural domains in the protein linked by exposed hinge regions. Some functions have already been defined for the c a r b o x y l -t e~i n~ domains, and the identification of discrete structural domains in the amino terminus suggests this region may have specific biological function(s), which await further exploration. A novel property associated with the amino terminus, oligomerization, will provide an additional insight into the structure/function relationships of this tumor suppressor gene product and perhaps give some clues as to its mechanism of action in cell growth regulation.