Differential regulation of retinoblastoma protein function by specific Cdk phosphorylation sites.

The retinoblastoma tumor suppressor protein, RB, contains at least three distinct protein binding domains. The A/B pocket binds proteins with the LXCXE motif, the C pocket binds the nuclear c-Abl tyrosine kinase, and the large A/B pocket binds the transcription factor E2F. Dissociation of RB from its targets is observed as RB becomes phosphorylated during G1/S progression. There are 16 Cdk consensus phosphorylation sites in RB. It was previously unknown whether the many phosphorylation sites had redundant or distinct functions in the regulation of RB. Using RB mutant proteins lacking specific phosphorylation sites, we show that each of the binding domains is inhibited by different sites. Thr-821/826 phosphorylation is required to inhibit the binding to LXCXE containing proteins. Mutation of these two sites does not interfere with the hyperphosphorylation of RB. However, this phosphorylated mutant retains the ability to bind T-Ag, E7, and Elf-1, all of which contain the LXCXE motif. In contrast, Ser-807/811 phosphorylation is required to disrupt c-Abl binding. Mutation of Ser-807/811 and Thr-821/826 does not abolish the regulation of E2F binding. Taken together, these results show that the protein binding domains of RB are each regulated by distinct Cdk phosphorylation sites.

The retinoblastoma susceptibility gene, Rb-1, was cloned on the basis of its biallelic inactivation in human retinoblastomas (1,2). Rb-1 was subsequently observed to be mutated in a variety of tumor types, indicating that it may play a general role in the inhibition of the transformed phenotype (3,4). Reintroduction of Rb-1 into some RbϪ/Ϫ cells can lead to decreased tumorigenicity in nude mice or inhibition of growth in culture (5). Furthermore, ectopic expression of Rb-1 or microinjection of RB protein can block cell cycle progression at G 1 /S (6 -8).
RB forms complexes with many proteins, and this protein binding activity is required for growth suppression. RB binds to its target proteins by several different mechanisms (1). Viral oncoproteins such as the SV40 large T-antigen (T-Ag), 1 and several cellular proteins, e.g. D-type cyclins and Elf-1, contain the LXCXE motif that is important for binding to the A/B pocket of RB (9 -11). The E2F transcription factors do not contain the LXCXE motif and their binding requires the A/B pocket and C-terminal amino acids, and this E2F binding site is called the "large A/B pocket" (7,(12)(13)(14). The C-terminal region of RB also contains an A/B pocket-independent binding domain, the C pocket, which binds to the nuclear c-Abl tyrosine kinase (15). The large A/B pocket and the C pocket of RB do not overlap, because RB can simultaneously bind to E2F and c-Abl in vitro and in vivo (16). In addition, complexes containing T-Ag/RB/c-Abl and cyclin D2/RB/c-Abl have been detected (15,16). The protein assembly function of RB is known to be required for growth suppression, since overexpression of the individual domains can inactivate RB biological function (16).
The protein binding function of RB is regulated by phosphorylation (1). RB phosphorylation is observed as cells progress from G 1 into S phase of the cell cycle, and this is correlated with the disruption of RB-assembled protein complexes. RB contains 16 Ser/Thr-Pro motifs which are potential Cdk phosphorylation sites. At least seven of these sites (Ser-249, -807, -811, and Thr-252, -373, -821, and -826) have been shown to be phosphorylated in vivo (17)(18)(19). A number of other sites are also phosphorylated in vivo, but the exact identity of these sites is unknown (17)(18)(19). In mitotic cells, cdc2/cyclin B is the principle RB kinase (19). In interphase cells, RB is phosphorylated by other Cdk-cyclin complexes, including Cdk4/cyclin D and Cdk2/ cyclin A (2). Several Cdk consensus phosphorylation sites have been mutated in human and murine RB (9, 20 -22). Elimination of eight Cdk consensus phosphorylation sites in murine RB abolishes phosphorylation in vivo, and correlates with a more effective suppression of E2F or Elf-1 activity in transfected cells (11,22). The previous analysis of phosphorylation site mutants of RB did not address the question of whether the multiple Cdk phosphorylation sites of RB had redundant regulatory function.
Since RB can bind its target proteins through at least three different mechanisms, we hypothesized that each protein binding function might be regulated by distinct phosphorylation sites (1). This hypothesis would suggest that the multiple phosphorylation sites can establish a collection of functional states of RB, depending on the specific sites phosphorylated. To address this hypothesis, we developed in vitro binding assays to examine the requirement of specific phosphorylation sites on RB protein binding activity. We show here that different phosphorylation sites are required to inhibit the binding to T-Ag, c-Abl, and E2F. Furthermore, specific sites are also required for the efficient phosphorylation of RB in vivo and the reversal of RB-mediated growth suppression by cyclin A.

MATERIALS AND METHODS
Cell Culture-C33-A, HeLa, SAOS-2, and COS cells were obtained from the American Type Culture Collection. C33-A and SAOS-2 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10 and 15% fetal calf serum, respectively, at 37°C. COS and HeLa cells were cultured in Dulbecco's modified Eagle's medium supple-mented with 10% calf serum at 37°C. Sf9 cells were cultured at 25°C in Grace's medium containing 10% heat inactivated fetal calf serum, yeastolate, lactalbumin hydrolysate, and gentamicin.
Transfections-The BES-calcium phosphate method was used to transfect all cells (27). Cells were washed three times with phosphatebuffered saline 14 -20 h after the addition of DNA. In transient assays, cells were harvested approximately 48 h after washing. All transfections were carried out using 16 g of total DNA/100-mm dish. For the expression of FL-RB in C33-A cells, 10 g of RC-CMV-CycA and 6 g of the CMV-RB constructs were transfected. For co-immunoprecipitation of RB with T-Ag we used 8 g of CMV-CycA, 4 g of RSV-T-Ag, and 4 g of CMV-RB. For COS transfections 16 g of CMV-RB alone were transfected. For SAOS-2 transfections, 9 g of effector (either RC-CMV-CycA or RSV-T-Ag), 6 g of CMV-RB, and 1 g of pBABE-Puro was used. For the flat cell formation assays, transfected SAOS-2 cells were split into Dulbecco's modified Eagle's medium supplemented with 15% fetal calf serum and puromycin at 0.5 mg/ml approximately 24 h after washing. Selection was carried out for 7-9 days, plates were stained with crystal violet, and the number of flat cells on the plate was determined by counting random fields as described previously (8).
Immunoprecipitation and Immunoblotting-In vivo phosphate labeling of RB was carried out essentially as described previously (20). For the immunoprecipitation of phosphate-labeled RB, transfected cells were harvested by scraping and lysed in RIPA buffer (25 mM Tris, pH 7.5, 50 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, and 0.2% Nonidet P-40) supplemented with protease inhibitors (10 g/ml 1,10phenanthroline, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride), and phosphatase inhibitors (10 mM sodium fluoride, 10 mM sodium phosphate, 10 mM sodium pyrophosphate). Lysates were clarified by centrifugation at 15,000 ϫ g for 10 min at 4°C followed by filtration through a 0.45-m syringe filter (Nalgene). Clarified lysates were then pre-cleared by incubation with 9 g of rabbit anti-mouse antibody (Cappel) and 15 l of protein A-Sepharose (Pharmacia) at 4°C for 30 min, followed by centrifugation at 2000 ϫ g, then subjected to specific immunoprecipitation. For immunoprecipitation from unlabeled cells, transfected cells were harvested by scraping and lysed in NNT-N buffer (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA). Lysates were clarified by centrifugation at 15,000 ϫ g for 10 min at 4°C. Proteins were immunoprecipitated from the clarified lysates using the following antibodies: for RB, 1.5 g of G3-245 monoclonal antibody (Pharmingen) and for T-Ag, 1.5 g of pAB 114 monoclonal antibody (Pharmingen). Immunoprecipitations were carried out for 4 h at 4°C using 4.5 g of rabbit anti-mouse and protein A-Sepharose in a volume of 0.5-1.0 ml. Immunocomplexes were recovered by centrifugation at 2000 ϫ g, and washed 4 -8 times with 1.0 ml of NNT-N or RIPA buffer. Immunoprecipitated proteins were separated by SDS-PAGE and transferred to Immobilon-P membrane (Millipore).
Immunoblotting was carried out with the following antibodies: fulllength RB proteins were detected with the G3-245 monoclonal antibody (Pharmingen), GST-RB proteins were detected with either 851 serum (15) or the XZ91 monoclonal antibody (Pharmingen), and T-Ag was detected with the pAB 114 monoclonal antibody (Pharmingen). Immunoblots were developed using either alkaline phosphatase-conjugated secondary antibody for colorimetric development (Fisher), or horseradish peroxidase-conjugated secondary antibodies (Cappel, Life Technologies, Inc.) for enhanced chemiluminescent (ECL) development (Amersham).
In Vitro Kinase Reactions-Active Cdk/cyclin kinases used to phosphorylate RB were obtained from two sources. Cdk2/cycA was purified from recombinant baculovirus-infected Sf9 cells (J. W. Harper). Alter-natively, active cdc2 was immunopurified from nocodazole-arrested mitotic HeLa cells as described before (19). Kinase reactions with either active cdc2 (immune complexes from 5 to 10 ϫ 10 5 mitotic cells) or purified Cdk2/cycA (final concentration 1-5 nM) were carried out in 50 l of kinase buffer using 150 M ATP, 80 -200 Ci of [␥-32 P]ATP, and 10 -30 ng of RB as substrate. Reactions were allowed to proceed for 30 -45 min with shaking at room temperature. The resulting phosphorylated protein was then either resolved by SDS-PAGE, transferred to Immobilon-P, and detected by autoradiography and immunoblotting, used in the determination of stoichiometry of phosphorylation (see below), or utilized in in vitro binding assays (see below).
Quantitation-For the determination of the stoichiometry of phosphorylation with in vitro phosphorylated proteins, the phosphorylated 32 P-labeled proteins were first resolved by SDS-PAGE. The incorporated phosphate was then determined by liquid scintillation counting of the labeled bands, from which the phosphate molecules per protein molecule were determined. For the determination of the relative phosphorylation in vivo, immunoprecipitated metabolically labeled proteins were resolved by SDS-PAGE and detected by autoradiography, followed by immunoblotting. The autoradiograms ( 32 P) and immunoblots (protein) were quantitated with an LKB densitometer, and the 32 P/protein ratio was normalized to WT, which was arbitrarily set to 100.
Binding Assays-GST fusion proteins were expressed in bacteria and purified on glutathione-agarose as described previously (15). The binding of RB to c-Abl was also carried out as described previously (15). Approximately 500 ng of T-Ag in 0.5 ml of NNT-N were added to the immobilized GST-AE and incubated at 4°C for 45 min. The GST-AE beads were then washed five times with NNT-N. The bound protein was solubilized by boiling in SDS buffer and resolved by 8% SDS-PAGE, and T-Ag was visualized by immunoblot. The E2F supershift assay was carried out using 30 ng of soluble of GST-AE and E2F purified from HeLa cells as described previously (16).

Construction and Characterization of RB Phosphorylation
Site Mutants-We focused our analysis on four Cdk consensus phosphorylation sites in RB, Ser-807, Ser-811, Thr-821, and Thr-826. These four sites were chosen because they are phosphorylated to a higher level than other sites in HeLa and Molt4 cells, suggesting that they are phosphorylation sites of physiological relevance (19). These four sites can also be phosphorylated in vitro by cdc2/cyclin B or Cdk2/cyclin A (19). We constructed double mutants of Ser-807/811 or Thr-821/826, because we were intrigued by the close proximity of the two Ser and the two Thr consensus sites. Both Ser-807 and -811 were mutated in Phosphorylation Site Mutant 2S (PSM.2S, Fig. 1). In PSM.2T, Thr-821 and -826 were mutated (Fig. 1). The PSM.4 combines the mutations in PSM.2S and PSM.2T. The WT and PSM proteins were expressed in bacteria as GST fusions in two forms: the GST-SE, which only contained the C pocket; or the GST-AE, which contained the A/B and C pockets of RB (Fig. 1). They were also expressed in mammalian cells as full-length RB proteins.
Stoichiometric Phosphorylation of RB in Vitro-The in vitro phosphorylation of WT RB with Cdk/cyclin could inhibit its protein binding activity as shown below. However, the mere incorporation of phosphate into RB did not necessarily cause an inhibition of binding, because some of the 32 P-labeled WT RB was found to bind to its target proteins (data not shown). This result suggested that some of the phosphorylation sites might be irrelevant for the inhibition of protein binding. Therefore, we developed conditions to achieve the quantitative phosphorylation of RB in vitro at all Cdk consensus sites, using mitotic cdc2 kinase or purified Cdk2/cycA. Quantitative phosphorylation was indicated by the complete conversion of RB to a single band with slower electrophoretic mobility on SDS-PAGE (Fig.  2, compare lanes 1 and 2). The stoichiometry of phosphorylation was measured for WT and PSM RB, and found to be in agreement with the number of Cdk consensus phosphorylation sites present in each protein (Table IA). GST is not phosphorylated by Cdk/cyclin (18). Along with the lower stoichiometry, the phosphorylated PSM proteins also exhibited a reduced shift in their electrophoretic mobility. The phosphorylated WT RB migrated to a higher position than its corresponding phosphorylation site mutants (Fig. 2, compare lanes 2-5 with lanes 6 -8). Thus, the stoichiometry of phosphorylation appears to dictate the degree of mobility shift.
In Vitro Binding Assays-The quantitatively phosphorylated WT and PSM RB were used in two in vitro assays to examine the regulation of protein binding activity. In the first assay (Fig. 3), GST-RB proteins were labeled with 32 P. Each labeled sample was combined with a known amount of its unphosphorylated counterpart to give a constant 32 P/protein ratio, and then applied to immobilized c-Abl or T-Ag. The 32 P/protein ratio of the input and the bound fraction was compared for each sample. If phosphorylation inhibited binding, the 32 P/protein ratio of the bound fraction would be lower than that of the input. However, if phosphorylation did not affect binding, the 32 P/protein ratio of the bound fraction would be equal to that of the input. In the second assay (Fig. 4), GST-RB proteins were quantitatively phosphorylated. The RB proteins were then used either in an immobilized form to bind soluble c-Abl or T-Ag (Fig. 4, B and C), or in a soluble form to interact with E2F in gel shift assays (Fig. 4D).  1-5). The SE proteins were resolved on 15% SDS-PAGE (lanes 6 -9). Proteins were transferred to Immobilon-P, and detected by autoradiography ( 32 P, upper panels) followed by anti-RB immunoblotting (␣RB, lower panels) a The Cdk consensus phosphorylation site is defined as any S/T-P motif, within the given RB sequence.
b Each given GST fusion was phosphorylated under conditions of quantitative conversion to a single hyperphosphorylated band (Fig. 2A). The stoichiometry for each hyperphosphorylated species was determined as described under "Materials and Methods." The stoichiometry for GST-AE proteins was determined in two independent experiments. ND, not determined. c The given RB proteins were coexpressed with cyclin A in C33-A cells and labeled with phosphate. The 32 P to protein ratio for each protein was determined as described under "Materials and Methods," and normalized to the WT protein which was set to be 100. The relative phosphorylation of the proteins was determined in two independent experiments.
Ser-807/811 Are Required to Inhibit c-Abl Binding-Quantitative phosphorylation of the WT SE fragment of RB could be shown to inhibit c-Abl binding in vitro (Fig. 3A, lane 1). This inhibition was indicated by a 15-20-fold reduction in the 32 P/ protein ratio of the bound fraction (Fig. 3A, compare lane 1 of the two panels). Unlike WT SE, the phosphorylated PSM.2S could bind to c-Abl, and the 32 P/protein ratios of the input and bound fractions were similar (compare lane 2 of the two panels). PSM.4 behaved in a fashion identical to PSM.2S (Fig. 3A,  lane 3). The ME fragment of RB did not bind to c-Abl, demonstrating specificity in the binding reaction (lane 4). This result suggested that phosphorylation at Ser-807/811 was necessary for the inhibition of RB/c-Abl interaction.
The same result was obtained with the second in vitro assay (Fig. 4B). Equal amounts of unphosphorylated (Fig. 4A, lanes  1-4) and phosphorylated WT and PSM GST-AE (Fig. 4A, lanes  5-8) were used in the binding reactions. The unphosphorylated WT or PSM proteins bound the same amount of c-Abl, which was translated in vitro and labeled with [ 35 S]methionine (Fig.  4B, lanes 1-4). Phosphorylation of WT RB reduced c-Abl bind-ing to background level (compare lanes 1 and 5). Phosphorylation of PSM.2S and PSM.4, again, had no effect on RB/c-Abl complex formation (Fig. 4B, compare lanes 2 and 4 with 6 and  8). Phosphorylation of PSM.2T did cause an inhibition of c-Abl binding (Fig. 4B, lanes 3 and 7). Taken together, these results show that phosphorylation of Ser-807/811 is required to inhibit the C pocket function of RB, while the phosphorylation of Thr-821/826 is not required. In the absence of Ser-807/811, phosphorylation of the remaining eight Cdk consensus sites in the AE fragment of RB cannot disrupt the RB/c-Abl interaction.
Thr-821/826 Are Required to Inhibit T-Ag Binding-The inhibition of T-Ag binding by phosphorylation could also be demonstrated in vitro, as shown by a 10 -15-fold reduction in the 32 P/protein ratio of the WT GST-AE fraction bound to immobilized T-Ag (Fig. 3B, lane 1). The RB mutant C706F failed to bind to T-Ag under the experimental conditions (lane 4), showing specificity for the binding reactions. The phosphorylated PSM.2S showed a reduced affinity for T-Ag, comparable to that of the WT (Fig. 3B, compare lanes 1 and 2). In contrast, phosphorylated PSM.4 was capable of binding to T-Ag, as indicated by the similar 32 P/protein ratio in the input and the bound fractions (Fig. 3B, lane 3). The same result was again obtained with the second in vitro assay. The unphosphorylated WT or PSM proteins bound equal amounts of T-Ag (Fig. 4C,  lanes 1-4). Phosphorylation of WT and PSM.2S led to an inhibition of T-Ag binding (Fig. 4C, lanes 5 and 6). However, the phosphorylated PSM.2T protein retained full activity to complex with T-Ag (Fig. 4C, compare lanes 3 and 7). As expected, PSM.4 behaved like PSM.2T (Fig. 4C, compare lane 4 with 8). These results showed that phosphorylation at Thr-821/826, but not at Ser-807/811, is necessary for the inhibition of T-Ag binding.
Ser-807/811 and Thr-821/826 Are Not Required to Inhibit E2F Binding-To assay the E2F binding activity, the unphosphorylated or phosphorylated GST-AE fragments were added to purified E2F and an E2F-oligonucleotide (Fig. 4D). In the unphosphorylated form, the WT and the PSM fragments tested all bound to E2F, as indicated by the "supershift" of the E2F⅐DNA complex (lanes 1-4). All four proteins failed to complex with E2F⅐DNA when phosphorylated (lanes 5-8). As a negative control, the GST-SE fragment of RB did not supershift the E2F⅐DNA complex (lane 10). These results showed that Cdk sites within the AE fragment of RB were sufficient to inhibit E2F binding. However, phosphorylation of Ser-807/811 and Thr-821/826 are not required to inhibit E2F binding.
In Vivo Phosphorylation of PSM RB-With the in vitro quantitatively phosphorylated RB, we showed that different Cdk sites were required to regulate binding to c-Abl, T-Ag, and E2F. Because not all of the Cdk sites were quantitatively phosphorylated in vivo, we wished to determine the protein binding activity of in vivo phosphorylated WT and PSM RB. The WT and PSM proteins were transiently expressed in the human cervical carcinoma cell line, C33-A, which contains a truncated RB that is unstable. Phosphorylation of the exogeneously expressed RB was enhanced by the cotransfection with cyclin A. The relative extent of phosphorylation was determined by the appearance of slower migrating bands on anti-RB immunoblots and by 32 P labeling (Fig. 5). The WT and PSM.2T were phosphorylated in vivo, as indicated by the slower migrating bands which were 32 P-labeled (Fig. 5, lanes 1 and 3, upper and lower  panels). The PSM.2S and PSM.4 proteins were labeled with 32 P in vivo (Fig. 5, lanes 2 and 4, upper panel), but did not generate the characteristic mobility shift (Fig. 5, lanes 2 and 4, lower  panel). The relative level of in vivo phosphorylation was determined for each protein (Table IB). PSM.2T was phosphorylated to 80 -90% that of WT, consistent with the mutation of two out 1-4) obtained by thrombin cleavage were quantitively phosphorylated with mitotic cdc2 kinase. Each phosphorylated protein was combined with a 10-fold excess of its unphosphorylated counterpart, and this was utilized as the Input. Each sample was assayed for binding to GST-A1 (the ATP-binding lobe of c-Abl fused with GST (15)). The Input (upper panels) represent 10% of the sample applied to the binding reaction. Protein fractions were resolved by 15% SDS-PAGE and transferred to Immobilon-P. Phosphorylated SE fragments were visualized by autoradiography ( 32 P, left panels), and the unphosphorylated SE or ME were revealed by anti-RB immunoblot (␣RB, right panels). B, phosphorylation inhibits binding to T-Ag. Soluble GST-AE proteins were quantitatively phosphorylated with mitotic cdc2 kinase. Each phosphorylated protein was combined with a 25-fold excess of its unphosphorylated counterpart, and this was assayed for binding to SV40 large T-antigen immobilized by immunoprecipitation with protein A-Sepharose. The input and the bound fraction of WT (lane  (Table IB). The relative phosphorylation for PSM.2S or PSM.4 was only about 40% that of WT (Table IB). The inefficient phosphorylation of the Ser-807/811 mutant RB was consistent with a previous report which showed that mutation of the murine equivalent of Ser-807/811 prevented the hyperphosphorylation of murine RB in vivo (21).

FIG. 3. In vitro phosphorylation inhibits the protein binding function of RB. A, phosphorylation inhibits binding to c-Abl. RB-SE or ME fragments (lanes
Hyperphosphorylated PSM.2T Binds T-Ag-Because WT and PSM.2T were both hyperphosphorylated in vivo (Fig. 6A,  lanes 1 and 2), we compared their binding to c-Abl, E2F, and T-Ag (Fig. 6, B-D). Since PSM.2S was not hyperphosphorylated in vivo, it was not used in the subsequent assays. The hyperphosphorylated WT and PSM.2T could not bind to c-Abl (Fig.  6B), nor did they bind to E2F-1 (Fig. 6C). In contrast, the hyperphosphorylated bands of PSM.2T could bind to T-Ag (Fig.  6D). These results were consistent with those obtained with in vitro phosphorylated RB, showing that phosphorylation at Thr-821/826 is only required for the inhibition of T-Ag binding.
Since T-Ag interacts with RB through an LXCXE motif, we tested whether phosphorylation of Thr-821/826 was required to regulate the binding of other LXCXE containing proteins, e.g. E7 and Elf-1. As with T-Ag, the phosphorylated WT RB did not bind to E7 or Elf-1 (Fig. 6, E and F, lane 1), while the phosphorylated upper bands of PSM.2T did bind to these two proteins (Fig. 6, E and F, lane 2). Thus, phosphorylation cannot inhibit the LXCXE binding function of RB when Thr-821/826 are mutated.
To determine whether the GST pull-down assays reflected the regulation of T-Ag binding inside the cell, we performed coimmunoprecipitation. In C33-A cells cotransfected with T-Ag and RB, only the underphosphorylated form of WT was coprecipitated with T-Ag (Fig. 7A, lane 1). In contrast, the hyper-  1 and 5), PSM.2S (lanes 2 and 6), PSM.2T (lanes 3 and 7), and PSM.4 (lanes 4 and 8), were immobilized on GSH-agarose and incubated with Cdk2/cycA in the absence (lanes 1-4) or presence (lanes 5-8) of 150 M ATP. A portion of each sample was resolved by 8% SDS-PAGE, transferred to Immobilon-P, then immunoblotted with anti-RB antibody. B, phosphorylation of Ser-807/811 is required to inhibit c-Abl binding. Equal amounts of GST-AE proteins, either unphosphorylated (lanes 1-4) or phosphorylated (lanes 5-9) as shown in A, were incubated with in vitro translated c-Abl, labeled with [ 35 S]methionine. The c-Abl bound to immobilized GST-AE, WT (lanes 1 and 5), PSM.2S (lanes 2 and 6), PSM.2T (lanes 3 and 7), or PSM.4 ( lanes  4 and 8), was recovered and resolved by 6.5% SDS-PAGE. The amount of c-Abl bound was determined by autofluorography. C, phosphorylation of Thr-821/826 is required to inhibit T-Ag. binding. The immobilized GST-AE proteins, either unphosphorylated (lanes 1-4) or phosphorylated (lanes 5-8) as shown in A, were incubated with lysate of Sf9 cells infected with a recombinant baculovirus expressing T-Ag. The T-Ag bound to GST-AE, WT (lanes 1 and 5), PSM.2S (lanes 2 and 6), PSM.2T (lanes 3 and 7), or PSM.4 (lanes 4 and 8), was resolved by 8% SDS-PAGE and transferred to Immobilon-P. The amount of bound T-Ag was determined by immunoblotting with antibodies for T-Ag. D, phosphorylation at Ser-807/811 and Thr-821/826 is dispensible for inhibiting binding to E2F. The unphosphorylated GST-AE proteins (lanes 1-4 and  9), the phosphorylated GST-AE proteins (lanes 5-8), or the GST-SE protein (lane 10) were purified and added to DNA binding reactions, as described under "Materials and Methods." To prevent the phosphorylation of E2F by Cdk2/cyclin A, excess EDTA was added to the GST-AE preparations after phosphorylation but prior to incubation with E2F.  1 and 8).  4). The transfected cells were metabolically labeled with [ 32 P]phosphoric acid, lysed, and RB was immunoprecipitated with anti-RB antibodies. Immunocomplexes were recovered, resolved by 7. 2% (lanes 1-4) SDS-PAGE, and transferred to Immobilon-P. RB was detected by autoradiography ( 32 P, upper panel), followed by immunoblotting with anti-RB (␣RB, lower panel) antibodies. phosphorylated PSM.2T bands were brought down by anti-T-Ag immunoprecipitation (Fig. 7A, lane 2). Similar results were also obtained when WT and PSM.2T proteins were expressed in COS cells and phosphorylated by the endogenous Cdk/cyclins (Fig. 7B, lanes 1 and 2). Thus, PSM.2T binds to T-Ag irrespective of its phosphorylation status.
Growth Suppression by PSM RB-Phosphorylation of RB has been correlated with the inactivation of its growth suppression function in SAOS-2 cells (8). Exogenous RB does not become phosphorylated in SAOS-2 cells and causes a cell cycle block at G 1 /S which generates growth-arrested flat cells. Cotransfection of cyclin A can drive RB phosphorylation in SAOS-2 cells and alleviate the cell cycle block (8). In keeping with previous reports, WT RB gave rise to numerous flat cells, and the number of flat cells was reduced 8-fold with the cotransfection of cyclin A (Table II). PSM.2T induced a similar number of flat cells and the number was reduced 6-fold by the cotransfection of cyclin A. With PSM.2S, however, cyclin A only caused a 1.6-fold reduction in the number of flat cells (Table II). To determine if PSM.2S could be inactivated by other means, we cotransfected it with T-Ag (Table II). The wild type RB and PSM.2S were both sensitive to T-Ag, which caused a 33-fold reduction in the number of flat cells. Cotransfection with cyclin A drove the phosphorylation of WT and PSM RB, and the relative level of phosphorylation was similar to that observed with C33-A cells (Fig. 5 and Table IB). PSM.2S was poorly phosphorylated in SAOS-2 cells, and this might account for its resistance to the inactivation by cyclin A. DISCUSSION Under conditions where every Cdk consensus site is phosphorylated, we have shown that in vitro phosphorylation does inhibit the protein binding function of RB. Using mutants lacking specific Cdk phosphorylation sites, we also demonstrated that different Cdk sites are required for inhibiting distinct RB protein binding activities. Specifically, Ser-807/811 is required for the inhibition of c-Abl binding, while phosphorylation of Thr-821/826 is required for the inhibition of binding to T-Ag, Elf-1, and E7 (all of which contain the LXCXE motif). However, none of these sites are required for the inhibition of E2F binding. We find that the mutation of Ser-807/811 prevents the efficient phosphorylation of RB in vivo, and cyclin A cannot overcome the growth suppressing activity of this mutant in SAOS-2 cells. In contrast, mutation of Thr-821/826 does not prevent RB hyperphosphorylation nor the inactivation of its growth suppressing activity by cyclin A.
Regulation of the A/B Pocket by Phosphorylation-Both T-Ag and E2F interact with RB through the A/B pocket. However, the LXCXE motif that mediates the T-Ag binding is not found in E2F (12,13). Furthermore, the CRII fragment of E1A, which contains the LXCXE motif, cannot disrupt the RB/E2F complex (28). Thus, the A/B pocket of RB may contain two binding sites, one for the LXCXE motif and another for E2F. This idea is consistent with our finding that the regulation of T-Ag and E2F binding to RB requires different Cdk phosphorylation sites. Phosphorylation of Thr-821/826 is required for the disruption of the T-Ag⅐RB complex. In addition to T-Ag, we have found that the inhibition of binding to Elf-1 and E7, both of which contain the LXCXE motif, also required Thr-821/826. This finding strongly supports that the LXCXE binding site is regulated by Thr-821/826 phosphorylation. We cannot determine if both threonine residues or only one of them is required for the regulation of LXCXE binding. Single site mutants would have to be constructed to resolve this issue. Since mutation of Thr-821/826 does not interfere with the regulation of E2F bind-  1 and 2) proteins coexpressed with cyclin A were used in binding reactions with GST-E7. The fraction bound to E7 was recovered and resolved by 7.2% SDS-PAGE. RB bands were revealed by immunoblotting with anti-RB antibodies. F, in vivo hyperphosphorylated PSM.2T binds Elf-1. The WT and PSM.2T (lanes 1 and 2) proteins produced in C33-A cells were assayed for binding to GST-Elf-1. RB proteins bound to Elf-1 were recovered and resolved by 7.2% SDS-PAGE. RB was detected by anti-RB immunoblot. ing, other phosphorylation sites must be sufficient to disrupt the A/B pocket function that mediates E2F interaction. It is interesting to find that the phosphorylation sites required for the inhibition of LXCXE binding are outside of the minimal A/B pocket domain. We could envision at least two possible mechanisms for these phosphothreonines to inhibit the A/B pocket activity. The phosphorylated threonine may directly bind to the A/B pocket and sterically hinder the binding of LXCXE containing proteins. Alternatively, phosphorylation at Thr-821/ 826 may result in a conformational change that can lead to the disruption of the LXCXE binding structure in the A/B pocket.
Inactivation of RB Growth Suppression by Phosphorylation-The growth suppression function of RB depends on its ability to assemble protein complexes (16). This concept is supported by three lines of evidence: (a) both the large A/B and the C pockets of RB are required for growth suppression, (b) the A/B and C domains of RB do not function in-trans, and (c) the A/B or C domain fragments can act as dominant interfering mutants to disrupt RB function. According to this model, any phosphorylation event that inhibits one of the three protein binding functions should be sufficient to inactivate the growth suppression function of RB. This could explain the observation that PSM.2T-RB, despite the lack of two major phosphorylation sites, can still be inactivated by cyclin A. The phosphorylated PSM.2T, as shown here, retains the ability to bind the LXCXE motif, but cannot suppress SAOS-2 growth. Presumably this is because the c-Abl and the E2F binding activities are still inactivated by phosphorylation in this mutant, and therefore the phosphorylated PSM.2T-RB cannot assemble protein complexes. Alternatively, the LXCXE binding function may not be relevant for the suppression of SAOS-2 cells. The present data does not rule out the possibility that PSM.2T-RB may act as a dominant growth suppressor in other cell types.
Role of Partially Phosphorylated RB-If partial phosphorylation of RB can disrupt its growth suppression function, why then design multiple phosphorylation sites in RB? One possibility is that the Cdk-mediated phosphorylation is directed at regulating what types of complexes RB can assemble. The protein complexes formed on partially phosphorylated RB may be important in fine-tuning regulatory events during cell cycle progression. There is evidence that RB phosphorylation occurs in a stepwise manner during G 1 progression. Two-dimensional phosphotryptic mapping of RB in synchronized cells has revealed sequential loading of specific phosphorylation sites as cells progress from quiescence to S-phase (29). A conceivable consequence of stepwise phosphorylation may be as follows: if phosphorylation occurs at sites that inhibit E2F binding but not at Thr-821/826, this event would release E2F and make more RB available for binding to the cellular LXCXE containing proteins. This might be a way for Cdk to orient the A/B pocket of RB toward specific cellular targets. Partial phosphorylation of RB might also lead to other modulations of RB binding proteins. For example, phosphorylation of Ser-807/811, but not at other sites, would activate the c-Abl tyrosine kinase which could then phosphorylate proteins still bound to the A/B pocket. This is a possible mechanism for Cdk to activate the tyrosine phosphorylation of a specific nuclear protein. The selective, stepwise phosphorylation of RB may provide a continuous modulation of RB-assembled complexes to alter their activities throughout the cell cycle.