Cell, Vol. 64, 521-532, February8, 1991,Copyright© 1991 by Cell Press
Identification of Cellular Proteins That Can Interact Specifically with the T/EIA-Binding Region of the Retinoblastoma Gene Product William G. Kaelin Jr.,* David C. Pallas,* James A. DeCaprio,* Frederic J. Kaye, t and David M. Livingston* * The Dana-Farber Cancer Institute and Harvard Medical School Boston, Massachusetts 02115 t National Cancer Institute-Navy Medical Oncology Branch and Uniformed Services University of the Health Sciences Bethesda, Maryland 20814
Summary The SV40 T antigen (T)/adenovirus EIA-binding domain of the retinoblaetoma gene product (pRB) has been fused to S. japonlcum glutathione S-transferase, and the chimera, bound to insoluble glutathione, was used to search for cellular proteins that can interact specifically with pRB. At least seven such proteins were detected in extracts of multiple human tumor cell lines. These proteins failed to bind to a family of pRB fusion proteins that harbor inactivating mutations in the T/EIA-binding domain and to the wild-type fusion protein in the presence of a peptide replica of the pRBbinding domain of T. Therefore, the binding of one or more of these proteins may contribute to the growthsuppressing function of pRB. Introduction Functional inactivation of both copies of the retinoblastoma susceptibility gene (RB) is an invariant feature of both sporadic and familial retinoblastomas (Cavenee et al., 1983; Horowitz et al., 1990; Murphree and Benedict, 1984) and may play a role in the evolution of some osteosarcomas, soft tissue sarcomas, small cell lung carcinomas, breast carcinomas, and genitourinary carcinomas (Bookstein et al., 1990; Friend et al., 1987; Harbour et al., 1988; Horowitz et al., 1989, 1990; Lee et al., 1988; Reissmann et al., 1989; TAng et al., 1988; Toguchida et al., 1989; Varley et al., 1989; Weichselbaum et al., 1988; Yokota et al., 1988). Indeed, RB expression has been detected in all human tissues examined to date (Bernards et al., 1989; Lee et al., 1987a), and there is indirect evidence to suggest that the RB gene product, a 110 kd nuclear phosphoprotein (Lee et al., 1987b; Ludlow et al., 1989), may function as a human cell cycle control element (Buchkovich et al., 1989; Chen et al., 1989; DeCaprio et al., 1989; Ludlow et al., 1990; Mihara et al., 1989; Xu et al., 1989). In keeping with its proposed function as a tumor suppressor gene, Lee and coworkers have demonstrated that reintroduction of a wild-type RB allele into RB-/cells, in some instances, led to a decrease in cell growth rate in culture and, where studied, to a reduction in the ability of the recipient cells to form tumors in nude mice (Bookstein et al., 1990; Huang et al., 1988).
How RB carries out its growth suppressor function(s) is currently unknown. The RB gene product (pRB) can bind to DNA-cellulose, although at the present time there is no evidence that pRB can recognize a specific DNA sequence (Lee et al., 1987b). Furthermore, at least two spontaneously occurring, loss-of-function pRB mutants can bind to DNA-cellulose with affinity comparable to the wildtype protein, suggesting that DNA binding is not sufficient for RB growth suppression function (Horowitz et al., 1989; Shew et al., 1990a). Additional clues to how pRB functions have come from the study of DNA tumor viruses. Transforming proteins of several unrelated DNA tumor viruses including the adenovirus E1A protein (E1A), the SV40 large T antigen (1-), other papovaviral T antigens, and the human papillomavirus E7 protein (E7) have been shown to bind to pRB in vitro (DeCaprio et al., 1988; Dyson et al., 1989b, 1990; Whyte et al., 1988). Mutational analysis of these viral proteins suggests that their ability to transform cells is linked, in part, to their ability to bind to pRB and that this interaction is mediated primarily by short, colinear amino acid sequences homologous to the adenovirus E1A transforming region designated CR2 (DeCaprio et al., 1988; Green, 1989; Jones et al., 1990; Moran, 1988; Phelps et al., 1988; Whyte et al., 1989). In addition, it has been demonstrated that T binds exclusively to the unphosphorylated form of pRB (Ludlow et al., 1989, 1990). This suggests that unphosphorylated pRB performs those elements of pRB growth suppressor function that T can perturb. The minimal region of pRB needed for the binding to two of these viral proteins has been mapped in several laboratories (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990). It constitutes approximately 40% of the protein sequence, extending from residue 379 to 792. As an intact subsegment of pRB, this sequence binds not only to E1A and T, but also interacts stably and specifically with a 14 residue peptide containing all of the sequence of the pRBbinding domain of T (Kaelin et al., 1990). To date, all spontaneously occuring RB gene mutations, which do not grossly compromise pRB stability, map to this region, and, as predicted, the altered RB gene products are inactive with respect to E1A and T binding (Bookstein et al., 1990; Horowitz et al., 1989, 1990; Kaye et al., 1990; Shew et al., 1990a, 1990b). Furthermore, this region can be predicted to contain a leucine zipper (Bernards et al., 1989; Hong et al., 1989; McGee et al., 1989), a structural motif that has been implicated in the ability of some proteins to form hetero- or homodimers (Kouzarides and Ziff, 1989; Mitchell and Tjian, 1989). These observations led us to hypothesize that normal pRB function requires it to interact with a cellular protein containing a region structurally similar to the pRB-binding motif found in these diverse viral proteins. In this model, one or more aspects of pRB function can be inactivated by alterations in the T/E1A-binding domain (either as a result of mutation or, possibly, protein modification, e.g., phosphorylation) or by binding to a pseudosubstrate or ef-
Cell 522
fector such as T or E1A. In short, the model predicts that cells produce one or more proteins that interact with the T/E1A-binding domain of pRB in a manner resembling that employed by T and E1A. To date, coimmunoprecipitation has failed to demonstrate the existence of a cellular pRB-binding protein. The use of this technique to identify protein-protein interactions has a number of limitations, at least some of which might account for the failure to detect such a protein(s) in the past, For example, if T and E1A interfere with the interaction between pRB and a specific cellular protein(s), one must presume that this putative cellular protein binds to pRB with a lower affinity than T and E1A and/or is present in amounts far less than the typical amount of T or E1A in SV40 or adenovirus-transformed cells. The binding of a given cellular protein to pRB might also be cell cycle dependent, as is the case for T (Ludlow et al., 1990). Thus, in an asynchronous population of cells, only a fraction of pRB molecules might exist in a protein-bound form. Furthermore, a large portion of the pRB sequence is devoted to T/E1A binding (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990), and, in principle, it could become masked by a bound protein. If so, an antibody that interacts with this region of the protein surface might, preferentially or exclusively, recognize the unbound form of the protein. This would be especially true if pRB ordinarily binds to a large, multimeric complex of proteins that effectively shields key epitopes from specific antibody-combining sites. We therefore devised a pRB affinity binding method that would allow one to circumvent the need for pRB antibodies and to make use of the genetic data cited earlier, which points to the existence of a discrete, independent pRB domain (hereafter referred to as "pocket") that can bind, specifically, to selected, short viral peptide sequences. The scheme incorporates the use of the prokaryotic expression vector pGEX-2T (Smith and Johnson, 1988) and the polymerase chain reaction (PCR) (Saiki et al., 1988) to generate chimeric glutathione S-transferase-pRB protein affinity chromatographic reagents capable of binding T and E1A. Such fusion proteins would be expected to have a relatively high affinity for reduced glutathione and can be recovered and purified in relatively nondenatured form as noncovalent glutathione-Sepharose conjugates (Smith and Johnson, 1988). Results of experiments in which these chimeras were used as affinity binding reagents have led to the identification of a family of cellular proteins that can interact in vitro with the T/E1A-binding domain of pRB with the same specificity as T, E1A, or the aforementioned peptide copy of the pRB-binding domain of T. Thus, cells produce proteins that have the potential of interacting in a physiologically relevant manner with pRB. Results Glutathione S-Transferaae Fusion Proteins Containing the pRB T/EIA-Binding Domain Can Interact Specifically with T or EIA In Vitro We and others have previously demonstrated that the pRB
Glutathione
S-Transferase
Fusion
Proteins
T/E1A-Sinding Region
pR8 1
pGT-RB(379-792;dl
573 645)
379
792
928
glutathlone-stransferase 573 645
pGT RB(403 816) 403
pGT RB(379 928:dl Exon 21)
8t6
91ulathlone s 703 737
pGT-RB(379-928:dl Exon 22)
glu,ath,ones 738 775
pGT-RB(379-928:706
C ~IPF)
I glutalhio~e s
cL~.F pGT-RB(379-928)
I g[utath~orles
pGT RB(379 792)
I g'°'~'~ ...... t~ .s,.~?Zq,~; × ,//.,×? tran~erase >~ I/'/;?//~,/?',/t× I
BB
In-Frame Deletion
Figure 1. GlutathioneS-Transferase-pRBFusion Proteins
region spanning residues 379-792 is sufficient for binding to T or E1A and must be relatively intact for this interaction to occur (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990). Furthermore, we have shown that this region appears to make direct contact with a short consensus sequence present in these otherwise unrelated viral proteins (Kaelin et al., 1990). To search for the existence of cellular proteins capable of interacting with pRB by a similar mechanism, we generated a series of glutathione S-transferase-pRB fusion proteins (see Figure 1) for use as affinity reagents when bound noncovalently to glutathioneSepharose. All contained an intact or mutant version of the T/E1A-binding domain of pRB. The wild-type T/EIA-binding domain of pRB has been shown previously to retain function as an independent unit (Kaelin et al., 1990). Two of these chimeric fusion proteins harbored in vitro generated mutations in this region. One, pGT-RB(379-792;dl 573-645), was constructed by sequential PCR as described previously (Higuchi et al., 1988). The pRB segment deleted from this construct, namely, residues 573-645, has been shown by others to serve a "spacer" function within the T/E1A-binding region, i.e., deletion of this region inactivated the binding function of the domain unless a "filler" sequence was reinserted (Hu et al., 1990; Huang et al., 1990). The second mutant, pGT-RB(403-816), contained an RB cDNA fragment previously demonstrated to contain a deletion mutation (see Figure 1), which inactivates T or E1A binding (Kaelin et al., 1990). Three chimeras, pGT-RB(379-928;dl exon 21), pGT-RB(379-928;dl exon 22), and pGT-RB(379-928;706 C~F), each contained a loss-of-function mutation previously described in a stable product of a spontaneously oc-
Cellular Proteins That Can Interact with pRB 523
I
2 3 4 5 6 7 8
9
10 11
12 13 14 15 16
205 116-97~ 66~
45~
29--
Figure 2. Expression and Purification of Gluthathione S-Transferase Fusion Proteins Whole-cell lysates of bacterial clones expressing the pGEX-2Tencoded glutathione S-transferase leader sequence (lane 1), pGTRB(379-792;dl 573-645) (lane 2), pGT-RB(403-816) (lane 3), pGTRB(379-928;dl sxon 21) (lane 4), pGT-RB(379-928;dl exon 22) (lane 5), pGT-RB(379-928;706 C~F) (lane 6), pGT-RB(379-928) (lane 7), and pGT-RB(379-792) (lane 8) were prepared as described in Experimental
Procedures.Foreachclonerepresentedin lanes1-8, a bacterialsonicats was also preparedand incubatedwith glutathione-Sepharose (lanes 9-16, respectively).The glutathione-Sepharosewas then washedand bound proteinswereelutedby boilingin the presenceof SDS. Proteinswereresolvedby electrophoresisin a 10% SDS-polyacrylamidegel and visualizedby Coomassieblue staining.
curring mutant RB allele (Bookstein et al., 1990; Horowitz et al., 1989, 1990; Kaye et al., 1990). The cDNA inserts for these three constructs were generated by PCR of reverse-transcribed mRNA from cell lines previously shown to be hemizygous or homozygous for each mutation. The presence of each mutation in the coding regions of the respective fusion proteins was confirmed by DNA sequence analysis. The two remaining chimeras, pGT-RB(379-928) and pGT-RB(379-792), contained intact T/E1A-binding domains. Expression and purification of the pRB fusion proteins are shown in Figure 2. Total protein from 500 p.I of broth containing bacteria transformed with parental pGEX-2T (lane 1) or with the various pGEX.2T recombinants (lanes 2-8) was resolved by SDS-polyacrylamide gel electrophoresis and visualized by Coomassie blue staining. For each of these bacterial clones, sonicates were also prepared from 10 ml of broth and passed over glutathione-Sepharose (lanes 9-16, respectively). The bound proteins, in each case, were eluted by boiling in an SDScontaining buffer and resolved by SDS--polyacrylamide gel electrophorasis. As can be seen from this figure, each pGEX-2T recombinant gave rise to a fusion protein that specifically bound to glutathione-Sepharose. Each pro-
tein had an apparent molecular weight consistent with the size of the pRB fragment encoded by its RB cDNA insert plus the 26 kd contributed by the glutathione S-transferase leader sequence. The results shown in this figure also imply a limited recovery of each fusion protein as a glutathione-Sepharose complex (i.e., ,~1% ). This is probably, in part, due to a decrease in solubility of the fusion proteins relative to the glutathione S-transferase leader polypeptide alone. We next investigated whether the pRB T/E1A-binding domain remained functional when expressed as a glutathione S-transferase fusion protein recovered from bacteria, as described above. Results of such an experiment are shown in Figure 3A. A whole-cell lysate was prepared from 293 cells, a human embryonic cell line transformed by a fragment containing the left-hand end of the adenovirus 5 genome and known to synthesize abundant quantities of E1A (Graham et al., 1977). Aliquots of the lysate were equilibrated with glutathione-Sepharose beads (lane 1), glutathione-Sepharose beads bound to glutathi0ne S-transferase (lane 2), or glutathione-Sepharose beads bound to each of a series of glutathione S-transferasepRB chimeras (lanes 3-9), prepared as shown in Figure 2, lanes 10-16, respectively. In this and all other experiments, identical amounts of each fusion protein were present in all reaction mixtures. Bound proteins were eluted by boiling in SDS-containing buffer and resolved by SDSpolyacrylamide gel electrophoresis, and a Western blot was performed using the monoclonal antibody M73 directed against E1A (Harlow et al., 1985). Lanes 10 and 11 served as controls in which immunoprecipitations were performed with either M58 (anti-E1A) (Harlow et al., 1985) or PAb 419 (anti-T) monoclonal antibodies (Harlow et al., 1981). As can be seen, E1A bound to the two fusion proteins in which pRB residues 379-792 remain intact (Figure 3A, lanes 8 and 9), but not to the others. Thus, the only two chimeras predicted to have specific E1A-binding activity displayed this function. To determine whether T was also able to bind to the appropriate glutathione S-transferase fusion proteins, an experiment similar to that described in Figure 3A was performed. A lysate was prepared from [3SS]methionine-labeled T1 cells, a CV1-P-derived cell line that overproduces T. Glutathione-Sepharose beads were loaded with the same set of fusion proteins described in Figure 3A. Each was then incubated with aliquots of the T1 cell lysate. Lanes 10 and 11 contain immunoprecipitates of this lysate performed with either PAb 419 (anti-T) or M58 (anti-E1A) monoclonal antibodies, respectively. Bound proteins were resolved by SDS-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane for anti-T Western blotting (Figure 3B). An autoradiogram of this blot is shown in Figure 3C. Again, as was the case for E1A, T was bound only by those fusion proteins containing a wild-type T/E1A-binding domain. Furthermore, this interaction appeared to be highly specific, as can be seen by noting the number of background bands in lanes 8 and 9 relative to lane 10 in Figure 3C. In a similar experiment, it was demonstrated that none of the fusion proteins recognized the T antigen point mutant K1
Cell 524
A 1
C 2
3
4
5
6
7
8
9
1o 11
1
2
3
4
5
6
7
8
9
1
10 11
2
3
4
5
6
7
8
;I
1o 11
180--
116~
18o--
180--
84-58--
116~
48--
4-T
E1A 36--
Identification of Cellular Proteins That Can Interact Specifically with the T/EIA-Binding Region of the Retinoblastoma Gene Product William G. Kaelin Jr.,* David C. Pallas,* James A. DeCaprio,* Frederic J. Kaye, t and David M. Livingston* * The Dana-Farber Cancer Institute and Harvard Medical School Boston, Massachusetts 02115 t National Cancer Institute-Navy Medical Oncology Branch and Uniformed Services University of the Health Sciences Bethesda, Maryland 20814
Summary The SV40 T antigen (T)/adenovirus EIA-binding domain of the retinoblaetoma gene product (pRB) has been fused to S. japonlcum glutathione S-transferase, and the chimera, bound to insoluble glutathione, was used to search for cellular proteins that can interact specifically with pRB. At least seven such proteins were detected in extracts of multiple human tumor cell lines. These proteins failed to bind to a family of pRB fusion proteins that harbor inactivating mutations in the T/EIA-binding domain and to the wild-type fusion protein in the presence of a peptide replica of the pRBbinding domain of T. Therefore, the binding of one or more of these proteins may contribute to the growthsuppressing function of pRB. Introduction Functional inactivation of both copies of the retinoblastoma susceptibility gene (RB) is an invariant feature of both sporadic and familial retinoblastomas (Cavenee et al., 1983; Horowitz et al., 1990; Murphree and Benedict, 1984) and may play a role in the evolution of some osteosarcomas, soft tissue sarcomas, small cell lung carcinomas, breast carcinomas, and genitourinary carcinomas (Bookstein et al., 1990; Friend et al., 1987; Harbour et al., 1988; Horowitz et al., 1989, 1990; Lee et al., 1988; Reissmann et al., 1989; TAng et al., 1988; Toguchida et al., 1989; Varley et al., 1989; Weichselbaum et al., 1988; Yokota et al., 1988). Indeed, RB expression has been detected in all human tissues examined to date (Bernards et al., 1989; Lee et al., 1987a), and there is indirect evidence to suggest that the RB gene product, a 110 kd nuclear phosphoprotein (Lee et al., 1987b; Ludlow et al., 1989), may function as a human cell cycle control element (Buchkovich et al., 1989; Chen et al., 1989; DeCaprio et al., 1989; Ludlow et al., 1990; Mihara et al., 1989; Xu et al., 1989). In keeping with its proposed function as a tumor suppressor gene, Lee and coworkers have demonstrated that reintroduction of a wild-type RB allele into RB-/cells, in some instances, led to a decrease in cell growth rate in culture and, where studied, to a reduction in the ability of the recipient cells to form tumors in nude mice (Bookstein et al., 1990; Huang et al., 1988).
How RB carries out its growth suppressor function(s) is currently unknown. The RB gene product (pRB) can bind to DNA-cellulose, although at the present time there is no evidence that pRB can recognize a specific DNA sequence (Lee et al., 1987b). Furthermore, at least two spontaneously occurring, loss-of-function pRB mutants can bind to DNA-cellulose with affinity comparable to the wildtype protein, suggesting that DNA binding is not sufficient for RB growth suppression function (Horowitz et al., 1989; Shew et al., 1990a). Additional clues to how pRB functions have come from the study of DNA tumor viruses. Transforming proteins of several unrelated DNA tumor viruses including the adenovirus E1A protein (E1A), the SV40 large T antigen (1-), other papovaviral T antigens, and the human papillomavirus E7 protein (E7) have been shown to bind to pRB in vitro (DeCaprio et al., 1988; Dyson et al., 1989b, 1990; Whyte et al., 1988). Mutational analysis of these viral proteins suggests that their ability to transform cells is linked, in part, to their ability to bind to pRB and that this interaction is mediated primarily by short, colinear amino acid sequences homologous to the adenovirus E1A transforming region designated CR2 (DeCaprio et al., 1988; Green, 1989; Jones et al., 1990; Moran, 1988; Phelps et al., 1988; Whyte et al., 1989). In addition, it has been demonstrated that T binds exclusively to the unphosphorylated form of pRB (Ludlow et al., 1989, 1990). This suggests that unphosphorylated pRB performs those elements of pRB growth suppressor function that T can perturb. The minimal region of pRB needed for the binding to two of these viral proteins has been mapped in several laboratories (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990). It constitutes approximately 40% of the protein sequence, extending from residue 379 to 792. As an intact subsegment of pRB, this sequence binds not only to E1A and T, but also interacts stably and specifically with a 14 residue peptide containing all of the sequence of the pRBbinding domain of T (Kaelin et al., 1990). To date, all spontaneously occuring RB gene mutations, which do not grossly compromise pRB stability, map to this region, and, as predicted, the altered RB gene products are inactive with respect to E1A and T binding (Bookstein et al., 1990; Horowitz et al., 1989, 1990; Kaye et al., 1990; Shew et al., 1990a, 1990b). Furthermore, this region can be predicted to contain a leucine zipper (Bernards et al., 1989; Hong et al., 1989; McGee et al., 1989), a structural motif that has been implicated in the ability of some proteins to form hetero- or homodimers (Kouzarides and Ziff, 1989; Mitchell and Tjian, 1989). These observations led us to hypothesize that normal pRB function requires it to interact with a cellular protein containing a region structurally similar to the pRB-binding motif found in these diverse viral proteins. In this model, one or more aspects of pRB function can be inactivated by alterations in the T/E1A-binding domain (either as a result of mutation or, possibly, protein modification, e.g., phosphorylation) or by binding to a pseudosubstrate or ef-
Cell 522
fector such as T or E1A. In short, the model predicts that cells produce one or more proteins that interact with the T/E1A-binding domain of pRB in a manner resembling that employed by T and E1A. To date, coimmunoprecipitation has failed to demonstrate the existence of a cellular pRB-binding protein. The use of this technique to identify protein-protein interactions has a number of limitations, at least some of which might account for the failure to detect such a protein(s) in the past, For example, if T and E1A interfere with the interaction between pRB and a specific cellular protein(s), one must presume that this putative cellular protein binds to pRB with a lower affinity than T and E1A and/or is present in amounts far less than the typical amount of T or E1A in SV40 or adenovirus-transformed cells. The binding of a given cellular protein to pRB might also be cell cycle dependent, as is the case for T (Ludlow et al., 1990). Thus, in an asynchronous population of cells, only a fraction of pRB molecules might exist in a protein-bound form. Furthermore, a large portion of the pRB sequence is devoted to T/E1A binding (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990), and, in principle, it could become masked by a bound protein. If so, an antibody that interacts with this region of the protein surface might, preferentially or exclusively, recognize the unbound form of the protein. This would be especially true if pRB ordinarily binds to a large, multimeric complex of proteins that effectively shields key epitopes from specific antibody-combining sites. We therefore devised a pRB affinity binding method that would allow one to circumvent the need for pRB antibodies and to make use of the genetic data cited earlier, which points to the existence of a discrete, independent pRB domain (hereafter referred to as "pocket") that can bind, specifically, to selected, short viral peptide sequences. The scheme incorporates the use of the prokaryotic expression vector pGEX-2T (Smith and Johnson, 1988) and the polymerase chain reaction (PCR) (Saiki et al., 1988) to generate chimeric glutathione S-transferase-pRB protein affinity chromatographic reagents capable of binding T and E1A. Such fusion proteins would be expected to have a relatively high affinity for reduced glutathione and can be recovered and purified in relatively nondenatured form as noncovalent glutathione-Sepharose conjugates (Smith and Johnson, 1988). Results of experiments in which these chimeras were used as affinity binding reagents have led to the identification of a family of cellular proteins that can interact in vitro with the T/E1A-binding domain of pRB with the same specificity as T, E1A, or the aforementioned peptide copy of the pRB-binding domain of T. Thus, cells produce proteins that have the potential of interacting in a physiologically relevant manner with pRB. Results Glutathione S-Transferaae Fusion Proteins Containing the pRB T/EIA-Binding Domain Can Interact Specifically with T or EIA In Vitro We and others have previously demonstrated that the pRB
Glutathione
S-Transferase
Fusion
Proteins
T/E1A-Sinding Region
pR8 1
pGT-RB(379-792;dl
573 645)
379
792
928
glutathlone-stransferase 573 645
pGT RB(403 816) 403
pGT RB(379 928:dl Exon 21)
8t6
91ulathlone s 703 737
pGT-RB(379-928:dl Exon 22)
glu,ath,ones 738 775
pGT-RB(379-928:706
C ~IPF)
I glutalhio~e s
cL~.F pGT-RB(379-928)
I g[utath~orles
pGT RB(379 792)
I g'°'~'~ ...... t~ .s,.~?Zq,~; × ,//.,×? tran~erase >~ I/'/;?//~,/?',/t× I
BB
In-Frame Deletion
Figure 1. GlutathioneS-Transferase-pRBFusion Proteins
region spanning residues 379-792 is sufficient for binding to T or E1A and must be relatively intact for this interaction to occur (Hu et al., 1990; Huang et al., 1990; Kaelin et al., 1990). Furthermore, we have shown that this region appears to make direct contact with a short consensus sequence present in these otherwise unrelated viral proteins (Kaelin et al., 1990). To search for the existence of cellular proteins capable of interacting with pRB by a similar mechanism, we generated a series of glutathione S-transferase-pRB fusion proteins (see Figure 1) for use as affinity reagents when bound noncovalently to glutathioneSepharose. All contained an intact or mutant version of the T/E1A-binding domain of pRB. The wild-type T/EIA-binding domain of pRB has been shown previously to retain function as an independent unit (Kaelin et al., 1990). Two of these chimeric fusion proteins harbored in vitro generated mutations in this region. One, pGT-RB(379-792;dl 573-645), was constructed by sequential PCR as described previously (Higuchi et al., 1988). The pRB segment deleted from this construct, namely, residues 573-645, has been shown by others to serve a "spacer" function within the T/E1A-binding region, i.e., deletion of this region inactivated the binding function of the domain unless a "filler" sequence was reinserted (Hu et al., 1990; Huang et al., 1990). The second mutant, pGT-RB(403-816), contained an RB cDNA fragment previously demonstrated to contain a deletion mutation (see Figure 1), which inactivates T or E1A binding (Kaelin et al., 1990). Three chimeras, pGT-RB(379-928;dl exon 21), pGT-RB(379-928;dl exon 22), and pGT-RB(379-928;706 C~F), each contained a loss-of-function mutation previously described in a stable product of a spontaneously oc-
Cellular Proteins That Can Interact with pRB 523
I
2 3 4 5 6 7 8
9
10 11
12 13 14 15 16
205 116-97~ 66~
45~
29--
Figure 2. Expression and Purification of Gluthathione S-Transferase Fusion Proteins Whole-cell lysates of bacterial clones expressing the pGEX-2Tencoded glutathione S-transferase leader sequence (lane 1), pGTRB(379-792;dl 573-645) (lane 2), pGT-RB(403-816) (lane 3), pGTRB(379-928;dl sxon 21) (lane 4), pGT-RB(379-928;dl exon 22) (lane 5), pGT-RB(379-928;706 C~F) (lane 6), pGT-RB(379-928) (lane 7), and pGT-RB(379-792) (lane 8) were prepared as described in Experimental
Procedures.Foreachclonerepresentedin lanes1-8, a bacterialsonicats was also preparedand incubatedwith glutathione-Sepharose (lanes 9-16, respectively).The glutathione-Sepharosewas then washedand bound proteinswereelutedby boilingin the presenceof SDS. Proteinswereresolvedby electrophoresisin a 10% SDS-polyacrylamidegel and visualizedby Coomassieblue staining.
curring mutant RB allele (Bookstein et al., 1990; Horowitz et al., 1989, 1990; Kaye et al., 1990). The cDNA inserts for these three constructs were generated by PCR of reverse-transcribed mRNA from cell lines previously shown to be hemizygous or homozygous for each mutation. The presence of each mutation in the coding regions of the respective fusion proteins was confirmed by DNA sequence analysis. The two remaining chimeras, pGT-RB(379-928) and pGT-RB(379-792), contained intact T/E1A-binding domains. Expression and purification of the pRB fusion proteins are shown in Figure 2. Total protein from 500 p.I of broth containing bacteria transformed with parental pGEX-2T (lane 1) or with the various pGEX.2T recombinants (lanes 2-8) was resolved by SDS-polyacrylamide gel electrophoresis and visualized by Coomassie blue staining. For each of these bacterial clones, sonicates were also prepared from 10 ml of broth and passed over glutathione-Sepharose (lanes 9-16, respectively). The bound proteins, in each case, were eluted by boiling in an SDScontaining buffer and resolved by SDS--polyacrylamide gel electrophorasis. As can be seen from this figure, each pGEX-2T recombinant gave rise to a fusion protein that specifically bound to glutathione-Sepharose. Each pro-
tein had an apparent molecular weight consistent with the size of the pRB fragment encoded by its RB cDNA insert plus the 26 kd contributed by the glutathione S-transferase leader sequence. The results shown in this figure also imply a limited recovery of each fusion protein as a glutathione-Sepharose complex (i.e., ,~1% ). This is probably, in part, due to a decrease in solubility of the fusion proteins relative to the glutathione S-transferase leader polypeptide alone. We next investigated whether the pRB T/E1A-binding domain remained functional when expressed as a glutathione S-transferase fusion protein recovered from bacteria, as described above. Results of such an experiment are shown in Figure 3A. A whole-cell lysate was prepared from 293 cells, a human embryonic cell line transformed by a fragment containing the left-hand end of the adenovirus 5 genome and known to synthesize abundant quantities of E1A (Graham et al., 1977). Aliquots of the lysate were equilibrated with glutathione-Sepharose beads (lane 1), glutathione-Sepharose beads bound to glutathi0ne S-transferase (lane 2), or glutathione-Sepharose beads bound to each of a series of glutathione S-transferasepRB chimeras (lanes 3-9), prepared as shown in Figure 2, lanes 10-16, respectively. In this and all other experiments, identical amounts of each fusion protein were present in all reaction mixtures. Bound proteins were eluted by boiling in SDS-containing buffer and resolved by SDSpolyacrylamide gel electrophoresis, and a Western blot was performed using the monoclonal antibody M73 directed against E1A (Harlow et al., 1985). Lanes 10 and 11 served as controls in which immunoprecipitations were performed with either M58 (anti-E1A) (Harlow et al., 1985) or PAb 419 (anti-T) monoclonal antibodies (Harlow et al., 1981). As can be seen, E1A bound to the two fusion proteins in which pRB residues 379-792 remain intact (Figure 3A, lanes 8 and 9), but not to the others. Thus, the only two chimeras predicted to have specific E1A-binding activity displayed this function. To determine whether T was also able to bind to the appropriate glutathione S-transferase fusion proteins, an experiment similar to that described in Figure 3A was performed. A lysate was prepared from [3SS]methionine-labeled T1 cells, a CV1-P-derived cell line that overproduces T. Glutathione-Sepharose beads were loaded with the same set of fusion proteins described in Figure 3A. Each was then incubated with aliquots of the T1 cell lysate. Lanes 10 and 11 contain immunoprecipitates of this lysate performed with either PAb 419 (anti-T) or M58 (anti-E1A) monoclonal antibodies, respectively. Bound proteins were resolved by SDS-polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride (PVDF) membrane for anti-T Western blotting (Figure 3B). An autoradiogram of this blot is shown in Figure 3C. Again, as was the case for E1A, T was bound only by those fusion proteins containing a wild-type T/E1A-binding domain. Furthermore, this interaction appeared to be highly specific, as can be seen by noting the number of background bands in lanes 8 and 9 relative to lane 10 in Figure 3C. In a similar experiment, it was demonstrated that none of the fusion proteins recognized the T antigen point mutant K1
Cell 524
A 1
C 2
3
4
5
6
7
8
9
1o 11
1
2
3
4
5
6
7
8
9
1
10 11
2
3
4
5
6
7
8
;I
1o 11
180--
116~
18o--
180--
84-58--
116~
48--
4-T
E1A 36--
