Transforming Growth Factor-1 Modulates p107 Function in Myeloid Cells CORRELATION WITH CELL CYCLE PROGRESSION

Transforming growth factor-β1 (TGF-β1) is a potent inhibitor of hematopoietic cell growth. Here we report that TGF-β1 signals inhibition of IL-3-dependent 32D-123 murine myeloid cell growth by modulating the activities of cyclin E and cyclin-dependent kinase 2 (cdk2) proteins and their complex formation in the G phase of the cell cycle. Whereas the cyclin E protein was hyperphosphorylated in TGF-β1-treated cells, TGF-β1 decreased both the phosphorylation of cdk2 and the kinase activity of the cyclin E-cdk2 complex. Decreased cyclin E-cdk2 kinase activity correlated with decreased phosphorylation of the retinoblastoma-related protein p107. In support of these observations, transient overexpression of p107 inhibited the proliferation of the myeloid cells, and expression of antisense oligodeoxynucleotides to p107 mRNA blocked TGF-β1 inhibition of myeloid cell growth. Furthermore, as reported previously, in 32D-123 TGF-β1-treated cells, c-Myc protein expression was decreased. TGF-β1 increased the binding of p107 to the transcription factor E2F, leading to decreased c-Myc protein levels. p107 inhibited E2F transactivation activity and was also found to bind the c-Myc protein, suggesting p107 negative regulation of c-Myc protein function. These studies demonstrate the modulation of p107 function by TGF-β1 and suggest a novel mechanism by which TGF-β1 blocks cell cycle progression in myeloid cells.

D type cyclins appear earlier in G 1 than cyclin E. Cyclin D expression oscillates minimally throughout the cell cycle. Cdk4, which is thought to act at the G 1 /S restriction point, is associated exclusively with cyclin D (6 -8). The expression of cyclin E and its associated kinase, cdk2, reaches a peak late in the G 1 phase (9 -11). The S phase is controlled by the expression of cyclin A, which associates with both cdc2 and cdk2, whereas the G 2 /M transition is controlled by cyclin B and cdc2 (2,3,12,13).
Cdc2, cdk2, and their activating cyclins also interact with and/or phosphorylate other proteins believed to be important in cell cycle regulation and DNA synthesis, including the tumor suppressor gene product RB, the RB-related protein p107, and the transcription factor E2F (14 -17). The RB protein negatively regulates E2F. E2F is known to play a role in the activation of genes required for cell cycle progression. Genes such as c-myc, c-myb, thymidine kinase, and cdc2 contain E2F binding sites in their promoters (18,19). The biological activity of the RB protein is regulated by its phosphorylation state throughout the cell cycle. Hypophosphorylated RB (G 0 /G 1 phase) has been shown to bind and sequester the E2F protein, thereby inhibiting E2F transactivation (20,21). The cyclin E-cdk2 and cyclin D-cdk4 complexes are known to phosphorylate and inactivate RB at the G 1 /S boundary, thus allowing cell cycle progression (7,10,22). The c-myc proto-oncogene, whose expression is transactivated by E2F, is also known to play an important role during the G 1 /S transition of the cell cycle (23,24).
The p107 tumor suppressor protein shares many structural and biochemical features with RB. Similar to RB, p107 is a potent inhibitor of E2F-mediated transactivation, and overexpression of p107 can inhibit the proliferation of certain cell types by arresting the cells in G 1 (17,25). More recent studies have provided evidence for p107 binding to the c-Myc protein, suggesting for the first time a possible role for p107 in the regulation of c-Myc function (26). p107 is usually underphosphorylated during G 1 and is phosphorylated in the S, G 2 , and M phases (27). However, the regulation of p107 function during growth suppression remains to be elucidated.
TGF-␤1 is a multifunctional regulator of cell growth and differentiation in many cell types, including epithelial, endothelial, and hematopoietic cells (28,29). In epithelial cells, it has been shown that the TGF-␤1-induced proliferative block occurs in late G 1 and thus prevents DNA synthesis (30 -33). TGF-␤1 inhibition of keratinocyte proliferation involves the suppression of c-myc transcription via hypophosphorylation of RB. It has been shown that one of the steps by which TGF-␤1 exerts its inhibitory effects on cell cycle progression is by preventing the hyperphosphorylation of RB (34,35). This TGF-␤1mediated block in RB hyperphosphorylation also prevents the transactivation of the c-myc gene, which causes c-Myc protein levels to decline rapidly (36). In addition, in Mv1Lu cells TGF-␤1 also has been shown to disturb the assembly of cyclin E-cdk2 complexes and cyclin E-associated kinase activity (32).
The growth of interleukin 3 (IL-3)-dependent myeloid leukemic cell lines is inhibited by TGF-␤1 (37,38). The molecular mechanisms involved in TGF-␤1-induced growth inhibition and cell cycle progression in myeloid cells remain largely unknown and may be different from those found in keratinocytes and other epithelial cells. Because TGF-␤1 plays a critical role in hematopoiesis, the molecular mechanisms by which TGF-␤1 inhibits myeloid cell proliferation, specifically at the G 1 /S transition, were studied. These results indicate that in myeloid cells, TGF-␤1 modestly inhibits the association of the cyclin E-cdk2 complexes and markedly inhibits the activity of the complex, as well as markedly stimulating phosphorylation of cyclin E and dephosphorylation of cdk2. However, phosphorylation of RB was unchanged in synchronized IL-3-dependent 32D-123 myeloid cells treated with TGF-␤1. In contrast, we found that in response to TGF-␤1 treatment of the 32D-123 myeloid cells, p107 was hypophosphorylated, and p107 binding to E2F and to c-Myc was increased. This modulation of p107 by TGF-␤1 together with the ability of antisense to p107 to block TGF-␤1 inhibition of growth suggests a novel mechanism by which TGF-␤1 blocks cell cycle progression in myeloid cells.
Cell Sychronization-Cells were synchronized at the G 1 /S boundary of the cell cycle by aphidicolin treatment (40) or at the G 0 by IL-3 and serum deprivation. Aphidicolin treatment was used to define the different phases of the cell cycle, whereas IL-3 deprivation allowed synchronization of the cells at the G 0 phase. Logarithmically growing cells were incubated for 12 h at 37°C in the presence of 2 g/ml aphidicolin. After synchronization, cells were washed twice with fresh serum-free medium, resuspended at 5ϫ 10 5 cells/ml in complete medium without aphidicolin, and then reincubated at 37°C in Corning culture flasks. For IL-3 deprivation, logarithmically growing cells were washed twice with serum-free medium followed by incubation in serum-free medium at 37°C for 18 h. After G 0 arrest, cells were treated with IL-3 (1,000 units/ml) and 10% fetal calf serum. Synchrony was monitored by serial cell counts, measurement of [ 3 H]thymidine incorporation, and cell cycle analysis. For [ 3 H]thymidine incorporation, the cells were pulsed with 0.2 Ci of [ 3 H]thymidine/well for 6 h, and radioactivity was determined by liquid scintillation counting. Cell cycle analysis was performed by staining DNA with 0.05 mg/ml propidium iodide (Sigma) for 1 h and then analyzing DNA by fluorescence-activated cell sorting (FACS IV flow cytometer; Becton Dickinson, Mountain View, CA). For detection of propidium iodide, cells were excited at 488 nm, and their emissions were collected between 515 and 545 nm.
In Vivo Phosphorylation and Immunoprecipitation-Cells were cultured and synchronized by aphidicolin treatment or IL-3 deprivation as described earlier, except that 4 h before each time point, aliquots of 1 ϫ 10 7 cells were gently pelleted and resuspended in labeling medium (phosphate-free RPMI 1640 containing 10% (v/v) dialyzed fetal calf serum). After incubation at 37°C for 30 min to deplete the intracellular pool of phosphate, cells were pelleted, resuspended, and incubated in 1 ml of labeling medium containing 0.4 mCi of 32 P (8,500 -9,120 Ci/mmol; DuPont) for 3.5 h at 37°C. Cells were then washed in phosphatebuffered saline and lysed with RIPA buffer containing phosphatase inhibitors (10 mM sodium pyrophosphate, 10 mM sodium fluoride, and 2 mM sodium orthovanadate). The cell lysates were centrifuged at 15,000 rpm for 15 min to remove cellular debris and precleared with protein A/G-agarose. Proteins were then immunoprecipitated by incubation with antibodies (anti-cyclin E, cdk2, RB, or p107, Santa Cruz Biotechnology) and protein A/G-agarose at 4°C for 4 h with rotation. Immunoprecipitates were washed once with RIPA buffer containing 1% bovine serum albumin (RIPA/BSA), twice with RIPA, boiled for 5 min at 100°C in SDS-sample buffer, and analyzed by SDS-PAGE and autoradiography. For the phosphatase treatments, cell lysates were immunoprecipitated with an anti-p107 antibody, washed once with RIPA/BSA, twice with RIPA, and twice with phosphatase buffer (10 mM MOPS, pH 7.5, 60 mM 2-mercaptoethanol, 0.1 M NaCl, 1 mg/ml BSA). Washed immunoprecipitates were incubated for 15 min at 37°C in 100 l of phosphatase buffer with protein phosphatase 1A (0.5 unit; Upstate Biotechnology Inc., Lake Placid, NY). Following incubation, the supernatant was aspirated, SDS-sample buffer was added to stop the reaction, and the immunoprecipitates were analyzed by SDS-PAGE (4 -15%) and Western blotting with an anti-p107 antibody.
Plasmids-The mouse RB expression vector (pmRB) was kindly provided by Dr. Paul D. Robbins (University of Pittsburgh School of Medicine). pmRB contains an EcoRI (position Ϫ100) to OxaNI (position ϩ2,766) mouse cDNA fragment cloned into the EcoRI and ClaI sites of pJ3⍀ (41). The mouse p107 expression vector (mp107) is a 4.8-kilobase full-length cDNA that was cloned from a pre-B cell line (22D6) cDNA gt10 library and then subcloned into the NotI site of pBluescript SK. 2 After amplification in RB404 (dam-) bacteria, the clone was cut with NotI and BclI enzymes. (The BclI site is located 12 base pairs after the stop codon.) The 3.25-kilobase fragment was subcloned into the NotI and BamHI sites of the Rous sarcoma virus long terminal repeat-driven expression vector, pREP 9 (Invitrogen, San Diego). A 500-base pair NotI-EcoRV fragment of mouse p107 cDNA comprising the 5Ј-noncoding region, the ATG translation initiation codon, and a 440-base pair coding sequence was inserted at the NheI of the pMAM-Neo-CAT vector (Clontech, Palo Alto, CA) by standard linker ligation. Antisense and sense clones (pMAM-107 AS and pMAM-107 S) with reference to the murine mammary tumor virus long terminal repeat were identified by HindIII restriction analysis.
GAL4-E2F was constructed by inserting a polymerase chain reaction-generated fragment that contains murine E2F sequences into the SaII and XbaI sites in the polylinker of pBluescript SKϩ (42). The GAL4-E2F construction used in these experiments contains the complete 440 amino acids of the E2F protein fused to the DNA binding domain of GAL4 (amino acids 1-147). The reporter construction, G1B, contains one GAL4 binding site upstream of the adenovirus E1b box driving CAT (41).
Transfections and Proliferation Assay-The 32D-123 cells (1 ϫ 10 7 / 200 l) were transfected by electroporation as described previously (43). Briefly, after the addition of DNA (50 g of mp107, 50 g of pmRB, or control pUC-18 (50 g)), the cells were chilled on ice for 10 min. Cells were then electroporated in a Bio-Rad gene pulser system (0.3 kilovolts, 960 microfarads) and incubated on ice for 10 min. The cells were cultured at a concentration of 1 ϫ 10 6 /ml for 24 h. At this time point, cells were counted, and cell viability was determined by trypan blue exclusion. To determine cell proliferation, 5 ϫ 10 3 viable cells/well were incubated for 16 h in a 96-well plate, and cell proliferation was measured by [ 3 H]thymidine incorporation as described above or using the MTT assay kit (Promega, Madison, WI).
CAT Assays-The murine myeloid 32D-123 cells were transfected using the electroporation method as described previously (43). Equal amounts of DNA were transfected in each case by the addition of pBR322 when necessary. After 24 h at 37°C the cells were harvested, and the protein concentration of the cell lysates was determined using the Bio-Rad protein assay kit. CAT enzyme activity was normalized for transfection efficiency by cotransfection of secreted growth hormone expression plasmid, pSV6H, and determination of secreted growth hormone in the medium, prior to harvesting for CAT activity (Nichols Institute, San Juan Capistrano, CA). Equal amounts of protein were used to assay for the CAT enzyme, and each assay was repeated at least three times.

RESULTS
Cell Synchrony and TGF-␤1 Treatment-Previous studies by our laboratory, using exponentially growing nonsynchronized cells, have demonstrated the ability of TGF-␤1 to inhibit 32D-123 cell proliferation (44,45). To define more specifically the phase of the cell cycle where TGF-␤1 exerts its inhibitory effects, we synchronized the 32D-123 cells using two methods: 1) aphidicolin treatment, which reversibly inhibits DNA polymerase-␣ activity and DNA synthesis and synchronizes the cells at the G 1 /S phase, and 2) IL-3 deprivation, which causes the cells to arrest at the G 0 phase (34,46). Synchronized cells were analyzed by FACS at different time points. Aphidicolin-synchronized 32D-123 cells were arrested at the G 1 /S boundary (61.7%) and showed continuous cell cycle progression upon release from the block (Fig. 1A). IL-3-deprived cells were ar-rested at the G 0 cell cycle phase (73.3%, Fig. 1B). The aphidicolin-treated cells were better at defining the phases of the cell cycle, as indicated in Fig. 1A. However, IL-3-deprived cells were found to have a longer G 1 phase (0 -16 h, Fig. 1B), which allowed a more accurate determination of the optimal time for TGF-␤1 regulation of cell growth.
To determine the optimal time point for TGF-␤1 treatment (maximal growth inhibition) and because 32D-123 cells require 12 h to complete the cell cycle following aphidicolin treatment ( Fig. 1A), at 2-h intervals (0, 2, 4, 6, 8, 10, and 12 h) after aphidicolin treatment cells were treated with TGF-␤1 (10 ng/ ml), and proliferation was analyzed by [ 3 H]thymidine incorporation. As shown in Fig. 1C using aphidicolin-treated cells maximal inhibition of [ 3 H]thymidine incorporation (45%) was observed by 8 h. Table I shows the FACS analysis of cells treated in the presence and absence of TGF-␤1 at 8 h after aphidicolin treatment. By 24 h after TGF-␤1 treatment 64% of the cells arrested in the G 1 phase (control 39.6%, Table I). Longer incubation times in the presence of TGF-␤1 increased the number of cells arrested in G 1 (data not shown). In Fig. 1C time zero refers to the addition of aphidicolin. There was no significant inhibition of 32D-123 cell proliferation when TGF-␤1 was added after 12 h of aphidicolin treatment because the cells are at the G 1 /S restriction point (as indicated in Fig.  1A). 32D-123 cell growth inhibition was seen at 8 h, when a significant number of cells are found in the G 1 phase (a point of the cell cycle when they are most sensitive to TGF-␤1) moving toward the restriction point (34). At earlier time points low levels of inhibition are seen because of the lack of synchronization of the cells. Therefore, the following TGF-␤1 studies were done with cells treated with aphidicolin for 12 h. TGF-␤1 was added to the cells 8 h after the initial treatment with aphidicolin, for a total of 24 h. Aphidicolin and TGF-␤1 treat-

TGF-␤1 Inhibits Cell Cycle Progression via p107
ment overlapped only for a period of 4 h.
IL-3-deprived cells were treated with IL-3 and serum followed by TGF-␤1 treatment (for a total of 24 h) at 2-h intervals (0, 2, 4, 6, 8, and 10 h). Maximal inhibition of [ 3 H]thymidine incorporation (76%) was observed when 32D-123 cells were treated with TGF-␤1 at 6 h after IL-3 addition (Fig. 1D). Lower levels of 32D-123 growth inhibition by TGF-␤1 treatment at 2 and 4 h following IL-3 addition were not unexpected. Most, if not all growth factors up-regulate TGF-␤ cell surface receptors (45,47), and previous studies by our laboratory have indicated the expression of a low number of TGF-␤ receptors in the 32D-123 cells (45). Specifically, TGF-␤ ligand binding studies previously indicated the absence of TGF-␤1 receptors in IL-3deprived cells (data not shown). However, whether low or high number of receptors are bound by TGF-␤1 ligand followed by internalization, the reappearance of the TGF-␤ receptors in the outer cell membrane only occurs 72 h later (45). Consequently, since inhibition of proliferation by TGF-␤1 occurs only after threshold levels of the intracellular regulators are reached, 6 h after IL-3 addition is when maximal expression/internalization of receptor-ligand complexes of TGF-␤ takes place. The TGF-␤1-induced maximum growth inhibition of IL-3-deprived cells at 6 h indicated optimal expression of cell surface TGF-␤ receptors. FACS analysis of these cells identified that by 14 h, 64.3% of the cells were growth arrested at the G 1 phase (Table  II). This FACS analysis confirms our earlier observation that aphidicolin synchronization defines the phases of the cell cycle more clearly than IL-3 deprivation (Tables I and II).
Expression of Cyclins and Cdks in Synchronized Myeloid Cells-In mammalian cells, cyclins A, B, D, and E are expressed and have defined roles as regulatory subunits of the cdks during specific phases of the cell cycle (8,11,48,49). To study the expression pattern of these cyclins and cdks on 32D-123 myeloid cells, we used aphidicolin-treated cells; these cells were lysed at each phase of the cell cycle and analyzed by Western blotting (Fig. 2A). Cyclin A and cyclin B protein expression peaked at the S and G 2 /M phases and decreased in the G 1 phase. The steady-state levels of cyclin D remained constant during the entire cell cycle, as has been reported previously for mammalian cells (6,50). Cyclin E protein expression showed two bands that are known to be regulated independently (9). Increased expression of the lower band and decreased expression of the higher band were observed at the G 1 phase. During the G 2 /M transition, the cyclin E bands were expressed at equal levels. Cdc2 protein expression was abundant in the S and G 2 /M phases, whereas cdk2 protein levels were abundant in the G 1 and S phases. Cdk4 protein expression remained constant throughout the cell cycle. Cdk5 protein expression was high during the S phase in the 32D-123 cells, and the levels decreased throughout G 2 , M, and G 1 phases. Thus, the expression levels of the different cyclins and cdks are representative con-trols for each cell cycling population defining the cell cycle phases in 32D-123 aphidicolin-treated cells. The levels of expression of these proteins were found to be consistent with the regulation reported for other cell systems (4,5,11,51).
Effect of TGF-␤1 on the Expression of G 1 Cyclins-Since cyclins D and E are known to play a critical role in the G 1 /S transition (50), we investigated the effect of TGF-␤1 (for 24 h) on the regulation of the steady-state levels of these cyclins in the 32D-123 myeloid cells at the G 1 /S transition. TGF-␤1 treatment of aphidicolin-treated cells did not affect cyclin D protein levels (Fig. 2B). Cyclin A protein expression decreased in aphidicolin-treated 32D-123 cells treated with TGF-␤1 (Fig.  2B). Although the combined level of the two cyclin E proteins did not change, the addition of TGF-␤1 resulted in an increased expression of the higher band; the lower band was slightly decreased (Figs. 2B and 3A). In aphidicolin-treated myeloid cells, TGF-␤1 treatment decreased the steady-state levels of cdk2 and cdk5, but not cdk4 (Fig. 2B).
TGF-␤1 Modulation of Cyclin E-cdk2 Kinase Activity-To investigate the association of cyclin E with cdk2, immunoprecipitation studies were performed with 32D-123 cells (aphidicolin-treated) treated with TGF-␤1 for 24 h (G 1 /S). Cell lysates were then immunoprecipitated with anti-cyclin E antibody followed by Western analysis with anti-cyclin E and cdk2 antibodies. Although the combined level of the two cyclin E proteins did not change, the addition of TGF-␤1 increased the higher band and decreased the lower band (Fig. 3A). The cyclin E-associated cdk2 protein was decreased in the TGF-␤1-treated cells (Fig. 3A), suggesting that the association between cyclin E and cdk2 was modestly diminished in the TGF-␤1-treated myeloid cells. In vivo 32 P-phosphorylated cell lysates from TGF-␤1-treated and untreated cells (aphidicolin-treated) were immunoprecipitated with an anti-cyclin E antibody and analyzed by SDS-PAGE. As shown in Fig. 3B, phosphorylation of the two cyclin E bands was markedly increased by TGF-␤1 treatment at the G 1 /S transition. Immunoprecipitation of the in vivo 32 Pphosphorylated cell lysates with an anti-cdk2 antibody followed by SDS-PAGE analysis indicated that phosphorylated cdk2, the more rapidly migrating form (30,52), was markedly decreased in TGF-␤1 treated cell lysates (Fig. 3C). The decreased phosphorylation of cdk2 and/or the increased phosphorylation of cyclin E (Fig. 3B) could lead to the decreased association of cyclin E and cdk2 proteins. To study the effect of TGF-␤1 on the activity of cyclin E-cdk2, we performed histone H1 kinase assays using anti-cyclin E immunoprecipitates. The immunoprecipitates were incubated in the presence of [␥-32 P]ATP and histone H1 and were then analyzed by SDS-PAGE. TGF-␤1 was found to reduce markedly cyclin E-associated kinase activity, but not the activity of the control, cyclin B (Fig. 3D). These results indicate that in myeloid cells, TGF-␤1 modestly inhibits the association of cyclin E-cdk2 complexes a Cells were treated with aphidicolin (2 g/ml) for 12 h and washed twice with serum-free medium. b Eight h after aphidicolin treatment cells were treated in the absence (Ϫ) and presence (ϩ) of TGF-␤1 for the time periods indicated and subjected to flow cytometry analysis to determine the cell cycle profiles. The data are expressed as the percentage of total cells.

TGF-␤1 Inhibits Cell Cycle Progression via p107
and markedly inhibits the activity of the complex, as well as markedly stimulating the phosphorylation of cyclin E and the dephosphorylation of cdk2.
TGF-␤1 Effect on RB and p107 Protein Expression-TGF-␤1 inhibition of cell growth is known to occur via the dephosphorylation of the RB protein during G 1 in Mv1Lu cells (32,35). Cyclin E-cdk2 is one of the major kinases that phosphorylates RB (10,30). In aphidicolin-treated myeloid cells, the cyclin E-cdk2 kinase activity decreased as a result of TGF-␤1 treatment (Fig. 3D). Therefore, we next investigated TGF-␤1 regulation of RB phosphorylation in myeloid cells. To study the phosphorylation state of RB, cell lysates from G 0 synchronized 32D-123 cells (IL-3-deprived) were analyzed by Western blot with an anti-RB antibody. Hypophosphorylated RB was observed at the G 0 phase (0 h after IL-3 addition, Fig. 4A, lane 1) and hyperphosphorylated RB was observed at the G 1 /S phase (20 h after IL-3 addition), regardless of TGF-␤1 treatment (Fig.   4A, lanes 5 and 6). RB protein phosphorylation was not regulated by TGF-␤1 in the 32D-123 cells (Fig. 4A). The lack of TGF-␤1 regulation of the RB protein led us to consider the possibility that TGF-␤1 could signal the regulation of an alternate RB-related protein. The ability of TGF-␤1 to regulate the expression of p107 in myeloid cells was investigated using Western analysis with an anti-p107 antibody. The expression of p107 was found to be decreased at the G 1 /S phase after TGF-␤1 treatment, with a marked decrease seen in the level of the lower band (Fig. 4B, lanes 5 and 6). p107 is known to be phosphorylated in a manner similar to RB by cyclin E-cdk2 kinase (27). Consequently, to investigate the effect of TGF-␤1 on the activity of the RB-and the p107-associated kinases, we performed an in vitro kinase assay. RB and p107 immunoprecipitates from TGF-␤1-treated and untreated 32D-123 cell lysates (G 1 /S phase) were incubated in the presence of [␥-32 P]ATP and histone H1 and then analyzed for their kinase activity. This was followed by SDS-PAGE. A decrease in p107associated kinase activity was seen in TGF-␤1-treated cells a Cells were starved of serum and IL-3 for 18 h, and then IL-3 and 10% FCS were added. b Six h after IL-3 and serum addition cells were treated in the absence (Ϫ) and presence (ϩ) of TGF-␤1 for the time periods indicated and subjected to flow cytometry analysis to determine the cell cycle profiles. The data are expressed as the percentage of total cells.

FIG. 2. Expression of cyclins and effect of TGF-␤1 treatment on the levels of G 1 cyclins in 32D-123 cells.
Panel A, cell aliquots were taken at the indicated times after release from an aphidicolin block, and cell lysates (150 g of protein) were subjected to immunoblot analysis. Blots were probed with antibodies to cyclin A, B, D, E, cdc2, cdk2, cdk4, or cdk5 as indicated. Panel B, cells were aphidicolin synchronized for a total of 12 h. Before the end of the aphidicolin treatment TGF-␤1 was added to the cells for a total of 24 h (aphidicolin and TGF-␤1 treatment overlap for 4 h). Cell aliquots were taken at the G 1 /S boundary (24 h after TGF-␤1 treatment), lysed, and protein samples (150 g) were subjected to immunoblot analysis. Blots were probed with antibodies to cyclins D, E, and A and cdks 2, 4, and 5 as indicated.

FIG. 3. TGF-␤1 modulation of cyclin E protein levels and associated protein kinase activity in lysates from 32D-123 myeloid cells synchronized at the G 1 /S phase (24 h after TGF-␤1 treatment)
. 32D-123 cells were synchronized and treated as described in the legend to Fig. 2. Panel A, immunoblot analysis of anti-cyclin E immune complexes. Immunoblots of proteins immunoprecipitated with anticyclin E were probed with antibodies to cyclin E and cdk2. Panels B and C, TGF-␤1 modulation of cyclin E and cdk2, respectively, as measured by in vivo phosphorylation. Cells were labeled with [ 32 P]orthophosphate for 3.5 h, lysed, and cyclin E or cdk2 proteins were analyzed by immunoprecipitation and SDS-PAGE (12% for cyclin E and 15% for cdk2) followed by autoradiography. Panel D, histone H1 kinase assay was carried out using anti-cyclin E and B immune complexes prepared from 32D-123 cells treated with TGF-␤1. The immune complexes were incubated under the appropriate conditions in the presence of [␥-32 P]ATP and histone H1. This was followed by SDS-PAGE and autoradiography. (Fig. 4C, lane 2), whereas the RB-associated kinase activity remained unchanged (Fig. 4C, lane 4). Inhibition of p107-associated kinase activity is at least partially a consequence of the TGF-␤1-induced decrease of cyclin E-cdk2 kinase activity (Figs.  3D and 4C, lanes 1 and 2).
Characterization of p107 Phosphorylation-The phosphorylation state of p107 is important for the regulation of the cell cycle (27). We investigated the TGF-␤1 regulation of RB and p107 phosphorylation by in vivo 32 P labeling, immunoprecipitation, and SDS-PAGE analysis. In exponentially growing control cells, immunoprecipitation of in vivo 32 P-labeled p107 and Rb (Fig. 5A, lanes 1-4) indicated the absence of cross-reactivity between the anti-p107 and anti-RB antibodies. Furthermore, characterization of the phosphorylated form of p107 was carried out using in vivo 32 P-phosphorylated lysates and immunoprecipitation with p107 antibodies in the presence of protein phosphatase 1A. In vivo 32 P-labeled p107 (Fig. 5B, lanes 1 and  2) was dephosphorylated by protein phosphatase 1A treatment (Fig. 5B, lanes 3 and 4).
Mechanism of Cell Cycle Inhibition by TGF-␤1-The finding that TGF-␤1 inhibits 32D-123 cell proliferation by regulating p107 phosphorylation led us to consider other possible mechanisms by which p107 functions in the TGF-␤1 inhibitory pathway. Subsequent experiments involved the overexpression of p107 using a p107 expression vector (mp107) and a p107 antisense expression vector. Specifically, the expression vectors used include mp107, pMAM-p107 AS (a construction that express antisense RNAs complementary to 500 nucleotides at the 5Ј end of the murine p107 mRNA, including the first methionine ATG site), and pMAM-p107 S (a control construction that expresses sense RNA). TGF-␤1 inhibited 32D-123 mock transfected cell proliferation to 52.1% (Table III). Overexpression (transfection) of mp107 inhibited 32D-123 cell proliferation to 35.8%, and TGF-␤1 treatment further decreased proliferation to 0.29% compared with untreated mock transfected cells (Table III). As a control, overexpression of pmRB lead to 58.1% proliferation of 32D-123 cells (data not shown). These results show mp107 expression to regulate negatively 32D-123 cell proliferation and TGF-␤1 to signal the activation of p107. As expected, RB also mediated negative regulation of myeloid cell growth.
Expression (transfection) of pMAM-p107 AS and pMAM-p107 S had little effect on 32D-123 cell proliferation (89.8 and 101.9% proliferation, respectively) compared with mock transfected cells (Table III). In addition, TGF-␤1 treatment of pMAM-p107 AS transfected cells did not inhibit cell proliferation compared with untreated pMAM-p107 AS transfected cells (82.6 and 89.8% proliferation, respectively, Table III). However, TGF-␤1 treatment of pMAM-p107 S transfected cells resulted in an inhibition of proliferation similar to that of the control (Table III). The ability of antisense to p107 mRNA to reverse the TGF-␤1 inhibitory effect is evidence that p107 is a mediator of TGF-␤1 inhibition of myeloid cell proliferation, suggesting a novel mechanism that requires p107 regulation for TGF-␤1 function.
Exogenous p107 Functions in the Repression of E2F Transactivation Activity-E2F transcriptional activity is regulated in a cell cycle-dependent manner by RB and E2F (14,53). Next, we investigated the ability of p107 to repress the E2F ability to transactivate a heterologous promoter G1B by using the GAL4-E2F fusion activator that contains the DNA binding domain of the yeast transactivator GAL4 fused to the entire E2F protein.
These constructions were transiently transfected together with a reporter construction containing one GAL4 binding site upstream of the E1b TATA box driving the CAT gene (41) into the 32D-123 cells (Fig. 6). Activation of the GAL4-E2F construction occurs in the 32D-123 cells and requires the E2F moiety of the GAL4-E2F fusion protein and the presence of the GAL4 binding site in the promoter of the reporter gene (Fig. 6). Cotrans-  (6 h, lane 2), late G 1 (10 h, lanes  3 and 4), and G 1 /S (14 h, lanes 5 and 6) were separated by SDS-PAGE (7.5%), transferred to Immobilon membrane, and probed with antibodies to RB or p107. Panel C, analysis of histone H1 kinase activity associated with RB (lanes 1 and 2) and p107 (lanes 3 and 4) was carried out using anti-RB and p107 immune complexes prepared from 32D-123 cells treated with TGF-␤1. The complexes were incubated under the appropriate conditions in the presence of [␥-32 P]ATP and histone H1, followed by SDS-PAGE and autoradiography. Data are representative of three experiments.

TABLE III
Proliferation assay of transfected 32D-123 cells Cells were transfected with the indicated plasmids and subjected to proliferation assay in the absence (Ϫ) and presence (ϩ) of TGF-␤1. The data are expressed as the percentage of maximal proliferation as measured by the MTT assay (the absorbance at 570 nm for untreated control transfected cells was 0.381) from a representative experiment (n ϭ 3). Analysis of variance followed by Duncan's multiple comparison test showed that transfection with p107 expression vector significantly inhibited 32D-123 cell proliferation (p Ͻ 0.005).

TGF-␤1 Inhibits Cell Cycle Progression via p107
fection experiments of the p107 expression vectors with the above described constructions consistently repressed the activity driven by the GAL4-E2F construction by 4-fold. This level of repression was greater than that observed for the RB expression vector. These results indicate that p107 is a potent inhibitor of E2F-mediated transcription in 32D-123 myeloid cells. TGF-␤1 Signals p107/E2F Binding Correlation with Growth Inhibition-Regulation of the c-myc gene has been shown to occur at the level of transcription, when RB binds E2F and inhibits the ability of E2F to transactivate c-myc expression (54,55). To study whether p107 can also regulate the signaling of c-myc in TGF-␤1-treated cells, we investigated the ability of p107 to associate with E2F, a transactivator of c-myc expression. We used IL-3-deprived cell lysates treated in the presence and absence of TGF-␤1 for 14 h. Lysates were immunoprecipitated with an anti-E2F antibody and were analyzed by Western blotting with an anti-p107 antibody. E2F bound to p107 was found to be increased in the TGF-␤1-treated lysates (Fig. 7A), suggesting that binding of E2F by p107 results in the inactivation of E2F, as has been reported previously for RB (14,21,53). This inactivation may, in turn, lead to the down-regulation of c-myc expression. In support of this observation, when we performed Western analysis of the lysates with an anti-c-Myc antibody, our analysis indicated down-regulation of the different molecular weight forms of c-Myc in the TGF-␤1-treated cells (Fig. 7B). Furthermore, using the anti-c-Myc antibody and the corresponding c-Myc peptide, the p65-p68 c-Myc proteins were identified by competition studies (data not shown). We also investigated p107 regulation at the level of c-Myc protein function by immunoprecipitating the cell lysates with an anti-p107 antibody and then analyzing c-myc expression by Western blotting with an anti-c-Myc antibody. The p107-bound c-Myc protein was increased in the lysates from the TGF-␤1-treated cells (Fig. 7C). Thus, even though low levels of c-Myc can be detected by Western analysis in lysates from TGF-␤1-treated myeloid cells, an increase in the amount of c-Myc protein bound to p107 is observed in the same lysates. These results further suggest a new mechanism for TGF-␤1 inhibition of cell proliferation which occurs via regulation of the p107 gene product.
Our results also suggest that during TGF-␤1 growth inhibition of 32D-123 cells, c-myc transcription and function are inhibited as a result of 1) p107 binding to E2F, and 2) p107 binding to c-Myc protein, respectively.

DISCUSSION
Previous studies have indicated that TGF-␤1 inhibits cell proliferation by affecting cell cycle progression (10,32,34,56). As in the case of fibroblasts and lymphocytes, we found TGF-␤1 to suppress myeloid cell growth in the G 1 phase, therefore our studies focused on the molecular mechanisms affected by TGF-␤1 during the G 1 and G 1 /S transition of the myeloid cell cycle. In 32D-123 murine cells and in human TF-1 (data not shown) TGF-␤1 markedly inhibits the activity of the cyclin E-cdk2 complex by increasing both the phosphorylation of cyclin E and the dephosphorylation of cdk2. As a result of the decrease in cyclin E-cdk2 kinase activity by TGF-␤1, the p107 protein but not the RB protein is dephosphorylated, and the binding of p107 to E2F and c-Myc proteins is increased, thus regulating c-Myc protein expression and function. The TGF-␤1 modulation of p107 function was confirmed further when transient overexpression of p107 inhibited the proliferation of the myeloid cells, and expression of antisense oligodeoxynucleotides to p107 mRNA blocked TGF-␤1 inhibition of myeloid cell growth.
First, we examined the effect of TGF-␤1 on the cyclins and cdks in the G 1 phase. Recent observations of fibroblasts have shown that TGF-␤1 inhibits the association of cyclin E-cdk2 complexes (30). This study prompted us to examine the regulation of cyclins and cdk2 by TGF-␤1 in myeloid cells. Results indicated that TGF-␤1 targets the regulation of the G 1 cyclincdk complex, cyclin E-cdk2. In 32D-123 cells, TGF-␤1 treatment increased the levels of the 48-kDa form and decreased the levels of the 45-kDa form of cyclin E and its associated cdk2 partner, resulting in a modest decrease in cyclin E-cdk2 complex formation. The immunoprecipitation studies suggest that the 45-kDa form of cyclin E has the highest affinity for cdk2, whereas the 48-kDa form has lower affinity. It is not known whether these two forms of cyclin E have different functions.
Studies using in vivo 32 P labeling have demonstrated that markedly increased phosphorylation of both forms of cyclin E occurs in response to TGF-␤1 treatment. The kinase activity of cdk2, which is determined by its phosphorylation state, has been shown to vary during the different phases of the cell cycle (30,57). TGF-␤1 markedly decreased the active phosphorylated FIG. 6. p107 represses GAL4-E2F transactivation activity. 32D-123 cells were cotransfected with 10 g of G1B reporter plasmid and 10 g of GAL4-E2F, or 1, 5, and 10 g of p107 or RB expression vector or pBR322 control. CAT activity was measured 24 h later, and values were normalized as described under "Experimental Procedures." The CAT activity of cells transfected with GAL4-E2F and without the p107 expression vector was set equal to 1.0 (control). The relative CAT activity expressed here represents the averages from at least three individual transfections. Standard deviations were approximately 10% of the mean values obtained.  3 and 4). Immunoblots were probed with a polyclonal antibody to p107. The p107 control (lane 1; 150 g) obtained from the above described lysates was analyzed by Western blotting with the anti-p107 antibody (lane 1). Panel B, the TGF-␤1treated (lane 1) and untreated (lane 2) lysates described in Fig. 5 were analyzed by Western blotting with anti-c-Myc antibody and by immunoprecipitation (panel C) with an anti-p107 antibody followed by Western blotting with anti-c-Myc antibody.
form of cdk2 in 32D-123 cells. Previous studies have shown that TGF-␤1 prevents phosphorylation of p33 cdk2 in epithelial cells, thus maintaining cdk2 in its inactive form (30,58). In vitro analysis of cyclin E-cdk2 kinase activity in 32D-123 lysates from TGF-␤1-treated cells indicated a reduction in cyclin E-associated kinase activity compared with the untreated control. Thus, in 32D-123 cells, TGF-␤1 alters the phosphorylation state of cyclin E and cdk2, decreases the association of the complex, and decreases the kinase activity of the complexes formed. More recent studies have shown that stimulation of C3H fibroblastic cell growth by TGF-␤1 results in increased phosphorylation of cyclin E. 3 Therefore, the role of cyclin E phosphorylation and dephosphorylation during TGF-␤1 signal transduction is not clear. The phosphorylation of cyclin E could be an epiphenomenon and may not be involved directly in cell cycle control by TGF-␤1.
Our studies suggest that the cyclin E-cdk2 complex is less stable in TGF-␤1-treated cells as a result of the changes in the phosphorylation states of their components. The kinases involved in the phosphorylation of the cyclins, particularly cyclin E, have not yet been identified. Thus far only the tyrosine phosphatase 25 (cdc25) and protein phosphatases 1 and 2A have been implicated in the dephosphorylation of cdc2 (61)(62)(63). In 32D-123 cells, the down-regulation of cdk2 phosphorylation in TGF-␤1-treated cells results from increased phosphatase activity. TGF-␤1 treatment of the 32D-123 cells may render cdk2 intrinsically incompetent for cyclin E activation. This regulation may also occur via a post-translational modification or a block imposed by the binding of a third protein to cdk2, like the association that has been reported for cyclin D-cdk4 and the p27 protein (64).
In 32D-123 myeloid cells, TGF-␤1 did not significantly affect the phosphorylation state of RB during the G 1 /S phase. Other studies support this observation; for example, TGF-␤1 treatment of murine B-cell lymphomas did not affect the phosphorylation state of RB in growth-arrested cells (65), and overexpression of RB did not affect the inhibition of cell proliferation in cervical carcinoma cells (14,25). A more defined role for RB is found in proliferating Mv1Lu cells, where TGF-␤1 has been shown to prevent RB phosphorylation during G 1 , thus retaining RB in a hypophosphorylated active state that may suppress cell progression into the S phase (35,60,66). Studies by Clarke et al. (67) with RB Ϫ/Ϫ knockout mice indicated abnormal differentiation or a failure to differentiate, particularly in the hematopoietic and neuronal lineages (67)(68)(69), suggesting that the role of RB may be lineage-specific and may also be related to the differentiation state of these cells.
Because p107 has been shown to be phosphorylated by cyclin E-cdk2 in a manner similar to RB, we next investigated whether p107 was a target for TGF-␤1 regulation (27). TGF-␤1 was found to decrease the phosphorylation of p107 at the G 1 /S boundary. Furthermore, TGF-␤1 treatment decreased histone H1 kinase activity in p107 immunoprecipitates. p107 is known to be physically associated with cyclin E and cdk2 in a variety of human cell lines (17,70). Our studies suggest that decreased p107 phosphorylation in response to TGF-␤1 treatment is a result of decreased cyclin E-cdk2 activity.
Binding of the transcription factor E2F to the c-myc promoter is required for transcriptional activity to occur (71,72); this transcriptional activity is modulated by other proteins that bind E2F, such as RB and p107 (14,15,17). In the 32D-123 cells, p107 was found to be associated with E2F, and TGF-␤1 treatment increased the E2F-associated p107 protein, resulting in decreased c-Myc protein expression. In support of this, transient transfection assays with NIH-3T3 cells have indicated that p107 inhibits E2F-dependent transcription, and free E2F accumulates as cells leave G 1 and enter S phase (15,18). In addition, recent studies have identified the growth-inhibitory block of TGF-␤ to be located close to the G 1 /S border in the cell cycle in human embryonic lung fibroblasts (73).
The functional relevance of p107 mediating TGF-␤1 regulatory effects in the 32D-123 cells was established in several ways. Overexpression of p107 led to suppression of cell proliferation, presumably by blocking cell cycle progression. In addition, by using GAL4-E2F chimeric constructions, p107 was shown to inhibit directly E2F transactivation activity. More recent studies have identified the regulation of the c-Myc protein function by the binding of p107 to the c-Myc transactivation domain. The binding of c-Myc by p107 resulted in inhibition of c-Myc transactivation activity (26). In our studies, the steady-state levels of c-Myc in 32D-123 cells were significantly decreased by TGF-␤1 treatment, whereas the levels of c-Myc bound to p107 were increased. Moreover, the ability of the p107 antisense to reverse the TGF-␤1 inhibition of 32D-123 cell proliferation confirms the direct role of p107 as a mediator of TGF-␤1 function.
Our studies indicate that some of the inhibitory effects of TGF-␤1 on cell cycle progression in myeloid cells occur via regulation of the p107 protein. p107 was demonstrated to be important for inhibition of cell proliferation and TGF-␤1-induced p107 activation (dephosphorylation), allowing the p107-E2F complexes to form. TGF-␤1 also modulated cyclin E and cdk2 at the post-translational level (phosphorylation). As a result, the G 1 /S cyclin E-cdk2 transitional complex activity was decreased, p107 was activated, and cell cycle progression was blocked. These studies provide a greater understanding of the molecular mechanisms by which TGF-␤1 regulates cell cycle progression in myeloid cells.