Constitutive overexpression of CDK2 inhibits neuronal differentiation of rat pheochromocytoma PC12 cells.

Changes in the levels of cyclins A, D, and E, p21, and cyclin-dependent kinase 2 (CDK2) were examined in rat pheochromocytoma PC12 cells during neuronal differentiation induced by nerve growth factor (NGF). Expression of cyclin A decreased to an undetectable level after 5 days of exposure to NGF, while expression of CDK2 decreased gradually after day 3. In contrast, the levels of cyclins D1 and E increased gradually through day 10, yet the amount of cyclin E associated with CDK2 decreased concomitant with a decrease in the CDK2 protein level. p21 expression increased gradually after day 7, while the level of CDK2-associated p21 remained unchanged. When human cDNAs encoding cyclins and CDK2 were introduced into PC12 cells, only CDK2 overexpression inhibited NGF-induced differentiation. The cell lines overexpressing CDK2 showed stable and high levels of CDK2 kinase activity during differentiation, whereas parental and vector-transfected cell lines displayed a marked decline in CDK2 kinase activity 1 day after NGF treatment. In cell lines overexpressing cyclins A, D, and E, this reduction of the kinase activity was not apparent until day 3. These results suggest that down-regulation of CDK2 activity is a crucial event for the neuronal differentiation of PC12 cells.

The commitment to cellular differentiation is a highly controlled stochastic process consisting of successive steps that require specific signals for survival and simultaneous loss of proliferative potential. The determination as to whether cells continue to proliferate or differentiate appears to be executed during the G 1 phase of the cell cycle when several types of G 1 cyclins and cyclin-dependent kinases (CDKs) 1 interact in various combinations (1)(2)(3)(4)(5)(6)(7). The D-type cyclins assemble primarily with CDK4, whose activity is detectable at mid-G 1 and which increases as cells approach the G 1 -S boundary (8 -10). Cyclin E associates with CDK2 and induces maximal levels of kinase activity at the G 1 -S transition (11)(12)(13). In addition, cyclin A expression peaks at the G 1 -S boundary and accumulates in early S phase, activating both CDK2 and Cdc2 (14 -16).
Cyclin-CDK complexes are believed to play an essential role in the G 1 -S transition. Constitutive ectopic expression of cyclin D or E in normal fibroblasts has been reported to shorten G 1 and reduce the dependence of cells on growth factors (17)(18)(19)(20). In addition, microinjection of antibodies against cyclin D1 during G 1 phase prevents cells from entering S phase (17). Inhibition of the function of CDK2 has also been reported to prevent the entry of cells into the S phase (21)(22)(23). The fact that cyclin-CDKs play a crucial role in the G 1 phase implies that the regulation of their functions is also critical for the commitment to cell differentiation. Indeed, ectopic expression of D-type cyclins and CDK4 was reported to inhibit granulocyte colonystimulating factor-induced differentiation of murine myeloid cells and erythroleukemia cells, respectively (24,25).
In the present study, we investigated the roles of cyclins and CDK2 in the process of the neuronal differentiation of PC12 cells. In addition, we examined the expression of p21, a negative regulator of several cyclin-CDK complexes, including cyclin D-CDK4, cyclin E-CDK2, and cyclin A-CDK2 (26 -28). Rat pheochromocytoma PC12 cells have proven to be a good model for neuronal differentiation due to their characteristic responsiveness to NGF. This response consists of partial growth arrest and acquisition of phenotypic properties typical of sympathetic ganglion, such as prominent neurite outgrowth (29). Herein, we demonstrated that suppression of CDK2 activity is a critical step in the NGF-induced differentiation of PC12 cells.

EXPERIMENTAL PROCEDURES
Cell Lines-PC12 cells were grown on plastic dishes without collagen coating in Dulbecco' modified Eagle's medium, containing 10% heatinactivated horse serum and 5% fetal bovine serum. For NGF treatment, cells were cultured on collagen-coated plastic dishes in Dulbecco' modified Eagle's medium supplemented with 2% horse serum and 1% fetal bovine serum. NGF purified from mouse submaxillary glands was purchased from Wako Ltd. Cells that have one or more neurites with the length more than twice the diameter of the cell body were defined as "differentiated" according to the previously described criteria (30). Alternatively, cells that have neurites with a length of more than three times the diameter of the cell body were defined as "well differentiated." The number of such "differentiated" cells per 500 cells was counted and used to compare the extent of differentiation among different samples.
Plasmids and Transfection-cDNA of human cyclin A was subcloned into pMKITneo (gift from Dr. Maruyama, Tokyo Medical and Dental Universtity), and cDNAs of cyclins D1 and E, CDK2, and p21 were subcloned into the pME18 vector. All of these cDNAs are under the control of the SR␣ promotor and were used for transfection into PC12 cells. PC12 cells (6 ϫ 10 5 cells) were transfected with 5 g of cDNAs along with (cyclin D1, cyclin E, CDK2, and p21) or without (cyclin A) 0.5 g of the neomycin-resistant gene, pSV2neo, by use of Lipofectamine (Life Technologies, Inc.) followed by selection for G418 (Life Technologies, Inc.) resistance (500 g/ml). Three weeks after transfection, 50 clones were picked up for each cDNA transfectant and were maintained in medium containing 100 g/ml G418. For each cDNA-transfectant, three clones showing the highest expression were selected by immunoblotting analysis and used for further experiments.
Antibodies-Polyclonal antibodies against cyclin A and cyclin E were obtained from Upstate Biochemicals Incorporated. Polyclonal antibodies against cyclin D1, CDK2, and p21 were prepared by immunizing rabbits with synthetic peptides corresponding to the 15, 12, and 17 carboxyl-terminal amino acid residues, respectively, of these proteins.
In Vitro Kinase Reactions-Lysates of each cell line (500 g of protein) were prepared as described above and were subjected to immunoprecipitation with anti-CDK2 antibody diluted 1:150 in a total volume of 300 l. A bacterially expressed fragment of the RB protein (amino acids 385-928) fused to glutathione S-transferase was purified by affinity chromatography on glutathione-Sepharose and was used as a substrate (0.5 g of protein) in 50 l of kinase reaction buffer containing 50 mM Tris-HCl (pH 7.2), 10 mM MgCl 2 , 1 mM dithiothreitol, 20 mM [␥-32 P]ATP (5 Ci; 1 Ci ϭ 37 kBq) (31). After incubation for 10 min at 25°C, the sample was analyzed by SDS-polyacrylamide gel electrophoresis followed by autoradiography.
Densitometric Analysis-Densitometric quantification of the data obtained by immunoblotting analysis (nitrocellulose filters) and in vitro kinase assay (x-ray films) was done with a Dual wavelength flying spot scanner (Shimadzu Ltd.). Since more than one band was detected in immunoblotting analysis of cyclin E and CDK2, the most slowly migrating band was subjected to quantification.

Expression of Endogenous Cyclins, CDK2, and p21 during
Differentiation of PC12 Cells-PC12 cells were cultured in medium supplemented with 2% horse serum, 1% fetal bovine serum, and 20 ng/ml NGF for 10 days. At days 0, 1, 3, 7, and 10, were prepared from cells that had been cultured in medium containing 20 ng/ml of NGF and harvested at the indicated times and were subjected to immunoblotting analysis to detect cyclins A, D1, and E, CDK2, and p21. b, histograms were generated by quantitating the intensity of the bands on nitrocellulose filters in panel a with a dual wavelength flying spot scanner. All values are indicated by ratios relative to those obtained on day 0. In the case of cyclin E and CDK2, the most slowly migrating bands were subjected to quantification. cells were lysed, and the expression of cyclins A, D1, and E, CDK2, and p21 was evaluated by immunoblotting analysis using specific antibodies. As shown in Fig. 1, expression of cyclin A was dramatically suppressed to an undetectable level after day 5. The expression of CDK2 showed a gradual but significant decrease to 21% of the initial level after NGF treatment for 7 days. In contrast, the expression of cyclins D1 and E increased gradually, reaching 4-and 6-fold higher levels, respectively, through the 10 days of observation. The level of p21 remained unchanged for at least 3 days and increased to 1.6fold thereafter.
Introduction of the cDNAs Encoding Human Cyclins, CDK2, and p21-To evaluate the significance of the changes in the expression of cyclins, CDK2, and p21 described above, we asked whether forced expression of these molecules alters the sensitivity of PC12 cells to NGF-induced neuronal differentiation. Expression constructs containing the cDNAs of the human cyclins A, D1, and E as well as CDK2 and p21 were transfected separately into PC12 cells, and clonally derived cell lines were established. Immunoblotting analysis of the cloned cell lines showed that the bands corresponding to the ectopically overexpressed gene products of cyclins A and D, CDK2, and p21 were almost identical in size to the endogenous gene products (Fig.  2). The levels of the overexpressed proteins were approximately 5-, 5.5-, 12-, and 20-fold higher, respectively, than that of the endogenous proteins. As reported previously with Rat-1 cells (19,20,33), ectopically expressed human cyclin E could be detected as a protein of 50 kDa, while the endogenous rat cyclin E was detected as three more slowly migrating bands (Fig. 2), consistent with the fact that the human and rat cyclin E genes encode proteins of 395 and 491 amino acids, respectively (11,34). The level of the exogenous 50-kDa cyclin E was 3-fold higher than that of the endogenous 55-kDa cyclin E. The 50-kDa human cyclin E species has been shown to be competent to regulate cell cycle progression as well as CDK2 kinase activity (19,20).

Effects of NGF Stimulation on the Growth of PC12 Cells-
NGF-induced differentiation of PC12 cells is known to be accompanied by a reduction in the growth rate (29). Thus, the effects of overexpressing cyclins A, D, and E, CDK2, and p21 on the growth of PC12 cells were examined by culturing these cell lines in medium supplemented with 10% horse serum and 5% fetal bovine serum, in the absence or presence of NGF (20 ng/ml). In the absence of NGF, all of the cell lines exhibited similar growth rates (Fig. 3a). After 3 days of culture in the presence of NGF, the growth of cell lines overexpressing cyclin A (clone A26), cyclin D1 (clone D3), cyclin E (clone E4), and p21 (clone P7) as well as the parental and vector-transfected cells declined significantly. However, in striking contrast to these cells, cell lines overexpressing CDK2 (clones K7, K15, and K20) did not show a significant reduction in cell growth for at least 5 days (Fig. 3b).
Morphological Changes Associated with NGF-induced Differentiation of PC12 Cells-All of the established cell lines showed morphologies very similar to the parental PC12 cells in the absence of NGF; cells were round or polygonal in shape and were loosely adherent on the dish (Fig. 4a). When the parental cells and vector-transfected cells were stimulated with NGF, approximately 40% of the cells displayed prominent neurite outgrowth at day 1 (Fig. 4b). Cyclin A-, D, and E-expressing cells began to show neurites from day 3 (Fig. 4c). By contrast, all of the three independent CDK2-expressing cell lines did not show significant neurite outgrowth through the 7 days of observation (Fig. 4d). The neurites developed in cyclin A-, D-, and E-expressing cells at day 3 were still shorter than those in parental cells and vector-transfected cells at day 1. For example, neurite outgrowth was observed in 34% of cyclin A-expressing cells at day 3 (Fig. 4e). Furthermore, only 12% of cyclin A-expressing cells developed neurites having a length more than 3 times the diameter of the cell body, which we defined as "well differentiated" (Fig. 4f). By contrast, parental cells and vector-transfected cells developed neurites in about 60% of the entire population at day 3, and 40% were well differentiated (Fig. 4, e and f).
Changes in CDK2 Kinase Activity-We next examined the changes in CDK2 kinase activity following immunoprecipita-tion from the lysates of the cell lines during NGF-induced differentiation. CDK2 immunoprecipitates were assayed for their ability to phosphorylate a fragment of bacterially produced pRB in the presence of [␥-32 P]ATP. As shown in Fig. 5, a rapid decline in kinase activity to about 55% of starting levels at day 1 and to 2% at day 3 was detected in parental cells and a vector-transfected cell line (clone neo5). In cell lines overexpressing cyclins A, D1, and E and p21, an apparent decrease in the activity of CDK2 to 75% and 15% of the initial levels was detected at day 3 and day 7, respectively ( Fig. 5b; only the data obtained with cyclin A-expressing cells are shown). However, CDK2 kinase activity in all of the cell lines overexpressing CDK2 (clones K7, K15, K20) remained high throughout the 10 days of observation.
Immunoblotting Analysis of CDK-associated Cyclin A, Cyclin E, and p21-To evaluate the levels of cyclin A, cyclin E, and p21 associated with CDK, we subjected p13 suc1 precipitates and CDK2 immunoprecipitates prepared from NGFstimulated cells to immunoblotting analysis using antibodies against cyclins A and E and p21, respectively. As shown in Fig. 6, the amount of cyclin A associated with CDK that bound to p13 suc1 -Sepharose decreased after day 3, in parallel with the previously observed levels of cyclin A expression (Fig. 1). However, the levels of CDK-associated cyclin E gradually decreased to 60% of the initial level at day 7, despite the fact that the level of cyclin E expression gradually increased during the time course (Fig. 6; compare to Fig. 1). The level of p21 associated with CDK2 was minimally affected by the presence of NGF. DISCUSSION In the present study, we examined alterations in the expression levels of cyclins, CDK2, and p21 during the NGF-induced neuronal differentiation of PC12 cells. Furthermore, we examined the effects of overexpression of these cell cycle regulators on the differentiation of PC12 cells. The results obtained from these experiments suggest that CDK2 is a key regulator of neuronal differentiation. The activity of CDK2 dramatically decreased following the addition of NGF, and constitutive overexpression of CDK2, but not of any cyclins tested, significantly blocked differentiation of PC12 cells. Ectopically expressed CDK2 probably exerted its effect by forming complexes with endogenous cyclins E and A and possibly with cyclins D2 and D3, i.e. those cyclins whose expression may be high enough to interact with CDK2 (2,8,32,35). On the other hand, overexpression of cyclins A, D1, and E had a weaker effect. These findings are different from those previously reported using myeloid and erythroleukemia cell lines. Ectopic overexpression of cyclins D2 and D3 was reported to have inhibited granulocyte colony-stimulating factor-induced differentiation of murine myeloid cells. The mechanism of these results was attrib- FIG. 5. Kinase activities associated with CDK2. a, CDK2-associated kinase activity in parental PC12 cells and derived cell lines expressing cyclins A, D1, and E, CDK2, and p21 during NGF treatment (20 ng/ml) are shown. Lysates prepared at the indicated times were immunoprecipitated with anti-CDK2 antibody. The immunocomplex was assayed for kinase activity using GST-RB fusion protein as a substrate. Since the changes in kinase activity of the cell lines overexpressing cyclins A, D1, and E and p21 were quite similar, only the results of a representative cyclin A-expressing cell line are shown. b, The histograms were generated by quantitating the intensity of radioactive signals on the x-ray films in panel a with a dual wavelength flying spot scanner. All values are indicated by ratios relative to those obtained on day 0. uted to the interaction of ectopically expressed cyclin D2 or D3 with CDK2, which was persistently expressed in that cell line (24). Kiyokawa et al. (25) determined that the suppression of CDK4 expression is a critical event in the pathway of terminal differentiation of the erythroleukemia cell line MEL. In addition, Jahn et al. (36) reported that CDK2 activity does not change during differentiation of the mouse skeletal myogenic cell line C2C12. These differences may be due to cell typespecific molecular mechanisms of cell cycle modulation and differentiation.
Although the levels of CDK2 and cyclin A were found to be suppressed during NGF-induced differentiation, neurite outgrowth apparently preceded this decrease. Hence, suppression of the expression levels of CDK2 or cyclin A may not be of primary importance in the induction of differentiation.
In contrast to CDK2 and cyclin A, the levels of cyclins D1 and E were found to increase gradually during differentiation of PC12 cells. Although overexpression of G 1 cyclins has been reported to accelerate G 1 phase in several cultured cell lines (17)(18)(19)(20), the increase in cyclin D1 and E expression observed at the later stages of differentiation of PC12 cells may not be involved in the acceleration of the G 1 /S transition; indeed, our experiments using p13 suc1 precipitates showed a gradual decrease in the amount of cyclin E associated with CDK. Similar up-regulation of cyclins has been reported to occur during the differentiation of various cell lineages. For example, the expression of cyclin D1 has been reported to increase in neurons at the onset of rat brain maturation as well as during differentiation of PC12 h cells (37,38). During 12-O-tetradecanoylphorbol-13-acetate-induced differentiation of human promyelocytic leukemia HL60 cells into macrophage-like cells, cyclin D1 expression is also up-regulated, although its expression is down-regulated in Me 2 SO-induced granulocytic cells (39). The expression of cyclin D3 is also known to increase during differentiation of mouse erythroleukemia MEL cells (25) and rat myoblast L6 cells (40). In the latter case, the kinase activity of the cyclin D3 complex, which presumably includes CDK2 and CDK4, has been shown to be markedly suppressed in the differentiated myotubes. A similar phenomenon was also observed in senescent human diploid fibroblasts (41). Cyclin Eassociated CDK2 activity is very low in senescent cells, although the amounts of cyclins D1 and E are 10 -15-fold higher than observed in quiescent early passage cells. Taken together, these findings raise the possibility that G 1 cyclins FIG. 6. Levels of CDK-associated cyclin A, cyclin E, and p21 during differentiation. a, lysates prepared from PC12 cells treated with NGF for the indicated time periods were precipitated with p13 suc1 (for cyclin A and cyclin E) or immunoprecipitated with anti-CDK2 antibody (for p21). The immunocomplex was subjected to immunoblotting analysis with antibodies against cyclin A, cyclin E, and p21. b, the histograms were generated by quantitating the intensity of the bands on nitrocellulose filters in panel a with a dual wavelength flying spot scanner. All values are indicated by ratios relative to those obtained on day 0. In the case of cyclin E, the most slowly migrating bands were subjected to quantification. may play some roles other than G 1 /S acceleration during cellular differentiation. For example, it has been observed that the G 1 cyclins can form complexes with pRB or p107, although the significance of these complexes is still unclear (2,(42)(43)(44).
Since the changes in the expression levels of CDK2 and cyclin E do not seem to be of primary importance in the downregulation of CDK2 activity during differentiation of PC12 cells, an alternative mechanism may involve CDK inhibitor proteins (6,7). However, expression of p21 was not significantly changed during differentiation of PC12 cells. Furthermore, overexpression of p21 neither induced differentiation nor growth retardation. Consistent with these results, no significant suppression of CDK2 kinase activity was observed in p21-overexpressing cells. One possible reason may be that the levels of ectopically expressed p21 in these established cell lines was not high enough to block the function of CDK2, although there was an approximately 20-fold increase in p21 level as compared with the parental cells. Presumably, cell lines overexpressing p21 at higher levels were not established due to its cell proliferation-inhibitory activity. In addition to p21, a growing number of CDK inhibitors, such as p16 INK4 , p15 INK4B , p27 KIP1 and p57 KIP2 , have recently been reported (6,7,45,46). Interestingly, the TGF-␤-mediated cell cycle block has been shown to involve p15 INK4B and p27 KIP1 (6,7,47). Additionally, it has been reported that a protein, SNT, that appears as a doublet of 78 -90 kDa in SDS-polyacrylamide gel electrophoresis gels and which coprecipitates with p13 suc1 -agarose is rapidly phosphorylated on tyrosine in neurons and PC12 cells treated with differentiation factors but not in those treated with mitogens (48). Since p13 suc1 associates with cyclin-CDK, this finding raises the possibility that SNT may link the differentiation signal mediated by receptor tyrosine kinases to the cell cycle regulator CDKs; i.e. SNT may act as a negative regulator of CDKs. Thus, the contribution of CDK inhibitors as well as of SNT to the NGF-induced inhibition of CDK2 activity remains to be elucidated in future studies.
Cyclin-CDK is believed to phosphorylate cellular proteins important for cell cycle regulation (6,7). One of the main targets of cyclin-CDK is pRB, which negatively regulates cell cycle progression through the G 1 phase (31,43). pRB also plays a crucial role in early neuronal and hematopoietic development as demonstrated by the analysis of mice carrying a targeted mutation in the RB gene (49 -51). The amount of underphosphorylated pRB, which is believed to be the active form of pRB, is increased by extracellular signals, which induce cell cycle arrest and differentiation (52,53). The results obtained in this study suggest that the NGF-induced reduction in CDK2 activity may be responsible for the accumulation of the underphosphorylated form of pRB during differentiation of PC12 cells. Accordingly, constitutive overexpression of CDK2 may block differentiation by driving the phosphorylation of pRB. Consistent with this notion, ectopic overexpression of adenovirus E1A, which associates with and inactivates pRB as well as p107, has been demonstrated to inhibit NGF-induced neuronal differentiation of PC12 cells (54 -56). Further detailed analysis of the function of pRB, as well as the identification of other CDK2 substrates, may prove important for our understanding of the precise mechanism by which PC12 cells differentiate.