Increased expression of the P27KIP1 protein in human esophageal ...

Carcinogenesis vol.18 no.6 pp.1139–1148, 1997

Increased expression of the P27KIP1 protein in human esophageal cancer cell lines that over-express cyclin D1

Yuichiro Doki, Masaya Imoto, Edward Kyu-Ho Han, Alessandro Sgambato and I.Bernard Weinstein1 Herbert Irving Comprehensive Cancer Center, Columbia University, College of Physicians and Surgeons, New York, NY 10032, USA 1To

whom correspondence should be addressed

In the present study we have characterized eight human esophageal squamous carcinoma cell lines for levels of expression of cyclins D1, E, A and B1; CDKs 1, 2 and 4; the CDK inhibitors p16INK4, p21WAF1 and p27KIP1; the retinoblastoma (Rb) protein; and in vitro CDK2- and CDK4-associated kinase activity; and also compared the growth properties of these cell lines. The level of the cyclin D1 protein varied by over 30-fold amongst the eight cell lines. The high level in two cell lines was associated with amplification of this gene, but in three cell lines it was due to post-transcriptional events. Amongst the eight cell lines there was a significant correlation between the levels of cyclin D1, Rb and p27KIP1 proteins, and CDK4-associated kinase activity. Furthermore, when an exogenous cyclin D1 cDNA was over-expressed in the EC109 cell line by transfection, this led to increased expression of both Rb and p27KIP1. There was, however, no correlation between the level of cyclin D1 expression and the cell doubling times, duration of the G1 phase, or colony-forming efficiency in agar. Two of the cell lines displayed a high level of the cyclin E protein, low levels of cyclin D1, lacked expression of the Rb protein and expressed high levels of the p16INK4 protein. One of these cell lines displayed amplification of the latter gene. There was no correlation between the levels of cyclins E or A and in vitro CDK2 kinase activity, but CDK2 kinase activity was inversely correlated with the duration of the G1 phase of the cell cycle. Taken together, these studies indicate marked heterogeneity in the expression of cell cycle-related proteins amongst a series of esophageal carcinoma cell lines. The correlation between the levels of the cyclin D1, Rb and p27Kip1 proteins suggest the existence of a homeostatic feedback loop between positive and negative acting components of the cell cycle machinery. Introduction In mammalian cells progression of the cell cycle is regulated by the interactions of cyclins, cyclin dependent kinases (CDKs*) and CDK inhibitors (CDKIs). At least 10 distinct cyclins, seven CDKs and seven CDKIs have been identified (1–4). Their functions have been analysed in detail especially with respect to G1 progression and the G1/S checkpoint. Cyclin D1 binds to and activates its major catalytic partners CDK4 and CDK6 (5,6) which can then phosphorylate the Rb tumor suppressor protein as well as two Rb-related *Abbreviations: CDK, cyclin-dependent kinase; CDKI, cyclin-dependent kinase inhibitor. © Oxford University Press

proteins p107 and p130 (for review see 1,2,4). Cyclin E binds to and activates CDK2, which also phosphorylates the Rb protein. The cyclin E/CDK2 complex can also phosphorylate the Rb-related proteins p107 and p130 in vitro (7,8), but it has not been established that this also occurs in vivo. These kinase activities are controlled negatively by the binding of the CDKI proteins p21WAF1 (9), p27KIP1 (10) and p57KIP2 (11) to cyclin D1/CDK4 and CDK6, and cyclin E/CDK2 complexes, and also by the CKDIs p15INK4 (12), p16INK4 (13) and p18INK4 (14) which bind directly to CDK4 and CDK6. In early S phase, the cyclin A/CDK2 complex shows maximal activity, and in the G2/M phase cyclin B/CDC2 (also called CDK1) and cyclin A/CDC2 complexes are active. The critical substrates for the cyclin A and cyclin B/CDK complexes have not yet been identified with certainty in mammalian cells (3). Ectopic over-expression of cyclins D1 and E in some cell cultures shortens the G1 phase, but curiously, does not shorten the doubling time of these cultures (15–18). Stable overexpression of cyclin D1 in some cell lines can also enhance cell growth and malignant cell transformation (15,16). Therefore, cyclin D1 can function as an oncogene. On the other hand, ectopic over-expression of cyclin D1 (19–21) or cyclin E (22) in some cell systems can inhibit cell cycle progression and suppress rather than stimulate cell growth suggesting that cell context can markedly influence the phenotypic effects of these cyclins. There is also evidence for the existence of feedback control loops that might influence the actions of cyclins D1 and E. For example, increased levels of Rb induce the expression of cyclin D1 (23,24) and cells that lack expression of Rb have elevated levels of the p16INK4 (25,26). Various external factors can also influence the actions of cyclins D1 and E. Thus, transforming growth factor β can induce decreased expression of CDK4 (27) and cyclins A and E (28), induce the expression of p15INK4 (12), and increase the inhibition by p27KIP1 of cyclin D- and cyclin E-CDK activities (12,29). Furthermore, DNA damage (30,31) and differentiation (32,33) can arrest the G1 to S progression through the action of p21WAF1 and other inhibitory proteins. There is accumulating evidence that the carcinogenic process frequently involves abnormalities in the expression of cyclins and other cell cycle-related genes (for review see 1). The earliest examples were mutations that inactivate the functions of the p53 and Rb tumor suppressor genes, which normally function as inhibitors at the G1/S checkpoint. Other quite frequent abnormalities are amplification and/or over-expression of the cyclin D1 gene (1) and deregulation in the expression of cyclin E (34). Our laboratory (35) and other investigators (36,37) previously reported that 30–50% of primary human squamous carcinomas of the esophagus display amplification and/or over-expression of the cyclin D1 gene. However, a detailed analysis of possible correlations between this abnormality and several other factors that control cell cycle progression had not been examined. Therefore, in the present study we analysed 1139

Y.Doki et al.

in parallel eight human esophageal cancer cell lines for levels of expression of cyclins D1, E, A and B1; CDK4, CDK2 and CDC2; CDKIs p16INK4, p21WAF1, and p27KIP1; and Rb. We also assessed extracts of these eight cell lines for various in vitro CDK-associated protein kinase activities, and compared the growth properties and cell cycle parameters of these cell lines. We found marked variations between these cell lines for some of these parameters but at the same time two striking correlations: (i) a positive correlation between the levels of cyclin D1, Rb, p27KIP1 and CDK4-associated kinase activity; and (ii) an inverse correlation between Rb and p16INK4 expression. These findings provide further evidence for the existence of feedback control loops that regulate G1/S progression in the cell cycle. The possible implications of these feedback loops with respect to the carcinogenic process are discussed. Materials and methods Cell culture and viral transduction Eight human esophageal cancer cell lines, TE-3, TT, TTn, TE-2 (provided by Dr H.Shiozaki, Osaka University, Japan) and HCE4, HCE7, EC17 and EC109 (provided by Dr C.Harris, NCI, Bethesda, MD) (35) were grown in RPMI 1640 medium supplemented with 5% fetal calf serum, and maintained in a 37°C incubator with 5% CO2. The construction of the cyclin D1 retroviral expression plasmid PMV7CCND1 and the methods used for virus packaging and transduction have been previously described (15,19). Virus supernatants containing the control PMV7 vector or the PMV7-CCND1 construct were used to infect EC109 cells. Cells were selected in the presence of 1 mg/ml of G418. The surviving cells were collected as a pool 3 weeks after infection, and then the vector control pool of cells (designated ‘EC109-V’) and the cyclin D1 over-expressor pool (designated ‘EC109-D1’) were further analysed (see Figure 6). The following antibodies were used in this study: polyclonal rabbit IgG antibodies against human cyclin A, cyclin E, cyclin D, CDK2 and CDK4, which were purchased from UBI (Lake Placid, NY). Monoclonal mouse IgG antibodies against human cyclin B1 and human Rb protein were purchased from Pharmingen (San Diego, CA). Polyclonal rabbit IgG antibodies against mouse p21WAF1, human p27KIP1 and human p16INK4 were from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal mouse IgG antibodies against human CDC2 were from Transduction Laboratories (Lexington, KY). A monoclonal mouse IgG antibody against human E-cadherin was provided by Dr H. Shiozaki (Osaka University, Japan). Cell extraction and Western blotting Four 3106 cells were seeded on 15-cm plastic dishes and 48 h later exponentially growing cells were collected. They were lysed in lysis buffer (50 mM HEPES, 150 mM NaCl, 2.5 mM EGTA, 1 mM EDTA, 1 mM DTT, 10 mM β-glycerophosphate, 1 mM NaF, 0.1 mM Na3VO4, 0.1% Tween, 10% glycerol, 1 mM PMSF, 10 µg/ml aprotinin, and 10 µg/ml leupeptin, adjusted to pH 7.5), then sonicated twice for 10 s on ice with a Sonifer Cell Disruptor. Protein concentration were determined with the Bio-Rad protein assay (BioRad, Hercules, CA). Fifty µg of protein from each cell line were resuspended in an equal volume of SDS–sample buffer (50 mM Tris, pH 6.8, 2% SDS, 5% 2-mercaptoethanol, 10% glycerol, 0.2 mg/ml bromophenol blue), boiled for 5 min and subjected to 10% SDS–PAGE (or a 6% gel for RB). The separated proteins were transferred to immobilon-P membranes (Millipore, Bedford, MA) and blocked with blocking buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.2% Tween 20, 3% bovine serum albumin) for 60 min at room temperature. The membranes were incubated with 1 µg/ml of the indicated antibodies for 1 h at room temperature, washed with washing buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.2% Tween 20) and incubated with a 1:5000 dilution of horseradish peroxidase conjugated anti-mouse or anti-rabbit IgG antibodies (Amersham, Arlington Heights, IL) for 1 h at room temperature. After a second washing, visualization was performed using the ECL Western detection system (Amersham). Southern and Northern blot analyses Genomic DNA was isolated from cells as described (19). Ten µg of DNA previously digested with EcoRI were electrophoresed on a 1% agarose gel and transferred to Hybond-N membranes (Amersham). Total RNAs from exponentially growing cells were prepared using RNAzol (Biotecx Laboratories, Houston, TX) following the company’s instructions. Briefly, cells grown on 15-cm dishes were washed twice with ice cold PBS and then collected with 2 ml of RNAzol. After mixing with 0.2 ml chloroform, the

1140

RNA was separated from the phenol–chloroform layer by centrifugation (12,000 g, 15 min). The extracted total RNA was precipitated with isopropanol, washed using 75% ethanol and resuspended in TE buffer (10 mM Tris, pH 8.0, 1 mM EDTA). Fifteen µg of total RNA were then electrophoresed in 1% agar–6% formaldehyde gels and transferred onto Hybond-N membranes (Amersham). Membranes containing either DNA or RNA were preincubated in Church buffer at 65°C and then hybridized with 32P-labeled probes to human cyclins D1, cyclin E, Rb, p16INK4 or p27KIP1. The membranes were washed with 23SSC containing 0.1% SDS for 20 min at room temperature, followed by washing for 20 min at 65°C with the same washing solution. Finally the membranes were exposed to Kodak XAR-5 film with an intensifying screen at –70°C. Doubling time and flow cytometry Subconfluent cultures of each cell line were trypsinized, counted and then seeded in six-well (35-mm diameter) plates at 2 3105 and 1 3105 per well. The culture media were changed every day, and after 24 and 72 h cell numbers were counted, and the doubling times were calculated. Cells were 60–80% confluent after 72 h. These experiments were repeated three times, with similar results. Similar exponentially growing cells were analysed by flow cytometry to determine their distributions in each phase of the cell cycle. The cells were trypsinized, washed twice in PBS and then fixed in 80% ethanol at 4°C for 1 h. After again washing twice with PBS, the fixed cells were resuspended in 0.2 mg/ml of propidium iodide containing 0.6% NP-40 and 1 mg/ml of RNAase A (Sigma, St. Louis, MO) and incubated for 1 h at room temperature in the dark. The cell suspensions were then filtered through 50 µm nylon mesh and analysed on a FACStar flow cytometer (Beckton Dickinson, San Jose, CA). The percentage of cells in different phases of the cell cycle was determined using a MODFit 5.2 computer program. Colony-forming assay in soft agar Growth in soft agar assays were performed as described (19). Four 3104 cells were resuspended in 1.0 ml of culture medium containing 0.33% agar and placed on a 2.5 ml bottom layer containing 0.5% agar, in six-well plates. After 14 days, colony formation was monitored by microscopy and the numbers of colonies whose diameter were larger than 0.05 mm or 0.1 mm were counted in at least 10 different fields. CDK and CDK inhibitor assays CDK assays were performed as previously described (38), with minor modifications. Twenty-five µg of a total cell lysate was precleared with 10 µl of 50% protein A sepharose (Sigma, St. Louis, MO) and then incubated with 1 µg of anti-CDK2 antibody (alternatively 2 µg of anti-cyclin A or 2 µg of anti-cyclin E antibody) for 1 h at 4°C. All of the following steps were done at 4°C, unless otherwise described. Immunocomplexes were precipitated with 15 µl of protein A sepharose, washed four times with washing buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 2.5 mM EGTA,1.0 mM EDTA, 1 mM DTT, 0.1% Tween 20) and twice with reaction buffer (50 mM HEPES, pH 7.5, 10 mM MgCl2, 1 mM DTT, 2.5 mM EGTA, 10 mM β-glycerophosphate, 1 mM NaF and 0.1 mM Na3VO4). Kinase reactions were performed in 50 µl reaction buffer containing 2 µg of H1 histone (Boehringer Mannheim, Germany) and 5 µCi of [γ-32P]ATP (3 Ci/mM) (Amersham) and incubated for 15 min at 30°C. Reactions were terminated by addition of an equal volume of SDS sample buffer. The kinase activity was assessed by autoradiography after SDS–PAGE. For the CDC2 kinase assay, 1 µg of anti-CDC2 was used instead of the CDK2 antibody, and this was followed by incubation with 1 µg of goat antimouse IgG for 1 h before incubation with protein A sepharose. For the CDK4 assay, 150 µg of total cell lysate was used and 2 µg of anti-CDK4 antibody or 2 µg of anti-cyclin D1 antibody was used. The kinase reaction was performed with 1 µg of GST-Rb fusion protein as the substrate (38), instead of histone H1, and with 10 µCi of ATP, for 1 h at 30°C. In order to evaluate heat stable CDK inhibitory activity, 25 µg of total cell lysate was boiled for 3 min, then clarified by centrifugation at 12,000 g for 15 min. The supernatant fraction was then mixed and incubated with 25 µg of untreated EC109 total cell lysate for 20 min at 30°C. CDK2immunoprecipitates were then prepared from this mixture and assayed for histone H1 kinase activity as described above. The ‘inhibitor activity’ for each cell line was calculated by subtracting the value of the CDK2-associated kinase activity obtained in the presence of the heat stable fraction from the value obtained in the absence of the heat stable fraction (i.e. the control). Statistical analysis The intensities of each of the bands on the Western, Northern and Southern blots, and on the SDS–PAGE gels used to assess kinase activity were quantitated by densitometric scanning using an Image Quant instrument (Molecular Dynamics, Sunnyvale, CA). Correlations between two parameters

Correlation between cyclins and associated proteins

Fig. 1. Western blot analysis of cell cycle related proteins in eight human esophageal squamous carcinoma cell lines. Fifty µg of total cell extracts were separated in 10% SDS–PAGE (or 6% for Rb) and the indicated proteins detected with the respective antibodies. The figure shows: cyclin A, 58 kD (arrow); cyclin B1, 61 kD; cyclin D1, 34 kD; multiple bands of cyclin E, 50, 45 and 42 kD (arrows); underphosphorylated Rb protein (pRb); 110 kD and phosphorylated Rb protein (pRb-p), 114 kD; Rb protein status in cells starved of serum for 48 h; CDK2, 33 kD and a phosphorylated form (threonine 160) (CDK2-p); CDK4, 34 kD; CDC2, 34 kD and a phosphorylated form (threonine 161) (CDC2-p); p16INK4, 16 kD; p21WAF1, 21 kD; and p27KIP1, 27 kD. For additional details see Materials and methods.

were evaluated by linear regression, and judged as statistically significant when the coefficient of determination (r2 value) was .0.696 (corresponding to P,0.01) (39).

Results Expression of various cyclins, CDKs and CDKIs at the protein level In this study we characterized eight human cancer cell lines. Our results indicated that, although all of these cell lines were originally derived from squamous carcinomas of the esophagus, they revealed markedly different levels of expression of cell cycle related genes, when examined by Western blot analysis (Figure 1). Thus, the 34 kD cyclin D1 protein was expressed at a very high level in exponentially dividing cultures of the TE3 cell line, at a moderate level in the TT, TTn, HCE4 and HCE7 cell lines, and at only a very low level in the TE2, EC17 and EC109 cell lines. On the other hand, 50 and 45 kD cyclin E protein bands were present at high levels in the TE2 and EC17 cell lines, even though these two cell lines had only a trace amount of cyclin D1. These cyclin E proteins were present at a moderate level in the TE-3, TT and HCE7 cell lines, and at a low level in the TTn, HCE4 and EC109 cell lines. Thus, the EC109 cell line expresses only trace amounts of both of the two G1 cyclins, cyclins D1 and E. None of the cell lines expressed very high levels of both cyclins D1 and E. All of the eight cell lines expressed appreciable amounts of a 58 kD cyclin A protein (the higher molecular weight band appears to be a cross reactive protein) and a 61 kD cyclin B1 protein (Figure 1). Variations between

the cell lines in the expression of cyclins A and B were less striking than the variations seen with cyclins D1 and E. Figure 1 indicates that the eight cell lines also differed considerably in their levels of expression of the following cell cycle related proteins: the Rb protein, which displayed a hypophosphorylated band at 110 kD and a hyperphosphosphorylated band at about ~114 kD; the 34 kD CDK4 protein; and the CDKIs p16INK4, p21WAF1 and p27Kip1. On the otherhand, they expressed fairly similar levels of an ~33 kD CDK2 protein doublet, the lower band presumably representing a more phosphorylated form. They also expressed fairly similar amounts of an ~34 kD CDC2 protein. A slightly slower, presumably more phosphorylated form of this protein, was present at variable levels amongst the eight cell lines (Figure 1). We also examined the level of the Rb protein after starving the cell lines of serum for 48 h. The total amounts of Rb were, in general, similar to those seen in exponentially dividing cultures. However, more of the underphosphorylated form was observed in the serum-starved cells (Figure 1). To evaluate possible positive or negative correlations between the levels of expression of the above proteins amongst the eight cell lines, the relative abundance of these proteins on the Western blots shown in Figure 1 was determined by densitometry. The significant correlations are shown graphically in Figure 2, and are based on simple linear regression analyses. We found that there was a strong positive correlation between the levels of the cyclin D1 and Rb protein (r2 5 0.969) and between cyclin D1 and p27KIP1 (r2 5 0.827). Although the two cell lines, TE-2 and EC17, that expressed high levels of cyclin E lacked Rb expression (Figure 1), 1141

Y.Doki et al.

amongst the eight cell lines there was not a significant negative correlation between the levels of expression of cyclin E and Rb (r2 5 0.120). It is of interest that only the TE-2 and EC17 cell lines expressed detectable, in fact, rather high levels of p16INK4 (Figure 1). There was a statistically significant positive correlation between the levels of cyclin E and p16INK4 expression, although this correlation is based on a very limited number of samples (r2 5 0.956). No significant correlations were seen between: the levels of expression of cyclin A and either cyclins D1 or E; cyclins D1 and E; CDC2, CDK2 and CDK4 with either cyclins D1, E, A or B; or p21WAF1 and any of the cyclins (data not shown). Southern and Northern blot analyses In view of the above described marked variations among the eight cell lines in the levels of expression of cyclins D1 and E and other cell cycle related proteins, we did Southern and Northern blot analyses to examine the underlying mechanisms. Southern blot analysis (Figure 3A) indicated that the cyclin D1 gene was amplified in four of the esophageal carcinoma cell lines, i.e. the cell lines TT, TTn, HCE4 and HCE7. Densitometric analysis indicated that the amplification ranged from ~10-fold in the TTn cells to ~5-fold in the HCE7 cells, when compared to normal human fibroblasts or the remaining four cell lines (data not shown). On the other hand, Northern blot analysis (Figure 3B) indicated that only the HCE4 and HCE7 cells demonstrated an appreciable increase in cyclin D1 mRNA when compared to the other cell lines. Furthermore, a comparison of Figures 1, 3A and 3B indicate that the level of expression of the cyclin D1 protein did not always correlate with cyclin D1 gene amplification or the level of expression of cyclin D1 mRNA. For example, although TE-3 cells had the highest level of cyclin D1 protein they did not display amplification of the cyclin D1 gene or a very high level of cyclin D1 mRNA. These results suggest that post-transcriptional events, possibly at the level of translational control, can play an important role in determining cellular levels of the cyclin D1 protein. None of the eight cell lines displayed evidence of amplification or gross deletions of the Rb gene, by Southern blot analysis (Figure 3A), although point mutations might exist in the Rb gene. The TE-2, EC17 and EC109 cell lines displayed diminished levels of Rb mRNA (Figure 3B), which correlated with an absence or a marked reduction of the Rb protein in these three cell lines (Figure 1). It is of interest that there was a significant positive correlation between the level of Rb mRNA and the level of cyclin D1 protein amongst the eight cell lines (r2 5 0.861) (Figures 1 and 3B). On the other hand, there was no significant correlation between the levels of cyclin D1 mRNA and Rb protein or mRNA. With respect to cyclin E, it is of interest that the EC17 cell line which expressed the highest level of the cyclin E protein (Figure 1) also displayed amplification of the corresponding gene (Figure 1A) and a relatively high level of the corresponding mRNA (Figure 3B). On the other hand, the TE-2 cell line which expressed a moderate increase in cyclin E protein and mRNA did not display amplification of this gene. With respect to the p16INK4 gene, two cell lines, TE-3 and EC109, showed homozygous deletion of this gene (Figure 3A), which explains the absence of the corresponding mRNA and protein in these cell lines (Figures 1 and 3B). In contrast, the EC17 cell line displayed ~3.5-fold amplification of the p16INK4 gene, and high levels of expression of the correspond1142

ing mRNA and protein. The TE-2 cell line did not show amplification of p16INK4, but displayed a relatively high level of the corresponding mRNA and protein. The remaining four cell lines, TT, TTn, HCE4 and HCE7 did not show any gross change in p16INK4 at the DNA level and expressed fairly abundant amounts of the corresponding mRNA, even though they did not express detectable amounts of this protein (Figures 1 and 3). None of the eight cell lines displayed amplification or gross deletions of the p27KIP1 gene. They expressed variable amounts of the corresponding mRNA which did not correlate with the relative abundance of this protein amongst these cell lines. Thus, the high level of p27KIP1 protein in the TE-3 cells appears to reflect a post- transcriptional effect. CDK activity and CDK inhibitory activity Since cyclins act by binding to and activating a series of CDKs we evaluated the in vitro kinase activities of the appropriate immunoprecipitates prepared from the eight cell lines (Figure 4). The major partner for cyclin D1 is CDK4 (38). When cyclin D1 or CDK4 immunoprecipitates were assayed for their ability to phosphorylate a Rb-GST fusion protein, both activities varied considerably amongst the eight cell lines and showed a significant correlation with each other (r2 5 0.722) and also with the level of expression of the cyclin D1 protein (r2 5 0.756), amongst the eight cell lines (Figure 2). CDK2-associated kinase activity was assayed using histone H1 as the in vitro substrate. The TE-3 cells had the highest activity and the other cell lines had fairly similar activities. Since CDK2 binds to and is activated by both cyclins E and A (2,3), we specifically assayed cyclin E- and cyclin Aassociated kinase activities. We found that the two cell lines that expressed the highest levels of the cyclin E protein, TE2 and EC17 (Figure 1), had high cyclin-E associated kinase activity, but had low cyclin-A associated kinase activity. Curiously, TE-3 cells which had high cyclin D1-CDK4 kinase activity also had both high cyclin E- and cyclin-A associated kinase activity. CDC2 binds to cyclins A and B1, and the TE-3 cell line also had high CDC2-associated histone H1 kinase activity. As expected, there was a positive correlation between CDC2-associated kinase activity and the level of cyclin B1 protein, but this correlation was not very strong (r2 5 0.610). However, there was a strong positive correlation between CDC2- and CDK2-associated kinase activity (r2 5 0.986). This correlation is of interest in view of a recent study indicating that the CDK2 kinase is a positive regulator of cyclin B/CDC2 kinase activity (40). The known CDKIs are heat stable (29,41), therefore we prepared heat denatured extracts from the eight cell lines and assessed them for their ability to inhibit CDK2-associated histone kinase activity. As a source of the kinase we used CDK2-immunoprecipitates from the EC109 cell line, since extracts of this cell line did not have detectable CDKIs (Figure 1) or CDK2-inhibitory activity (Figure 4). Marked inhibitory activity was seen with the heat denatured extract of the TE-3 cell line and variable amounts of inhibition were seen with extracts of the other cell lines. The extent of this inhibitory activity correlated positively with the level of expression of p27KIP1 (r2 5 0.747) but not with the level of expression of p21WAF1 or p16INK4, suggesting that the inhibitory activity reflected, at least in part, the p27KIP1 protein. It is of interest that, despite this, there was a positive correlation between this inhibitory activity, p27KIP1 expression and CDK2 kinase activity (r2 5 0.755).

Correlation between cyclins and associated proteins

Fig. 2. Correlations between levels of expression of various cell cycle-related proteins, in vitro CDK enzyme activities and cell growth. The levels of the various proteins and kinase activities were quantitated by densitometric analysis of the data shown in Figures 1 and 4. Simple linear correlations between two parameters were calculated. Correlations which had a coefficient of determination (r2) .0.696 were considered to be statistically significant and are shown in this figure. The values of both axes are arbitrary units, except for the duration of the G1 phase. ‘pRb-p’ represents the ratio of the phosphorylated form of Rb protein to the non-phosphorylated form of Rb protein.

Studies on cell growth and cell cycle parameters It was of interest to determine whether any of the above mentioned parameters correlated with the growth properties or cell cycle parameters of the respective cell lines. The exponential doubling times of the eight cell lines when grown in monolayer culture in complete growth medium varied from ~20 to 35 h and their total colony-forming efficiencies in agar varied from ~4 to 28% (Figure 5). However, neither of these growth properties correlated with the levels of expression of Rb or the various cyclins, CDKs, CDKIs, or CDK kinase

activities. The results of cell cycle analyses on exponentially dividing cultures are also shown in Figure 5. The length of the G1 phase varied from ~8 to 15 h amongst the eight cell lines and there was a significant negative correlation with CDK2 kinase activity (r2 5 0.731) (Figure 2), but there was no significant correlation with CDK4 activity (r250.226) or other parameters. No significant correlation was seen between the duration of the S or G2/M phases and CDK activities or other parameters. It is of interest that although the TE-3 cell line had the highest level of cyclin D1 protein and cyclin D11143

Y.Doki et al.

CDK4-associated kinase activity, and relatively high levels of cyclin E-associated kinase activity, its growth properties and cell cycle parameters when grown in complete medium containing 5% fetal calf serum, were not remarkably different from those of the other seven cell lines (Figure 5). It is of interest, however, that the TE-3 cell line can grow in medium lacking serum (42). All of the above-described studies on the expression of cyclins and related genes (Figures 1–5) were done on exponentially dividing cells. Since, for the most part, we did not see a simple correlation between the doubling times or cell cycle parameters (Figure 5) and the levels of expression of these genes, the differences in gene expression observed between these cell lines is not simply due to differences in the durations of individual phases of the cell cycle. Effects of ectopic over-expression of cyclin D1 One of the most striking correlations in the above studies was the positive correlation between the levels of expression of the cyclin D1, Rb and p27KIP1 proteins (Figures 1 and 2). These findings raised the possibility that high expression of the cyclin D1 protein might enhance the expression of the latter two proteins. Therefore, we used a previously described retrovirus vector containing the human cyclin D1 cDNA sequence to develop derivatives of the EC109 cell line that stably over-expressed relatively high levels of the exogenous cyclin D1. This cell line was chosen since it initially displayed low levels of the cyclin D1, Rb, p27KIP1 and cyclin E proteins (Figure 1). As shown in Figure 6, EC109-D1 cells, which expressed ~6-fold increase in cyclin D1 protein when compared to a vector control EC109 cells, displayed a marked increase in the levels of the Rb and p27KIP1 proteins. On the other hand, the level of expression of E-cadherin, used as a control, was not altered (Figure 6) and there was only a slight increase in the level of cyclin A (data not shown) in the EC109-D1 cells. Discussion In the present study we found that five of eight human esophageal squamous carcinoma cell lines expressed relatively high levels of the cyclin D1 protein and four of these cell lines had a 5- to 10-fold amplification of the cyclin D1 gene (Figures 1 and 3). Parallel Southern, Northern and Western blots on the eight cell lines indicated, however, that there was not a consistent correlation between the levels of expression of the cyclin D1 protein and mRNA, or the extent of cyclin D1 gene amplification (Figures 1 and 3). Similar findings in a different series of esophageal squamous carcinoma cell lines were reported by Nakagawa et al. (37). Thus, it appears that in some cell lines the increased expression of cyclin D1 is due, at least in part, to alterations at the level of translation or protein stability, but the underlying mechanisms are not known. It is of interest that ~40% of primary human breast (1,43,44) and colon carcinomas express increased levels of cyclin D1 (45,46) in the absence of gene amplification (45–47). It is known that in some cell types specific growth factors can stimulate the expression of cyclin D1 (2) and esophageal carcinomas often display increased expression of the EGF receptor (EGFR) (48). We found that four of our eight cell lines (TT, TTn, TE-2 and HCE4) expressed relatively high levels of EGFR (data not shown), but this did not correlate with increased cyclin D1 over-expression. There is also evidence that in some cell types an activated ras oncogene can induce expression of cyclin D1 (49–51), as well as Rb 1144

(51), but this is an unlikely explanation for our findings since activating mutations in ras genes do not occur, or are very rare, in human squamous carcinomas of the esophagus (52,53). We found a strong positive correlation between the levels of expression of the cyclin D1 and Rb proteins amongst the eight cell lines (Figures 1 and 2). This finding is consistent with the original observation from our laboratory that in a series of primary esophageal cancers those tumors that have high expression of cyclin D1 always express the Rb protein, whereas those that have lost the expression of Rb have low or negligible amounts of the cyclin D1 protein (35). A similar relationship between cyclin D1 and Rb expression was subsequently reported in human breast carcinomas (54) and non-small cell lung carcinomas (55). In addition, ectopic expression of Rb in osteosarcoma and breast carcinoma cells induced the expression of both cyclin D1 mRNA and protein (23,24). However, in the present study, although the level of the Rb protein correlated with the level of cyclin D1 protein, it did not correlate with the level of cyclin D1 mRNA, suggesting a post-transcriptional effect in the esophageal cancer cells. To our surprise we also found a strong positive correlation between cyclin D1 and Rb expression and the level of expression of the cyclin dependent kinase inhibitory (CDKI) protein p27KIP1 (Figures 1 and 2). The level of cyclin D1 expression did not, however, correlate with the levels of two other CDKIs, p16INK4 or p21WAF1. During the preparation of this paper Kitahara et al. (55,56) also reported a correlation between the levels of expression of cyclin D1 and p27KIP1 in a series of human squamous esophageal carcinoma cell lines, thus further indicating the importance of this correlation. In DNA transfection studies we also found that ectopic overexpression of cyclin D1 in the EC109 cell line, which normally expressed low levels of cyclin D1, Rb and p27KIP1, led to increased expression of both the Rb and p27KIP1 proteins (Figure 6). In separate studies we have also found that ectopic over-expression of cyclin D1 in certain mammary epithelial cell lines is also associated with increased levels of the p27KIP1 protein (21), but this is not the case in R6 rat fibroblasts (Imoto et al., unpublished). Thus, it appears that in some cell types there is a homeostatic feedback loop between cyclin D1, Rb and p27KIP1. Since the level of p27KIP1 mRNA (Figure 3) did not correlate with the level of cyclin D1 protein (Figure 1) in the eight esophageal cancer cell lines, the increase in p27KIP1 protein appears to be due to post-transcriptional effects of cyclin D1 on p27KIP1 expression. Post-transcriptional alterations in the level of the p27KIP1 protein have been observed by other investigators (57). During the cell cycle p27KIP1 mRNA and protein synthesis are fairly constant. Nevertheless, the amount of this protein decreases during G1 phase progression (22,57), and there is evidence that the ubiquitin–proteasome pathway regulates the abundance of this protein (58). Additional studies are required to determine the precise mechanism responsible for the increased level of p27KIP1 protein in esophageal cancer cells that express high levels of cyclin D1. In any case, the paradoxical increase in p27KIP1 protein in cancer cells is not confined to esophageal cancer cell lines, since we have also found relatively high levels of the p27KIP1 protein in several human breast cancer cell lines and primary breast tumors (21,22; unpublished studies). In vitro assays for kinase activity, using a Rb-GST fusion protein as the substrate, indicated that cyclin D1-associated and CDK4-associated activities correlated with the level of

Correlation between cyclins and associated proteins

Fig. 3. Southern blot (A) and Northern blot (B) analyses of extracts of the eight esophageal cancer cell lines. Ten µg of DNA previously digested with EcoR I (A) or 15 µg of total RNA (B) were loaded in each lane, separated by 1% agar gel and then hybridized with the indicated 32P-labeled probes. The 28S ribosomal RNA band stained with ethidium bromide is shown as a control for RNA loading (B). The approximate sizes of the mRNAs were calculated from the positions of the two ribosomal RNAs and were as follows: cyclin D1, 4.5 kb; Rb, 3.9 kb; p16, 1.6 kb; cyclin E, 2.0 kb; and p27, 2.4 kb. For additional details see Materials and methods.

expression of cyclin D1 in the respective cell lines (Figure 4). This was true even though the cyclin D1 over-expressing cells had relatively high levels of the p27KIP1 inhibitory protein and they did not necessarily express high levels of CDK4 (Figure 1). Furthermore, immunoprecipitation studies indicated that the p27KIP1 protein was bound to the cyclin D1/CDK4 complex (data not shown). A notable example was the TE-3 cell line which expressed the highest levels of cyclin D1 and p27KIP1, and had the highest cyclin D1- and CDK4-associated kinase activities. Presumably, this reflects the fact that cyclin D1 is more rate-limiting than CDK4, as well as the fact that this cell line lacked expression of the CDK4 inhibitor p16INK4. It appears that these, and perhaps other factors, over-ride the potential inhibitory effects of p27KIP1. Western blot analyses also indicated a correlation between in vitro CDK4 activity and increased in vivo phosphorylation of the Rb protein (Figures 1 and 4). Two of these eight cell lines, TE-2 and EC17, also expressed relatively high levels of the cyclin E protein, with a prominent 50 kD band and a less prominent 45 kD band (Figure 1). Increased expression of cyclin E, and the presence of lower molecular weight forms of this protein, have been previously reported in a variety of primary human tumors and tumorderived cell lines (34). The EC17 cell line, which expressed the highest level of this protein, also displayed about a four fold amplification of the cyclin E gene (Figure 3A).

Amplification of the cyclin E gene appears to be infrequent in human tumors although it has been previously described in a single human colon cancer cell line (47). The TE-2 and EC17 cell lines also had high cyclin E-associated in vitro kinase activity when assayed with histone H1 as the substrate and expressed high levels of the p16INK4 protein. Since these cells are Rb-negative the in vivo target of this cyclin E-associated kinase may be the Rb related proteins p107 (7) or p130 (8) or other yet unidentified substrates. For reasons that are not apparent, extracts of the TE-3 cell line also had high cyclin E-associated in vitro kinase activity (Figure 4) even though they did not express unusually high levels of cyclin E or CDK2, expressed relatively high levels of p27KIP1 and contained a heat stable inhibitor of CDK2 activity (Figures 1 and 4). We also observed a correlation between the lack of expression of the Rb protein and an increase in expression of the CDK4 and CDK6 inhibitor p16INK4 in this series of eight esophageal cancer cell lines (Figure 1). Furthermore, the TE-2 and EC17 cell lines that express high levels of p16INK4 lacked cyclin D1- and CDK4-associated in vitro kinase activity (Figure 4). These findings are consistent with previous reports indicating that Rb-negative cell lines often express high levels of p16INK4 (25,26) and lack the cyclin D-CDK complex, presumably because of the direct binding of p16INK4 to CDK4 and CDK6 (25,26). Indeed, immunoprecipitates of CDK4 from extracts of the TE-2 and EC17 cells contained the p16INK4 1145

Y.Doki et al.

Fig. 4. In vitro assays for CDK activities, employing the indicated immunoprecipitates. Twenty-five µg of total cell extract were immunoprecipitated with the indicated antibodies. Two µg of histone H1 as the substrate were incubated with each immunoprecipitate plus 5 µCi [µ-32P]ATP, for 15 min at 30°C. The reaction mixture was then subjected to SDS–PAGE and the extent of histone H1 phosphorylation was detected by autoradiography. For the CDK4 and cyclin D1 associated kinase assays, 150 µg of the cell extract were employed and 1 µg of GST-Rb protein was used as the substrate instead of histone H1. For the CDK2 inhibitor assay, 25 µg of a total cell extract from each cell line was heat-denatured, and the supernatant fraction was mixed with 25 µg of a total cell extract of EC109 cells and this mixture subjected to the CDK2 assay, as described above. CDK2 activity of 25 µg of the EC109 extract mixed with an equal volume of lysis buffer was used as the control (lane C).

protein, but lacked the cyclin D1 protein, whereas CDK4 immunoprecipitates of the TE-3 cells contained the cyclin D1 protein, but not the p16INK4 protein (data not shown). We also found that the two cell lines, TE-2 and EC17, that expressed high levels of the p16INK4 protein also expressed correspondingly high levels of the related mRNA (Figure 3B). Southern blot analysis revealed that the EC17 cell line, which expressed the highest level of p16INK4 protein and mRNA, also displayed ~3.5-fold amplification of this gene. To our knowledge this is the first report of amplification of the p16INK4 gene in cancer cells. The TE-3 and EC109 cells lacked expression of p16INK4 at both the protein and mRNA levels which is consistent with the fact that they displayed homozygous deletions of this gene, a finding frequently seen in human cancer cell lines (59). Indeed, homozygous deletions of the p16 gene were seen with high frequency in previous studies of human esophageal squamous carcinoma cell lines (60–62). The remaining cell lines in the present study that did not express the p16INK4 protein (TT, TTn, HCE4 and HCE7), expressed p16INK4 mRNA, even though these cell lines are Rb-positive (Figure 3B). This finding was unexpected 1146

Fig. 5. Cell cycle distribution (A), doubling times (B) and colony-forming efficiency in soft agar (C). (A) The cell cycle distribution of exponentially dividing cultures was determined by flow cytometry. (B) The doubling times of the eight esophageal cancer cell lines were determined in exponentially dividing monolayer cultures. (C) Single cell suspensions were grown in 0.33% soft agar in complete medium for 14 days and the number of colonies .0.1 mm in diameter (white bar) and .0.05 mm in diameter (dark bar) were evaluated by microscope, and expressed as the percentage of the original number of cells plated. For additional details see Materials and methods.

since previous studies suggested that functional Rb suppresses p16INK4 expression at the level of transcription (25). The failure of these four cell lines to express detectable amounts of the p16INK4 protein might indicate that in some cells Rb suppresses this protein at a post-transcriptional level or that increased expression of cyclin D1 can cause this phenotype, since all of these four cell lines expressed high levels of cyclin D1. Point mutations in the p16INK4 gene are frequently observed in human esophageal cancers (63), and this could also play a role in impairing expression of the p16INK4 protein. The biological significance of the above-described variations in the expression of cyclin D1 and related proteins in this series of esophageal cancer cell lines is not known, since none of the individual changes correlated with the doubling time, colony-forming efficiency in agar or cell cycle parameters of the respective cell lines, although in vitro CDK2 kinase activity

Correlation between cyclins and associated proteins

factors. At the same time, the cells are partially protected from ‘excessive’ stimulation by virtue of these homeostatic feedback loops. Variations amongst cell types in the ‘set-point’ for the induction of some of these feedback responses might explain why ectopic over-expression of cyclin D1 enhances the growth of some cell lines but inhibits the growth of others (15–21). Further studies are in progress to examine these hypotheses and their relevance to the clinical behavior of human tumors. Acknowledgements This research was supported by an award from Fuso Pharmaceutical Company (Osaka, Japan) to Y.D., and a National Cancer Institute grant (R01CA63467) and an award from the National Foundation for Cancer Research to I.B.W.

References

Fig. 6. Western blot analysis of the levels of expression of the cyclin D1, Rb and p27KIP1 proteins in EC109 vector control cells (‘C’) and EC109 cyclin D1 over-expressing cells (‘D’). Human E-cadherin (124 kD) was expressed equally in both cell types. For additional details see Figure 1, and Materials and methods.

did display a significant negative correlation with the duration of the G1 phase of the cell cycle (Figure 5). It remains to be determined whether any of the variations described in the present study influence the in vivo growth properties of tumors. Certain correlations observed in the present study, together with previous findings (21,22,24,25,35) suggest that mammalian cells normally contain several homeostatic feedback loops that play a critical role in cell cycle control. These include a positive correlation between the expression of cyclin D1, Rb and p27KIP1, and a negative correlation between the expression of Rb and p16INK4. These correlations seem paradoxical since cyclin D1 enhances the G1 to S progression of the cell cycle, whereas Rb, p27KIP1 and p16INK4 have the opposite effect. It appears, therefore, that normal cells are designed to maintain a balance between these positive and negative-acting factors, perhaps because an excessive change in one direction might be toxic. For example, an excessive increase in cyclin D1-associated kinase activity might lead to excessive activation of E2F transcription factors, or the phosphorylation of additional substrates, which might have a cytotoxic effect. Indeed, the fact that loss of Rb expression induces the expression of p16INK4, a specific inhibitor of CDK4 and CDK6, suggests that cyclin D1-associated kinases may have targets in addition to Rb. Thus, the functional significance of this induction of p16INK4 would be to limit the phosphorylation of these targets. According to this model, during the evolution of tumors the constitutive over-expression of cyclin D1 or the loss of Rb expression, enhances transformation by distorting this balance between these positive and negative

1. Hunter,T. and Pines,J. (1994) Cyclins and cancer II: cyclin D and Cdk inhibitors come of age. Cell, 79, 573–582. 2. Sherr,C.J. (1994) G1 phase progression: cycling on cue. Cell, 79, 551–555. 3. Nurse,P. (1994) Ordering S phase and M phase in the cell cycle. Cell, 79, 547–550. 4. Sherr,C.J. and Roberts,J.M. (1995) Inhibitors of mammalian G1 cyclindependent kinases. Genes Dev., 9, 1149–1163. 5. Kato,J., Matsushime,H., Hiebert,S.W., Ewan,M.E. and Sherr,C.J. (1993) Direct binding of cyclin D to the retinoblastoma product (pRb) and pRb phosphorylation by the cyclin D-dependent kinase CDK4. Genes Dev., 7, 331–342. 6. Matsushime,H., Quelle,D.E., Shurtleff,S.A., Shibuya,M., Sherr,C.J. and Kato,J.-Y. (1994) D-type cyclin-dependent kinase activity in mammalian cells. Mol. Cell. Biol., 14, 2066–2076. 7. Beijersbergen,R.L., Carlee,L., Kerkhoven,R.M. and Bernards,R. (1995) Regulation of the retinoblastoma protein-related p107 by G1 cyclin complexes. Genes Dev., 9, 1340–1353. 8. Li,Y., Graham,C., Lacy,S., Duncan,A.M.V. and Whyte,P. (1993) The adenovirus E1A-associated 130-kD protein is encoded by a member of the retinoblastoma gene family and physically interacts with cyclins A and E. Genes Dev., 7, 2366–2377. 9. Harper,J.W., Adami,G., Wei,N., Keyomarsi,K. and Elledge,S.J. (1993) The p21kd Cdk interacting protein Cip1 is a potent inhibitor of G1 cyclindependent kinases. Cell, 75, 805–816. 10. Toyoshima,H. and Hunter,T. (1994) p27, a novel inhibitor of G1-cyclincdk protein kinase activity, is related to p21. Cell, 78, 67–74. 11. Lee,M.H., Renisdottir,I. and Massague,J. (1995) Cloning of p57KIP2, a cyclin-dependent kinase inhibitor with unique domain structure and tissue distribution. Genes Dev., 9, 639–649. 12. Hannon,G.J. and Beech,D. (1994) p15INK4B is a potential effector of TGFβ induced cell cycle arrest. Nature, 371, 257–261. 13. Serrano,M., Hannon,G.J. and Beach,D. (1993) A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4. Nature, 366, 704–707. 14. Guan,K.L., Jenkins,C.W., Nichols,Y., Wu,X., O’Keefe,C.L., Matera,A.G. and Xiong,Y. (1994) Growth suppression by p18, a p16INK4/MTS1– and p14INK4B/MTS2– related CDK6 inhibitor, correlates with wild-type pRb function. Genes Dev., 8, 2939–2952. 15. Jiang,W., Kahn,S.M., Zhou,P., Zhang,Y.-J., Cacace,A.M., Infante,A.S., Doi,S., Santella,R.M., and Weinstein,I.B. Over-expression of cyclin D1 in rat fibroblasts causes abnormalities in growth control, cell cycle progression and gene expression. Oncogene, 8, 3447–3457. 16. Quelle,D.E., Ashmun,R.A., Shurtleff,S.A., Kato,J.Y., Bar-Sagi,D., Roussel,M.F. and Sherr,C. J. (1993) Over-expression of mouse D-type cyclins accelerates G1 phase in rodent fibroblasts. Genes Dev., 7, 1559– 1571. 17. Ohtsubo,M. and Roberts,J.M. Cyclin-dependent regulation of G1 in mammalian fibroblasts. Science, 259, 1908–1912. 18. Resnitzky,D., Gossen,M., Bujard,H. and Reed,S. I. (1994) Acceleration of the G1/S phase transition by expression of cyclins D1 and E with an inducible system. Mol. Cell. Biol., 14, 1669–1679. 19. Han,E.K., Sgambato,A., Jiang,W., Zhang,Y.J., Santella,R.M., Doki,Y., Cacace,A.M., Schieren,I. and Weinstein,I.B. (1995) Stable over-expression of cyclin D1 in a human mammary epithelial cell line prolongs the Sphase and inhibits growth. Oncogene, 10, 953–961.

1147

Y.Doki et al. 20. Atadja,P., Wong,H., Veillete,C. and Riabowol,K. (1995) Over-expression of cyclin D1 blocks proliferation of normal diploid fibroblasts. Exp. Cell Res., 217, 205–216. 21. Han,E.-K-H., Begemann,M., Sgambato,A., Sohn,J.-W., Doki,Y., Xing,W.Q., Liu,W. and Weinstein,I.B. (1996) Increased expression of cyclin D1 in a murine mammary epithelial cell line induces p27KIP1, inhibits growth and enhances apoptosis. Cell Growth Diff., 7, 699–710. 22. Sgambato,A., Han,E.K., Zhou,P., Schieren,I. and Weinstein,I.B. (1996) Over-expression of cyclin E in the HC11 mouse mammary epithelial cell line inhibits growth and induces expression of the p27KIP1 inhibitor protein. Cancer Res., 56, 1389–1399. 23. Lukas,J., Mu¨ller,H., Bartkova,J., Spitkovsky,D., Kjerulff,A.A., JansenDu¨r,P., Strauss,M. and Bartek,J. (1994) DNA tumor virus oncoproteins and retinoblastoma gene mutations share the ability to relieve the cell’s requirement for cyclin D1 function in G1. J. Cell Biol., 125, 625–638. 24. Mu¨ller,H., Lukas,J., Schneider,A., Warthoe,P., Bartek,J., Eilers,M. and Strauss,M. (1994) Cyclin D1 expression is regulated by the retinoblstoma protein. Proc. Nat. Acad. Sci., 91, 2945–2949. 25. Parry,D., Bates,S., Mann,D.J. and Peters,G. (1995) Lack of cyclin D-Cdk complex in Rb-negative cells correlates with high levels of p16INK4/MTS1 tumor suppressor gene product. EMBO J., 14, 503–511. 26. Aagaard,L., Lukas,J., Bartkova,J., Kjerulef,A.-A., Strauss,M. and Bartek,J. (1995) Aberrations of p16INK4 and retinoblastoma tumor-suppressor gene occur in distinct sub-sets of human cancer cell lines. Int. J. Cancer, 61, 115–120. 27. Ewen,M.E., Sluss,H.K., Whitehouse,L.L. and Livingston,D.M. (1993) TGFβ inhibition of Cdk4 synthesis is linked to cell cycle arrest. Cell, 74, 1009–1020. 28. Geng,Y. and Weinberg,R.A. (1993) Transforming growth factor beta effects on expression of G1 cyclins and cyclin-dependent protein kinases. Proc. Natl Acad. Sci USA, 90, 10315–10319. 29. Polyak,K., Kato,J.-Y., Solomon,M.J., Sherr,C.J., Massague,J., Roberts,J.M. and Koff,A. (1994) p27KIP1, a cyclin-cdk inhibitor, links transforming growth factor-β 5 beta and contact inhibition to cell cycle arrest. Genes Dev., 8, 9–22. 30. Dulic,V., Kaufmann,W.K., Wilson,S.J., Tlsty,T.D., Lees,E., Harper,J.W., Elledge,S.J. and Reed,S.I. (1994) P53-dependent inhibition of cyclin dependent kinase activities in human fibroblasts during radiation-induced G1 arrest. Cell, 76, 1013–1023. 31. Li,R., Waga,S., Hannon,G.J., Beach,D. and Stillman,B. (1994) Differential effects by p21 CDK inhibitor on PCNA-dependent DNA replication and repair. Nature, 371, 534–537. 32. Lucibello,F., Sewing,A., Brusselbach,S., Burger,C. and Muller,R. (1993) Deregulation of cyclin D1 and E and suppression of Cdk2 and Cdk 4 in senescent human fibroblasts. J. Cell Sci., 105, 123–133. 33. Horiguchi-Yamada,J., Yamada,H., Nakada,S., Ochi,K. and Nemoto,T. (1994) Changes of G1 cyclins, cdk2, and cyclin A during the differentiation of HL60 cells induced by TPA. Mol. Cell. Biochem., 132, 31–37. 34. Keyomarsi,K., O’Leary,N., Molnar,G., Lees,E., Fingert,H.J. and Pardee,A.B. (1994) Cyclin E, a potential prognostic marker for breast cancer. Cancer Res., 54, 380–385. 35. Jiang,W., Zhang,Y.-J., Kahn,S.M., Hollestein,M.C., Santella,R.M., Lu,S.H., Harris,C.C., Montesano,R. and Weinstein,I.B. (1993) Altered expression of the cyclin D1 and retinoblastoma genes in human esophageal cancer. Proc. Natl Acad. Sci. USA, 90, 9026–9030. 36. Adelaide,J., Monges,G., Derderian,C., Seitz,J.F. and Birnbaum,D. (1995) Oesophageal cancer and amplification of the human cyclin D gene CCND1/ PRAD1. Br. J. Cancer, 71, 64–68. 37. Nakagawa,H., Zunkerberg,L., Togawa,K., Meltzer,S.J., Nishihara,T. and Rustagi,A.K. (1995) Human cyclin D1 oncogene and esophageal squamous cell carcinoma. Cancer, 76, 541–549. 38. Matsushime,H., Quelle,D.E., Shurtleff,S.A., Shibuya,M., Sherr,C.J. and Kato,J.-Y. (1994) D-type cyclin-dependent kinase activity in mammalian cells. Mol. Cell. Biol., 14, 2066–2076. 39. Zar,J.H. (1974) Biostatistical Analysis. Prentice-Hall, Inc., Englewood Cliffs, NJ. 40. Guadagno,T.M. and Newport,J.W. (1996) Cdk2 kinase is required for entry into mitosis as a positive regulator of Cdc2-cyclin B kinase activity. Cell, 84, 73–82. 41. Slingerland,J.M., Hengst,L., Pan,C.H., Alexander,D., Stampfer,M.R. and Reed,S.I. (1994) A novel inhibitor of cyclin-Cdk activity detected in transforming growth factor β-arrested epithelial cells. Mol. Cell. Biol., 14, 3683–3694. 42. Iihara,K., Shiozaki,H., Oku,K., Tahara,H., Doki,Y., Oka,H., Kadowaki,T., Iwazawa,T., Inoue,M. and Mori,T. (1993) Growth-regulatory mechanism of two human esophageal-cancer cell lines in protein-free conditions. Int. J. Cancer, 55, 364–370.

1148

43. Bartkova,J., Lukas,J., Strauss,M. and Bartek,J. (1995) Cyclin D1 oncoprotein aberrantly accumulates in malignancies of diverse histogenesis. Oncogene, 10, 775–778. 44. Zhang,S.Y., Caamano,J., Cooper,F., Guo,X. and Klein-Szanto,A.J. (1994) Immunohistochemistry of cyclin D1 in human breast cancer. Am. J. Clin. Path., 102, 695–698. 45. Bartkova,J., Lukas,J., Strauss,M. and Bartek,J. (1994) The PRAD-1/cyclin D1 oncogene product accumulates aberrantly in a subset of colorectal carcinomas. Int. J. Cancer, 58, 568–573. 46. Arber,N., Hibshoosh,S.F., Moss,T., Sutter,Y., Zhang,M., Begg,S., Wang,S., Weinstein,I.B. and Holt,P.R. (1996) Increased expression of cyclin D1 is an early event in multistage colorectal carcinogenesis. Gastroenterology, 110, 669–674. 47. Leach,F.S., Elledge,S.J., Sherr,C.J., Willson,J.K., Markowitz,S., Kinzler,K.W. and Vogelstein,B. (1993) Amplification of cyclin genes in colorectal carcinomas. Cancer Res., 53, 1986–1989. 48. Yoshida,T., Sakamoto,H. and Terada,M. (1993) Amplified genes in cancer in upper digestive tract. Sem. Cancer Biol., 4, 33–40. 49. Liu,J.J., Chao,J.R., Jiang,M.C., Ng,S.Y., Yen,J.J. and Yang-Yen,H.F. (1995) Ras transformation results in an elevated level of cyclin D1 and acceleration of G1 progression in NIH 3T3 cells. Mol. Cell. Biol., 15, 3654–3663. 50. Filmus,J., Robles,A.I., Shi,W., Wong,M.J., Colombo,C.L. and Conti,C.J. (1994) Induction of cyclin D1 over-expression by activated ras. Oncogene, 9, 3627–3633. 51. Arber,N., Sutter,T., Miyake,M., Kahn,S., Venkatraj,V.S., Sobrino,A., Warburton,D., Holt,P.R. and Weinstein,I.B. (1996) Increased expression of cyclin D1 and the Rb tumor suppressor gene in c-K-ras transformed rat enterocytes. Oncogene, 12, 1903–1908. 52. Wei,J., Kahn,S.M., Guillem,J.G., Lu,S.-H. and Weinstein,I.B. (1989) Rapid detection of ras oncogene in human tumors: applications to colon, esophageal, and gastric cancer. Oncogene, 4, 923–928. 53. Galiana,C., Fusco,A., Martel,N., Nishihira,T., Hirohashi,S. and Yamasaki,H. (1993) Possible role of activated ras genes in human esophageal carcinogenesis. Int. J. Cancer, 54, 978–982. 54. Bartkova,J., Lukas,J., Muller,H., Lutzhoft,D., Strauss,M. and Bartek,J. (1994) Cyclin D1 protein expression and function in human breast cancer. Int. J. Cancer, 57, 353–361. 55. Schauer,I.E., Siriwardana,S., Langan,T.A. and Sclafani,R.A. (1994) Cyclin D1 over-expression vs. retinoblastoma inactivation: implications for growth control evasion in non-small cell and small cell lung cancer. Proc. Natl Acad. Sci., USA, 91, 7827–7831. 56. Kitahara,K., Yasui,W., Yokozaki,H., Semba,S., Hamamoto,T., Hisatsugu,T. and Tahara,E. (1996) Expression of cyclin D1, CDK4, and p27KIP1 is associated with the p16MST1 gene status in human esophageal carcinoma cell lines. J. Expt. Therapeut. Oncol., 1, 7–12. 57. Hengst,L. and Reed,S.I. (1996) Translational control of p27KIP1 accumulaation during the cell cycle. Science, 271, 1861–1863. 58. Pagano,M., Tam,S.W., Theodoras,A.M., Beer-Romero,P., Sal,G.D., Chau,V., Yew,P.R., Draetta,G.F. and Rolfe,M. (1995) Role of the ubiquitinproteasome pathway in regulation abundance of the cyclin-dependent kinase inhibitor p27. Science, 269, 682–685. 59. Igaki,H., Sasaki,H., Tachimori,Y., Kato,H., Watanabe,H., Kimura,T., Harada,Y., Sugimura,T. and Terada,M. (1995) Mutation frequency of the p16/CDKN2 gene in primary cancers in the upper digestive tract. Cancer Res., 55, 3421–3423. 60. Qingyun,L., Yan,Y.-X., McClure,M., Nakagawa,H., Fujimura,F. and Rustgi,A.K. (1995) MTS-1 (CDKN2) tumor suppressor gene deletions are a frequent event in esophagus squamous cancer and pancreatic adenocarcinoma cell lines. Oncogene, 10, 619–622. 61. Zhou,X., Suzuki,H., Shimada,Y., Imamura,M., Yin,J., Jiang,H.-Y., Tarmin,L., Abraham,J.M. and Meltzer,S.J. (1995) Genomic DNA and messenger RNA expression alterations of the CDKN2B and CDKN2 genes in esophageal squamous carcinoma cell lines. Genes, Chromosomes Cancer, 13, 285–290. 62. Enders,G.H., Koh,J., Missero,C., Rustgi,A.K. and Harlow,E. (1996) p16 inhibition of transformed and primary squamous epithelial cells. Oncogene, 12, 1239–1245. 63. Mori,T., Miura,K., Aoki,T., Nishihira,T., Mori,S. and Nakamura,Y. (1994) Frequent somatic mutation of the MTS1/CDK4I (multiple tumor suppressor/ cyclin-dependent kinase 4 inhibitor) gene in esophageal squamous cell carcinoma. Cancer Res., 54, 3396–3397. Received on November 22, 1996; revised on February 7, 1997; accepted on February 20, 1997