Reduced Expression of the Cell Cycle Inhibitor p27^Kip1 in Non-Small Cell Lung Carcinoma: A Prognostic Factor Independent of Ras¹

Altered regulation of the cell cycle is a hallmark of human cancers (1). The cell cycle is governed by cdks,³ the activity of which is regulated by binding of positive effectors, the cyclins (2, 3), by negative regulators, the cdk inhibitors (4, 5), and by phosphorylation and dephosphorylation events (6, 7). The cdks integrate mitogenic and growth inhibitory signals and coordinate cell cycle transitions (2, 3). Passage through G₁ into S phase is regulated by the activities of cyclin D-, cyclin E-, and cyclin A-associated kinases. B-type cdks regulate the G₂-M phases. Two families of cdk inhibitors have been identified (4, 5). INK4 family members p15, p16, p18, and p19 bind cyclin D-dependent cdks, specifically cdk4 and cdk6. KIP family members, including p21^Cip1, p27^Kip1, and p57^Kip2, bind and inhibit their target cyclin/cdks. The cdk inhibitors regulate entry into and progression through the cell cycle by modulating the activity of cdks in response to mitogenic and antimitogenic stimuli. The cyclins, cdks, and cdk inhibitors are frequently altered in cancer or disrupted secondarily by other oncogenic events (1).

p27^Kip1 is a cdk inhibitor, which regulates progression from G₁ into S phase (4). Regulation of this protein occurs primarily at the posttranslational level by proteasomal degradation (8, 9, 10). p27^Kip1 binds and inhibits its target cdks in a stoichiometric manner; therefore, the availability of p27^Kip1 is critical in the regulation of cyclin/cdk activity. Although mutations in the p27^Kip1 gene are rare in human tumors (11, 12, 13, 14), decreased p27^Kip1 protein levels are found in aggressive cancers of the breast (15, 16, 17), colon (18), prostate (19), and gastric mucosa (20), suggesting that its loss may both reflect and participate in the process of tumor progression. NSCLC represents ∼75% of all lung cancers. Its overall prognosis is poor, with a 5-year survival rate of 15%. Even for stage I, the survival rate after complete surgical resection is only 60–70%, and for stage II it falls to 40–50%. Patients whose resected lung tumor shows certain distinct genetic and biological features seem at even greater risk of relapse and death from this disease (21, 22). An important genetic/biological prognostic factor in ADCs is activation of the Ki-ras oncogene (23, 24, 25). We have recently evaluated the prevalence of ras mutations, microvessel and vascular endothelial factor CVE of NSCLC. The present study was undertaken to assess p27^Kip1 levels in this cohort of NSCLC specimens to evaluate its potential use as a prognostic factor, and to assess the relationship between alteration in p27^Kip1 levels and ras mutations. We show that p27^Kip1 protein levels are frequently reduced in primary NSCLCs and provide evidence to suggest that loss of p27 may be an important indicator of poor prognosis in this group of patients.

Frozen tissues and formalin-fixed paraffin blocks of primary NSCLC and corresponding normal lung parenchyma were obtained from lobectomy or pneumonectomy specimens. These were sampled within 2 h after resection. The tissues were snap-frozen in liquid nitrogen and stored at −80°C until RNA extraction. Paraffin blocks were processed according to routine surgical pathological technique. A total of 136 NSCLC tumors were examined in this study; sixty-four were from specimens resected at the Montreal General Hospital, and 72 cases were obtained from the Canadian Lung Tumor Bank, established in association with the National Cancer Institute of Canada-Clinical Trial Group BR.10 protocol. Tumors were classified according to the WHO classification (26) and were staged according to the tumor-node-metastasis staging system adopted by the American Joint Committee on Cancer and the Union Internationale Contre le Cancer (27). The entire study cohort included 92 ADCs, 26 SQCCs, and 18 LCUCs. Clinical follow-up data were available for 67 BR.10 patients who entered the study between July 1994 and January 1996. Their median follow-up was 25 months (range, 16 days-3.1 years).

Genomic DNA was extracted from frozen normal lung and tumor tissues, as described previously (28). The presence of mutations in Ha-ras, Ki-ras, and N-ras were identified using PCR, followed by hybridization with allele (mutant)-specific oligonucleotide probes (29). The PCR amplification primers (amplimers) for exons 1 and 2 of Ki-ras, Ha-ras, and N-ras genes were purchased from Clontech Laboratories, Inc. (Palo Alto, CA).

The oligonucleotide probes for the wild type and mutant alleles of codons 12, 13, and 61 of Ha-ras, Ki-ras, and N-ras were those contained in the MUTA-LYZER Clontech kit. Each codon-specific kit contains a panel of either seven or eight individual oligonucleotide probes that correspond either to the wild type or to pertinent mutations of the indicated ras gene. End-labeling of the probe and dot-blotting were carried out as previously described (29). Ras mutations were identified using a two-step screening procedure that allowed evaluation of all 57 potential mutations. Positive ras mutations were confirmed by sequencing, as described (29).

For each case, five micron sections were prepared from representative formalin-fixed and paraffin-embedded tumor tissue from the surgical pathology archival blocks. Expression of the p27^Kip1 protein was determined immunohistochemically. Paraffin sections of tumor blocks were deparaffinized with xylene, rehydrated, and microwaved in 10 mm citrate buffer (pH 6.0). Sections were blocked with 3% hydrogen peroxide in methanol, followed by normal horse serum (1:20 dilution) in 0.1 M PBS (pH 6.0), and then incubated overnight at 4°C with anti-p27^Kip1 monoclonal antibody (Transduction Laboratories, Lexington, KY) diluted 1:1000 (0.25 μg/ml) in PBS. Slides were then reacted with biotin-labeled antimouse immunoglobulin G and incubated with preformed avidin-biotin-peroxidase complex (Vector Laboratories, Burlingame, CA). Metal-enhanced diaminobenzidine substrate (Pierce Chemical Co., Rockford, IL) was then added in the presence of horseradish peroxidase. Sections were counterstained with hematoxylin, dehydrated, and mounted. Tumor histology, grade, and p27^Kip1 staining were scored independently by two pathologists (C. C. and M-S. T.). Between 15 and 20 high power fields of tumor were scored for the percentage of nuclei showing positive p27^Kip1 staining. The scoring system used was as in our previous studies (15): 1 = 0–25%, 2 = 25–50%, 3 = 50–75%, and 4 = >75% positive nuclear staining.

Calculations were performed using SAS (release 6.12; SAS Institute Inc., Cary, NC). Associations between p27^kip1 and other tumor covariates were assessed with Spearman correlation coefficients. Survival curves were calculated by the Kaplan-Meier method (30) and compared with the logrank test (31).

Strong nuclear p27^Kip1 staining was consistently seen in normal bronchial epithelial cells and in infiltrating lymphocytes, and these provided internal staining controls in most tumors. Similar patterns of p27^Kip1 staining were seen with a polyclonal anti- p27^Kip1 antibody, and the immunoreactivity was lost when the antibody was blocked with a control peptide. Of the 136 cases in the study, p27^Kip1 staining was technically not feasible in 3 cases, and these were excluded from further analysis. There was a consistent reduction in the nuclear p27^Kip1 staining in all of the histological subtypes (ADC, SQCC, and LCUC). Examples of p27^Kip1 staining in normal mucosa and tumor samples are shown in Fig. 1, A–C. Within the SQCCs, p27^Kip1 levels were lowest at the periphery of tumor nodules and often high at the center of tumor nodules adjacent to areas of necrosis (Fig. 1, D and E).

The distribution of histological subtype, degree of differentiation (well, moderately, and poorly differentiated), size, tumor and node stage, p27^Kip1 scores, and ras mutations are indicated in Table 1. Sixty-three tumors (47%) were poorly differentiated, 46 (35%) were moderately differentiated, and 24 (18%) were well differentiated. The mean tumor diameter of the entire group was 3.9 cm. The tumors were largely confined to tumor stages T₁ and T₂ (97%) and nodal stages N₀ and N₁ (92%). Ras mutations were present in 30 tumors, all of which were ADCs (33% of ADCs).

We analyzed the correlations between the above clinicopathological variables and p27^Kip1 scores (1, 2, 3, 4) and also p27^Kip1 scored as low or high (staining in <50% and >50% of tumor nuclei, respectively). These data are shown in Table 2. There was no correlation between p27^Kip1 score and tumor stage, nodal status, or tumor histology (ADC versus SQCC or LCUC). There was no correlation between p27^Kip1 staining levels and the presence or absence of ras mutations in either the group as a whole or in the adenocarcinoma subset. In contrast, there was statistically significant inverse correlation between tumor grade and p27^Kip1 status within the whole group (n = 133), with high-grade, poorly differentiated tumors showing more frequent p27 loss (P = 0.008). Tumors with low p27^Kip1 scores tended to be larger than those with high p27^Kip1 scores (mean diameters, 4.04 cm, standard deviation 1.77, and 3.11, standard deviation 2.16, respectively; t test, P = 0.0423).

Clinical follow-up data were available on 67 patients in the BR10 cohort. The median duration of follow-up was 2.1 years (minimum, 16 days; maximum, 3.1 years). All 16 patients who died and 21 of 22 who relapsed had tumors with low p27 scores. All but one of the patients who had tumors with high p27 scores were alive and relapse-free at last follow up. The effect of p27 on overall survival, however, did not reach statistical significance (P = 0.22).

Our data yield several conclusions: (a) p27^Kip1 levels are frequently reduced in NSCLC; (b) loss of p27^Kip1 is correlated with increasing tumor grade; and (c) ras mutations arise independently of changes in p27^Kip1 levels.

We have shown that levels of the cell cycle inhibitor p27^Kip1 are frequently reduced in non-small cell carcinomas of lung compared with levels seen in normal human bronchial epithelium. Low p27^Kip1 levels (staining in <50% of tumor nuclei) were seen in 86% of tumors in the entire group (n = 133). Loss of the p27^Kip1 protein has been observed in other cancers, including breast, colon, prostate (15, 16, 17, 18, 19, 32), and ovary.⁴ This suggests that p27^Kip1 is important in the regulation of epithelial cell proliferation and that it is a common target for alteration in human tumors. As such, reduced levels of this protein may play an important role in tumor progression. The loss of p27^Kip1 may reflect changes in the proliferative fraction, although there is no clear correlation between decreased p27^Kip1 and tumor proliferation at present. In breast and colon cancers, low p27^Kip1 did not correlate with an increased percentage of S phase or with increased expression of the proliferation marker, MIB-1 (15, 16, 17, 18). As a result, it could reflect changes in other properties such as cell-cell adhesion and increased metastatic potential (33, 34, 35, 36, 37). Decreased p27^Kip1 levels can occur in human tumors in the presence of abundant p27^Kip1 mRNA expression (15, 17, 18, 32), consistent with recent studies that have demonstrated that reduced p27^Kip1 levels are posttranslationally regulated, at least in part, by proteasomal degradation (8, 9, 10, 18, 32). The molecular factors responsible for reduced p27^Kip1 levels in tumor progression are not clear at present.

In this study, p27^Kip1 levels showed a statistically significant inverse correlation with tumor grade. High levels of p27^Kip1 (staining in >50% of tumor nuclei) were seen in one-third of well differentiated (low grade) tumors, but only in 8% of the poorly differentiated group. A similar inverse correlation between p27^Kip1 and tumor grade has been observed in breast, prostate, and colon carcinomas (15, 16, 17, 18, 19). This may in part be explained by recent observations that p27^Kip1 plays an important part in normal differentiation (38, 39, 40), and that p27^Kip1 levels are regulated in part by E-cadherin (41), a key molecule involved in epithelial cell-cell adhesion, which is frequently reduced during tumor progression.

The association between low p27^Kip1 and poor tumor grade suggests that low p27^Kip1 may have prognostic potential in NSCLC. In particular, high p27^Kip1 may define a more favorable prognostic group. Esposito et al. (32) recently reported that low p27^Kip1 was an indicator of poor prognosis in a cohort of 108 patients with NSCLC on univariate analysis. Yatabe et al. (42) reported similar findings. We did not have access to follow-up data on the cohort of patients from the Montreal General Hospital. We did, however, have follow-up data on 67 patients from the BR.10 study in which patients with resectable NSCLC were randomized to receive or to not receive adjuvant chemotherapy. The median duration of follow-up in this study, which began in July 1994, was 25 months. The immaturity of follow-up in this smaller cohort of patients precluded significant analysis of p27^Kip1 on disease outcome. However, it is worth noting that all of the 16 individuals who died during the study period had tumors with low p27^Kip1. Low p27 levels were also observed in 21 of the 22 cases that suffered a relapse during the follow-up period, and all but one patient with high tumor p27 levels were alive and relapse-free at last follow-up. These observations suggest that p27^Kip1 levels may be an important indicator of prognosis in NSCLC and indicate the need for a larger study with longer follow up to test this hypothesis.

Although a number of chemotherapeutic agents have shown activity in NSCLC, both response rates and duration have been disappointing. The distribution of p27^Kip1 staining in the SQCC tumor may shed light on the chemoresistance of this cancer. Recent studies of tumor spheroids grown in three-dimensional culture have shown that tightly adhesive spheroids have both reduced proliferative rates and reduced responsiveness to alkylating chemotherapy (33). It is possible, using antisense oligonucleotides, to decrease p27^Kip1 levels (33) and to increase the proliferative fraction (percentage of cells in S phase) and reduce cell-cell adhesion. Antisense p27-treated spheroids are, thus, converted from tight aggregates to loose aggregates that are more responsive to chemotherapeutic agents. In the present in vivo study of SQCC lung tumors, as in our previous studies of cultured spheroids, we observed reduced p27^Kip1 immunostaining at the periphery with higher levels at the center (Fig. 1, D and E). The increased p27^Kip1 in cells adjacent to necrotic tumor in vivo may reflect the cytostasis induced by hypoxia and low pH. Nonproliferating cancer cells are typically resistant to cytotoxic agents that rely on DNA synthesis and/or tumor proliferation for activity (43). The population of cells expressing high p27^Kip1 protein in SQCC may identify a chemoresistant fraction that is recruited into cycle and repopulates after the more proliferatively active portion of the tumor is killed by cytotoxic agents. New drugs that recruit this population into cycle before treatment may improve the efficacy of chemotherapy; alternatively, drugs that prevent reentry of such cells into the cell cycle (i.e., mechanisms to maintain dormancy by manipulating p27^Kip1 levels) are both potential avenues for novel treatment strategies in NSCLC.

Finally, we have observed that ras mutations and loss of p27^Kip1 seem to occur independently in ADC. Ras mutations occur almost exclusively in ADC and, by an overwhelming majority (>90%), involve codon 12 of the Ki-ras gene (29). Although recent reports link ras activation with changes in p27^Kip1 regulation (44, 45, 46), the two show no association in ADC and, thus, have the potential to contribute independently to patient outcome. The fact that there was no particular influence of tumor type on p27^Kip1 levels is consistent with the observation that p27^Kip1 levels and ras mutations are independent, because ras mutations are largely confined to ADC (23, 29, 47, 48).

In summary, we have shown that p27^Kip1 levels are reduced in the majority of cases of NSCLC and that loss of p27 is statistically correlated with increased tumor size and with reduced tumor differentiation or high tumor grade. We also present evidence to suggest that p27^Kip1 levels may be an important prognostic factor in NSCLC, which indicates the need for larger studies with longer follow-up to assess this question.

We thank M. Viscardi for assistance in preparation of the manuscript. The establishment of the Canadian Lung Tumor Bank is partially supported by the Glaxo-Wellcome Pharmaceutical Company.