Purpose: Overall survival of head and neck squamous cell cancer (HNSCC) patients has not improved despite advances in our understanding of the biology and molecular features of this disease. In particular, patients with advanced HNSCC have the poorest prognosis. To understand more about the contribution of cell cycle alterations to HNSCC development and their possible value in predicting prognosis and response to chemotherapy, we evaluated the levels of proteins involved in cell cycle control in patients diagnosed with advanced HNSCC.
Experimental Design: A tissue microarray was made with 122 HNSCC specimens obtained from biopsy material. Protein expression was evaluated by immunohistochemistry and correlated with clinical and pathological characteristics.
Results: Multiple alterations at various checkpoints of cell cycle progression were observed. Loss of P16 protein was less common in oropharyngeal tumors than at other HNSCC locations (P = 0.02). Evaluation of the simultaneous expression of different proteins highlighted direct correlations (P < 0.05) such as that of the cyclin-dependent kinases with their cyclin-partners, and the Ki-67 protein with cyclin-dependent kinases 1, cyclin A (CA) and cyclin B1. Median overall survival and time-to-progression were longer in patients with CA-expressing tumors (not reached versus 34.4 months, P = 0.02) and (47.3 versus 14.6 months, P = 0.006), respectively. Moreover, expression of CA in tumors predicted a better response to chemotherapy. Positive expression of cyclin E in tumors was also associated with an increased median time-to-progression (14.6 versus 25.8 months, P = 0.04). Finally, patients with cyclin D1-expressing tumors had shorter median overall survival (29.6 months versus not reached, P = 0.05) and shorter median time-to-progression (21.5 months versus not reached, P = 0.06). However, in a multivariate analysis a CA-negative–expressing tumor was the only independent poor prognostic factor in the entire cohort of HNSCC patients [odds ratio, 2.3; 95% confidence interval (CI) = 1.2–4.5; P = 0.01].
Conclusions: Our results provide detailed information on the molecular profile of cell cycle components in HNSCCs and identify CA-negative–expressing tumors as an independent marker of tumor progression and poor response to chemotherapy in patients diagnosed with advanced HNSCC.
Squamous cell carcinomas are the sixth most common cancer in developed countries, with >500,000 new cases diagnosed worldwide each year (1) . Whereas surgical therapy is the preferred treatment for early-stage head and neck squamous cell carcinomas (HNSCC), chemotherapy and/or radiotherapy are used for patients with advanced disease. The overall 5-year survival rate for advanced-stage HNSCC patients is <40% and to date there have been no reliable predictors of survival or response to therapy for these patients. Thus, it is essential to investigate the molecular alterations of advanced HNSCCs that may help to identify patients who are at the greatest risk of progression and who might require a more definitive therapy (2) .
Deregulation of cell division is one of the most common abnormalities of cancer cells. Cell cycle progression is dependent on the timely formation and dissociation of protein kinase complexes that enable the cell to pass through several regulatory transition points (3 , 4) . Among the components of cell cycle regulation are cyclins, cyclin-dependent kinases (CDKs), and cyclin-dependent kinase inhibitors. Gene abnormalities in components of the cell cycle (e.g., chromosomal deletions, gene mutations, and promoter hypermethylation), such as p53 and p16/INK4A, are very common in most tumor types, including HNSCCs (5, 6, 7) . Moreover, amplification of the cyclin D1 gene has been identified as a mechanism of cyclin D1 (CD1) protein overexpression in HNSCCs (6) . In HNSCCs cell cycle components frequently feature aberrant amounts of protein, although they are rarely altered at the genetic level. Whereas alterations on the protein expression of several cyclin-dependent kinase inhibitors, such as P21 and P27 (8 , 9) and cyclins (10, 11, 12, 13) , have commonly been observed in HNSCCs, there are few studies on the levels of the different CDKs at distinct HNSCC locations, and there is little information on the combined expression profiles of all these proteins.
To investigate the combined pattern of alterations of cell cycle regulators in advanced HNSCCs and to identify factors with potential clinical implications, we have used tissue microarrays to detect immunohistochemically the levels of a considerable number of proteins involved in cell cycle control. Molecular data were correlated with tumor location, clinical outcome and response to therapy in a well-characterized cohort of advanced HNSCC patients.
MATERIALS AND METHODS
Patients and Tumor Specimens.
We examined tumor specimens from 122 patients with HNSCCs diagnosed at five different Spanish National Health Centers (Hospital de Alicante, Hospital Doce de Octubre, Hospital del Mar, Hospital Marqués de Valdecilla, Hospital Sant Pau) between 1998 and 2000. According to the World Health Organization classification, all cases were stages III to IV. Patients and tumor characteristics are presented in Table 1⇓ . The median age of the patients was 56 years (range 37–74). The median overall survival for the entire cohort was 37.7 months [95% confidence interval (CI) = 26.3–49.1] and the median time-to-progression was 29.1 months (95% CI = 9.3–48.9). All patients underwent chemotherapy as a primary treatment and then chemo-radiotherapy as a consolidation treatment (induction chemotherapy with cisplatin plus 5-fluouracil and chemo-radiotherapy with cisplatin and radiotherapy in primary tumor and neck of up to 66 Gy). The clinical follow-up of the patients included a clinical examination every 3 months, a computed tomography scan of the head and neck region two times per year and one chest radiograph every year. Recurrences or new primary tumors were determined by clinical examination, computed tomography scan, and biopsy. Tumor specimens were at-diagnosis biopsies taken before the onset of treatment, fixed in formol, and embedded in paraffin.
Tissue Microarray Design.
A tissue arrayer (Beecher Instruments, Silver Spring, MD) was used to construct the tissue microarrays. Slides were reviewed by expert pathologists (M. R-P and J. L. R-P) who selected areas containing tumor cells but who avoided those featuring necrosis, inflammation, and keratinization. To assess reproducibility, 0.6-mm diameter cylinders from two separate areas were taken from each tumor. Our study thus contained at least one duplicate of each tumor sample. We constructed two different tissue microarray blocks each of 142 cylinders, containing the 122 cases and 20 controls. The control specimens consisted of reactive tonsil tissues and the normal epithelium adjacent to the tumor tissue in some cases.
Three-μm–thick sections were cut from the tissue microarray and transferred to positively charged surface glass slides. The sections were dried for 16 hours at 56°C before being dewaxed in xylene and rehydrated through a graded ethanol series and washed with PBS. To retrieve antigens, they were treated in a pressure cooker for 2 minutes in 10 mmol/L citrate buffer (pH 6.5). We did immunohistochemical staining as described previously (14) in an automated immunostainer, using the biotin-streptavidin-peroxidase procedure with diaminobenzidine chromogen as a substrate. Two different pathologists (M. R-P. and J. L. R-P.) evaluated immunostaining using uniform criteria and without prior knowledge of the clinical and pathologic characteristics of the patients. The antibodies used were as follows: CDK1 (CDC2 or p34), antibody 1 (1:1500 dilution, Transduction Labs, Lexington, KY); CDK2, antibody 8D4 (1:500 dilution, Neomarkers, Freemont, CA); CDK6, antibody BD (1:600 dilution, BD Biosciences PharMingen, San Diego, CA); cyclin A (CA), antibody GEG (1:100 dilution, Novocastra Laboratories Ltd., Newcastle upon Tyne, United Kingdom); cyclin B1 (CB1), antibody 7A9 (1:25 dilution, Novocastra); CD1, antibody DCS-6 (1:100 dilution, DAKO, Glostrup, Denmark); cyclin D3 (CD3), antibody DCS-22 (1:10 dilution, Novocastra), cyclin E (CE), antibody 13A3 (1:10 dilution, Novocastra); HDM2, antibody IF2 (1:10 dilution, Oncogene, La Jolla, CA); Ki-67, antibody M1B1 (1:15,000 dilution, DAKO); P16, antibody F-12 (1:50 dilution, SantaCruz Biotechnology, Santa Cruz, CA); P21, antibody EA10 (1:50 dilution, Oncogene); P27, antibody 57 (1:1,000 dilution, Transduction Lab); P53, antibody DO-7 (1:50 dilution, Novocastra); retinoblastoma protein (RB), antibody G3–245 (SD, 1:250 dilution, PharMingen). The cutoff criteria and control data are summarized on Table 2⇓ .
Frequencies were compared either by Fisher’s exact test or the χ2 contingency test, as appropriate. We estimated survival and time-to-progression curves using the Kaplan-Meier method and compared them using a two-sided log-rank test. The Brookmeyer-Crowley method was used to calculate the 95% CI. Multiple logistic-regression that used a Cox proportional hazards model was used to determine whether the molecular characteristics of the tumors independently predicted survival in our cohort of advanced HNSCC patients. Differences of P < 0.05 were considered statistically significant. Analyses were carried out with the SSPS program, version 10.0.5. (SSPS Inc, Chicago, IL). We drew graphics using the S Plus 6 statistics package.
The frequency of positive and negative expression for each protein marker is indicated in Table 3⇓ . Examples of immunostaining are depicted in Fig. 1⇓ . The concordance of the analysis in both core samples from the same biopsy ranged from 92 to 98%, depending on the marker, indicating a high degree of intra-sample reproducibility. To our knowledge, this is the first report on the immunohistochemistry of CDK1 and CDK6 in primary HNSCCs. Positive expression of CDK1 was present in normal epithelia in cells from basal and parabasal layers, whereas loss of CDK1 was observed in 62% of HNSCC specimens. CDK6 protein was detected in cells from all layers in normal epithelium, whereas it was lost in 59% of tumor specimens.
Correlation of protein expression with the distinct tumor locations revealed that the distribution of abnormal protein levels was similar at the distinct HNSCC locations, except for P16. Tumors from oropharynx had less frequent loss of P16 protein than did tumors from other locations (P = 0.02; Table 3⇓ ).
Fisher’s exact test revealed many significant associations between the different markers (Fig. 2)⇓ . Remarkably, direct statistically significant associations (P < 0.05) were observed between proteins acting in complexes such as the cyclin-dependent kinases (CDKs), their regulatory subunits (cyclins CDK2-CE, CDK2-CA, CDK1-CA, CDK1-CB1, and CDK2,6-CD1,3), and their cyclin-dependent kinase inhibitors, P21 or P27. There was also a strong correlation between high levels of Ki-67 protein and an increase in CDK1, CA, and CB1 proteins. These act during the late G2 and mitosis phases of the cell cycle and thus indicate that these tumors are probably highly proliferative.
Intriguingly, we observed that low CA expression was associated with decreased overall patient survival (P = 0.02; Fig. 3A⇓ ). Median survival time of patients with positive-CA–expressing tumors was longer (34.4 months, 95% CI = 18.9–49.9) than that of patients with negative-CA–expressing tumors (not reached). CA expression was also a good predictor of patient response to the treatment. As shown in Table 4⇓ , patients with CA-expressing tumors had an increased probability of complete response or >80% partial response relative to patients with negative-CA–expressing tumors (P = 0.005). Moreover, the median time-to-progression was shorter in the latter (14.6 months, 95% CI = 10.1–22.7) than in the former (47.3 months) group of HNSCC patients (P = 0.006; Fig. 3B⇓ ). Similarly, the negative-expressing CE tumors had a shorter time-to-progression (14.6 months, 95% CI = 13.7–16.1) than did positive-expressing CE tumors (25.8 months, 95% CI = 9.8–33.2; P = 0.04). In contrast, negative-CD1–expressing tumors correlated with better overall patient survival and longer time-to-progression. Median survival time of patients with negative-CD1–expressing tumors was longer (not reached) than that for patients with positive-CD1 tumors (29.6 months, 95% CI = 17.6–41.5; P = 0.05). Moreover, median time-to-progression in patients with negative- and positive-CD1–expressing tumors was not reached versus 21.5 months (95% CI = 12.9–30.1), respectively (P = 0.06).
Logistic regression was used to determine whether any of these markers were an independent predictor of survival in advanced HNSCC patients. Gender, age, performance status, tumor stage, tumor location, and expression of all of the markers evaluated in the present study were included in the stepwise model-selection process. The analysis revealed that the only independent marker that was a predictor of death in advanced HNSCC patients was CA (P = 0.01), with an odds ratio of 2.3 (95% CI = 1.2–4.5; Table 5⇓ ).
We have taken a tissue-microarray approach to screen for the expression pattern of several cell cycle components in tumors from advanced HNSCC patients. Previous reports have shown that altered expression of the positive regulators CDK2, CA, CB1, CD1, CD3, CE, Ki-67, and HMD2 (11, 12, 13, 14, 15, 16, 17, 18 , 19, 20 , 25, 26) and negative regulators P16, P21, P27, and RB are common hallmarks of HNSCC development (5, 6 , 8 , 10 , 15, 16, 17, 18 , 24, 25, 26, 27) . Our present findings agree with these observations in HNSCCs and highlight a concomitant altered expression of the CDKs with their associated cyclins, i.e., overexpression of the CDK1 and 2 and their respective associate cyclins (CA, CB, and CE), implying a functional relevance for these alterations. Our present data reports for the first time that abnormal levels of the CDK1 and CDK6 proteins in HNSCCs are also common events in HNSCC development. Most cell cycle components are not apparently altered at genetic or epigenetic levels and the mechanisms leading to the abnormal expression of these proteins in HNSCCs are still unknown. Alterations in upstream pathways that regulate the activity of CDKs, cyclins, or cyclin-dependent kinase inhibitors are among the most plausible hypothesis for these molecular abnormalities. Nevertheless, because many small compounds that inhibit CDKs are being developed, the identification of tumors carrying increased expression of proteins acting in specific checkpoints of the cell cycle may help to select patients for specific and individualized therapies.
Tumors from the oropharynx featured the less frequent loss of P16 protein expression than did those from the oral cavity, hypopharynx, and larynx. Different molecular characteristics in oropharyngeal and nonoropharyngeal tumors have been reported previously. For example, we have shown that oropharyngeal tumors have more frequent promoter hypermethylation at the O6-methylguanine-DNA methyltransferase and death-associated protein kinase genes compared with HNSCC tumors from other locations (7) . In addition, tumors from the oropharynx are more likely to be positive for the human papilloma virus than are those from nonoropharyngeal locations (28) . Taken together, these differences imply that a subset of oropharyngeal tumors may represent a different molecular and pathologic disease entity among HNSCCs.
Some authors have reported a correlation of positive expression of CA in tumor tissue with advanced stages and worse prognosis in non–small-cell lung cancer (29) , although others found no such association in oral cancer patients (16) . Here, we have shown that absence of CA protein in tumors is an independent negative prognostic factor in a retrospective analysis of advanced HNSCC patients who received chemotherapy. Patients with CA-expressing tumors had a longer time-to-progression and better response to treatment compared with their counterparts. Similar observations were made in patients diagnosed with soft-tissue sarcoma (30) . CA, in conjunction with CDK2, regulates entry into and progression through the S phase and remains active until the beginning of mitosis. CA has therefore been proposed as a marker of cell proliferation. Supporting this, our present and other results published previously (30) identify a direct association between high levels of CA and Ki-67, a cell proliferation marker. Proliferating cells are more sensitive to chemotherapeutical agents, and a positive correlation between proliferation rate and quality of prognosis after chemotherapy treatment has already been noted in other malignancies such as breast cancer, melanoma, and lymphoma (31, 32, 33) . Moreover, in vitro experiments have shown that CA correlates with the sensitivity of human cancer cells to the cytotoxic effect of 5-fluorouracil (34) , further supporting our observations. A univariate analysis of our data revealed increased time-to-progression in those patients with CE-expressing tumors. CE forms a complex with, and functions as a regulatory subunit of, CDK2, the activity of which is required to ensure the G1-S transition in the cell cycle. This supports the hypothesis of a positive relationship between the proliferation index and the success of response to chemotherapy.
In conclusion, our results further attest to the presence of multiple abnormal levels of proteins that participate in cell cycle regulation and identify CA-protein–expressing tumors as a marker of favorable chemotherapeutic response in patients with advanced HNSCC.
We thank Maria Jesus Acuña and Raquel Pajares from the Immunohistochemistry and Histology Unit of the CNIO for meticulous technical support. We also want to thank the following collaborators from the Pathology Departments at several Spanish Hospitals: F. Sancho-Poch (Hospital Sant Pau, Barcelona); F. Fernandez and M. Mayorga (Hospital Valdecilla, Santander) and M. Niveiro (Hospital de Alicante, Alicante).
Grant support: in part by the Spanish Ministerio de Ciencia y Tecnología (SAF2002-01595) and by the Comunidad de Madrid (CAM 08.1/0032/2003 1). M. Sanchez-Cespedes is a supported by the Ramon y Cajal Program of the Ministerio de Ciencia y Tecnología, Spain.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: Montserrat Sanchez-Cespedes, Molecular Pathology Program, Spanish National Cancer Centre (CNIO), Melchor Fernandez Almagro 3, 28029 Madrid, Spain. Phone: 34-912246954; Fax: 34-912246923. E-mail:
- Received April 23, 2004.
- Revision received July 31, 2004.
- Accepted September 10, 2004.