Purpose: The INK4A-ARF locus at chromosome 9p21 is frequently altered in head and neck squamous cell carcinoma (SCC) and encodes two distinct tumor suppressors, p16INK4A and p14ARF. This study addressed the role of p14ARF as a potential prognostic marker in this disease.

Experimental Design: p14ARF protein expression was assessed by immunohistochemistry in a cohort of 140 patients with SCC of the anterior tongue. Using univariate and multivariate Cox's proportional hazards models, the outcomes examined were time to disease recurrence or death, with or without clinicopathologic covariates, including nodal status, disease stage, treatment status, Ki-67 staining, and molecular markers with known functional or genetic relationships with p14ARF (p16INK4A, p53, pRb, p21WAF1/CIP1, E2F-1).

Results: On multivariate analysis, p14ARF positivity (nucleolar p14ARF staining and/or nuclear p14ARF staining in ≥30% of tumor cells) was an independent predictor of improved disease-free survival (DFS; P = 0.002) and overall survival (OS; P = 0.002). This was further enhanced when p14ARF positivity was cosegregated with positive (≥1%) p16INK4A staining (DFS, P < 0.001; OS, P < 0.001). Patients whose cancers were p14ARF negative and p53 positive (>50%) had the poorest outcome (DFS, P < 0.001; OS, P < 0.001) of any patient subgroup analyzed.

Conclusions: These data show that in patients with SCC of the tongue, combined nuclear and nucleolar expression of p14ARF protein predicts for improved DFS and OS independent of established prognostic markers.

Head and neck cancers account for 3.8% of all newly diagnosed cancers in men in Australia (1). Almost all (90%) head and neck cancers are squamous cell carcinomas (SCC) and >50% arise in the oral cavity, with the majority representing SCC of the anterior tongue. Improvements in cancer treatment have not translated into improved disease-free survival (DFS) and overall survival (OS) in head and neck SCCs (HNSCC) over the past 20 years (2). The identification of prognostic markers with potential clinical utility may assist in clinical management and suggest new targets for therapy, as well as provide insights relevant to a better understanding of the biology of HNSCC.

Dysregulation of the normal cell cycle regulatory machinery is integral to the neoplastic process and there is now compelling evidence implicating loss of cell cycle control in the development and progression of most human cancers (3). Consequently, cell cycle regulatory molecules are attractive potential prognostic markers. The INK4A-ARF locus is frequently disrupted in human cancers and consists of two overlapping genes that encode two unrelated proteins, p16INK4A and p14ARF (4). p16INK4A and p14ARF act through the retinoblastoma protein (pRb) and p53 pathways, respectively (3), to regulate cell cycle progression, and these important oncogenic pathways are frequently dysregulated in HNSCC.

The pRb pathway comprises INK4A proteins, such as p16INK4A, the cyclin D/E-dependent kinases, and pRb family proteins. Cyclin D1 and cyclin-dependent kinases 4 and 6 (Cdk4/6) act in concert with cyclin E-Cdk2 to phosphorylate pRb, leading to release of E2F-1 in late G1 phase. E2F-1 then initiates transcription of genes essential for DNA synthesis. The p16INK4A protein specifically inhibits the cyclin D–dependent kinases, at least in part by preventing cyclin D1 association with Cdk4/6. pRb-E2F-1 complexes actively repress gene expression and, consequently, derangement of the pRb pathway releases this inhibitory control resulting in deregulated S-phase entry and uncontrolled cell proliferation (3). Alterations to this pathway through frequent Cdk6 hyperactivation, cyclin D1 overexpression, and p16INK4A mutations have been shown in tongue cell lines and carcinomas (5, 6); overall, aberrations in the pRb pathway are found in >80% of HNSCC (7).

A second critical oncogenic pathway involves the tumor suppressor p53, its negative regulator Hdm2 (Mdm2 in mice), and p14ARF. Whereas mutated p53 is found in >50% of HNSCC (8, 9), the roles of other components of this pathway are not well understood in this disease. p53 accumulates in response to DNA damage, hypoxia, and oncogene activation. Nuclear p53 is then stabilized and activated, which may result in cell cycle arrest (via the Cdk inhibitor p21WAF1/CIP1) or apoptosis (10, 11). Hdm2/Mdm2 binds to p53 and reduces nuclear p53 levels by transporting p53 to the cytoplasm where it is degraded (12, 13). This results in the inhibition of p53-mediated cell cycle arrest and apoptosis. p14ARF abrogates this function of Hdm2/Mdm2 through direct binding and nucleolar sequestration of Hdm2/Mdm2 (14, 15). Excess p14ARF restrains cell growth by not only inhibiting Hdm2/Mdm2 function but also by independently stabilizing p53, triggering a p53-dependent transcriptional response (16). p14ARF activity has mainly been attributed to this p53-dependent mechanism. However, p14ARF is also able to suppress growth independent of p53 (17) through delaying S-phase progression (18). Although the precise mechanisms remain to be elucidated, it is possible that these p53-independent effectors of p14ARF may include either Mdm2 or E2F-1. Moreover, ARF is an E2F-1 target gene (19, 20) and thereby provides a link between the pRb and p53 pathways (3).

The majority of previous studies of p16INK4A and p14ARF in HNSCC have focused on genetic aberrations rather than alterations in protein expression. The overlapping INK4A-ARF genes are altered through homozygous deletion, promoter methylation of exon 1α, or mutation of 1β in up to 50% of HNSCC (21, 22). Extensive analysis of INK4A methylation in human HNSCC has also shown a high level of inactivation that correlated with a lack of p16INK4A protein expression by immunohistochemistry (23). We have previously identified p16INK4A as an independent prognostic indicator in SCC of the anterior tongue (6).

The ARF gene has an increased frequency of inactivation in recurrent HNSCC lesions compared with the primary lesion (22). Expression of p14ARF protein has not been investigated in HNSCC despite frequent aberrations in both the INK4A-ARF (21, 22) and TP53 gene loci (8, 9) and, thus, the goal of this study was to determine the relationship between p14ARF protein expression and disease recurrence or death in SCC of the anterior tongue. Because p14ARF seems to play pivotal roles in both the pRb and p53 pathways, independently as well as through its interaction with E2F-1 (24, 25), we did subgroup analyses to further assess the interaction of p14ARF with either p16INK4A or p53 protein expression in SCC of the anterior tongue.

Patient and tissue samples. Following Institutional Review Board approval (St. Vincent's Hospital Ethics Committee H01/28), we identified patients with primary operable SCC of the anterior tongue, with complete survival data and tissue previously used by Bova et al. (6) for cyclin D1 (n = 147), p16INK4A (n = 143), p53 (n = 143), and Ki-67 (n = 148) immunohistochemistry analyses. Three of these 148 patients (6) were excluded from pRb (n = 145) and E2F-1 (n = 145) immunohistochemistry analysis because of lack of invasive cancer specimens in Kwong et al.'s study (26) and because one further patient lacked tissue for the present study. Thus, p14ARF expression was examined on 144 patients. Overall, 140 patients had a complete clinicopathologic and molecular staining data set, and it is these 140 patients whose characteristics are presented in Table 1 and further analyzed in this study. All patients were treated with curative intent in the Departments of Head and Neck Surgery at St. Vincent's Hospital and Westmead Hospital, Sydney, New South Wales, Australia. Clinical follow-up was recorded to a minimum of 2 years or when the patient was diagnosed with recurrent disease. Updated survival data on all eligible patients increased the median duration of follow-up to 61 months (range: 6-161 months). Forty-one patients (29.3%) had recurrence of their disease, corresponding to an annual incidence of 6.8% [95% confidence interval (95% CI), 4.6-8.7]. The median time to disease recurrence was 41 months, with the 25th and 75th percentiles being 16 and 77 months, respectively. During the follow-up period, 33 patients (23.6%) died from their disease. The median OS was 49 months, with the 25th and 75th percentiles being 24 and 87 months, respectively.

Table 1.

Clinicopathologic, treatment, and outcome features of 140 patients with SCC of the anterior tongue

Clinicopathologic parameterNo. patients (n = 140)Percentage of patients (%)
Tumor stage*   
    I 75 53.6 
    II 53 37.9 
    III 10 7.1 
    IV 1.4 
Overall stage*   
    I 72 51.4 
    II 37 26.4 
    III 16 11.4 
    IV 15 10.8 
Lymph node stage*   
    N0 119 85.0 
    ≥N1 21 15.0 
Tumor grade   
    Well differentiated 37 26.4 
    Moderately differentiated 72 51.4 
    Poorly differentiated 31 22.1 
Adjuvant radiotherapy   
    Yes 53 37.8 
    No 87 62.1 
Recurrence   
    Yes 41 29.3 
    No 99 70.7 
Died of disease   
    Yes 33 23.6 
    No 107 76.4 
Clinicopathologic parameterNo. patients (n = 140)Percentage of patients (%)
Tumor stage*   
    I 75 53.6 
    II 53 37.9 
    III 10 7.1 
    IV 1.4 
Overall stage*   
    I 72 51.4 
    II 37 26.4 
    III 16 11.4 
    IV 15 10.8 
Lymph node stage*   
    N0 119 85.0 
    ≥N1 21 15.0 
Tumor grade   
    Well differentiated 37 26.4 
    Moderately differentiated 72 51.4 
    Poorly differentiated 31 22.1 
Adjuvant radiotherapy   
    Yes 53 37.8 
    No 87 62.1 
Recurrence   
    Yes 41 29.3 
    No 99 70.7 
Died of disease   
    Yes 33 23.6 
    No 107 76.4 
*

Tumor and lymph node stages were determined by pathologic analysis according to the tumor node metastases system. Data from 140 patients who had complete clinicopathologic and molecular staining data sets are presented.

Tumor grade was determined by pathologic analysis.

Tissue microarray. Tissue microarrays were constructed as outlined by Horvath et al. (27). H&E-stained template sections of the original paraffin-embedded tissue block were marked up for areas of invasive SCC of the tongue by a histopathologist (J.G. Kench). Verification of the pathology of each tissue core was also similarly done. Each patient case was represented by a mean of three (range, two to six) cylindrical 1.5 mm tissue biopsy cores distributed randomly over at least 2 of the 14 tissue microarray blocks.

Immunohistochemistry. Four-micrometer tissue sections were cut from formalin-fixed, paraffin-embedded tissue microarrays, dewaxed in xylene, and rehydrated through graded alcohol concentrations. For p14ARF, unmasking was achieved using EDTA citrate buffer (pH 8.0) solution at high temperature and pressure for 30 minutes, followed by rapid cooling within 2 minutes. Endogenous peroxidase activity was then quenched by 0.3% H2O2 treatment and the sections were treated with protein blocking solution (DAKO, Carpinteria, CA) to reduce background staining. The samples were processed using a DAKO autostainer with p14ARF antibody (FL-132, Santa Cruz Biotechnologies, Santa Cruz, CA) at a dilution of 1:75. The primary antibody was visualized with a secondary detection system, using Envision Plus antirabbit antibody, for 10 minutes. Color development was achieved using 3,3-diaminobenzidine (DAB kit, Vector Laboratories, Burlingame, CA) and hematoxylin was used as a counterstain. Further immunohistochemistry staining was done in the control cell lines, with an additional p14ARF antibody (p14ARF C-18, Santa Cruz Biotechnologies) that confirmed the specificity of the p14ARF FL-132 antibody (data not shown). HBL100 and T-47D breast carcinoma cell lines were used as positive controls. INK/ARF-deleted breast cancer cell lines MCF-7 and MDA-MB-231 (28) were used as negative controls. Positive and negative controls for p14ARF are shown in Fig. 1A to D. p21WAF1/CIP1 was processed using techniques outlined by Bova et al. (6) with p21WAF1/CIP1 antibody (1:200; Transduction Laboratories, Lexington, KY). Immunohistochemistry staining for p16INK4A, cyclin D1, E2F-1, pRb, p53, and Ki-67 on specimens in this database was published previously (6, 26).

Fig. 1.

Expression of p14ARF in SCC of the tongue. Original magnification, ×400, unless otherwise specified. p14ARF positive control: HBL100 (A) and T-47D (B) breast cancer cell lines (×200); negative control: MCF-7 (C) and MDA-MB-231 (D) breast cancer cell lines. E, positive p14ARF nuclear staining; F, negative p14ARF nuclear staining. G, positive nucleolar and negative nuclear staining of tongue SCC tissue (×1,000, oil immersion). H, negative nucleolar and positive nuclear staining of tongue SCC tissue (×1,000, oil immersion). NCL, nucleolus. NUC, nuclear.

Fig. 1.

Expression of p14ARF in SCC of the tongue. Original magnification, ×400, unless otherwise specified. p14ARF positive control: HBL100 (A) and T-47D (B) breast cancer cell lines (×200); negative control: MCF-7 (C) and MDA-MB-231 (D) breast cancer cell lines. E, positive p14ARF nuclear staining; F, negative p14ARF nuclear staining. G, positive nucleolar and negative nuclear staining of tongue SCC tissue (×1,000, oil immersion). H, negative nucleolar and positive nuclear staining of tongue SCC tissue (×1,000, oil immersion). NCL, nucleolus. NUC, nuclear.

Close modal

Immunohistochemical scoring. Each immunohistochemistry slide stained for p14ARF (FL-132) and p21WAF1/CIP1 was scored by two independent blinded observers (R.A. Kwong and L.H. Kalish for p14ARF and R.A. Kwong and R.J. Bova for p21WAF1/CIP1) in the same fashion as used for p16INK4A, cyclin D1, E2F-1, pRb, p53, and Ki-67 in the previous studies (6, 26). For each tissue microarray core, a score was accepted if it was within ≤10% for the two observers; however, if scores were significantly dissimilar (>10%), then the core was reviewed by both scorers simultaneously on a double-headed microscope until a consensus was reached. These finalized core scores were then collated for each tumor sample (represented by two to six tissue cores; i.e., 4-12 scores) and averaged to yield the final patient tumor score that was used for statistical analysis.

p14ARF was scored according to the percentage of positively stained tumor cell nuclei and whether positively staining nucleoli were present. A positive p14ARF score indicated positively staining nucleoli and/or an average nuclear expression of ≥30% in the tumor cores (Fig. 1E). Conversely, a negative score indicated tumor samples that had <30% positively stained nuclei and lacked nucleolar staining (Fig. 1F). An example of a cell demonstrating nucleolar positivity with intensely stained nucleoli is presented in Fig. 1G, whereas a nuclear positive cell with no nucleoli staining is shown in Fig. 1H. A summary of immunohistochemistry scoring for the molecular markers studied is shown in Table 2.

Table 2.

Immunohistochemical scoring system for molecular markers

CovariateScoring*(percentile)Range (median)
p14ARF (negative vs positive) Positive: 0-90% (33; for p14ARF nuclear staining) 
 1: Any nucleolar positive staining  
 2: Any nucleolar positive and <30% nuclear staining  
 3: Any nuclear staining ≥30%  
 Negative:  
 1: Only nucleolar negative and <30% nuclear staining (∼50th percentile for p14ARF nuclear staining)  
Cyclin D1 (positive vs negative) >10% vs ≤10% (∼25th percentile) 0-90% (50) 
p16INK4A (negative vs positive) 0% vs ≥1% (50th percentile) 0-95% (1) 
pRb (positive vs negative) ≥50% vs <50% (∼50th percentile) 0-95% (45) 
E2F-1 (negative vs positive) ≤35% vs >35% (75th percentile) 0-78% (15) 
p53 (+10% to ≤10% vs >10% to ≤70% vs >70% (33rd and 66th percentiles) 0-100% (50) 
p21WAF1/CIP1 (+10% to ≤25% vs >25% to ≤60% vs >60% (33rd and 66th percentiles) 0-90% (40) 
Ki67 (+10% to ≤20% vs >20% to ≤60% vs >60% (33rd and 66th percentiles) 5-90% (30) 
p14ARF/p53 (+1p14ARF-negative and >50% p53 nuclear expression vs p14ARF-positive and >50% p53 nuclear expression vs p14ARF-negative and ≤50% p53 nuclear expression vs p14ARF-positive and ≤50% p53 nuclear expression  
p14ARF/p16INK4A (+1p14ARF positive and p16INK4A positive vs p14ARF positive and p16INK4A negative or p14ARF negative and p16INK4A positive vs p14ARF negative and p16INK4A negative  
CovariateScoring*(percentile)Range (median)
p14ARF (negative vs positive) Positive: 0-90% (33; for p14ARF nuclear staining) 
 1: Any nucleolar positive staining  
 2: Any nucleolar positive and <30% nuclear staining  
 3: Any nuclear staining ≥30%  
 Negative:  
 1: Only nucleolar negative and <30% nuclear staining (∼50th percentile for p14ARF nuclear staining)  
Cyclin D1 (positive vs negative) >10% vs ≤10% (∼25th percentile) 0-90% (50) 
p16INK4A (negative vs positive) 0% vs ≥1% (50th percentile) 0-95% (1) 
pRb (positive vs negative) ≥50% vs <50% (∼50th percentile) 0-95% (45) 
E2F-1 (negative vs positive) ≤35% vs >35% (75th percentile) 0-78% (15) 
p53 (+10% to ≤10% vs >10% to ≤70% vs >70% (33rd and 66th percentiles) 0-100% (50) 
p21WAF1/CIP1 (+10% to ≤25% vs >25% to ≤60% vs >60% (33rd and 66th percentiles) 0-90% (40) 
Ki67 (+10% to ≤20% vs >20% to ≤60% vs >60% (33rd and 66th percentiles) 5-90% (30) 
p14ARF/p53 (+1p14ARF-negative and >50% p53 nuclear expression vs p14ARF-positive and >50% p53 nuclear expression vs p14ARF-negative and ≤50% p53 nuclear expression vs p14ARF-positive and ≤50% p53 nuclear expression  
p14ARF/p16INK4A (+1p14ARF positive and p16INK4A positive vs p14ARF positive and p16INK4A negative or p14ARF negative and p16INK4A positive vs p14ARF negative and p16INK4A negative  
*

Scoring was based on the percentage of positively staining nuclei unless otherwise specified.

Scoring was assessed for each unit increase.

Statistical analysis. Both univariate and multivariate analyses were done to assess the association between positive p14ARF expression and DFS and OS in relation to covariates. The clinicopathologic covariates considered were tumor stage, nodal stage, grade, pathologic stage, and treatment. The biochemical markers considered were p16INK4A (6), cyclin D1 (6), E2F-1 (26), pRb (26), p53 (6), and Ki-67 (6) protein expression. Preliminary analyses indicated that the distributions of a number of covariates were highly skewed and logarithmic transformation failed to normalize the distribution; therefore, for some covariates, the data were reduced into mutually exclusive categories. As a result, tumor stage, nodal stage, pathologic stage, and grade were each reduced into categories according to the clinical severity of cancer in addition to being assessed as continuous variables. Expression was analyzed as outlined on Table 2. Briefly, in line with previous publications, p16INK4A was assessed for complete absence of staining (0%) or positive nuclear expression (≥1%), whereas >10% and ≤10% positive tumor nuclei was used for cyclin D1 (6). Similarly, E2F-1 and pRb were examined using ≤35% and >35%, and ≥50% and <50% positive nuclear expression, respectively (26). For p21WAF1/CIP1, Ki-67, and p53, dichotomizations at each 10th percentile had no impact on outcome variables or multivariate models. To maximize the data analysis, trichotomizations (at the 33rd and 66th percentiles) were done for p53, p21WAF1/CIP1, and Ki-67 at 0% to ≤10%, >10% to ≤70%, and >70%; 0% to ≤25%, >25% to ≤60%, and >60%; 0% to ≤20%, >20% to ≤60%, and >60%, respectively. The outcome variables were assessed as time-to-event, which was defined as the difference between the time of diagnosis and the time of disease recurrence or death. In the univariate analysis, survival curves were constructed using the Kaplan-Meier method.

The differences in survival times between the categories were compared using the two-tailed log-rank statistic. Subsequent analysis involved the use of the Cox's proportional hazards model, with the SAS procedure of PHREG (29), to estimate the hazards ratio (HR) and 95% CI associated with each risk factor and covariate. In this analysis, variables meeting an entry criterion of P < 0.15 in the univariate analyses and for which data were complete (n = 140) were analyzed in a multivariate model. A backward stepwise selection procedure was used to confirm the “final” result, with variables being removed from the model according to a partial likelihood ratio test. All models met the proportional hazards assumption for Cox proportional test. No significant interactions were detected. All statistical tests were two sided.

p14ARF immunohistochemistry staining and immunoscoring. In our examination of p14ARF expression in the cohort of 140 tongue SCCs, we assessed nucleolar as well as nuclear p14ARF staining because p14ARF sequestration of Hdm/Mdm2 in the nucleolus has been implicated in p53 activation and, therefore, may be indicative of the growth inhibitory potential of p14ARF overexpression. Human p14ARF localizes predominantly in the nucleolus, which was difficult to accurately assess in patients with high nuclear p14ARF levels. When p14ARF is highly expressed, it is possible that enhanced nuclear accumulation could mask the presence of a minor nucleoplasmic pool that is nonetheless capable of executing the physiologic functions of p14ARF (30). Therefore, in this study, positive p14ARF immunostaining included all patients with nucleolar p14ARF staining and/or p14ARF nuclear staining in ≥30% of HNSCC cells (Fig. 1E and G).

Association of p14ARF expression with clinicopathologic parameters. The relationships between clinicopathologic parameters, previously assessed molecular markers, and p14ARF are shown on Table 3. Twenty percent (n = 29) of patients were negative for p14ARF staining. There were no significant associations between p14ARF negativity and tumor grade, overall stage, or lymph node stage, nor with the other molecular markers examined. However, p14ARF positivity was significantly associated with low tumor stages (I and II; P = 0.0388).

Table 3.

Association of clinicopathologic, treatment, and outcome features with p14ARF

Clinicopathologic parameterp14ARF nucleolar and/or ≥30% nuclear expression (n = 111)p14ARF no nucleolar and <30% nuclear expression (n = 29)χ2P
Tumor stage    
    I 60 15 0.0388 
    II 42 11  
    III  
    IV  
Overall stage    
    I 58 14 0.9296 
    II 29  
    III 13  
    IV 11  
Lymph node stage    
    N0 96 23 0.3352 
    ≥N1 15  
Tumor grade    
    WD 33 0.0910 
    MD 52 20  
    PD 26  
Radiotherapy    
    Yes 67 20 0.3949 
    No 44  
Recurrence    
    Yes 27 14 0.0116 
    No 84 15  
Died of disease    
    Yes 21 12 0.0112 
    No 90 17  
pRb    
    <50% 73 15 0.1635 
    ≥50% 38 14  
    E2F-1   0.1635 
    >35% 82 25  
    ≤35% 29  
    p16INK4A   0.3758 
    ≥1% 60 13  
    0% 51 16  
Cyclin D1   0.6787 
    >10% 72 20  
    ≤10% 39  
p53   0.5122 
    0% to ≤10% 41 11  
    >10% to ≤70% 36 12  
    >70% 34  
Ki67   0.7048 
    0% to ≤20% 44 11  
    >20% to ≤60% 51 12  
    >60% 16  
p21WAF1/CIP1   0.1144 
    0% to ≤25% 38 16  
    >25% to ≤60% 53 10  
    >60% 20  
Clinicopathologic parameterp14ARF nucleolar and/or ≥30% nuclear expression (n = 111)p14ARF no nucleolar and <30% nuclear expression (n = 29)χ2P
Tumor stage    
    I 60 15 0.0388 
    II 42 11  
    III  
    IV  
Overall stage    
    I 58 14 0.9296 
    II 29  
    III 13  
    IV 11  
Lymph node stage    
    N0 96 23 0.3352 
    ≥N1 15  
Tumor grade    
    WD 33 0.0910 
    MD 52 20  
    PD 26  
Radiotherapy    
    Yes 67 20 0.3949 
    No 44  
Recurrence    
    Yes 27 14 0.0116 
    No 84 15  
Died of disease    
    Yes 21 12 0.0112 
    No 90 17  
pRb    
    <50% 73 15 0.1635 
    ≥50% 38 14  
    E2F-1   0.1635 
    >35% 82 25  
    ≤35% 29  
    p16INK4A   0.3758 
    ≥1% 60 13  
    0% 51 16  
Cyclin D1   0.6787 
    >10% 72 20  
    ≤10% 39  
p53   0.5122 
    0% to ≤10% 41 11  
    >10% to ≤70% 36 12  
    >70% 34  
Ki67   0.7048 
    0% to ≤20% 44 11  
    >20% to ≤60% 51 12  
    >60% 16  
p21WAF1/CIP1   0.1144 
    0% to ≤25% 38 16  
    >25% to ≤60% 53 10  
    >60% 20  

NOTE: Statistically significant results are shown in bold.

Abbreviations: WD, well differentiated; MD, moderately differentiated; PD, poorly differentiated.

In a univariate Cox survival analysis (Table 4), p14ARF negativity was highly significantly associated with decreased DFS (HR, 2.865; 95% CI, 1.488-5.525; P = 0.0017) and decreased OS (HR, 3.115; 95% CI, 1.517-6.410; P = 0.0020). Kaplan-Meier survival analysis showed that patients with p14ARF-negative and p14ARF-positive cancers have a 41.4% and 78.3% 5-year DFS, respectively (Fig. 2A). Similar benefits were found for 5-year OS rates: Patients with p14ARF-negative cancers had a 5-year OS of 42.9% compared with an OS of 81% for patients with positive p14ARF staining (Fig. 2B). On univariate analysis, p16INK4A, cyclin D1, treatment, grade, lymph node status, and overall pathologic stage were significant on both DFS and OS (Table 4). With updated survival data, E2F-1 (26) was a significant predictor of increased OS (P = 0.0476), whereas pRb displayed a trend toward significance for DFS (P = 0.0633).

Table 4.

Univariate Cox's proportional hazards analysis for clinicopathologic parameters and molecular markers

DFS
OS
CovariateHR*95% CIPHR*95% CIP
p14ARF (negative vs positive) 2.865 1.488-5.525 0.0017 3.115 1.517-6.410 0.0020 
Cyclin D1 (>10% vs ≤10%) 2.985 1.321-6.711 0.0085 4.329 1.520-12.346 0.0061 
p16INK4A (0% vs ≥1%) 3.077 1.613-5.952 0.0008 3.559 1.667-7.634 0.0012 
pRb (≥50% vs <50%) 1.798 0.968-3.338 0.0633 1.782 0.897-3.536 0.0991 
E2F-1 (≤35% vs >35%) 1.377 0.637-2.985 0.4166 3.333 1.020-10.869 0.0476 
p53 (0% to ≤10% vs >10% to ≤70% vs >70%) 1.063 0.727-1.559 0.7529 1.221 0.800-1.862 0.3555 
p21WAF1/CIP1 (0% to ≤25% vs >25% to ≤60% vs >60%) 1.044 0.674-1.611 0.8468 1.262 0.787-2.021 0.3347 
Ki-67 (0% to ≤20% vs >20% to ≤60% vs >60%) 1.244 0.804-1.925 0.3264 1.339 0.836-2.140 0.2250 
Treatment (surgery alone vs surgery and radiotherapy) 2.058 1.114-3.800 0.0211 2.938 1.460-5.913 0.0025 
N stage§ (+14.204 2.160-8.182 <0.0001 6.876 3.408-13.875 <0.0001 
Overall stage§ (+11.461 1.100-1.939 0.0088 1.874 1.388-2.530 <0.0001 
T stage§ (+11.256 0.816-1.933 0.3007 1.687 1.091-2.609 0.0188 
Grade§ (+11.716 1.100-2.678 0.0174 1.978 1.195-3.274 0.0079 
DFS
OS
CovariateHR*95% CIPHR*95% CIP
p14ARF (negative vs positive) 2.865 1.488-5.525 0.0017 3.115 1.517-6.410 0.0020 
Cyclin D1 (>10% vs ≤10%) 2.985 1.321-6.711 0.0085 4.329 1.520-12.346 0.0061 
p16INK4A (0% vs ≥1%) 3.077 1.613-5.952 0.0008 3.559 1.667-7.634 0.0012 
pRb (≥50% vs <50%) 1.798 0.968-3.338 0.0633 1.782 0.897-3.536 0.0991 
E2F-1 (≤35% vs >35%) 1.377 0.637-2.985 0.4166 3.333 1.020-10.869 0.0476 
p53 (0% to ≤10% vs >10% to ≤70% vs >70%) 1.063 0.727-1.559 0.7529 1.221 0.800-1.862 0.3555 
p21WAF1/CIP1 (0% to ≤25% vs >25% to ≤60% vs >60%) 1.044 0.674-1.611 0.8468 1.262 0.787-2.021 0.3347 
Ki-67 (0% to ≤20% vs >20% to ≤60% vs >60%) 1.244 0.804-1.925 0.3264 1.339 0.836-2.140 0.2250 
Treatment (surgery alone vs surgery and radiotherapy) 2.058 1.114-3.800 0.0211 2.938 1.460-5.913 0.0025 
N stage§ (+14.204 2.160-8.182 <0.0001 6.876 3.408-13.875 <0.0001 
Overall stage§ (+11.461 1.100-1.939 0.0088 1.874 1.388-2.530 <0.0001 
T stage§ (+11.256 0.816-1.933 0.3007 1.687 1.091-2.609 0.0188 
Grade§ (+11.716 1.100-2.678 0.0174 1.978 1.195-3.274 0.0079 

NOTE: Statistically significant results are shown in bold.

*

Hazard ratios for p53 and cyclin D1 differ from previous publications (6, 26) due to changes in cutoffs; see text for details.

Hazards ratios were determined for both p16INK4A and E2F-1 as negative (i.e., 0% and ≤35%, respectively) versus positive (≥1% and >35%, respectively) rather than positive versus negative in the previous publication (25).

p53, p21WAF1/CIP1, and Ki-67 were not statistically significant predictors of outcome when dichotomized at each 10th percentile, in line with data for p53 and Ki-67 presented in prior publications (6, 25). Data were thus trichotomized at the 33rd and 66th percentiles.

§

N stage, overall stage, tumor stage, and grade were assessed as continuous variables rather than dichotomized as in the previous publications (6, 25).

Scoring was assessed for each unit increase.

Fig. 2.

Kaplan-Meier survival curves according to p14ARF status. Cumulative DFS (A) and cumulative OS (B) for p14ARF-positive (▪; n = 29) and p14ARF-negative (• with broken line; n = 111) patients.

Fig. 2.

Kaplan-Meier survival curves according to p14ARF status. Cumulative DFS (A) and cumulative OS (B) for p14ARF-positive (▪; n = 29) and p14ARF-negative (• with broken line; n = 111) patients.

Close modal

Multivariable models were constructed to determine the independence of p14ARF as a predictor of prognosis in SCC of the anterior tongue. Overall, p14ARF negativity was not only an independent predictor after adjusting for all variables but also maintained its HR of ∼2.5 in all three multivariate models tested. The strongest multivariate model for both DFS and OS included p14ARF, p16INK4A, cyclin D1, histopathologic grade, and lymph node status and is shown on Table 5. Decreased DFS was significantly associated with p14ARF negativity (HR, 2.506; 95% CI, 1.274-4.926), p16INK4A negativity (HR, 2.809; 95% CI, 1.439-5.495), and positive lymph node status (HR, 3.325; 95% CI, 1.655-6.681). Similarly, decreased OS was significantly associated with p14ARF negativity (HR, 2.832; 95% CI, 1.340-5.988), p16INK4A negativity (HR, 2.994; 95% CI, 1.368-6.579), positive lymph node status (HR, 5.459; 95% CI, 2.570-11.592), and each unit increase in the histopathologic grade (HR, 1.973; 95% CI, 1.110-3.508). When p14ARF was assessed against cyclin D1, p16INK4A, grade with either treatment (p14ARF DFS: HR, 2.571; 95% CI, 1.302-5.076; OS: HR, 2.778; 95% CI, 1.330-5.814), or overall stage (p14ARF DFS: HR, 2.437; 95% CI, 1.252-4.808; OS: HR, 2.660; 95% CI, 1.267-5.556), p14ARF remained an independent prognostic predictor of outcome for all models examined.

Table 5.

Multivariate Cox's proportional hazards analysis for clinicopathologic parameters and molecular markers

DFS
OS
Covariate (n = 140)HR95% CIPHR95% CIP
p14ARF (negative vs positive) 2.506 1.274-4.926 0.0078 2.832 1.340-5.988 0.0064 
Cyclin D1 (>10% vs ≤10%) 2.101 0.898-4.926 0.0872 2.525 0.847-7.519 0.0965 
p16INK4A (0% vs ≥1%) 2.809 1.439-5.495 0.0025 2.994 1.368-6.579 0.0061 
Grade (+1*) 1.604 0.977-2.631 0.0616 1.973 1.110-3.508 0.0206 
N stage (+1*)
 
3.325
 
1.655-6.681
 
0.0007
 
5.459
 
2.570-11.592
 
<0.0001
 
DFS
OS
Covariate (n = 140)HR95% CIPHR95% CIP
p14ARF (negative vs positive) 2.506 1.274-4.926 0.0078 2.832 1.340-5.988 0.0064 
Cyclin D1 (>10% vs ≤10%) 2.101 0.898-4.926 0.0872 2.525 0.847-7.519 0.0965 
p16INK4A (0% vs ≥1%) 2.809 1.439-5.495 0.0025 2.994 1.368-6.579 0.0061 
Grade (+1*) 1.604 0.977-2.631 0.0616 1.973 1.110-3.508 0.0206 
N stage (+1*)
 
3.325
 
1.655-6.681
 
0.0007
 
5.459
 
2.570-11.592
 
<0.0001
 

NOTE: Statistically significant results are shown in bold.

*

Scoring was assessed for each unit increase.

Association of combined p14ARF/p53 expression with clinicopathologic parameters. p14ARF expression is regulated by pRb through the E2F transcription factor family (19, 31), and thus p14ARF has the capacity to integrate both pRb- and p53-dependent mechanisms. Moreover, p53 represses transcription from the human ARF promoter and p14ARF and p53 status are inversely correlated in human tumor cell lines (32). To investigate the relationship between p14ARF and p53 in SCC of the tongue, we examined the correlations between the expression of these molecules in our cohort. Although there was no significant association between p14ARF and p53 staining by immunohistochemistry (P = 0.5122), on Kaplan-Meier analysis, there was a significant survival disadvantage apparent in patients who were negative for p14ARF and had high p53 expression (>50% positive nuclear staining representing the 50th percentile; n = 13) compared with p14ARF-negative patients who had low p53 expression (≤50% positive nuclear staining; n = 17) and p14ARF-positive patients, irrespective of p53 expression (n = 110; DFS: P = 0.0030; OS: P < 0.0001; Fig. 3A and B). This p14ARF-negative, high p53 expression subgroup of patients had a 12% 5-year survival rate for both DFS and OS compared with 77% to 80% for the remaining patients. When further investigated on multivariate analyses (Table 6), this combination of molecular markers was the strongest independent predictor of DFS and OS in multivariate models, including lymph node status, treatment, or overall stage. The p14ARF-negative/high p53 (>50% positive nuclear staining) subgroup with lymph node status (DFS: HR, 6.002; 95% CI, 2.275-15.832; OS: HR, 7.319; 95% CI, 2.418-22.156), with treatment (DFS: HR, 5.747; 95% CI, 2.126-15.353; OS: HR, 5.961; 95% CI, 2.621-13.557), or with overall stage (DFS: HR, 4.979; 95% CI, 1.925-12.879; OS: HR, 6.268; 95% CI, 2.692-14.594) had a higher HR than most established clinicopathologic parameters in this cohort. These data support a potential biological relationship between p14ARF and p53 in SCC of the tongue, but interpretation is limited by the small patient numbers in these subgroup analyses.

Fig. 3.

Kaplan-Meier survival curves according to p14ARF/p53 status and p14ARF/p16INK4A. Cumulative DFS (A) and cumulative OS (B) for p14ARF/p53 status: p14ARF positive and p53 >50% (▾ with broken line; n = 56); p14ARF negative and p53 ≤50% (▴ with dashed line; n = 17); p14ARF positive and p53 ≤50% (▪; n = 54); and p14ARF negative and p53 >50% (•; n = 13) patients. Cumulative DFS (C) and cumulative OS (D) for p14ARF/p16INK4A status: for both p14ARF and p16INK4A positive (▾ with broken line; n = 54); either p14ARF or p16INK4A (▴ with dashed line; n = 70); and both p14ARF and p16INK4A negative (•; n = 16) patients.

Fig. 3.

Kaplan-Meier survival curves according to p14ARF/p53 status and p14ARF/p16INK4A. Cumulative DFS (A) and cumulative OS (B) for p14ARF/p53 status: p14ARF positive and p53 >50% (▾ with broken line; n = 56); p14ARF negative and p53 ≤50% (▴ with dashed line; n = 17); p14ARF positive and p53 ≤50% (▪; n = 54); and p14ARF negative and p53 >50% (•; n = 13) patients. Cumulative DFS (C) and cumulative OS (D) for p14ARF/p16INK4A status: for both p14ARF and p16INK4A positive (▾ with broken line; n = 54); either p14ARF or p16INK4A (▴ with dashed line; n = 70); and both p14ARF and p16INK4A negative (•; n = 16) patients.

Close modal
Table 6.

Multivariate Cox's proportional hazards analysis for clinicopathologic parameters and molecular markers for either p14ARF/p53 or p14ARF/p16INK4A

DFS
OS
Covariate (n = 140)HR95% CIPHR95% CIP
p14ARF/p53 [p14ARF (−) and p53 (>50%) vs other*] 6.002 2.275-15.832 0.0003 7.319 2.418-22.156 <0.0001 
p14ARF/p16INK4A (negative vs other4.386 2.105-9.174 <0.0001 4.048 1.808-4.049 0.0009 
Cyclin D1 (>10% vs ≤10%) 1.821 0.776-4.274 0.1687 2.646 0.904-7.752 0.0756 
Grade (+11.524 0.931-2.494 0.0937 1.868 1.080-3.231 0.0253 
N stage (+13.768 1.888-7.519 0.002 7.367 3.465-15.663 <0.0001 
DFS
OS
Covariate (n = 140)HR95% CIPHR95% CIP
p14ARF/p53 [p14ARF (−) and p53 (>50%) vs other*] 6.002 2.275-15.832 0.0003 7.319 2.418-22.156 <0.0001 
p14ARF/p16INK4A (negative vs other4.386 2.105-9.174 <0.0001 4.048 1.808-4.049 0.0009 
Cyclin D1 (>10% vs ≤10%) 1.821 0.776-4.274 0.1687 2.646 0.904-7.752 0.0756 
Grade (+11.524 0.931-2.494 0.0937 1.868 1.080-3.231 0.0253 
N stage (+13.768 1.888-7.519 0.002 7.367 3.465-15.663 <0.0001 

NOTE: Statistically significant results are shown in bold.

*

Patients were considered in four categories: p14ARF-negative and p53 nuclear expression >50% vs p14ARF-positive and p53 nuclear expression >50% vs p14ARF-negative and p53 nuclear expression ≤50% vs p14ARF-positive and p53 nuclear expression ≤50%.

Patients were considered in two categories: patients with both p14ARF and p16INK4A negative vs patients with both p14ARF and p16INK4A positive and patients with either p14ARF or p16INK4A negative but not both.

Scoring was assessed for each unit increase.

Association of combined p14ARF/p16INK4A expression with clinicopathologic parameters. We further investigated whether the influence of p14ARF could be a consequence of influences in the pRb pathway via the intimate relationship of the overlapping INK4A and ARF loci. Surprisingly, there was no correlation between p14ARF and p16INK4A immunoreactivity (P = 0.3758), with p16INK4A negativity equally likely in both p14ARF-positive and p14ARF-negative groups. However, because loss of each protein is associated with deregulation of distinct pathways, we further analyzed the data to assess the effect of loss of one or both proteins. Sixteen patients (11.4%) were negative for both p14ARF and p16INK4A immunostaining. Although this analysis must be interpreted with caution because of the small size of this subgroup, loss of both p14ARF and p16INK4A resulted in significantly decreased DFS (P < 0.0001) and OS (P < 0.0001) on univariate analysis compared with patients who retained either or both p14ARF or p16INK4A. Of the 16 patients with combined p14ARF and p16INK4A loss, 11 patients (69%) had disease recurrence within the first 15 months following surgery and 9 (56%) had died of disease by 25 months. There was no correlation between combined p14ARF/p16INK4A-negative scoring and overall stage (P = 0.373), lymph node status (P = 0.180), treatment (P = 0.856), or histopathologic grade (P = 0.130). To fully assess the independence of this combination of molecular markers, multivariate analysis was done (Table 6). Combined p14ARF/p16INK4A negativity was a strong independent predictor of both poor DFS and OS in all three models assessed; when assessed with cyclin D1, tumor grade and either lymph node status (DFS: HR, 4.386; 95% CI, 2.105-9.174; OS: HR, 4.048; 95% CI, 1.808-4.049), or treatment (DFS: HR, 4.695; 95% CI, 2.247-9.804; OS: HR, 4.329; 95% CI, 1.961-9.524), or overall stage (DFS: HR, 4.444; 95% CI, 2.137-9.259; OS: HR, 3.891; 95% CI, 1.748-8.621). Kaplan-Meier analysis was done comparing combined p14ARF/p16INK4A negativity with combined p14ARF/p16INK4A positivity or positive staining for either p14ARF or p16INK4A alone (Fig. 3C and D). This revealed a significant association between worsening DFS (P < 0.0001) and OS (P < 0.0001) and loss of expression of one or both proteins. Assessment of combined loss, loss of one protein alone, or expression of both proteins showed a 67%, 26%, and 15% 5-year incidence of disease recurrence and a 53%, 24%, and 12% 5-year incidence of disease-specific death, respectively.

We have examined the relationship between p14ARF protein expression and outcome in SCC of the anterior tongue, demonstrating p14ARF as an independent predictor of DFS and OS in the presence of concurrent clinicopathologic predictors, including regional lymph node status. Thus, a predictor of outcome, p14ARF has potential advantages over the existing tumor node metastases staging system (33) where limited predictive ability is illustrated by varied recurrence rates within each tumor node metastases substage, and additionally confounded by the fact that the majority of patients present with no nodal metastasis at the time of diagnosis (34, 35).

We found significant advantages in 5-year outcome for both DFS and OS for patients with positive p14ARF immunoscoring (i.e., displaying nucleolar p14ARF and/or >30% nuclear p14ARF staining). This strong relationship was not observed when p14ARF nuclear positivity alone was considered, whereas for patients with positive p14ARF expression there was no difference in survival in patients with only nuclear staining compared with those with only nucleolar staining (data not shown). The nucleolus is best understood as the site for synthesis of rRNA, ribosome biogenesis, and shuttling of some mRNA species, and some recent studies have provided evidence that p14ARF may regulate ribosome biogenesis (36, 37). The predominant nucleolar localization of p14ARF and the murine homologue, p19ARF (14, 32), in contrast with the localization of its targets Hdm2/Mdm2 and p53 in the nucleoplasm, has raised questions over the mechanism by which p14ARF activates p53. In SCC of the cervix (HeLa) cells, levels of endogenous p14ARF and its distribution between the nucleolus and nucleoplasm are sensitive to changes in cell morphology, cell cycle, and nucleolar function (38). However, the role of p14ARF localization in its function remains unclear, with studies showing that nucleolar localization is not sufficient for growth inhibition (17) and is dispensable for relocalization and stabilization of Hdm2/Mdm2 and p53 (16, 39). This contrasts with the earlier view that sequestration of Hdm2/Mdm2 in the nucleolus is an essential component of p14ARF function. Our observation that incorporating the localization of p14ARF into immunohistochemistry scoring is important in identifying the effect of p14ARF on outcome suggests that the subcellular localization of p14ARF may be of functional significance in some aspects of HNSCC biology in vivo. Furthermore, the demonstration that patients with high p14ARF expression had a significantly improved DFS and OS independent of other molecules in the pRb and p53 pathways that impact on prognosis, including cyclin D1, E2F-1, or p53, implicates p14ARF as having a distinct role as a tumor suppressor in SCC of the anterior tongue.

The biological and clinical consequences of genetic defects in the INK4A/ARF locus in HNSCC are unclear and p14ARF protein expression has not been analyzed previously in HNSCC, although increased alterations to ARF through promoter methylation and/or point mutations have been shown in tumor recurrences of HNSCC (22). One unanswered question is whether the two gene products of INK4A/ARF are simultaneously affected or whether they can be selectively inactivated in HNSCC. We found that few tumors were negative for p14ARF alone (n = 13) compared with p16INK4A alone (n = 51) or both (n = 16). Patients displaying loss of p14ARF in these cancers (DFS: 46%; OS: 45%) had a trend toward a shorter 5-year OS and DFS than those lacking only p16INK4A (DFS: 73%; OS: 76%). Mice lacking p16INK4A, either through specific inactivating mutations or methylation of its promoter, develop spontaneous tumors but at a lower frequency and latency than those that lack p19ARF or both p19ARF and p16INK4A (40). Mice hemizygous for p19ARF do not develop tumors unless p16INK4A is also lost, suggesting that p19ARF and p16INK4A cooperate in tumorigenesis (41) and that p53 pathway control, which utilizes p19ARF, is critical in murine cells. Data from studies of human cells provide a different perspective, suggesting that p16INK4A is the major tumor suppressor of the INK4A/ARF locus. For example, p16INK4A is more commonly altered by point mutation or deletion than p14ARF in melanoma (40) and recent in vitro studies indicate that a reduction in p14ARF expression enhances growth but is not tumorigenic, whereas loss of p16INK4A expression does not enhance growth but can cause transformation when p53 is also inactivated (42). Whatever the implications of loss of either p14ARF or p16INK4A alone, loss of both confers very poor outcome. The predictive value of combined p14ARF/p16INK4A negativity was stronger for both DFS and OS than that of either p14ARF or p16INK4A alone (Tables 5 and 6).

The TP53 and INK4A/ARF loci encompass the most frequently inactivated genes in human cancer (43, 44) and numerous studies support a model where p14ARF and p53 have interdependent roles. Because p14ARF stabilizes p53, it has been proposed that the loss of p14ARF may be functionally equivalent to a p53 mutation. Consistent with this idea, many human tumors that retain wild-type p53 suffer loss of p14ARF and are, therefore, unable to activate p53 in response to abnormal proliferative signals. Conversely, cells lacking p53 are refractory to p14ARF-induced arrest (32, 41). However, in our cohort of HNSCC, high expression of p53 occurred at similar frequency in the p14ARF-positive and p14ARF-negative subgroups. When the cohort was dichotomized at the median into low and high p53 expressors, it was clear that there was an interaction between this parameter and p14ARF that had implications for outcome. Patients with both reduced p14ARF expression and high expression of p53 had the worst prognosis and reduced 5-year DFS and OS (12%) compared with patients who retained p14ARF expression and low p53 expression (77-80%). Previous studies have indicated that p53 mutation is usually associated with increased expression of immunoreactive protein (45); therefore, increased p53 expression in our cohort may reflect p53 mutation. Although this analysis is based on a small number of patients and must therefore be interpreted with caution, these data not only support the critical involvement of p14ARF in the p53 pathway but also suggest that p14ARF may have significant antineoplastic effects independent of p53 in SCC of the anterior tongue.

Activation of p53 by p14ARF is associated with up-regulation of the p53-responsive gene, p21WAF1/CIP1 (32, 46). As a potent Cdk inhibitor, p21WAF1/CIP1 is a major effector of p53-mediated cell cycle arrest. Because p14ARF does not directly interact with cyclins or Cdks (4), a prevailing concept is that p21WAF1/CIP1 mediates p14ARF-induced growth arrest. Thus, in addition to the induction of p14ARF by E2F-1, connections between p14ARF and the pRb pathway are indicated by the observations that p14ARF inhibits growth through pRb, based on the up-regulation of p21WAF1/CIP1, inhibition of Cdks, and the presence of hypophosphorylated pRb in wild-type p14ARF-arrested cells (4, 32, 47). However, other studies have shown that p14ARF can function independent of pRb (48) and p21WAF1/CIP1 is not required for p14ARF- and p53-mediated growth arrest, indicating that multiple downstream effectors exist that mediate the growth-suppressive functions of p14ARF (49). The observed molecular associations are consistent with proposed biological molecular interactions that support a role for p14ARF as a tumor suppressor, having effects in both pRb and p53 pathways. Outcome in this cohort of SCC of the anterior tongue was affected by interactions between p14ARF and the pRb pathway (via p16INK4A) as well as between p14ARF and p53.

In summary, we have shown, for the first time, the importance of p14ARF as an independent prognostic indicator in SCC of the anterior tongue. Until now, the role of p14ARF in HNSCC has been largely overshadowed by its counterpart, p16INK4A. However, our data indicate that loss of p14ARF expression is associated with poor outcome independent of other clinicopathologic markers in multivariate analysis. Furthermore, the biochemical activity of p14ARF as an activator of p53 and its biological role as an inducer of apoptosis in tumor cells suggest that reintroduction of p14ARF may have therapeutic utility. The importance of p14ARF in the pathogenesis of SCC of the anterior tongue, given its possible therapeutic and prognostic value, warrants further investigation. The suggestion from our data that interactions between p14ARF and p16INK4A or p53 status can have substantial effects on outcome particularly merits further attention.

Grant support: Garnett Passe and Rodney Williams Memorial Foundation, National Health and Medical Research Council of Australia, Cancer Council New South Wales, and R.T. Hall Trust.

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.

Note: R.A. Kwong and L.H. Kalish contributed equally to this work and are recipients of National Health and Medical Research Council postgraduate research fellowships. R.A. Kwong is also a recipient of a Garnett Passe and Rodney Williams research fellowship. E.A. Musgrove is a Cancer Institute NSW Fellow.

1
Registry NCC. Cancer in NSW incidence and mortality 2001 report. Sydney (Australia): NSW Cancer Council; 2003. p. 288.
2
Kleihues P, Stewart BW. World cancer report. Geneva (Switzerland): WHO and IARC; 2003.
3
Sherr CJ, McCormick F. The RB and p53 pathways in cancer.
Cancer Cell
2002
;
2
:
103
–12.
4
Quelle DE, Zindy F, Ashmun RA, Sherr CJ. Alternative reading frames of the INK4a tumor suppressor gene encode two unrelated proteins capable of inducing cell cycle arrest.
Cell
1995
;
83
:
993
–1000.
5
Piboonniyom SO, Timmermann S, Hinds P, Munger K. Aberrations in the MTS1 tumor suppressor locus in oral squamous cell carcinoma lines preferentially affect the INK4A gene and result in increased cdk6 activity.
Oral Oncol
2002
;
38
:
179
–86.
6
Bova RJ, Quinn DI, Nankervis JS, et al. Cyclin D1 and p16INK4A expression predict reduced survival in carcinoma of the anterior tongue.
Clin Cancer Res
1999
;
5
:
2810
–9.
7
Michalides RJ, Van De Brekel M, Balm F. Defects in G1-S cell cycle control in head and neck cancer: a review.
Head Neck
2002
;
24
:
694
–704.
8
van Houten VM, Tabor MP, van den Brekel MW, et al. Mutated p53 as a molecular marker for the diagnosis of head and neck cancer.
J Pathol
2002
;
198
:
476
–86.
9
Weber A, Bellmann U, Bootz F, Wittekind C, Tannapfel A. Expression of p53 and its homologues in primary and recurrent squamous cell carcinomas of the head and neck.
Int J Cancer
2002
;
99
:
22
–8.
10
Bates S, Hickman ES, Vousden KH. Reversal of p53-induced cell-cycle arrest.
Mol Carcinog
1999
;
24
:
7
–14.
11
Bates S, Vousden KH. Mechanisms of p53-mediated apoptosis.
Cell Mol Life Sci
1999
;
55
:
28
–37.
12
Kamijo T, Weber JD, Zambetti G, Zindy F, Roussel MF, Sherr CJ. Functional and physical interactions of the ARF tumor suppressor with p53 and Mdm2.
Proc Natl Acad Sci U S A
1998
;
95
:
8292
–7.
13
Tao W, Levine AJ. P19(ARF) stabilizes p53 by blocking nucleo-cytoplasmic shuttling of Mdm2.
Proc Natl Acad Sci U S A
1999
;
96
:
6937
–41.
14
Weber JD, Taylor LJ, Roussel MF, Sherr CJ, Bar-Sagi D. Nucleolar Arf sequesters Mdm2 and activates p53.
Nat Cell Biol
1999
;
1
:
20
–6.
15
Pomerantz J, Schreiber-Agus N, Liegeois NJ, et al. The Ink4a tumor suppressor gene product, p19Arf, interacts with MDM2 and neutralizes MDM2's inhibition of p53.
Cell
1998
;
92
:
713
–23.
16
Llanos S, Clark PA, Rowe J, Peters G. Stabilization of p53 by p14ARF without relocation of MDM2 to the nucleolus.
Nat Cell Biol
2001
;
3
:
445
–52.
17
Korgaonkar C, Zhao L, Modestou M, Quelle DE. ARF function does not require p53 stabilization or Mdm2 relocalization.
Mol Cell Biol
2002
;
22
:
196
–206.
18
Yarbrough WG, Bessho M, Zanation A, Bisi JE, Xiong Y. Human tumor suppressor ARF impedes S-phase progression independent of p53.
Cancer Res
2002
;
62
:
1171
–7.
19
Bates S, Phillips AC, Clark PA, et al. p14ARF links the tumour suppressors RB and p53.
Nature
1998
;
395
:
124
–5.
20
Khleif SN, DeGregori J, Yee CL, et al. Inhibition of cyclin D-CDK4/CDK6 activity is associated with an E2F-mediated induction of cyclin kinase inhibitor activity.
Proc Natl Acad Sci U S A
1996
;
93
:
4350
–4.
21
Shintani S, Nakahara Y, Mihara M, Ueyama Y, Matsumura T. Inactivation of the p14(ARF), p15(INK4B) and p16(INK4A) genes is a frequent event in human oral squamous cell carcinomas.
Oral Oncol
2001
;
37
:
498
–504.
22
Weber A, Bellmann U, Bootz F, Wittekind C, Tannapfel A. INK4a-ARF alterations and p53 mutations in primary and consecutive squamous cell carcinoma of the head and neck.
Virchows Arch
2002
;
441
:
133
–42.
23
Reed AL, Califano J, Cairns P, et al. High frequency of p16 (CDKN2/MTS-1/INK4A) inactivation in head and neck squamous cell carcinoma.
Cancer Res
1996
;
56
:
3630
–3.
24
Palmero I, Murga M, Zubiaga A, Serrano M. Activation of ARF by oncogenic stress in mouse fibroblasts is independent of E2F1 and E2F2.
Oncogene
2002
;
21
:
2939
–47.
25
Parisi T, Pollice A, Di Cristofano A, Calabro V, La Mantia G. Transcriptional regulation of the human tumor suppressor p14(ARF) by E2F1, E2F2, E2F3, and Sp1-like factors.
Biochem Biophys Res Commun
2002
;
291
:
1138
–45.
26
Kwong RA, Nguyen TV, Bova RJ, et al. Overexpression of E2F-1 is associated with increased disease-free survival in squamous cell carcinoma of the anterior tongue.
Clin Cancer Res
2003
;
9
:
3705
–11.
27
Horvath L, Henshall S. The application of tissue microarrays to cancer research.
Pathology
2001
;
33
:
125
–9.
28
Sutherland RL, Watts CKW, Lee CS, Musgrove EA. Human cell culture. Vol. 2. London (United Kingdom): Kluwer Academic Publishers; 1999. p. 79–106.
29
SAS Institute Inc. SAS technical report P-229, SAS/STAT software: changes and enhancements. Release 6.07 ed. Cary (NC): SAS Institute, Inc.; 1992. p 443–80.
30
Lindstrom MS, Klangby U, Inoue R, et al. Immunolocalization of human p14(ARF) to the granular component of the interphase nucleolus.
Exp Cell Res
2000
;
256
:
400
–10.
31
DeGregori J, Leone G, Miron A, Jakoi L, Nevins JR. Distinct roles for E2F proteins in cell growth control and apoptosis.
Proc Natl Acad Sci U S A
1997
;
94
:
7245
–50.
32
Stott FJ, Bates S, James MC, et al. The alternative product from the human CDKN2A locus, p14(ARF), participates in a regulatory feedback loop with p53 and MDM2.
EMBO J
1998
;
17
:
5001
–14.
33
Spiessl B, Beahrs OH, Hermanek P, et al. TNM atlas illustrated guide to the TNM/pTNM classification of malignant tumours. 3rd ed. Berlin (Germany): Springer-Verlag; 1992.
34
Mamelle G, Pampurik J, Luboinski B, Lancar R, Lusinchi A, Bosq J. Lymph node prognostic factors in head and neck squamous cell carcinomas.
Am J Surg
1994
;
168
:
494
–8.
35
Salesiotis AN, Cullen KJ. Molecular markers predictive of response and prognosis in the patient with advanced squamous cell carcinoma of the head and neck: evolution of a model beyond TNM staging.
Curr Opin Oncol
2000
;
12
:
229
–39.
36
Itahana K, Bhat KP, Jin A, et al. Tumor suppressor ARF degrades B23, a nucleolar protein Involved in ribosome biogenesis and cell proliferation.
Mol Cell
2003
;
12
:
1151
–64.
37
Sugimoto M, Kuo ML, Roussel MF, Sherr CJ. Nucleolar Arf tumor suppressor inhibits ribosomal RNA processing.
Mol Cell
2003
;
11
:
415
–24.
38
David-Pfeuty T, Nouvian-Dooghe Y. Human p14(Arf): an exquisite sensor of morphological changes and of short-lived perturbations in cell cycle and in nucleolar function.
Oncogene
2002
;
21
:
6779
–90.
39
Lin AW, Lowe SW. Oncogenic ras activates the ARF-p53 pathway to suppress epithelial cell transformation.
Proc Natl Acad Sci U S A
2001
;
98
:
5025
–30.
40
Sherr CJ. The INK4a/ARF network in tumour suppression.
Nat Rev Mol Cell Biol
2001
;
2
:
731
–7.
41
Kamijo T, Zindy F, Roussel MF, et al. Tumor suppression at the mouse INK4a locus mediated by the alternative reading frame product p19ARF.
Cell
1997
;
91
:
649
–59.
42
Voorhoeve PM, Agami R. The tumor-suppressive functions of the human INK4A locus.
Cancer Cell
2003
;
4
:
311
–9.
43
Hollstein M, Rice K, Greenblatt MS, et al. Database of p53 gene somatic mutations in human tumors and cell lines.
Nucleic Acids Res
1994
;
22
:
3551
–5.
44
Ruas M, Peters G. The p16INK4a/CDKN2A tumor suppressor and its relatives.
Biochim Biophys Acta
1998
;
1378
:
F115
–77.
45
Xu L, Chen YT, Huvos AG, et al. Overexpression of p53 protein in squamous cell carcinomas of head and neck without apparent gene mutations.
Diagn Mol Pathol
1994
;
3
:
83
–92.
46
Palmero I, Pantoja C, Serrano M. p19ARF links the tumour suppressor p53 to Ras.
Nature
1998
;
395
:
125
–6.
47
Kurokawa K, Tanaka T, Kato J. p19ARF prevents G1 cyclin-dependent kinase activation by interacting with MDM2 and activating p53 in mouse fibroblasts.
Oncogene
1999
;
18
:
2718
–27.
48
Weber JD, Jeffers JR, Rehg JE, et al. p53-independent functions of the p19(ARF) tumor suppressor.
Genes Dev
2000
;
14
:
2358
–65.
49
Modestou M, Puig-Antich V, Korgaonkar C, Eapen A, Quelle DE. The alternative reading frame tumor suppressor inhibits growth through p21-dependent and p21-independent pathways.
Cancer Res
2001
;
61
:
3145
–50.