Loss of Nuclear p16 Protein Expression Correlates with Increased Tumor Cell Proliferation (Ki-67) and Poor Prognosis in Patients with Vertical Growth Phase Melanoma¹

Alterations of the tumor suppressor gene CDKN2A(p16^INK4a) seem to play an important role in several tumor types (1, 2, 3), and the corresponding p16 protein acts by blocking progression through the cell cycle by its inhibitory effect on the cyclin D/CDK4-pRb pathway (4). Several studies of malignant melanoma have shown that the p16 gene might be inactivated by homozygous deletions (5), point mutations (6, 7, 8, 9), or methylation of the promoter region (10). The gene is frequently inactivated by germ-line mutations in familial melanoma kindreds (11). In sporadic melanomas, p16 mutations in primary tumors were reported to be rare (12, 13), although a recent study by Kumar et al. (9) found a mutation frequency of 26% in their series of microdissected cases. Loss of p16 protein expression by immunohistochemistry has been correlated with advanced stages of melanoma progression (14, 15, 16, 17, 18). In a recent study, we found that the subgroup of patients with p16 protein loss had a significantly reduced survival (19). This is in line with studies on other tumor types indicating that p16 alterations are associated with aggressive tumors and poor prognosis, e.g., pancreatic carcinoma (20, 21), squamous cell carcinoma of the esophagus (22),leukemia (23), and lung carcinoma (24). The aim of the present study was to further examine the frequency and prognostic significance of aberrant p16 protein expression in a large and consecutive series of 202 primary vertical growth phase melanomas and their metastases and to study the relationship with other prognostic markers, especially tumor cell proliferation (Ki-67 staining) and p53 protein expression.

Of all melanomas occurring in Hordaland County (10% of the Norwegian population) during 1981–1997, 97.5% were diagnosed at the Department of Pathology, The Gade Institute, Haukeland University Hospital. There were no differences in sex, anatomical site, or stage between these cases and the 2.5% of cases with a diagnosis from other laboratories, although the latter patients were 6 years younger (median age). The aim of this study was to focus on tumorigenic vertical growth phase melanomas without a radial growth phase. After microscopic review of all cases diagnosed and recorded as malignant melanoma of the nodular type or not otherwise specified during this period, 202 cases were included. There was no history of familial occurrence. The presence of a vertical growth phase and the lack of a radial growth phase, i.e., adjacent in situ or microinvasive component, were used as inclusion criteria (25). Altogether, 187 cases had sufficient material left for analyses of all markers (p16, p53, and Ki-67 proteins). In addition, 68 separate biopsies of local (skin; n = 17), regional (lymph nodes; n = 44), or distant (n = 7)metastases from 58 patients with recurrent disease were available for analyses. Biopsies from multiple sites were available for eight patients. In these cases, the largest biopsy (generally the lymph node)was selected for further analyses. Clinicopathological characteristics and some survival data have been reported previously for cases occurring in 1981–1989 (19, 26).

Complete information on patient survival and time and cause of death was available in all 202 cases, and information on recurrence-free survival was available in 167 patients.

The following variables were recorded: (a) date of histological diagnosis; (b) sex; (c) age at diagnosis; (d) anatomical site of the primary tumor; and(e) presence of local (skin), regional (lymph node), or distant metastases at diagnosis. The H&E-stained slides were reexamined, and the following histological features were also included:(a) tumor thickness according to Breslow (27);(b) histological tumor diameter (measured horizontally on a central cross-section); (c) level of invasion according to Clark et al. (28); (d) microscopic ulceration; and (e) vascular invasion (each case was examined closely using both H&E- and factor VIII-stained slides). Staging was done according to the AJCC³ pTNM(tumor-node-metastasis) staging system, including both clinical and histological data.

The immunohistochemical staining was performed on formalin-fixed and paraffin-embedded archival tissue. Thin sections (5 μm) were incubated with the polyclonal p16 antibody SC-468 (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at room temperature. For the cases from 1981–1989, we used the slides that had been stained previously with an antibody dilution of 1:500 (19). Using the same protocol for cases from 1990–1997, an antibody dilution of 1:200 was determined to obtain an equivalent staining reaction by comparing positive control slides from identical specimens. This difference is probably due to the fact that a different antibody batch was used for the last part of the series, although minor differences in other technical details like routine fixation procedures cannot be excluded. It has been suggested by others that antigen retrieval by microwave pretreatment might suppress p16 immunoreactivity (29), and protocols with or without retrieval were tested. Strong staining (in positive controls) was observed without retrieval pretreatment, whereas clearly weaker staining was present after microwave treatment, which was therefore not used in our study. Nuclear staining in stromal cells and adnexal glands was used as a positive internal control, whereas cases of endometrial carcinoma, known to be positive for p16 protein expression, were used as positive external controls.

The p53 staining protocol included microwave antigen retrieval (10 min at 750 W and 3 × 5 min at 500 W) and incubation for 1 h at room temperature with the DO-7 monoclonal antibody (M-700; DAKO,Copenhagen, Denmark) diluted 1:100 (30).

After microwave antigen retrieval (10 min at 750 W and 4 × 5 min at 500 W), the sections were incubated for 1 h at room temperature with the polyclonal Ki-67 antibody (A-047; DAKO) diluted 1:50 (30).

The staining procedures were performed on DAKO TechMate 500 slide processing equipment, using the standard avidin-biotin complex method. Finally, the peroxidase was localized by the 3-amino-9-ethylcarbazole (AEC) peroxidase reaction, using Harris hematoxylin as a brief (30 s) counterstain. Negative controls were obtained by omitting the primary antibody.

Immunohistochemical staining of p16 and 53 proteins was recorded as described previously (19), considering both the staining intensity and the proportion of positive tumor cells. Intensity was graded as follows: (a) 0, no staining;(b) 1, weak staining; (c) 2, moderate staining;and (d) 3, strong staining. Area was graded as follows:(a) 0, no positive tumor cells; (b) 1, <10%positive tumor cells; (c) 2, 10–50% positive tumor cells;and (d) 3, >50% positive tumor cells. Staining index was computed as follows: staining index = staining intensity ×positive area.

Combined examination of p16 expression in both nuclear and cytoplasmic compartments has been used by some researchers (14). Because cytoplasmic staining is controversial and regarded as nonspecific by others (15, 18, 29), we decided to examine cytoplasmic and nuclear staining separately.

Ki-67 staining was assessed according to the approach of Weidner et al. (31). Briefly, the tumors were scanned at low magnification (×40 and ×100) to identify the areas of most intense nuclear staining (hot spots). The percentage of immunoreactive tumor cell nuclei (the proliferative rate) was then calculated by counting at least 500 cells at ×1000 magnification within the selected areas.

Hospital records were used to obtain follow-up information,including the presence and type of recurrences as well as the time until recurrence. In some cases, general practitioners responsible for the follow-up of the patient were contacted. Information about the cause and date of death was obtained from the Cancer Registry of Norway and Statistics Norway. Complete information on patient survival and time and cause of death was available in all 202 cases. Last date of follow-up was December 18, 1998, and median follow-up time for all survivors (total survival) was 76 months (range, 13–210 months). Clinical follow-up (with respect to recurrences) was not carried out in 14 patients (predominantly older patients), and 21 patients were not treated with complete local excision. Thus, recurrence-free survival was available in 167 patients.

During the follow-up period, 69 patients (34%) died of malignant melanoma, and 39 (19%) died of other causes. Of the 167 radically treated patients with data on recurrence-free survival, 74 (44%) had recurrent disease.

Analyses were performed using the SPSS statistical package (32). Associations between different categorical variables were assessed by Pearson’s χ² test. Continuous variables not following the normal distribution were compared between two or more groups using the Mann-Whitney U or Kruskal-Wallis H tests. Wilcoxon’s signed ranks test and paired samples t test were used to compare related samples. The relationship between tumor thickness and Ki-67 expression was further assessed using linear regression analysis. Univariate analyses of time to death due to malignant melanoma or time to recurrence(recurrence-free survival) were performed using the product-limit procedure (Kaplan-Meier method), with date of histological diagnosis as the starting point. Patients who died of other causes were censored at the date of death. Differences between categories were tested by the log-rank test. The influence of covariates on patient survival and recurrence-free survival was analyzed by the proportional hazards method (33), including all variables with a P ≤ 0.15 in univariate analyses, and tested by the likelihood ratio test. In the multivariate analysis, categories that had a similar prognosis in univariate analysis or were too small to be analyzed separately were merged. Model assumptions were tested by log-minus-log plots. Estimated HR, 95% confidence interval for HR, and Ps are given in the tables.

The immunohistochemical staining for p16 protein (Fig. 1) was most often homogeneous throughout the tumor. Some cases showed a more heterogeneous staining pattern(mosaic type), with groups of positive and negative tumor cells side by side. Cytoplasmic staining was often observed in cases lacking nuclear staining, and the indices for cytoplasmic and nuclear staining are shown separately in Table 1. There was considerable variation in staining intensity for both compartments, and the most intensely stained sections most often showed combined cytoplasmic and nuclear staining in both primary tumors and metastases. Of all primary tumors, 45% showed no nuclear staining or minimal nuclear staining for p16 protein (staining index, 0–1), as compared with 77% of the metastases (Wilcoxon’s signed ranks test, P = 0.001; Table 1). A similar difference between primary tumors and corresponding metastases was observed for cytoplasmic staining (Table 1).

Absent or minimal nuclear p16 staining (staining index, 0–1) was significantly associated with the presence of ulceration(P = 0.001) and vascular invasion (P =0.03) in the primary tumors. The mean proliferative rate, as assessed by Ki-67 expression, was 35% in cases with absent or minimal nuclear p16 staining, compared with 24% in the remaining cases with moderate/strong nuclear p16 expression (t test, P < 0.0001). For the other clinicopathological variables, no significant associations were found with nuclear staining. Weak or absent cytoplasmic p16 staining (staining index,0–1) was significantly correlated with the presence of vascular invasion (P = 0.02) but not with other clinicopathological variables. Significantly reduced survival was present for cases showing absent or minimal nuclear p16 protein expression (P = 0.0003; Table 2; Fig. 2),and recurrence-free survival was also significantly decreased in this subgroup of patients (P = 0.007). No significant association was present between cytoplasmic p16 expression and patient outcome.

Nuclear staining for p53 protein was present in 84% of the primary tumors, whereas 16% were negative for p53 protein (Fig. 1; Table 1). A high p53 staining (index ≥ 6) was found in 11% of the primary tumors and was significantly more frequent in the head/neck area when compared with other sites (20% versus 8%; P = 0.04). There was no significant difference in p53 expression between primary tumors and corresponding metastases. No significant association was found between p53 and other clinicopathological variables as well as p16 expression. There was a tendency toward lower Ki-67 expression (median, 23%) in cases with an absence of nuclear p53 expression compared with all p53-positive cases(median, 28%), although this trend was not statistically significant(Mann-Whitney U test, P = 0.12). Furthermore, there was no significant difference in Ki-67 staining between cases with strong versus weak/negative p53 expression.

In this extended series, we found a significantly improved survival for those cases (16%) without any p53 protein expression in the melanoma cells (Table 2; Fig. 2; P = 0.0057) when compared with the rest of the cases, i.e., cases with weak, moderate, or strong p53 expression. In contrast, strong p53 expression (staining index ≥ 6) had no significant prognostic impact when compared with the remaining cases (staining index, 0–4; P =0.4). The presence of p53 expression (positive versusnegative) did not significantly predict reduced recurrence-free survival (P = 0.12), with 10-year recurrence-free survival rates of 35% and 60% for cases with and without p53 expression, respectively.

The staining pattern for Ki-67 varied in different lesions. Some of the tumors exhibited focal aggregates of immunoreactive nuclei, whereas other tumors showed diffuse, more homogeneous distribution of positive nuclei. However, it was possible to identify hot spots of stained nuclei in all cases with material available, and these hot spots were most often found in the periphery of the tumor, i.e., near the invasive front (tumor basis) or near the epidermis.

The median percentage of Ki-67-positive nuclei within the selected hot spot areas was 27% (range, 1–79%). The proportion of positively stained nuclei was significantly higher in the metastases than in the corresponding primary tumors, with a median count of 43% (paired t test, P < 0.0001). Increased Ki-67 expression was significantly correlated with increased tumor thickness(linear regression analysis, r = 0.32; P < 0.0001; Fig. 3),increased histological tumor diameter (linear regression analysis, r = 0.28; P < 0.0001), and increased Clark’s level of invasion (levels 2–4 versus level 5; t test, P = 0.02). Lesions located on the trunk and on the head and neck exhibited significantly higher tumor cell proliferation (mean, 32%) than lesions located on the extremities(mean, 24% Ki-67-positive nuclei; t test, P = 0.004). Furthermore, tumor ulceration (t test, P < 0.0001) and the presence of vascular invasion(t test, P = 0.02) were also significantly related to higher Ki-67 expression.

Based on the counts for Ki-67, the patients were divided into four categories by quartiles. Significantly decreased survival (Table 2; P < 0.0001) and recurrence-free survival(P = 0.0001) were observed in patients with the highest values. In later analyses, the patients were divided into two groups by the lower quartile, as illustrated in Fig. 2, because the upper three categories had a similar prognosis.

The following variables, which represent features of the primary tumor that were all either significant or of borderline significance in univariate analysis (P ≤ 0.15), were included in the multivariate analysis: (a) anatomical site;(b) tumor thickness; (c) Clark’s level of invasion; (d) tumor ulceration; (e) vascular invasion; and (e) p16, p53, and Ki-67 expression. As a well-established prognostic factor, tumor thickness was included as the standard measure of tumor size. Only cases with complete information on all variables were included in multivariate analysis (n = 187). Anatomical site, Clark’s level of invasion, vascular invasion,p16 expression, p53 expression, and Ki-67 expression remained as independent prognostic factors in the final multivariate model (Table 3). Tumor thickness (P =0.34) was not an independent prognostic variable when Clark’s level of invasion was included in the model. When Clark’s level of invasion was not included, both thickness (P = 0.03) and Ki-67 expression (P = 0.001) were significant in the model. All of these prognostic variables (except thickness) remained significant when AJCC stage was also included in the model [AJCC stage: HR (all stages) = 2.2; P = 0.0004] as well as when analyzed for cases with AJCC stage ≤ 2 only(n = 181).

Multivariate analysis of recurrence-free survival was performed using data from a total of 158 patients who were treated with complete local excision and for whom complete follow-up and information on the following variables were available: (a) anatomical site; (b) tumor thickness; (c) Clark’s level of invasion; (d) ulceration (absence versuspresence); (e) vascular invasion; (f) p16 expression; (g) p53 expression; and (h) Ki-67 expression (the variables were divided into groups as in Table 3). These variables were found to be significant or of borderline significance (P ≤ 0.15) in univariate analysis. Anatomical site (HR = 1.9; P = 0.02), tumor thickness (HR = 2.7; P = 0.0005), Clark’s level of invasion (HR = 2.6; P = 0.005), vascular invasion (HR = 3.1; P = 0.0002), p16 expression(HR = 2.0; P = 0.007), and Ki-67 expression(HR = 2.9; P = 0.01) had independent prognostic value in the final model, whereas p53 expression was not significant(HR = 2.1; P = 0.08).

In this series of 202 vertical growth phase melanomas of the nodular type, loss of nuclear p16 protein expression was found in 45%of the primary tumors. This is in agreement with other recent studies (14, 15, 17, 18), supporting the observation that p16 expression is frequently reduced in sporadic primary melanomas of the skin. In a previous report (19), we found that only 9% of the primary tumors showed loss of p16 protein expression using a combined examination of both nuclear and cytoplasmic compartments. In our present study, we focused mainly on nuclear p16 expression because cytoplasmic staining is controversial and regarded as nonspecific by some researchers (14, 15, 18, 29), although clinical importance has been reported for breast cancer (34). Absent or minimal nuclear p16 reactivity was significantly correlated with poor patient prognosis, and this was not found for cytoplasmic expression.

Lack of nuclear p16 staining was significantly associated with the presence of tumor ulceration and vascular invasion, but not with tumor thickness, histological diameter, or Clark’s level of invasion. These findings are in agreement with some previous studies (14, 17) and at odds with others (15, 18). Our data may indicate that p16 reduction might be an early event in a subgroup of melanomas and may not be related directly to tumor size at diagnosis.

In melanoma metastases, the frequency of p16 reduction is reported to be even higher than in primary tumors (14, 17). In our series, loss of nuclear p16 expression was found in 77% of the metastases, compared with 45% of the primary tumors. Pairwise analyses showed that p16 staining was significantly decreased in the metastases when compared with the corresponding primary melanomas. This may indicate that p16 reduction might also occur in metastases or,alternatively, that subclones of p16-altered tumor cells are more likely to escape from the primary tumor to survive and proliferate at distant sites. The latter explanation is supported by the significant association between p16 loss and vascular invasion. Our findings are in agreement with the conclusion of Morita et al. (35) reporting loss of p16 protein expression(but not p16 mutation) in analysis of primary melanomas and corresponding metastases.

The present data suggest that CDKN2A (p16^INK4a)might be inactivated in a relatively high proportion of cutaneous melanomas, although the mechanisms of this process are not clear. Previous studies have shown a high frequency of 9p21 (the site where the p16 gene is located) deletions (36, 37), and genetic alterations are particularly common in melanoma cell lines (1, 38). In contrast, studies on primary melanomas initially reported a low frequency of mutations (12, 13, 39), but a recent report by Kumar et al. (9) showed that 26% of microdissected cases had intragenic mutations. One study reported a low frequency (10%) of p16 promoter methylation (40). Thus, mechanisms other than genetic alterations might be operating.

There was a strong and independent association between absent or minimal p16 protein expression and patient prognosis, indicating a close relationship between p16 status of the primary tumor and its ability to spread. In line with this, a significant association between loss of nuclear p16 expression and vascular invasion was found in our series. These prognostic findings are in agreement with those of other reports (14, 15, 16, 17, 18, 19), although survival studies are few. Our results are similar to those found for other cancer types, such as pancreatic carcinoma (20, 21), squamous cell carcinoma of the esophagus (22), adult T-cell leukemia (23), and lung carcinoma (24, 41).

Loss of nuclear p16 protein expression was closely correlated with significantly increased tumor cell proliferation as measured by Ki-67 staining, supporting the role of intact p16 protein as an important cell cycle inhibitor. Increased proliferation rate was significantly associated with aggressive tumors, i.e., melanomas located on the trunk, tumors with increased thickness and histological diameter, and the presence of tumor ulceration or vascular invasion. High Ki-67 expression was also significantly and independently associated with decreased survival, in accordance with the findings of a few other studies (42, 43). Ki-67 expression was associated to some extent with tumor thickness, but proliferation rate was a stronger prognostic factor and was the only remaining variable of the two in the final multivariate model when Clark’s level of invasion was included.

The frequency and role of various p53 changes in cutaneous melanoma have been controversial, and the relation between mutations and protein overexpression is not clear (44, 45, 46). In a previous study,we found no consistent association between staining results and mutations in the frequently altered exons 7–8 in 30 cases studied (47). Furthermore, four of the five primary tumors that were mutated showed only weak/focal staining with the DO-7 antibody,whereas most cases with intermediate/strong immunostaining had no detected mutations. In the present study of vertical growth phase melanomas, 16% of all cases were completely p53 negative in the melanoma cells, whereas 11% showed strong nuclear staining; the other cases had intermediate nuclear reactivity (73%). No significant association was found between p53 and Ki-67 expression, and a major regulatory role of p53 for tumor cell proliferation in these advanced primary melanomas is therefore not likely. Correspondingly, no associations between p53 and measures of tumor size (thickness and diameter) and invasion (Clark’s level) were found. Other studies on cutaneous melanoma have suggested that increased p53 expression might be associated with tumor thickness and tumor cell proliferation (43, 48, 49).

Positivity for p53, either weak or strong, was significantly and independently associated with decreased patient survival in our study when compared with completely negative cases. It is known that positive staining for p53 protein might reflect either gene alterations or aberrant stabilization of the wild-type (nonmutant) protein (50), but the relative importance of different mechanisms of p53 changes, as well as their individual prognostic significance, is not known for melanomas.

Regarding traditional prognostic factors, multivariate analysis showed an independent importance of anatomical site, Clark’s level of invasion, and vascular invasion, whereas tumor thickness did not enter the final model. If Clark’s level was excluded as a variable, both thickness and tumor cell proliferation by Ki-67 staining were of independent importance. Features such as mitotic frequency and lymphocytic response (51) were not examined in the present study.

In conclusion, our findings indicate that loss of nuclear p16 protein expression identifies an aggressive subset of vertical growth phase melanomas and is associated with significantly increased tumor cell proliferation (Ki-67). Lack of p16 staining independently predicts decreased patient survival. Also, p16 reduction was significantly more frequent in metastases than in the corresponding primary tumors. Finally, p53 positivity, Ki-67 expression, and variables such as anatomical site, Clark’s level of invasion, and vascular invasion were also prognostic markers of independent importance in the final multivariate model.