Nuclear PTEN expression and clinicopathologic features in a population-based series of primary cutaneous melanoma

Abstract

Germline mutations of the PTEN tumor-suppressor gene, on 10q23, cause Cowden syndrome, an inherited hamartoma syndrome with a high risk of breast, thyroid and endometrial carcinomas and, some suggest, melanoma. To date, most studies which strongly implicate PTEN in the etiology of sporadic melanomas have depended on cell lines, short-term tumor cultures and noncultured metastatic melanomas. The only study which reports PTEN protein expression in melanoma focuses on cytoplasmic expression, mainly in metastatic samples. To determine how PTEN contributes to the etiology or the progression of primary cutaneous melanoma, we examined cytoplasmic and nuclear PTEN expression against clinical and pathologic features in a population-based sample of 150 individuals with incident primary cutaneous melanoma. Among 92 evaluable samples, 30 had no or decreased cytoplasmic PTEN protein expression and the remaining 62 had normal PTEN expression. In contrast, 84 tumors had no or decreased nuclear expression and 8 had normal nuclear PTEN expression. None of the clinical features studied, such as Clark's level and Breslow thickness or sun exposure, were associated with cytoplasmic PTEN expressional levels. An association with loss of nuclear PTEN expression was indicated for anatomical site (p = 0.06) and mitotic index (p = 0.02). There was also an association for melanomas to either not express nuclear PTEN or to express p53 alone, rather than both simultaneously (p = 0.02). In contrast with metastatic melanoma, where we have shown previously that almost two-thirds of tumors have some PTEN inactivation, only one-third of primary melanomas had PTEN silencing. This suggests that PTEN inactivation is a late event likely related to melanoma progression rather than initiation. Taken together with our previous observations in thyroid and islet cell tumors, our data suggest that nuclear–cytoplasmic partitioning of PTEN might also play a role in melanoma progression. © 2002 Wiley-Liss, Inc.

Germline mutations of PTEN, a tumor-suppressor gene on 10q23.3, cause Cowden syndrome, an underrecognized autosomal dominant multiple hamartoma syndrome with a high risk of breast, thyroid and endometrial cancers and, some believe, melanoma.1–4 Germline PTEN mutations also cause a subset of seemingly unrelated developmental disorders, such as Bannayan-Riley-Ruvalcaba syndrome and Proteus syndrome.5–7 PTEN expression is prominent in the neural crest and its derivatives during normal human embryonic and fetal development.8 PTEN is the major lipid 3-phosphatase which signals down the phosphatidylinositol 3-kinase/Akt proapoptotic pathway and effects cell cycle arrest and apoptosis.9–17

Initial evidence for the involvement of chromosome 10 in melanoma development came from karyotypic studies of nevi and melanomas, showing a wide variety of rearrangements but usually translocations or deletions of the long arm.18–22 Loss of heterozygosity (LOH) studies were subsequently carried out to localize the chromosomal region that harbors the 10q tumor-suppressor gene involved in melanoma.23–26 Meta-analysis of 10q LOH studies in melanoma (giving a total of 44/142 melanomas with LOH) indicated that only 2 melanomas have limited regions of LOH that do not encompass the 10q23 region to which the PTEN gene maps,23, 24 suggesting very strongly that PTEN is the major target tumor-suppressor gene for melanoma on this chromosome. Furthermore, Robertson et al.27 provided functional evidence for a tumor-suppressor gene involved in melanoma mapping to 10q23-qter, the chromosomal location of PTEN.

Several groups have examined the role of PTEN in melanoma-derived samples and have reported mutation/deletion rates of up to 57% in melanoma cell lines.28–35 In contrast, among noncultured melanomas, somatic intragenic mutations are rare, likely occurring in fewer than 10% of tumors, even in the metastatic setting.30–32, 35, 36 Hemizygous deletion of 10q involving the PTEN region can occur in up to one-third of these tumors. Taken together with epigenetic silencing, which is prominent in melanomas, PTEN inactivation occurs in two-thirds of noncultured tumors but mainly in the metastatic setting.36 To date, PTEN is the second most commonly mutated gene in melanoma cell lines after CDKN2A. While UV mutagenesis appears to have a causal role in the majority of mutations detected in the CDKN2A gene in melanoma, this is not true for mutations in PTEN, suggesting that PTEN mutations generally occur later in melanoma progression and/or upon selection in culture.36

To date, most studies exploring PTEN mutation and expression in melanomas have been based on selected series of cell lines. In the few studies where noncultured cutaneous melanomas were examined, the vast majority were metastatic lesions.30–32, 35, 36 Further, in the only study assessing PTEN protein expression by immunohistochemistry, mainly in metastatic samples, nearly two-thirds of samples had no or weak cytoplasmic staining while all samples had no or weak nuclear PTEN expression.36 Thus, in our current study, we examined PTEN expression in a sample of noncultured primary cutaneous melanomas from cases identified through a population-based register. Further, we identified whether loss of PTEN expression, in the cytoplasm or nucleus, was associated with any phenotypic and environmental risk factors known to predispose to melanoma or with histopathologic determinants of prognosis.

MATERIAL AND METHODS

Study population

The sample population has been described in detail previously.37 Briefly, we identified patients with newly diagnosed invasive melanoma from notifications to the Queensland Cancer Registry (notification of cancer is mandatory under Queensland legislation). Eligible cases were defined as male residents of southeast Queensland aged >50 years in whom a first diagnosis of primary cutaneous melanoma was made between 1 July 1993 and 30 June 1994. We excluded patients diagnosed with lentigo maligna melanoma or acral lentiginous melanoma. From the 422 notifications meeting these criteria, we selected a random sample of 179 patients, of whom 15 had died prior to contact for the study, 1 was unable to speak English and 5 were not contacted on the advice of the treating doctors. One further case could not be traced, and 7 refused to take part, leaving 150 participants and an overall participation rate among surviving eligible cases of 92%. Our study was approved by the Ethics Committee of the Queensland Institute of Medical Research and the Human Subjects Protection Committee of Ohio State University.

Data collection

Data were collected from participants during face-to-face interviews using a structured questionnaire, which asked about skin type and reaction to the sun, past medical history and occupational and recreational sun exposure. Each participant was examined by a single medically qualified investigator (DCW), who recorded hair and eye color and number of melanocytic nevi on the left upper limb, back and shoulders. From each subject, we sought paraffin-embedded tumor tissue from the diagnostic pathology laboratory; sufficient material was available from 92 participants for the analyses. Tumor sections were subjected to diagnostic review by a consultant histopathologist (MCC), who also measured tumor thickness, Clark's level, degree of tumor-infiltrating lymphocytes, mitotic index and level of ulceration.

Mitotic index was determined by direct microscopic examination of a hematoxylin and eosin–stained section of each tumor by a single consultant (attending) histopathologist. The entire tumor was assessed and scored on the most mitotically active area within the tumor, if applicable. The grading system used for this assessment was as follows: 1, none or rare mitotic figures; 2, occasional mitotic figures; 3, frequent mitotic figures observed within the tumor.

PTEN immunohistochemistry

Sections (3–5 μm) of each of the 92 available paraffin-embedded tumors from the cohort of 150 were subjected to immunohistochemistry using 6H2.1, a specific monoclonal antibody (MAb) raised against the terminal 100 C-terminal amino acids of human PTEN.36, 38, 39 Each section was deparaffinized and hydrated by passing it through xylene and a graded series of ethanol. Antigen retrieval was performed for 20 min at 98°C in 0.01 M sodium citrate buffer (pH 6.4) containing 0.3% hydrogen peroxide in a microwave oven. After blocking for 30 min in 0.75% normal horse serum, each section was incubated with 6H2.1 (1:50–1:200) overnight (approx. 16 hr) at 4°C. Each section was washed with PBS (pH 7.3) and then incubated with biotinylated horse antimouse IgG followed by avidin peroxidase (Vectastain ABC elite kit; Vector, Burlingame, CA). The chromogenic reaction was carried out with 3′,3′-diaminobenzidine, which gives a brown–black product. For sections with abundant melanin, 3′,3′,3′-triaminobenzidine, which gives a blue product, was used instead. After counterstaining with nuclear fast red (Rowley, Danvers, MA) and mounting in the standard fashion, each slide was evaluated under a light microscope. Immunostaining patterns and intensities were scored by 2 independent observers (X-PZ and CE). As previously described,36, 38, 40–42 the vascular endothelium served as a consistent internal positive control and PTEN expression was scored as ++, as previously described.36 For the evaluation of immunostaining in the nucleus and cytoplasm of noncultured primary melanomas, we scored immunostaining on a 3-point scale: –, no immunostaining (or the very faintest blush, for purposes of our study); +, decreased PTEN immunostaining; ++, “normal” PTEN expression equivalent to that of the internal positive control of the vascular endothelium, as previously described.36

p53 immunohistochemistry

We used the DO-7 mouse MAb directed against mutant and wild-type human p53 protein (Novocastra, Newcastle-upon-Tyne, UK) to detect expression of p53 in the sections. Labeled DO-7 (dilution 1:150) was applied to deparaffinized sections of tumors and stored overnight at 4°C in a humidified chamber. Sections were then washed, covered with secondary antibody, rinsed and covered with avidin-biotin complex as described above. For the DO-7 MAb, we used the enzyme substrate 9-amino 3-ethyl carbazole (AEC), producing a red color on oxidation. Sections were counterstained with freshly filtered Mayer's hematoxylin, mounted and dried at 80°C for at least 2 hr. With each batch of sections, a section from a strongly staining basal cell carcinoma was used as a positive control. Sections were viewed simultaneously by 2 investigators on a dual-observation microscope. As others have done,43, 44 we defined melanomas with >1% positively staining cells as p53-positive.

Statistical analysis

The principal outcome measure for these analyses was the presence of nuclear and cytoplasmic immunostaining for the anti-PTEN MAb 6H2.1. Because of the relatively small sample size, we sought to minimize the number of categories for statistical analysis and therefore dichotomized nuclear or cytoplasmic staining as follows: tumors with no PTEN expression or only the faintest blush of staining were categorized as PTEN-negative and the remainder were categorized as PTEN-positive. Other continuous variables (e.g., tumor thickness, age, number of nevi) were also categorized, to permit calculation of the χ² statistic and Fisher's exact test. All analyses were conducted using SAS (Cary, NC) version 8.0. Statistical significance was set at p < 0.05.

RESULTS

Description of histologic findings

Of 150 potential tumor specimens from the original series of cases, 92 (61%) had sufficient tissue remaining in paraffin blocks to allow sections to be cut for immunohistochemical analysis of PTEN expression. All sections had vascular endothelial cells present, which uniformly stained with an intensity graded at ++ in the cytoplasm and the nucleus, and/or normal melanocytes, which also stained with an intensity of ++; both served as positive internal controls, as previously described.36 PTEN expression in the cytoplasm and nuclei of melanoma cells ranged from – to ++, though in general nuclear PTEN expression was generally lower than that in the cytoplasm (Table I). Fifty-three samples uniformly expressed PTEN in the cytoplasm, with intensity rated at ++. One sample had no (–) cytoplasmic PTEN expression, and 29 had decreased PTEN immunostaining (Table I, Fig. 1). The remaining 9 sections had mixed-expression patterns, with 2 distinct patterns of PTEN immunostaining in each sample. In contrast, only 2 samples had ++ nuclear PTEN immunostaining; 33 had nuclear immunoreactivity graded as –, and 51 had decreased (+) immunostaining. The remainder had mixed populations with 2 levels of PTEN expression.

Table I. Distribution of Cytoplasmic and Nuclear Pten Immunostaining Among 92 Melanoma Cases

PTEN score	Total number of cases (n = 92)
Cytoplasmic
−	1
+	29
++	53
Mixed	9
Nuclear
−	33
+	51
++	2
Mixed	6

Figure 1. PTEN expression by immunohistochemistry using a specific MAb against PTEN. (a) High levels of PTEN expression (++), mainly in the nucleus, in normal melanocytes. Primary cutaneous melanomas have different levels of PTEN expression: high expression, ++ (b); decreased expression, + (c); completely devoid of PTEN protein, – (d).

Histopathologic features and PTEN expression

None of the histopathologic features was associated with cytoplasmic PTEN expression levels. In addition, we found no evidence that either Breslow thickness or Clark's level of melanoma was associated with nuclear expression of PTEN (Table II). Similarly, the prevalence of nuclear PTEN expression was comparable among tumors grouped according to the degree of ulceration and tumor-infiltrating lymphocytes. While statistical inference suggested that nuclear PTEN expression was associated with mitotic index (p = 0.02, Table II), immunoreactivity was unevenly distributed among tumors grouped according to mitotic index, with no clear trend of changing levels of expression with increasing numbers of mitotic bodies. The anatomical site of the tumor did appear to influence nuclear PTEN expression, which was observed more commonly among melanomas from the trunk compared to melanomas excised from the head or limbs (p = 0.06, Table II).

Table II. Distribution of Tumor Characteristics and Nuclear Pten Immunostaining Among 92 Melanoma Cases

Characteristics	PTEN-positive (n = 59)	PTEN-negative (n = 33)	p value
Clark level
II	21	11
III	8	9
IV	30	12
V	0	1	0.18
Breslow thickness (mm)
28	19
<0.76	19	5
0.76–1.50	7	2
1.51–2.25	3	5
2.26–3.00	2	2	0.18
>3.00
Ulceration
Nil	42	23
Moderate	11	6
Full	6	4	0.96
Lymphocytic infiltrate
Nil	8	5
Mild	29	13
Moderate	11	12
Heavy	10	4	0.53
Mitotic index
1	35	23
2	22	5
3	2	5	0.02
Site of lesion
Back/shoulders	32	11
Head/neck	6	8
Upper limb	8	6
Lower limb	8	8
Anterior trunk	5	0	0.06
p53 status
Negative	52	23
Positive	3	7	0.02

Tumor sections from 88 patients were available for both p53 and PTEN analyses. Although numbers were small, we found some evidence that primary melanomas were more likely to express either p53 alone (7%) or nuclear PTEN alone (25%) rather than to coexpress both proteins (8%) (p = 0.02, Table II).

Host characteristics, sun exposure and PTEN expression

We grouped patients according to age and phenotype but found no significant differences in PTEN expression related to age, skin type, eye color or number of nevi (Table III). Markers of intense intermittent sun exposure (e.g., numbers of painful, blistering or peeling sunburns) and chronic sun exposure (e.g., past history of nonmelanoma skin cancer) also showed no association with nuclear or cytoplasmic PTEN expression in this group of patients.

Table III. Distribution of Phenotypic Variables Among 92 Melanoma Cases According to Pten Nuclear Immunoreactivity

Risk factor	PTEN-positive (n = 59)	PTEN-negative (n = 33)	p value
Age group (years)
50–59	14	10
60–69	28	12
70–79	13	8
>80	4	3	0.77
Eye color
Brown	11	5
Green/hazel	20	10
Blue	28	18	0.80
Skin type
Burn then peel	28	19
Burn then tan	19	11
Always tan	12	3	0.35
Facial freckling
Nil	24	15
Few	18	8
Some	8	5
Many	9	5	0.93
Nevi
0–1	5	2
2–9	19	11
10–24	18	12
25+	17	7	0.84
Past history of skin cancer
No	30	17
Yes	29	16	0.95
Number of sunburns
0	22	16
1–5	23	11
>5	14	6	0.57

DISCUSSION

Based on the available data using melanoma cell lines and mainly metastatic tumors, many were led to believe that PTEN played a significant role in the initiation of sporadic cutaneous melanomas. Our current study constitutes the largest series of primary uncultured cutaneous melanomas ascertained from a population base. Using this unbiased sample set, we demonstrate that approximately one-third had no (–) or decreased (+) cytoplasmic PTEN protein expression and that the majority had no or decreased nuclear PTEN protein expression. In contrast, up to 65% of metastatic samples had no or decreased cytoplasmic PTEN expression and virtually all such samples had no or markedly decreased nuclear expression.36 In our earlier studies, we found that nonneoplastic melanocytes strongly express PTEN (graded ++) in both the nucleus and cytoplasm.7 Further, in uncultured metastatic melanoma samples, nuclear expression was virtually nonexistent and invariably so but cytoplasmic expression appeared to be germane and to correlate somewhat with PTEN genomic status.7 In the context of these observations, our study, which examined uncultured primary cutaneous melanomas, suggests that subcellular partitioning of PTEN might play some role in the progression of melanomas, a concept not previously considered for this tumor type. That PTEN can exist in the nucleus was initially felt to be impossible since it acts cytoplasmically, interacts with cytoplasmic proteins9, 10 and does not possess traditional nuclear localization signals.41, 45 However, there is now some histologic and biochemical evidence that PTEN can exist in the nucleus.40, 41, 46–48 In a subset of epithelial thyroid tumors and virtually all islet cell tumors examined, expression was predominantly in the nucleus in nonneoplastic cells; this expression became predominantly cytoplasmic with progression from normal nonneoplastic cells to neoplastic cells to aggressive malignant cells.36, 40 It appears that if epigenetic PTEN silencing is not involved, then the translocation of PTEN expression from nuclear predominance to cytoplasmic predominance also occurs from melanocyte to primary melanoma to metastatic melanoma. This suggests that, in melanomagenesis, in addition to PTEN silencing, inappropriate subcellular compartmentalization might play some role.

While strong associations were not observed between PTEN nuclear expression and the majority of tumor characteristics or phenotypic aspects of cases, 3 statistically significant associations were found. First, there is a clinical association between PTEN expression and melanomas from the trunk compared to those arising on the limbs or head region. Although this observation could suggest an association between lack of sun exposure and PTEN-mediated melanoma progression, this does not appear to be borne out when PTEN expression is compared to other standard parameters of sun exposure. Second, there was an association between lack of nuclear PTEN expression and a high mitotic index. Such an association is physiologically plausible given the function of PTEN in coordinating G₁ cell cycle arrest and apoptosis. This preliminary observation is particularly tantalizing when one considers that the correlation is with PTEN expression in the nucleus but not in the cytoplasm. However, the inconsistent trend suggests that other factors are involved in determining the mitotic index of primary cutaneous melanomas. Third, we found a statistically significant inverse association between nuclear PTEN expression and p53 expression. When p53 is not detectable immunohistochemically, it means one of two things. The most common reason is that p53 is present in the wild-type state. Rarely, lack of p53 protein expression can be secondary to either homozygous deletion or hemizygous deletion together with early truncation mutation of TP53.In vitro evidence has suggested that p53 partially regulates PTEN transcription.49 In this situation, p53-mediated apoptosis was PTEN-dependent. However, while there was evidence of a p53 binding site in the putative promoter of PTEN, there also were elements that would allow constitutive PTEN expression without p53. Our observations in melanoma provide some human in vivo evidence that this in vitro observation might be correct. Expression of wild-type p53, as reflected by tumors without p53 immunostain, should result in PTEN transcription. Our observations showing that 52 p53-negative tumors were also PTEN-positive and that 7 p53-positive tumors (i.e., dysfunctional) were also PTEN-negative (Table II) corroborate the in vitro data. Wild-type PTEN results in p53 transcription and, hence, lack of immunostaining. The remaining 26 tumors, which are either PTEN- and p53-negative (n = 23) or PTEN- and p53-positive (n = 3), likely represent a pathway independent of p53 transactivation of PTEN transcription.

Numerous studies have suggested that somatic PTEN mutations and deletions are early events in the development of a broad range of tumor types, consistent with the Knudson 2-hit theory of tumor-suppressor inactivation.50 However, these studies were generally performed on cancer cell lines and metastatic tumors. Studies performed on noncultured primary tumors demonstrated that, for most tumor types, disruption of PTEN is not an early event. A notable exception is endometrial cancer, where genetic and/or epigenetic inactivation of PTEN could be the initiating event, occurring even in histologically normal-appearing endometrial glands.39, 51 Our previous36 and current observations suggest that PTEN inactivation is not the initiating event in most melanomas and support the contention that the primary molecular target in most melanomas may be disruption of the CDKN2A/p16 pathway. The continued progression of melanomas is presumably dependent on further dysfunction of cell cycle checkpoints and/or apoptosis. Thus, disruption of PTEN function, which could occur by genomic disruption (mainly allelic loss), by epigenetic silencing (loss of transcript or loss of protein by whatever mechanism) and/or by inappropriate subcellular localization, is associated with progression and might be associated with highly aggressive, metastatic tumors.

Acknowledgements

We are grateful to Ms. M. Watson for assistance with data collection, the many pathologists who provided archival tumor specimens and the participants for their cooperation. This work was supported in part by grant RPG98-211-01 (to CE) from the American Cancer Society, grant P30CA16058 (to the Ohio State University Comprehensive Cancer Center as a seed grant to CE) from the National Cancer Institute, the Queensland Cancer Fund and the National Health and Medical Research Council of Canada.