Characterization of the neoplastic phenotype in the familial atypical multiple-mole melanoma–pancreatic carcinoma syndrome

Abstract

BACKGROUND

Previous studies suggest that the familial atypical multiple-mole melanoma (FAMMM) syndrome may predispose affected families to nonmelanoma carcinomas, including adenocarcinoma of the pancreas. It has been found that many of these families harbor mutations in the CDKN2A gene on chromosome 9p21. The phenotypic expression of CDKN2A mutations in these families has not been characterized fully.

METHODS

The authors studied eight families that appeared to inherit multiple nevi, cutaneous melanomas, and pancreatic carcinomas in association with a CDKN2A germline mutation. The expression of disease within these families was examined, and segregation ratios were estimated to assess the patterns of inheritance according to various definitions of phenotype.

RESULTS

Either multiple nevi or pancreatic carcinoma was diagnosed in 53% of first-degree relatives of the probands. The offspring of parents affected with multiple nevi, melanoma, or pancreatic carcinoma were significantly more likely to be affected themselves compared with the offspring of unaffected parents (48.9% vs. 16.7%; P = 0.004).

CONCLUSIONS

The current results provide additional evidence that multiple nevi, melanoma, or pancreatic carcinoma may be inherited as autosomal-dominant traits in families known to harbor CDKN2A mutations. Other malignancies may be a part of the phenotype in these families, although this hypothesis requires additional study. Cancer 2003;98:798–804. © 2003 American Cancer Society.

DOI 10.1002/cncr.11562

Pancreatic carcinoma (PC) is the fourth leading cause of cancer mortality in the U.S. among both males and females, with an estimated 29,700 deaths expected in the year 2002.1 The prognosis of patients with this disease remains poor because most patients have incurable disease and die within 1 year of diagnosis. The pathogenetic mechanisms that result in PC remain poorly understood, although it is estimated that approximately 10% of patients may have familial PC.2 The etiology of familial PC shows significant genotypic and phenotypic heterogeneity,3 although PC is an integral lesion in several hereditary cancer syndromes (Table 1). More recently, a PC susceptibility locus has been mapped to chromosome 4q32-34 in a kindred that appears to inherit PC as an autosomal-dominant trait.16 However, to our knowledge the etiology of the majority of familial PCs has yet to be determined. With improved understanding of the inherited forms of PC, it may be possible to discover the pathogenetic mechanisms underlying apparently sporadic malignancies and to implement strategies for screening, surveillance, and ultimately prevention of this highly fatal disease.

Table 1. Hereditary Syndromes and Known Genetic Mutations Associated with Familial Pancreatic Carcinoma

Syndrome	Known mutations	Reference
HNPCC: hereditary nonpolyposis colon carcinoma; FAP: familial adenomatous polyposis; FAMMM: familial atypical multiple-mole melanoma syndrome.
HNPCC	DNA mismatch-repair genes (MSH2, MLH1)	Lynch et al.4
Hereditary breast/ovarian carcinoma	BRCA2	Ozcelik et al.5
Peutz–Jeghers	Serine-threonine kinase (STK11/LKB1)	Giardiello et al.6
FAP	APC	Giardiello et al.7
FAMMM	CDKN2A	Lynch et al.,8 Goldstein et al.,9 Vasen et al.,10 Borg et al.,11 Goldsetin and Tucker,12 and Bartsch et al.13
Hereditary pancreatitis	Cationic trypsinogen (PRSS1)	Lowenfels et al.14
Cystic fibrosis	CFTR	Neglia et al.15

Members of families with familial atypical multiple-mole melanoma (FAMMM) syndrome inherit a predisposition to develop multiple atypical cutaneous nevi (> 50), although not all patients with melanoma in these families display this phenotype. These families also appear to be at increased risk of other malignancies, particularly adenocarcinoma of the pancreas.8–13 The most common known mutation in these melanoma-prone families involves the CDKN2A gene on chromosome 9p21. CDKN2A encodes p16, a low-molecular-weight protein that inhibits the cyclin D1-cyclin dependent kinase complex (CDK4). If it is not inhibited, the CDK4 complex, in turn, phosphorylates the retinoblastoma protein, allowing a cell to progress through the G1 phase of the cell cycle. Thus, p16 acts as a tumor suppressor protein, and mutations in CDKN2A can result in unregulated cell growth and neoplastic progression. Germ line CDKN2A mutations have been detected in up to 25% of melanoma-prone families worldwide.12 Mutations in the CDK4 gene on chromosome 12q13 have been detected in only a handful of melanoma-prone families, and the genetic defects in the other families are unknown.17

A melanoma kindred apparently predisposed to PC was reported first in 1968, and a number of additional kindreds have been identified subsequently.18, 19 Several studies of FAMMM kindreds have found an excess of nonmelanoma malignancies compared with the expected frequency of these malignancies in the general population. The risk of developing malignant disease in these kindreds appears to be increased 10-fold to 40-fold, and the cumulative risk of PC has been estimated at 17% by age 75 years.8–13 Studies of families who inherit the FAMMM phenotype have demonstrated functional mutations in CDKN2A, although other melanoma-prone families have mutations without clear functional significance. It is noteworthy that the increased risk of PC may be confined to kindreds with mutations that impair the function of the p16 protein.9 In addition, FAMMM kindreds may be at increased risk of developing other carcinomas, including breast tumors, lung tumors, sarcoma, and digestive tract tumors. Thus, the pleiotropic genetic effects of CDKN2A mutations and the resulting neoplastic phenotypes have yet to be defined fully.

To understand better the inheritance of neoplastic phenotypes in these families, we estimated segregation ratios for several neoplastic phenotypes using eight families that inherit multiple nevi, melanomas, and PC and that were characterized previously to harbor CDKN2A mutations. We hypothesize that a predisposition to atypical nevi, melanoma, PC, and possibly other malignancies is inherited as an autosomal trait consistent with the presence of mutations in the tumor suppressor function of p16.

MATERIALS AND METHODS

Eight families with both PCs and melanomas were ascertained from the Creighton University Familial Pancreatic Cancer Resource, as described previously.19 This registry includes 159 families with ≥ 2 relatives affected by PC who were either self-referred or referred by their physician. Nineteen of these families (12%) also manifest multiple nevi, as defined by the presence of > 50 atypical nevi on dermatologic examination. Family members with multiple nevi are at markedly increased risk of cutaneous malignant melanoma. However, a small number of high-risk family members without multiple nevi also develop melanomas.

We selected 8 of 19 families in which it was found that at least 1 member harbored a CDKN2A mutation. Questionnaires and personal interviews were administered previously to determine family history and to assemble pedigrees.19 Information was obtained from probands in all families and was corroborated by all other relatives who agreed to participate. Medical records, pathology reports, and death certificates were used to confirm the diagnoses. All participants provided written informed consent, and the study was approved by the Creighton University Institutional Review Board.

Several neoplastic phenotypes were defined, a priori, for this analysis. These included the following: multiple nevi (the presence of > 50 atypical nevi on dermatologic examination, with or without coexisting cutaneous melanoma), melanoma (with or without multiple nevi), PC, or the presence of any other malignancy. Combined phenotypes also were analyzed, because the CDKN2A mutation may have variable expression within families. The combined phenotypes analyzed included multiple nevi or melanoma, multiple nevi or PC, melanoma or PC, and multiple nevi or melanoma or PC. Subjects were considered affected if they expressed one or more phenotype within a given category.

The patterns of inheritance within the families were assessed by examining the proportion of the probands' relatives who were affected, according to the predefined neoplastic phenotypes. For this analysis, first-degree relatives included parents, siblings, and offspring. Second-degree relatives were limited to grandparents and aunts/uncles. Third-degree relatives included first cousins only. We calculated segregation ratios for the proportion of affected offspring20 according to the various definitions of neoplastic phenotypes. These analyses compared affected by unaffected matings for a given phenotype with matings between individuals with none of the neoplastic phenotypes. In such analyses, affected by unaffected matings should result in a greater number of affected offspring if genetic susceptibility is involved in the pathogenesis of disease. Because a number of sibships in the kindreds are quite young and still may manifest a given phenotype, all offspring younger than age 40 years were excluded from the primary analysis. Furthermore, the exact cause of death for many subjects from the earliest generations could not be determined, and the offspring of these matings also were excluded. The influence of these exclusion criteria was assessed by a sensitivity analysis of their effect on the proportion of affected offspring for the multiple nevi or melanoma or PC phenotype.

Descriptive statistics were calculated and expressed as means, medians, standard deviations, and ranges as appropriate. The number of affected offspring was calculated and expressed as a proportion. Differences in proportions were tested using a two-sample test of proportions, and differences were considered statistically significant at the P = 0.05 level. All analyses were performed using Stata software (version 7.0; Stata, Inc., College Station, TX).

RESULTS

The 8 families included a total of 370 members who were diagnosed with a substantial number of melanomas (n = 34 patients; 9.2%), PCs (n = 25 patients; 6.8%), and various other malignancies (n = 40 patients; 10.8%). There were 135 first-degree, second-degree, or third-degree relatives of the probands included in the study, including 15 relatives with PC, 17 relatives with melanoma, 25 relatives with multiple nevi, and 18 relatives with other malignancies (Table 2). The numbers of other malignancies were as follows: lung (n = 5 patients); sarcoma (n = 2 patients); colon (n = 2 patients); lymphoma (n = 2 patients); esophagus (n = 2 patients); cervix (n = 2 patients); and carcinoma of the breast, endometrium, parotid gland, and prostate (n = 1 patient each). The median age at the time of diagnosis was 59 years for patients with PC, 51 years for patients with melanoma, and 58 years for patients with the other types of malignancies. Patients with melanoma often had more than one melanoma diagnosed during their lifetime (mean, 1.6 melanomas per patient). Six of the probands' relatives had more than 1 cancer during their lifetime.

Table 2. Distribution of Neoplastic Phenotypes among Relatives of Probands from 8 Families with CDKN2A Mutations (n = 135 relatives total)a

Phenotype	No. of relatives with phenotype
	Family A	Family B	Family C	Family D	Family E	Family F	Family G	Family H	Total (range)
a These calculations included first-degree relatives (parents, siblings, offspring), selected second-degree relatives (grandparents, aunts/uncles), and selected third-degree relatives (cousins). b Types of nonpancreatic, nonmelanoma malignancies included lung carcinoma (n = 5 patients), sarcoma (n = 2 patients), colon carcinoma (n = 2 patients), lymphoma (n = 2 patients), esophageal carcinoma (n = 2 patients), cervical carcinoma (n = 2 patients), breast carcinoma (n = 1 patient), endometrial carcinoma (n = 1 patient), parotid carcinoma (n = 1 patient), and prostate carcinoma (n = 1 patient). One relative in Family F was diagnosed with both sarcoma and esophageal carcinoma. c Five relatives were diagnosed with two types of carcinoma and one relative was diagnosed with three types of carcinoma.
Total	14	27	15	8	21	9	23	18	135
Multiple nevi (> 50)
No. of relatives	3	2	7	1	0	3	5	4	25
Pancreatic carcinoma
No. of relatives	1	2	2	3	3	0	1	3	15
Median age at diagnosis (yrs)	47	52	69	70	51	—	58	50	59 (35–79)
Melanoma
No. of relatives	3	0	5	0	1	1	4	3	17
Median age at diagnosis (yrs)	60	—	45	—	59	38	61	41	51 (13–73)
Other malignanciesb
No. of relatives	3	4	1	1	1	3	1	4	18
Median age at diagnosis (yrs)	51	54	39	63	41	69	55	71	58 (39–77)
Individuals with more than one type of carcinomac
No. of relatives	1	0	1	0	0	1	1	2	6
Median age at diagnosis (yrs)	53	—	47	—	—	49	58	41	31 (50–58)

Although the probands were ascertained on the basis of their family history, a striking proportion of their relatives had neoplastic phenotypic expression of either carcinoma or multiple nevi (Table 3). Among first-degree relatives, the multiple nevi phenotype was most common (34.7%), although melanoma (20.4%) and PC also were quite frequent (18.4%). Greater than 50% of first-degree relatives were affected if the neoplastic phenotypic definition was broadened to include either multiple nevi or PC. Smaller proportions of second-degree and third-degree relatives were affected by any given phenotype. Overall, approximately one-third of probands' relatives were diagnosed with a malignancy. Furthermore, it is possible that the proportion of relatives with the multiple nevi phenotype may have been underestimated, because only 45% of first-degree relatives and 21% of second-degree relatives underwent dermatologic examinations,19 and additional relatives may manifest a phenotype at some point in the future. In addition, the cause of death could not be determined in two parents and in three siblings of the probands, and it is possible that these relatives also may have been affected.

Table 3. Proportion of Probands' Relatives (n = 135 relatives) Affected, According to Various Definitions of Phenotype

Neoplastic phenotype	First-degree relatives with phenotype (n = 49 relatives)		Second-degree relatives with phenotype (n = 46 relatives)a		Third-degree relatives with phenotype (n = 40 relatives)a		Total relatives with phenotype (n = 135 relatives)
	No.	%	No.	%	No.	%	No.	%
PC: pancreatic carcinoma. a Selected second-degree relatives (grandparents, aunts/uncles) and selected third-degree relatives (cousins) were included. b Multiple nevi: individuals with > 50 nevi, with or without a diagnosis of melanoma.
PC	9	18.4	4	8.7	2	5.0	15	11.1
Melanoma	10	20.4	5	10.9	2	5.0	17	12.6
Multiple nevib	17	34.7	4	8.7	4	10.0	25	18.5
Combined phenotypes
Multiple nevi or  melanoma	17	34.7	6	13.0	5	12.5	28	20.7
Melanoma or PC	19	38.8	9	19.6	4	10.0	32	23.7
Multiple nevi or PC	26	53.1	8	17.4	6	15.0	40	29.6
Multiple nevi or  melanoma or PC	26	53.1	11	23.9	7	17.5	44	32.6
Any carcinoma	23	46.9	18	39.1	8	20.0	49	36.3

Segregation ratios were consistent with an autosomal predisposition to FAMMM and PC; significantly more offspring of [phenotype present × phenotype absent] matings themselves were affected, compared with the offspring of [phenotype absent × phenotype absent] matings (Table 4). The multiple nevi or melanoma or PC phenotype was consistent with autosomal-dominant inheritance, because 49% of the offspring of [phenotype present × phenotype absent] matings also had the phenotype, compared with only 17% of the [phenotype absent × phenotype absent] offspring (P = 0.004). However, between 3% and 16% of apparently normal-normal matings also were affected with multiple nevi, melanoma, and/or PC phenotypes. This finding may be the result of incomplete penetrance of inherited mutations in the parents, misclassification of parental diagnoses, or possibly an environmental predisposition to these phenotypes.

Table 4. Segregation Ratios According to Various Definitions of Phenotype

Neoplastic phenotype	Mating typea						P value
	Phenotype present × phenotype absent: No. of offspring (%)			Phenotype absent × phenotype absent: No. of offspring (%)b
	Affected	Unaffected	Total	Affected	Unaffected	Total
PC: pancreatic carcinoma. a Figures for combined phenotypes do not necessarily sum as several parents and offspring had more than one phenotype. b Includes only matings between two parents with no malignancies or familial atypical multiple-mole melanoma syndrome (FAMMM). c Parental FAMMM phenotype was uncertain in most patients with melanoma, and the proportion may be an overestimate.
PC	6 (23.1)	20 (76.9)	26 (100)	4 (13.3)	26 (86.7)	30 (100)	0.23
Melanoma	9 (32.1)	19 (67.9)	28 (100)	3 (10.0)	27 (90.0)	30 (100)	0.04
Multiple nevic	7 (38.9)	11 (61.1)	18 (100)	1 (3.3)	29 (96.7)	30 (100)	0.001
Combined phenotypes
Multiple nevi or melanoma	9 (36.0)	16 (64.0)	25 (100)	3 (10.0)	27 (90.0)	30 (100)	0.02
Melanoma or PC	15 (34.9)	28 (65.1)	43 (100)	5 (16.7)	25 (83.3)	30 (100)	0.09
Multiple nevi or PC	17 (39.5)	26 (60.5)	43 (100)	4 (13.3)	26 (86.7)	30 (100)	0.02
Multiple nevi, or melanoma, or PC	22 (48.9)	23 (51.1)	45 (100)	5 (16.7)	25 (83.3)	30 (100)	0.004
Any malignancy	23 (46.0)	27 (54.0)	50 (100)	9 (30.0)	21 (70.0)	30 (100)	0.16

Family members were excluded from the analyses in Table 4 if they were younger than 40 years or if their parental phenotypes were unknown. However, neither the inclusion of members younger than 40 years nor the inclusion of matings with uncertain parental phenotype (assuming the parents were unaffected) appeared to have a substantial impact on the results for the multiple nevi or melanoma or PC phenotype. The segregation ratios when all offspring younger than 40 years were included were 31/81 (38.3%) for [phenotype present × phenotype absent] matings compared with 16/90 (17.8%) for [phenotype absent × phenotype absent] matings (P = 0.003). These segregation ratios were 32/76 (42.1%) and 7/62 (11.3%), respectively, when matings with uncertain parental phenotypes were included (P < 0.001).

DISCUSSION

Multiple members of these eight families were affected with multiple nevi, cutaneous melanoma, PC, and a variety of other malignancies; these phenotypes occurred substantially more often in these families than in the general population. Our results suggest that the multiple nevi or melanoma or PC phenotype is inherited in an autosomal-dominant fashion—half of the offspring of affected parents in these families manifest one of these phenotypes. Although there is substantial variation in the phenotypic distribution of malignancies within families, the findings of the current study suggest that PC may be inherited in association with CDKN2A mutations in these families. However, the study of additional families will be required to investigate the possible association between CDKN2A mutations and other malignancies and to refine the mode of inheritance and examine interactions with other genetic and environmental factors.

CDKN2A mutations can result in a variety of different types of malignant disease.21 The association between CDKN2A mutations and PC is plausible biologically, because most PCs are known to undergo inactivation of p16.22, 23 Thus, a germ line mutation in CDKN2A may place individuals at increased risk. The results of the current study are consistent with the results of previous studies: PC appears to be the second most common malignancy in families with a CDKN2A mutation.8–13, 17 The current analysis provides additional evidence of this association through the use of segregation ratios, and our study represents the largest such analysis of CDKN2A positive FAMMM kindreds.

We found that 23% of the offspring of parents with PC develop PC themselves. Although this proportion is less than the 50% that would be expected for an autosomal-dominant trait, this may reflect incomplete penetrance of a CDKN2A mutation.10 Furthermore, additional members of these families may develop PC at some point in the future, and other mutation carriers may have died of melanoma at a young age, before they developed PC. Indeed, the high prevalence of melanomas with a young age of onset may obscure the ability to detect other CDKN2A mutation-related carcinomas that develop later in life, and future studies should take this into account.

We were unable to detect a significant difference between the proportions of affected offspring for the PC phenotype alone, although there was a trend toward more PCs among the offspring of parents diagnosed with PCs (29% vs. 13%, respectively). Although the statistical power to detect such a difference was limited, it is also important to note that our results may overestimate the occurrence of PC in melanoma kindreds, because study participants were selected from a familial PC registry. Families with similar mutations who do not manifest PC as frequently may have been excluded inadvertently, and other genetic or environmental factors may interact with CDKN2A mutations or may act independently to result in PC in these families.

A substantial proportion (13%) of the offspring of unaffected parents had been diagnosed with PC, suggesting that some of these apparently unaffected parents may have been misclassified. Other sources of misclassification in the current study include the inability to determine the cause of death for some of the earliest generations and the possibility that some younger family members may have yet to manifest a given phenotype. Furthermore, the majority of family members did not undergo detailed dermatologic examinations, and additional affected family members may exist. To minimize potential misclassification in the segregation analysis, we included only normal-normal matings in which it was known that the parents did not manifest any of the phenotypes, and we excluded matings in which the parental phenotype was unknown as well as family members who were younger than 40 years. Although family members older than age 40 years still may be too young to manifest certain diseases, such as PC, we assumed that they were unaffected. Thus, our estimates were conservative, and the risk of PC and other phenotypes in these families may be even higher. Indeed, our attempt to minimize misclassification resulted in relatively few matings between completely unaffected parents that could be used for segregation analysis. However, we performed additional analyses that included parents with malignancies other than the phenotype in question as normal-normal matings, and the results of the study were unchanged (data not shown).

It is plausible that other malignancies may be a part of the phenotype in families with CDKN2A mutations. Sarcomas frequently harbor mutations in the retinoblastoma gene, which is part of the same regulatory pathway as p16, and preliminary results have suggested that CDKN2A mutations also may be associated with sarcomas.24 Although relatively few sarcomas were identified in the current study, the prevalence of sarcoma among these kindreds (1.5%) is greater than in the general population and also is elevated compared with the prevalence of other nonmelanoma, nonpancreatic malignancies in these kindreds. A previous study also suggested that breast carcinomas may be part of the phenotype in families with multiple melanomas and CDKN2A mutations,11 although we were unable to identify an excess of breast carcinomas in our kindreds. Larger studies should seek to examine the associations between melanomas and other malignancies, including sarcoma and breast carcinoma, particularly among families that carry mutations in tumor suppressor genes.

Several additional limitations of the current study deserve mention. Although we believe that the increased risk of multiple nevi, melanoma, and PC in these families is caused by CDKN2A mutations, we cannot exclude the possibly that additional mutations also may be segregating in these families and resulting in the phenotypes seen. It also is possible that shared environmental factors, such as smoking or sun exposure, may contribute to the risk in these families. However, members of these high-risk families may be somewhat less likely to engage in risky behavior because of their striking family history, and this theoretically may result in a decreased risk of developing malignant disease. The possibility of gene-environment interactions exists nonetheless. Finally, the generalizability of our results may be limited, because the study participants were recruited in a referral center, were mostly white, and were selected on the basis of their family history.

Despite these limitations, we believe the findings of the current study provide strong evidence that PC is part of the phenotypic expression of CDKN2A mutations in these and other kindreds predisposed to multiple nevi and melanomas. Some authors have begun to investigate the role of this mutation in other families who inherit PC,13 and studies of additional familial PC kindreds as well as patients with apparently sporadic PCs are needed. Future studies also should seek to identify additional genetic and environmental factors that influence pancreatic tumorigenesis in these families. Furthermore, it is possible that CDKN2A mutation status may be used to identify high-risk candidates for protocols to investigate primary preventive measures and screening methods aimed at early detection. We previously described methods for early detection of PC and precancerous pancreatic disease in family members who inherit PC; such early detection programs should be adopted at centers with extensive expertise in diseases of the pancreas.3 Finally, it would be prudent to advise patients with CDKN2A mutations not to smoke, because it appears that smoking is a particularly important environmental risk factor for the development of familial PC.25