Article Text
PDF
Abstract
Background Most familial pancreatic cancer (FPC) remains unexplained. The identification of individuals with a high genetic risk of developing pancreatic adenocarcinoma (PC) is important to elucidate its biological basis and is critical to better define emerging strategies for the detection of early pancreatic neoplasms.
Patients and methods A series of 225 consecutively enrolled patients with PC were tested for CDKN2A mutations. After personal and family cancer histories of all the patients had been reviewed, a subset of the patients were classified as FPC and were also tested for mutations in PALLD, PALB2, BRCA1 and BRCA2 as FPC candidate genes.
Results The CDKN2A mutation rate in the 225 PC cases was 5.7%. The CDKN2A founder mutations, p.E27X and p.G101W, were predominant, but the mutation spectrum also included p.L65P, p.G67R and two novel, potentially pathogenic variants, promoter variant c.-201ACTC>CTTT and p.R144C. None of the patients with FPC harboured germline mutations in PALLD, PALB2 or BRCA2. One family was positive for the BRCA1 UV variant p.P727L. Strikingly, five of 16 patients with FPC (31%) carried CDKN2A mutations.
Conclusion These findings suggest that a sizeable subset of Italian FPC families may carry CDKN2A mutations. This result may be of value for identifying the best candidates for future PC screening trials in Italy.
- Pancreatic cancer
- CDKN2A
- susceptibility gene
- PALB2
- BRCA
- melanoma
- genetic screening/counselling
- genetic epidemiology
- cancer: dermatological
- pancreas and biliary tract
- cancer: colon
- cancer: gastric
- cancer: oesophageal
- genetic screening/counselling
- clinical genetics
- genetics
Statistics from Altmetric.com
- Pancreatic cancer
- CDKN2A
- susceptibility gene
- PALB2
- BRCA
- melanoma
- genetic screening/counselling
- genetic epidemiology
- cancer: dermatological
- pancreas and biliary tract
- cancer: colon
- cancer: gastric
- cancer: oesophageal
- genetic screening/counselling
- clinical genetics
- genetics
Introduction
Pancreatic adenocarcinoma (PC) is the most deadly of the common cancers. It is estimated to be the fifth leading cause of cancer death in Italy.1 2 Owing to its rapid progression and almost certain outcome of death, identifying individuals at risk and detecting early lesions are likely to be important for improving outcomes. Yet, no screening strategy is available for the general population, and screening strategies for individuals with the highest risk of PC due to hereditary predisposition are only under investigation.3–5
PC has a familial basis in as many as 10% of patients. Some of the familial aggregation of PC is due to chance, and some to shared environmental exposures such as cigarette smoking.6 An inherited predisposition to PC is seen in different clinical settings. Several hereditary cancer syndromes are known to be associated with an increased risk of PC, mainly Peutz–Jeghers syndrome (PJS), pancreatic cancer–melanoma syndrome (PCMS) or familial atypical multiple mole melanoma (FAMMM), hereditary breast–ovarian cancer (HBOC) and hereditary non-polyposis colorectal cancer (HNPCC). In addition, an increased risk of PC is present in patients with hereditary pancreatitis or cystic fibrosis. The term familial pancreatic cancer (FPC) applies to families in which at least two first-degree relatives have been diagnosed as having PC but that do not meet the diagnostic criteria for the previous settings.3 7–9
Studies focusing specifically on FPC genes have not been very successful.8 PALLD, which encodes a cytoskeletal protein, has been proposed as a potential PC susceptibility gene,10 and PALB2, the partner and localiser of BRCA2,11 12 appears to account for 1–3% of FPC.13 15 Several genes underlying susceptibility to the cancer syndromes associated with PC have been investigated for their involvement in FPC predisposition. While the genes responsible for PJS14 and HNPCC syndromes15 16 do not seem to play a major role, BRCA2 and BRCA1 (to a lesser extent) are interesting candidates. BRCA2 has been considered an important PC predisposition gene since its discovery,17 and it is now estimated that it accounts for 6–12% of FPC families.8 18 19 BRCA1 gene mutations have been reported in a small number of patients with FPC.20 21
CDKN2A germline mutations account for 30–40% of patients with the PCMS and FAMMM/PC syndromes22–31 and have generally been considered to play a minor role in FPC.19 32–34 Most CDKN2A mutations are missense mutations located in the coding sequences of exons 1 and 2, common to both tumour suppressors encoded by this locus (p16INK4a and p14ARF), and a number of these mutations seem to derive from ancestral founders.35 We previously performed germline testing of CDKN2A in a series of unselected patients with PC. We found that 4% of these patients were CDKN2A-positive. Because we identified these mutations in unselected patients rather than in high-risk melanoma or melanoma/PC families, we concluded that there probably exists an unbiased association between these particular mutations and PC.36 37
In this study we reassessed the frequency of CDKN2A in a larger series of unselected patients with PC (which includes the previous sample), correlating this finding with the presence of PC and/or melanoma in the patients and their families. Furthermore, we investigated the contribution of germline mutations in PALLD, PALB2, BRCA1, BRCA2 and CDKN2A to the development of PC in the FPC kindreds we identified, with a view to facilitating the identification of individuals who are at high risk of developing PC because of their genetic background.
Methods
PC case–control series
This case–control study was approved by the ethics committees of the three participating hospitals (San Martino Hospital, Galliera Hospital and National Cancer Institute) in Genoa. It began in 1999 as the first Italian molecular genetic case–control study aimed at evaluating the contribution of candidate susceptibility genes to the development of PC. A subset of these cases (120) has previously been described for CDKN2A mutation frequency.37
Only patients with a confirmed PC diagnosis who provided written informed consent to the study, donated a blood sample, and answered a detailed questionnaire including their personal and family cancer history that was administered by trained interviewers were included. Physicians with expertise in PC (gastroenterologists, oncologists and surgeons) reviewed diagnoses for coding as adenocarcinoma and invited the patients to participate in the study. Between February 1999 and February 2011, we consecutively enrolled 225 patients with PC at the three hospitals and an equal number of controls. Recruitment began in 1999 at the Genoa National Cancer Institute where patients were referred from the nearby San Martino Hospital for diagnostic endoscopic retrograde cholangiopancreatography. As of 2002 patients have also been recruited at the San Martino Hospital (gastroenterology and mainly surgery and oncology units), and in 2006 recruitment started at the third main hospital in the town, the Galliera Hospital. A total of 443 patients with confirmed PC were diagnosed at the participating units during the overall study period; 250 were approached (56.4%) and 225 (90%) agreed to participate. Thus a selection of all patients with PC diagnosed at the three hospitals was included. PC diagnoses were confirmed by histology (73%), cytology (17%) or image studies/clinically (10%). Family pedigrees covering at least three generations were constructed to include all the information available on relatives, including the geographic origin of both branches of the family. Diagnoses of other cancers were confirmed in all of the patients with PC and, when possible, also among relatives, either through the local cancer registry or through medical records. Ninety per cent of the PC diagnoses reported in first-degree relatives were confirmed. The presence of a personal/family history of melanoma and of naevi was assessed with a specific question on surgically removed cutaneous lesions. All melanoma cases were confirmed.
Two control groups were recruited: patients hospitalised for diseases other than cancer and unrelated to digestive tract diseases or to the risk factors under investigation (n=135), who were identified at the same hospital where the matching cases were diagnosed; healthy volunteers who agreed to participate in the study as controls after a public enrolment campaign (n=90). All controls were matched to cases by gender, age (±5 years) and place of residence.
FPC kindreds
Patients with PC in our series were defined as affected by FPC when they had at least one first-degree relative with PC or when they belonged to families with two other relatives with PC who were first-degree relatives of each other, and did not fulfil the criteria for the main hereditary conditions known to be associated with PC, in accordance with the current definition for FPC.3 7 8
All of the 225 patients' family histories were examined for the presence of HBOC, PJS, hereditary pancreatitis and FAMMM/PCMS defined according to the criteria recommended by the Fourth International Symposium of Inherited Diseases of the Pancreas.3 We identified 10 patients who met the HBOC criteria, which are being analysed separately for BRCA1, BRCA2 and PALB2,38 no patients with PJS, and no patients with hereditary pancreatitis. A subset of patients who met the revised Bethesda criteria for HNPCC was also identified (n=23), and most have already been described39 showing a low MLH1 and MSH2 mutation rate. Absence of melanoma in the patients with PC and their first- and second-degree relatives was considered to rule out the presence of FAMMM/PCMS.
Mutational analysis
Standard PCR conditions, with an annealing temperature of 60°C for all primer sets using AmpliTaq Gold DNA polymerase (Applied Biosystems, Carlsbad, CA, USA), were applied to genomic DNA.
The CDKN2A coding region, including splice junctions, the 5′ and 3′-untranslated region (UTR), the intronic sequence described to contain the IVS2-105 A/G mutation and exon 1B, were entirely sequenced as previously described.40 Testing for the presence of the PALLD oncogenic mutation, P293S, was performed using primers 5′-ATGGAGATGGTGGGAGTCCCAG-3′ (forward) and 5′-TCCTGGAATACACGTTCCTGG-3′ (reverse) for both PCR amplification and sequencing. The 13 coding exons of PALB2 were completely sequenced using previously described primers.41 Mutational screening of BRCA1 and BRCA2 was carried out by direct sequencing, after PCR amplification of all coding exons and the corresponding splicing sites. DNA sequencing was performed directly on PCR products purified with AMPure XP-Agencourt (Beckman Coulter, Fullerton, CA, USA) using the Big Dye Terminator v3.1 Sequencing Kit (Applied Biosystems), and non-incorporated dideoxynucleoside triphosphates were eliminated by CleanSEQ-Agencourt (Beckman Coulter). Both purification steps were carried out on the laboratory automation workstation Biomek FX (Beckman Coulter).
Multiplex ligation-dependent probe amplification (MLPA) analysis
MLPA assay was performed for PALB2, to exclude large deletions or duplications using the SALSA MLPA Kit P057-A2 FANCD_PALB2 (MRC-Holland, Amsterdam, the Netherlands). The same technique was applied for CDKN2A in FPC probands negative for CDKN2A mutations using the SALPS MLPA Kit ME024-A1 9p21 CDKN2A/2B region (MRC-Holland). The assay was performed on samples with sufficient DNA of adequate quality as previously described.13
Statistical analysis
Differences between cases and controls in demographic and clinical characteristics and CDKN2A variant frequencies were evaluated by Fisher's exact test, and comparison of age between groups was performed using the Wilcoxon Mann–Whitney test. All statistical tests were two-sided and considered significant at the 0.05 level. Point estimates and 95% CIs of unadjusted ORs were calculated and used as the measure of association between melanoma risk and specific individual CDKN2A variants.
Results
Demographic and clinical characteristics of study population
Demographic and clinical characteristics of cases and controls are shown in table 1. Controls were age- and sex-matched to cases. The only difference found was with regard to family history of PC (p=0.0232).
Demographic and clinical characteristics of PC cases and controls
CDKN2A germline mutations in the case–control series: mutation frequency increases with the number of PC and melanoma occurrences in the family
A total of 225 cases and 225 controls were tested for the presence of germline mutations in CDKN2A. The results are reported in table 2 and correlated with occurrence of PC or melanoma in the patients and in their families (first- and second-degree relatives).
Distribution and percentage of CDKN2A mutations based on number of PC or melanoma occurrences in individuals or families
The CDKN2A mutation rate in the 225 PC cases was 5.7%, ranging from 2.6% in cases without a family history of PC or melanoma to 17% when cancer occurrences in the index patient or first-degree relatives were two, and to 45% when the cancer occurrences were three or more. Interestingly, 25% of the cases with one first-degree relative with melanoma were mutation-positive. In our previous set of 120 patients, the CDKN2A mutation frequency was 4.2% (two p.G101W, two p.E27X, one p.L65P); in the 105 new cases it is 7.6% (a novel promoter variant c.-201ACTC>CTTT, one p.G67R, one novel p.R144C mutation, four p.G101W and one p.E27X) (table 2). Neither of the novel variants was detected in the 225 healthy controls (online supplementary table S1) or in the 500 patients with melanoma in our Italian cohort.42 In silico analysis showed that p.R144C is potentially deleterious. Functional analysis is ongoing for the promoter variant, which is likely to lie in a region that is important for post-transcriptional control, as we described previously.42 Had we not considered these novel alterations as possibly disease-associated, the overall CDKN2A mutation frequency would have decreased from 5.7% to 4.8% in the full PC series. None of the known mutations identified among these patients were found in this control population (online supplementary table S1). The age at diagnosis of CDKN2A mutation-positive and mutation-negative patients did not differ (65.07 vs 66.42 years, p=0.66). One novel and one known intronic variant (c.150+27A/C and c.150+37C/G, respectively) were detected in two patients with PC without a family history of melanoma or PC but with early age of onset (50 years and 54 years). We did not classify these variants as disease-associated, but they were not detected in the control population and their functional impact has never been investigated and thus cannot be ruled out. The p.A148T polymorphism has been inconsistently associated with melanoma risk and rarely studied in PC.43 Recent studies found a 3.1% allele frequency.34 We found an allele frequency of 2.9%, which was not different from the control group (3.5%) (p=0.6) and in turn is comparable to other null studies,34 44 confirming that CDKN2A polymorphic variants are not associated with risk of PC. The same was observed for the known polymorphisms in the promoter (c.-191G>A) and the 3′-UTR (c.*69C>T) (60% vs 59.4%, p=0.87, and 5.3% vs 6.4%, p=0.54, respectively), while a modest association was found for the 3′-UTR polymorphism c.*29C>G (19% vs 13.6%, p=0.046, OR=1.485, 95% CI 1.012 to 2.178) (online supplementary table S1). Overall, we considered the germline substitutions found in cases and not in controls to be mutations when they had already been described as mutations in the literature or when we had indications from in silico evaluation or ongoing functional analysis.
CDKN2A germline mutations are prevalent in Italian FPC kindreds
Starting from the personal and family cancer histories of the 225 patients with PC, we identified 16 FPC families (7.5%). A detailed description of the relatives of PC cases in these families and their cause of death are provided in online supplementary table S2. No melanoma cases or any other hereditary syndromes were identified in these families. The 16 probands comprised nine men and seven women, with a mean age at PC diagnosis of 64.4 years (±10.4, range 47–81years). Age of diagnosis did not differ from that of the remaining 209 patients with PC (66.5±17.3, range 29–94) (p=0.48).
Deleterious or potentially deleterious CDKN2A mutations (c.-201ACTC>CTTT, one p.E27X, one p.G67R, two p.G101W) were found in five of 16 families (31%). A mutation was found in two of 10 (20%) families with two relatives with PC and in three of six (50%) families with three PC occurrences (table 3).
Mutational analysis and features of the 16 FPC kindreds
Fifteen of the 16 FPC index patients and their available relatives with PC (n=2) were tested for mutations in PALLD, PALB2, BRCA1 and BRCA2. None of these patients were found to harbour the P293S oncogenic mutation in PALLD or any disease-associated mutations in PALB2. MLPA was performed for PALB2 and also gave negative results.
No BRCA2 mutations were found, and a single BRCA1 variant was detected, which was previously described as a UV variant (P727L)45 (table 3). This variant was observed to co-segregate in the two affected members who were available for testing in a family with three affected relatives.
Discussion
The overall frequency of CDKN2A mutation in our series of 225 PC cases was 5.7% (13/225) regardless of family history (table 2). We classified the novel alterations (c.-201 ACTC>CTTT, p.R144C) as mutations based on allele frequency in controls and in patients with familial or non-familial melanoma, and on in silico prediction. Both are currently under functional investigation.42 Had we not taken these alterations into account, the CDKN2A mutation frequency would have decreased from 5.7% to 4.8% in the full PC series.
As shown in table 2, the mutations derived from common founders, p.G101W and p.E27X,40 46 are prevalent, but other known CDKN2A mutations were also identified (p.L65P, p.G67R) as well as two novel alterations. p.G101W is one of the most commonly reported CDKN2A mutations worldwide.26 35 47 We previously detected it, along with p.E27X, in melanoma-prone families in association with PC.28 40 p.G67R has been occasionally described as a deleterious mutation in melanoma-prone families and does not appear to be a founder mutation, nor does p.L65P, which is relatively rare.26 35
Strikingly, when we looked at the CDKN2A mutation frequency in the patients with PC who were classified as FPC, we found that 31% (5/16) were mutation-positive. The mutation frequency ranged from 20% in FPC families with two affected members to 50% in families with three, and is comparable to the mutation rate we observe in melanoma families.48
Anticipation within the families was observed, which is consistent with previous studies49 that reported anticipation for BRCA2 carriers in FPC families without CDKN2A mutations. Although we found no BRCA2 mutations, anticipation was observed in the three families with three affected members who were negative for CDKN2A mutations, but not in two of the three families with CDKN2A mutations.
A recent, large North-American study of 1537 unselected patients with PC found that 0.6% carried CDKN2A mutations.34 Among the 120 FPC cases in that study, four (3.3%) were CDKN2A-positive. The authors concluded that screening of patients with PC for CDKN2A mutations should not be performed, but also that these mutations are notably penetrant especially among smokers. The CDKN2A mutation rate in our FPC cases was nearly 10 times. This result indicates that a sizeable subset of Italian FPC families may carry CDKN2A mutations.
It has been proposed3 that individuals should be referred for CDKN2A testing when at least one of the following conditions is met: a personal history of melanoma and first-degree relative with melanoma; more than two confirmed primary melanomas; more than three (first-degree or second-degree) relatives with melanoma; personal or family history of pancreatic cancer and melanoma; personal history of melanoma and personal and/or family history of atypical moles. Other recommendations have included patients with more than three melanomas, or families with at least one melanoma and two other instances of melanoma or PC in the family, with mutation yields ranging between 20% and 40%.50
Following those criteria would have allowed us to identify two of five (40%) of our mutation-positive families with both melanoma and PC: one in which the PC patient had two relatives with melanoma and one in which the family also included a relative with PC and one with melanoma. However, as none of those criteria include FPC families, we would not have identified the CDKN2A-positive FPC kindreds. Interestingly, the North-American study came to the same conclusion, as the majority of their mutations were identified in FPC families, despite their low overall mutation frequency. Their finding is probably more generalisable, both because of their sample size and because it was not influenced by the presence of founder mutations.
Taken together, our results confirm that the occurrence of at least three cancer events (including PC and malignant melanoma) in the family is a good predictor of CDKN2A mutations (45%) (table 1). Importantly, however, the likelihood of identifying a CDKN2A mutation may also be high in families with two or more instances of PC or with one instance of PC and one of melanoma among first-degree relatives, as we found that 17% of such kindreds were positive for CDKN2A mutations. Conversely, we found no mutations in families with two events in more distant relatives (second-degree relative), or when the two cancer events occurred in the same individual. Others have found34 51 a low mutation rate in this subset of patients, and concluded that the underlying reason for an individual developing PC rather than malignant melanoma is still to be clarified.
To our knowledge, no other published study has investigated the presence of CDKN2A, PALB2, BRCA1 or BRCA2 germline mutations in Italian FPC families. We observed that none of the patients with FPC harboured the P293S oncogenic mutation in PALLD, the contribution of which to PC susceptibility is probably low.14 52 We observed no germline mutations or deletions in PALB2, or in BRCA2, which is consistently quoted as being the most frequently mutated gene in FPC.8 Few data are available on the prevalence of BRCA1 alterations in FPC. We identified one UV variant listed in the BIC Database in a family with three PC cases. Further studies are required to clarify whether this variant has any functional impact.
Overall our findings suggest that CDKN2A is the main FPC predisposition gene in Italian families.
A recent position paper from the Italian Familial PC registry indicated that high-risk individuals, defined as carriers of mutations in one of the candidate susceptibility genes (including CDKN2A) or members of families with more than three affected individuals (RR >10), as well as individuals belonging to families with two affected relatives, should be offered participation in surveillance programmes conducted by centres with specific experience and within clinical research protocols.2 In the frame of these future trials, our findings are of value for identifying criteria to select families for CDKN2A genetic testing in this country.
The strength of our conclusions may be limited by the small size of our sample and of our families. Moreover, as the majority of our patients with PC and their relatives did not undergo a skin examination, we cannot completely rule out the presence of preclinical or early melanoma lesions, although the questionnaire specifically elicited information on surgically removed skin lesions, number of naevi and presence of a family history of naevi and melanoma. Also, the median age at diagnosis of our patients with PC is higher than the average age at diagnosis of melanoma. Thus all efforts were made to exclude the melanoma/PC and FAMMM/PC phenotypes in the families defined as FPC. These results require confirmation in larger, independent series. Future research will focus on the evaluation of penetrance of CDKN2A mutations in carriers identified from this and additional series of Italian patients with PC, in order to contribute to the understanding of the impact of known PC risk factors (eg, cigarette smoking) on the risk of developing PC or other cancers.
Acknowledgments
We are indebted to the patients and families whose generous participation made this study possible.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online tables
Footnotes
Genoa Pancreatic Cancer Study Group (GPCSG) collaborators: P Ghiorzo, F Belli, L Bonelli, G Borgonovo, F De Cian, A Decensi, P Dulbecco, M Filauro G Fornarini, A Gozza, L Mastracci, F Grillo, S Sciallero, F Papadia, P Queirolo, C Parodi, P Romagnoli, G Sacchi, V Savarino and G Bianchi Scarrà.
Funding This study was funded by IRCSS 2007 Italian Ministry of Health DGRST.4/4235-P1.9.A.B, Fondazione CARIGE 2010, PRIN 2008 to GBS.
Competing interests None.
Patient consent Obtained.
Ethics approval This case–control study was approved by the ethics committees of the three participating hospitals (San Martino Hospital, Galliera Hospital and National Cancer Institute) in Genoa.
Provenance and peer review Not commissioned; externally peer reviewed.