Article Text
Abstract
Background Birt–Hogg–Dubé (BHD) syndrome is an autosomal dominant multisystem disorder with skin (fibrofolliculomas or trichodiscomas), lung (cysts and pneumothorax) and kidney (renal cell carcinoma) tumours. Although colorectal neoplasia was reported initially to be part of the BHD phenotype, some recent studies have not confirmed this association.
Methods A series of clinical and laboratory studies was undertaken to investigate possible relationships between colorectal neoplasia and the BHD gene (FLCN). The studies investigated whether individuals with familial colorectal cancer of unknown cause might have unsuspected germline FLCN mutations, looked for somatic FLCN C8 tract mutations in microsatellite unstable sporadic colorectal cancers, and assessed the risk of colorectal neoplasia and possible genotype–phenotype correlations in BHD patients.
Results Although it was found previously that germline FLCN mutations can be detected in ∼5% of patients with familial renal cell carcinoma, germline FLCN mutations were not detected in 50 patients with familial non-syndromic colorectal cancer. Analysis of genotype-phenotype correlations for two recurrent FLCN mutations identified in a subset of 51 families with BHD demonstrated a significantly higher risk of colorectal neoplasia in c.1285dupC mutation (within the exon 11 C8 mononucleotide tract) carriers than in c.610delGCinsTA mutation carriers (χ2=5.78, p=0.016). Somatic frameshift mutations in the FLCN exon 11 C8 mononucleotide tract were detected in 23% of sporadic colorectal cancers with microsatellite instability, suggesting that FLCN inactivation might contribute to colorectal tumourigenesis.
Conclusions These findings suggest that the previously reported clinical heterogeneity for colorectal neoplasia may reflect allelic heterogeneity and the risk of colorectal neoplasia in BHD syndrome requires further investigation.
- Gastroenterology
- clinical genetics
- cancer: colon
Statistics from Altmetric.com
Introduction
Colorectal cancer (CRC) is the third most common form of cancer in the USA and Europe. Genetic factors have an important role in the pathogenesis of CRC and may be implicated in about a third of cases.1 The identification of genes for familial colorectal cancer not only enhances clinical management of at risk families but can also provide important insights into the pathogenesis of familial and sporadic forms of CRC. Thus germline mutations in the APC tumour suppressor gene cause familial adenomatous polyposis and the APC gene is somatically inactivated in >80% of sporadic CRC.2 3 Similarly germline mutations in mismatch repair genes (most commonly MSH2, MLH1 and MSH6) are associated with Lynch syndrome (hereditary non-polyposis colon cancer syndrome (HNPCC)) which is characterised by the finding of microsatellite instability in colorectal polyps and tumours.4 Defects in mismatch repair can contribute to cancer development by predisposing to somatic mutations in colorectal cancer suppressor genes that contain short repeat coding sequences.5
Monogenic forms of colorectal cancer such as familial adenomatous polyposis and Lynch syndrome account for up to 5% of all cases of CRC.6 While there has been considerable recent progress in the identification of common low penetrance colorectal cancer susceptibility alleles (see Houlston et al 2008 and references within7), many cases of familial non-HNPCC clusters of colorectal cancer are unexplained. The delineation of additional inherited disorders associated with CRC susceptibility could lead to more accurate diagnosis of familial CRC and/or provide insights into molecular mechanisms of tumourigenesis in sporadic CRC.
Birt–Hogg–Dubé (BHD) syndrome is a dominantly inherited familial cancer syndrome characterised by the development of benign skin tumours (fibrofolliculomas or trichodiscomas) on the face and upper body, pulmonary cysts and pneumothorax and renal cell carcinoma (RCC) (see Menko et al8 and references within). BHD syndrome is caused by mutations in the folliculin gene (FLCN) at 17p11.2.9–12 More than 40% of germline FLCN mutations are accounted for by a hypermutable mononucleotide tract (C8) in exon 11.13 14 BHD displays variable expression and incomplete/age dependent penetrance and is underdiagnosed. However, molecular genetic analysis enables a diagnosis of BHD to be made in individuals who do not satisfy clinical diagnostic criteria. Recently we detected previously unsuspected germline FLCN mutations in ∼5% of patients with features of non-syndromic inherited RCC (familial RCC, multiple tumours or early onset).15 BHD was described in 1977 and early reports suggested an association with colorectal neoplasia.16–21 However, in a large study of 111 BHD patients, Zbar et al found no association between BHD and colonic polyps or CRC.22 Nevertheless, Khoo et al described a large family with BHD in which six of 20 affected individuals had developed colonic polyposis and two family members had died of probable gastrointestinal cancer.12 These observations suggested that some BHD families are at increased risk of colorectal neoplasia and that interfamilial differences might be related to allelic heterogeneity or modifier effects. Several studies have investigated the role of FLCN inactivation in colorectal tumourigenesis and somatic mutations in the exon 11 mononucleotide tract in CRC with microsatellite instability were identified in two studies.23 24 In order to evaluate further the potential role of folliculin in the pathogenesis of CRC, we investigated (1) whether individuals with familial colorectal cancer of unknown cause might have germline FLCN mutations, (2) the genotype–phenotype correlations for CRC in BHD patients, and (3) the frequency and clinicopathological associations of FLCN exon 11 mononucleotide tract mutations in microsatellite unstable CRC.
Patients and methods
Patients and samples
Three cohorts of patients were studied. (1) Blood DNA samples from 50 unrelated patients with familial colorectal cancer and no evidence of familial adenomatous polyposis or germline mismatch repair gene mutations (ascertained for the CORGI study25) were analysed for germline FLCN mutations. (2) Clinical data for colorectal neoplasia (colorectal cancer and adenomatous polyps) status was collected from 149 affected patients (from 51 families) with BHD (either known to have a germline FLCN mutation, or if mutation negative, clinically affected according to European BHD Consortium diagnostic criteria.8 A subset of patients (15 British and Dutch kindreds with two recurrent germline FLCN mutations) were analysed in order to identify genotype–phenotype correlations for colorectal neoplasia in BHD syndrome. (3) Tumour DNA (extracted from paraffin embedded pathological samples) from 30 patients with microsatellite unstable colorectal cancer was analysed for somatic mutations in the C8 mononucleotide tract in exon 11 of the FLCN gene, and at mononucleotide tracts in the TGFBR2, IGF2R and MSH6 genes. Participants gave informed consent. The study was approved by South Birmingham Local Research Ethics Committee and was performed in accordance with the Declaration of Helsinki.
Molecular genetic analysis
DNA was extracted from blood using standard methods and from paraffin embedded CRC samples using standard procedures.26 FLCN mutation analysis was performed for all coding exons and exon–intron boundaries by PCR amplification and direct sequencing of the PCR products. Primer sequences are shown in table 1. To test for the presence of frameshift mutations in microsatellite unstable (MSI) tumour samples small range, specific PCR reactions were designed. Primer sequences are shown in table 2. PCR was performed in 50 μl volumes using 20 mM MgCl2, 200 μM of each dNTP, 20 pmol primers and 1 unit of Faststart Taq DNA polymerase (Roche, Mannheim, Germany); 10 μl of product was cleaned using 5 units of Antarctic Phosphatase and 5 units of Exonuclease 1 (New England Biolabs, Hitchin, UK). The sequencing reaction consisted of 4 μl of cleaned PCR product, 1× ABI sequencing buffer, 20 pmol primer and 0.75 μl Big Dye terminator cycle sequencing mix (ABI Applied Biosystems, Warrington, UK) made up to 10 μl with clean H2O. Products were sequenced using an ABI 3730 automated sequencer (ABI Applied Biosystems).
Statistical analysis
Comparison of tumour characteristics for FLCN mutated and non-mutated sporadic colorectal cancer was undertaken using Fisher's exact test. Comparison of age related colorectal neoplasia risks in BHD patients was performed using Kaplan–Meier analysis and log rank testing. Statistical significance was taken at 5%.
Results
FLCN mutation analysis in non-syndromic familial colorectal cancer and colorectal cancer tumours
FLCN mutation analysis was undertaken in 50 unrelated affected individuals (mean age 52.2 years, range 30–72 years) with familial colorectal cancer (at least one relative with colorectal cancer) without evidence of familial adenomatous polyposis (16 patients had colorectal adenomas but no more than five) or Lynch syndrome (microsatellite stable tumours and/or no detectable mutation in MSH2 or MLH1). However, no FLCN mutations were detected (95% CI for FLCN mutations=0 to 8.4%).
Genotype–phenotype correlations for colorectal cancer in BHD patients.
The age related risk of colorectal cancer and colorectal neoplasia (cancer or polyps) was calculated for 149 BHD patients from 51 kindreds. The risk of colorectal cancer and colorectal neoplasia in the BHD patient cohort is shown in figures 1 and 2 respectively. Five patients had developed a CRC (mean age 57.4 years, range 48–64 years) and five patients had had symptomatic colorectal polyps(s) (mean age 52.0 years, range 42–68 years).
Germline FLCN mutations had been identified in 32 (containing 104 affected individuals/FLCN mutation carriers) of the 51 families. Six mutations occurred in two or more kindreds but only two mutations were present in >5 patients. Thus the frameshift mutation, c.1285dupC (formerly known as c.1733insC and c.1740dupC) was present in 37 individuals from nine families and c.610delGCinsTA (formerly known as c.1065-6delGCinsTA) was present in 32 individuals from six families. None of the c.610delGCinsTA mutation carriers developed a colorectal polyp or cancer but five individuals with a c.1285dupC mutation developed a colorectal neoplasm (three of which were malignant). Comparison of colorectal neoplasia risks in c.1285dupC and c.610delGCinsTA gene carriers revealed a significantly higher risk of colorectal neoplasia in the c.1285dupC mutation carriers (χ2=5.78, p=0.016) (figure 3).
FLCN C8 mononucleotide repeat mutation analysis in sporadic colorectal cancer tumours with microsatellite instability
Seven of 30 (23%) CRC with microsatellite instability demonstrated a frameshift mutation within the FLCN exon 11 mononucleotide repeat. In five cases there was a deletion (c.1285delC) and an insertion (c.1285dupC) was detected in two cases (figure 4). Mean % (±SD) of microsatellite instability (ie, % of microsatellite markers showing instability divided by number of microsatellite markers tested) was similar in FLCN C8 mutated and non-mutated tumours (83.5 (±8.38) and 64.2 (±5.86) respectively (t=1.57, p=0.128)). Similarly the mean percentage of mononucleotide microsatellite instability (tested using BAT25, BAT26 and/or BAT404) was similar for FLCN C8 mutated and non-mutated tumours (77.8 (±16.47) and 69.6 (±8.76), respectively (t=0.434, p=0.667)). Results of MSH2 and MLH1 protein expression (by routine immunohistochemistry) were available for 26 tumours. The frequency of FLCN C8 mutations was significantly higher in tumours that demonstrated loss of MLH1 or MSH2 protein expression than in those with no loss (43% (6/14) and 0% (0/12), respectively (p=0.017)) (table 3). All tumours tested demonstrated a mutation within the A10 tract in TGFBR2 and 2/30 (7%) harboured a mutation within the G8 tract in IGF2R. The two tumours with IGF2R mutations demonstrated instability for all microsatellite markers tested (6/6 and 7/7 markers). Immunohistochemistry for MSH2 and MLH1 expression was available for one of the IGF2R unstable cancers and loss of MLH1 expression was detected. No tumour demonstrated a mutation in both FLCN and IGF2R. MSH6 mononucleotide tract mutations were detected in 7/30 (23%) of the colorectal cancers tested (28.5% and 8%, respectively, of those with and without MSH2 or MLH1 protein expression loss; p=0.17) (table 4).
Discussion
The cumulative lifetime risk of developing colorectal cancer in the USA is about 6%.28 Assessing precise tumour risks in rare familial cancer syndromes is difficult because of the limited number of patients available and possible ascertainment bias. We undertook a retrospective study and none of the colorectal tumours or polyps that were diagnosed in our series were detected as a result of screening asymptomatic individuals. Although the risks of colorectal cancer were higher than in a UK general population cohort (figure 1), much larger numbers of patients would be required to obtain statistically significant results; in order to obtain more definitive data on colorectal neoplasia risks in BHD syndrome, we plan to perform a prospective multinational study. However, despite the limitations of the current study, it has provided several noteworthy findings. First, our results differ from those of Toro et al29 who did not detect a colorectal neoplasm in 152 patients with BHD syndrome. In a subsequent study, the same group reported three colorectal tumours in 111 patients with BHD syndrome, but concluded that this was not statistically significant and that there was not an increased risk of colorectal neoplasia in BHD.22 Such findings contrast with those of Khoo et al who described a high risk of colorectal neoplasia in a large French family with BHD syndrome.12
Differences in colorectal risk between different studies and families11 21 28 might result from interfamilial differences in exposure to environmental or genetic modifier effects or FLCN allelic heterogeneity (ie, different mutations in FLCN might be associated with differing risks of colorectal cancer). We found that BHD patients with an exon 11 mononucleotide tract mutation had a significantly higher risk of colorectal neoplasia than patients with a c.610_611delGGinsTA frameshift mutation. In addition, we note that the germline FLCN mutation in the high risk family described by Khoo et al also affected the exon 11 mononucleotide tract (c.1285delC (formerly known as c.1733delC). In the absence of nonsense mediated mRNA decay the c.1285dupC and c.1285delC mutations would be predicted to result in a protein with p.His429ProfsX27 (lacking 126 amino acids) and p.His429ThrfsX39 (lacking 114 amino acids), respectively. It could be hypothesised that, if p.His429ProfsX27 and p.His429ThrfsX39 are produced in colorectal cells, they might have a dominant negative effect on FLCN function that would not be associated with the c.610delGCinsTA mutation (this is predicted to result in a protein (p.Ala204X) lacking 377 amino acids). Alternatively, although both mutations would be predicted to result in proteins lacking the FNIP1 binding region, folliculin is likely to have multiple functions and so these might be differentially affected by the two different mutant proteins. Nevertheless, we note that 19 patients with exon 11 C8 frameshift mutations described by Toro et al14 did not develop colorectal neoplasia and further studies are required to confirm our genotype–phenotype findings in a larger dataset.
Somatic inactivation of familial cancer genes can play a major role in the pathogenesis of sporadic tumours as exemplified by the finding of somatic mutations of APC and VHL in most colorectal and clear cell RCC, respectively.30 31 In contrast, mutation analysis of FLCN has generally revealed a low frequency of mutations in colorectal cancer. Thus, in three studies in which the whole of the FLCN coding sequence was analysed in primary CRC the frequency of potential mutations (not involving the C8 tract) was 0/9 CRC,23 2/29 CRC (germline p. Arg320Gln and somatic p. Arg392Gly missense substitutions)24 and 2/30 microsatellite stable CRC (p.S79W and p.A445T).32 However, none of the four missense variants detected have been identified as germline mutations in BHD patients,33 and so the somatic changes may represent ‘passenger’ rather than ‘driver mutations’. In view of our finding of an association between colorectal neoplasia risk and a germline c.1285dupC mutation, we proceeded to investigate further whether there might be a link between colorectal neoplasia and exon 11 mononucleotide repeat region mutations. Such mononucleotide repeat regions are known to be hypermutable in microsatellite unstable CRC and although Kahnoski et al did not detect FLCN C8 mutations in 8 MSI+ CRC,32 Shin et al detected C8 frameshift mutations in 16% (5/32) sporadic MSI+ CRC (2 c.1285dupC and 1 c.1285delC and 1 c.1285delCC).23 We found FLCN C8 tract mutations in 23% of MSI+ CRC analysed. In each case, in addition to the frameshift mutation, normal wild-type sequence was also detected. Such an appearance might reflect absence of loss of heterozygosity and sequencing of the FLCN coding region in four MSI+ tumours with a FLCN C8 tract mutation did not detect a second truncating mutation (perhaps suggesting a dominant negative effect) or the presence of normal tissue in the tumour sample. Although the frequency of FLCN C8 mutations in MSI+ CRC was less than in TGFBR2, FLCN was more frequently mutated than IGF2R and the frequency was similar to that for MSH6. Somatic mutations in IGF2R and MSH6 have been considered to contribute to tumourigenesis in MSI+ CRC,34 35 suggesting that the frequency of FLCN C8 mutations in MSI+ CRC could be consistent with FLCN mutations undergoing selection during tumourigenesis.
The identification of genotype–phenotype correlations can provide important insights into the relationship between the specific functions of a disease associated protein and individual components of the disease phenotype. In such cases the effect of specific mutant proteins on gene function can be studied in vitro. However, the function of folliculin is, as yet, not well characterised. Baba et al demonstrated that folliculin interacts with FNIP1 (folliculin interacting protein 1), a poorly characterised protein that binds to 5′ AMP activated protein kinase (AMPK).36 It was also reported that FLCN phosphorylation was facilitated by FNIP1, and to be dependent on mTOR and AMPK activity, suggesting a functional relationship between FLCN/FNIP1 and mTOR/AMPK signalling and leading to suggestions that FNIP1 and FLCN may be downstream effectors of AMPK and mTOR.36 However, the effect of FLCN inactivation on the mTOR pathway has varied between studies. Thus whereas Baba et al found that an FLCN null RCC cell line has evidence of mTOR activation, Hartman et al reported lower levels of S6 (an indication of mTOR activity) in cysts and tumours from mice with targeted inactivation of the Bhd gene.37 Dysregulation of the mTOR pathway has been linked to intestinal tumourigenesis as gastrointestinal polyposis occurs in Cowden syndrome38 and rapamycin (an inhibitor of mTOR complex 1) therapy suppresses polyp formation in a mouse model of model for human familial adenomatous polyposis.39 Nevertheless, folliculin is likely to be implicated in the regulation of multiple signalling pathways and it may be that the risk of CRC in BHD is related to other pathways. Hence, in order to evaluate the possible functional basis of FLCN genotype–phenotype correlations, it will be necessary to first better characterise the function of the FLCN gene product.
Previously we identified clinically unsuspected germline FLCN mutations in individuals presenting with features of non-syndromic RCC.15 However, we did not identify any germline FLCN mutations in patients with features of non-syndromic CRC. This difference between the involvement of BHD in familial non-syndromic RCC and CRC may have several explanations. First, because BHD is a rare disorder and familial CRC is more frequent than familial RCC, it might be necessary to study a much larger group of familial CRC patients in order to identify cases with unsuspected BHD mutations. Second, whereas BHD is associated with early onset RCC, the mean age of colorectal cancer in our BHD patient series was 57.4 years. Many clinical criteria for the diagnosis of familial CRC cancer risk include a bias for earlier onset tumours (eg, the Amsterdam criteria for the diagnosis of HNPCC), which would seem to make it less likely that BHD patients might present in this group. As older patients with BHD are more likely to have fibrofolliculomas (enabling a clinical diagnosis of BHD), we suggest that in the absence of a previous medical history or family history of BHD associated clinical features, the frequency of BHD in patients with features of inherited CRC susceptibility is likely to be very low and FLCN mutation analysis is not indicated. The detection of a genotype–phenotype correlation for colorectal neoplasia risk in BHD provides a potential explanation for the reported heterogeneity in colorectal neoplasia risk in BHD. Although further studies are required to confirm and extend the correlation between FLCN mutation type and colorectal neoplasia risk, our findings suggest that when colorectal neoplasia does occur in BHD it does not occur at a very early age. If our findings are confirmed then mutation type might be used to determine CRC risk in BHD syndrome and so the need for colonoscopy surveillance. In the meantime we suggest that, in BHD families in which there is a history of colorectal neoplasia, colonoscopic screening should be offered to FLCN mutation carriers but, in view of the later age at onset of tumours, this could commence at age 45 years.
Acknowledgments
We thank the Myrovlytis Trust for financial support.
References
Footnotes
Funding The Myrovlytis Trust, First floor, 26 Cadogan Square London SW1X 0JP UK.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the South Birmingham REC.
Provenance and peer review Not commissioned; externally peer reviewed.