Article Text
Abstract
Background Gorlin syndrome (GS) is an autosomal dominant syndrome characterised by multiple basal cell carcinomas (BCCs) and an increased risk of jaw cysts and early childhood medulloblastoma. Heterozygous germline variants in PTCH1 and SUFU encoding components of the Sonic hedgehog pathway explain the majority of cases. Here, we aimed to delineate genotype–phenotype correlations in GS.
Methods We assessed genetic and phenotypic data for 182 individuals meeting the diagnostic criteria for GS (median age: 47.1; IQR: 31.1–61.1). A total of 126 patients had a heterozygous pathogenic variant, 9 had SUFU pathogenic variants and 46 had no identified mutation.
Results Patients with variants were more likely to be diagnosed earlier (p=0.02), have jaw cysts (p=0.002) and have bifid ribs (p=0.003) or any skeletal abnormality (p=0.003) than patients with no identified mutation. Patients with a missense variant in PTCH1 were diagnosed later (p=0.03) and were less likely to develop at least 10 BCCs and jaw cysts than those with other pathogenic PTCH1 variants (p=0.03). Patients with SUFU pathogenic variants were significantly more likely than those with PTCH1 pathogenic variants to develop a medulloblastoma (p=0.009), a meningioma (p=0.02) or an ovarian fibroma (p=0.015), but were less likely to develop a jaw cyst (p=0.0004).
Conclusion We propose that the clinical heterogeneity of GS can in part be explained by the underlying or SUFU variant.
- PTCH1
- SUFU
- Gorlin syndrome
- medulloblastoma
Statistics from Altmetric.com
Introduction
Gorlin syndrome (GS, MIM #109400), also known as nevoid basal cell carcinoma syndrome or basal cell nevus syndrome, is a dominantly inherited cancer predisposition syndrome. Gorlin and Goltz1 described a syndrome that included multiple basal cell carcinomas (BCC), jaw cysts and bifid ribs in 1960.1 The birth incidence of GS is approximately 1 in 15 000 births, with a prevalence of nearer 1 in 30 000.2 Affected individuals may show multiple phenotypic abnormalities, with characteristic facial features in over 50% of individuals that can include coarse facial features, macrocephaly with frontal bossing and hypertelorism.3 4 Diagnostic criteria for GS have been previously proposed by several groups.3 5–7 Approximately 70%–80% of individuals with GS have a first-degree relative with the syndrome, and in 20%–30% no family history is observed.5 Full diagnostic criteria are shown in box 1.8
Diagnostic criteria* as outlined in Jones et al
MAJOR CRITERIA
Lamellar or early (prior to age 20) calcification of the falx
Jaw keratocyst
Two or more palmar or plantar pits
Multiple basal cell carcinomas (>5 in a lifetime) or one prior to age 30
First-degree relative with Gorlin syndrome
MINOR CRITERIA
Medulloblastoma in childhood
Lymphomesenteric or pleural cysts
Macrocephaly (head circumference >97th percentile)
Cleft lip/palate
Vertebral/rib anomalies (eg, bifid/splayed/extra ribs or bifid vertebra)
Preaxial or postaxial polydactyly
Ovarian/cardiac fibromas
Ocular anomalies (cataract, developmental defects, pigment changes of retinal epithelium)
*Diagnosis made with two major or one major and two minor criteria fulfilled.
Patients with GS are at risk of developing benign and malignant neoplasms. Multiple basal cell (BCC) skin carcinomas are the hallmark feature most frequently occurring on sun-exposed areas such as the face, back and neck.8 Men and women are equally affected, and as of yet there has not been any clear PTCH1 genotype–phenotype correlation for the timing or number of BCCs that develop.8 Cardiac fibromas may develop in infants5 and ovarian fibromas in adolescent girls and women.5 Importantly, approximately 5% of individuals with GS develop medulloblastoma.5 9 Cases tend to present before 3 years of age, significantly younger than in sporadic cases, predominantly of the desmoplastic subtype,10 11 and are often the first manifestation of GS.9 11 12 In one review of 36 cases, 24 occurred aged ≤2 years of age, with all but one (97%) of the remaining cases occurring in less than 6 years.11 In addition, patients who are survivors of medulloblastoma treated with therapeutic radiation have a high risk of developing a large number of BCCs (>1000) in the radiation field.13 14
Germline pathogenic variants in genes of the sonic hedgehog (SHH) signalling pathway, including PTCH1, SUFU and in two case reports, PTCH2, have been found in individuals with GS,15–21 with PTCH1 variants being more common. Despite variants in PTCH1 having been known as the cause of GS for more than 20 years, no clear genotype–phenotype correlations have been described.
Methods
Patients with GS fulfilling the syndrome criteria (box 1) have been identified by the Manchester Centre for Genomic Medicine since the early 1980s. Syndromic features including those identified from a skeletal X-ray survey have been entered onto a bespoke FileMaker database. Numbers of BCCs including those previously removed were assessed through cutaneous examination and questionnaire. The majority of women had undergone a single ovarian ultrasound to detect ovarian fibroma. Jaw cysts were ascertained by orthopantogram screening, but the majority had presented symptomatically.
Affected individuals with GS (one from each family) were initially screened for germline PTCH1 pathogenic variants in lymphocyte DNA by Sanger sequencing and multiple ligation-dependent probe amplification (MLPA). Variant negative families were also screened for deep intronic pathogenic variants using RNA derived from cell lines.22 All negative families with available DNA then underwent Sanger sequencing and MLPA of SUFU. 18
Tests for significance were assessed by χ2 two-sided tests with Fisher’s exact correction.
Results
Clinical details on 232 individuals from 94 families with GS were available. No DNA sample was available for testing in 22 families. Of 182 patients who were genotyped (or where it was concluded from family testing), ages ranged from 0.5 to 90 years (median: 47.1; IQR: 31.1–61.1). PTCH1 pathogenic variants were identified in 43/72 families (60%) containing 126 affected individuals. SUFU pathogenic variants were found in 9 individuals from 3 (4%) families, and no pathogenic variant was identified in 26 families (36%) containing 47 affected individuals. As such a causative variant was identified in 46/72 families (64%) and 135/182 (74%) individuals. In isolated, apparently de novo, cases a pathogenic variant was found in 23/40 (57.5%) individuals. In contrast, a pathogenic variant was identified in 23/32 (72%) of second-generation familial cases. Overall, a PTCH1 or SUFU pathogenic variant was more likely to be found in an inherited than a de novo case (p=0.02). There were 50 people with truncating PTCH1 pathogenic variants, 26 with splicing variants (including one we have previously described with a deep intronic pathogenic variant (c.2561-2057A>G)),22 16 with exonic copy number variants detected on MLPA and 34 with missense PTCH1 pathogenic variants. The SUFU pathogenic variants previously described include a large multiexonic deletion.18 The proportion with a number of clinical features including age at diagnosis and age at last follow-up is shown in table 1. There was no significant difference in age at last follow-up, although those with missense variants had an older median age of 47.6 years. There were a number of clinical features that predicted the presence of a PTCH1 pathogenic variant (box 1). Patients with identified PTCH1 variants were more likely to be diagnosed earlier (median age: 19 vs 36 years; p=0.0008), have developed jaw cysts (62.7% vs 34.0%; p=0.002) and have bifid ribs (55.5% vs 34.2%; p=0.003) or any skeletal abnormality (74.3% vs 51.2%; p=0.003) than patients with no identified variant. Patients with missense variants in PTCH1 were diagnosed later (median: 26 years; p=0.03) and less likely to have developed at least 10 BCCs and jaw cysts than those with other PTCH1 variants (p=0.03) or to have developed at least 20 BCCs (p=0.05). Those with missense variants were also less likely to have all other GS features, including bifid ribs and jaw cysts, although this only reached statistical significance for the presence of any congenital skeletal anomaly (including vertebral defects) at 56.5% vs 76.5% for those with other PTCH1 gene variants (p=0.03). There was no other identifiable difference between the phenotypes of individuals with other PTCH1 variant types. All of the missense mutations were predicted to have some effect on the protein and were shown to segregate with disease when multiple affected family members were present. Three were also shown to have arisen de novo (table 2). None of the missense variants were reported in the ExAC database of around 121 200 alleles (http://exac.broadinstitute.org/gene/ENSG00000185920).
Patients with SUFU pathogenic variants were significantly more likely than those with PTCH1 mutations to develop a medulloblastoma (33% vs 2.4%; p=0.009) (as previously described),18 a meningioma (22.2% vs 1.6%; p=0.02) or an ovarian fibroma (42.9% vs 5.9%; p=0.015), but were less likely to develop a jaw cyst (0% vs 62.7%; p=0.0004).
Discussion
The present study has found a number of genotype–phenotype correlations in GS. A recent review identified no such correlations23 and we were not able to identify any from a PubMed review – December 2016. There are a number of key clinical features of GS that predict the presence of a PTCH1 pathogenic variant. These include the presence of skeletal anomalies (especially bifid ribs) and jaw cysts. The number of BCCs was not a predictor of the presence of a PTCH1 pathogenic variant as we have previously shown.8 It is of note nonetheless that a higher proportion of patients with GS without a pathogenic variant were sporadic (without a positive family history), and thus some may be mosaic for the underlying mutation. Although this is an extremely common mechanism in some other tumour prone syndromes such as neurofibromatosis type 2 (NF2),24 25 it has only reported once in GS.26 In theory, mosaicism should be relatively easy to prove with biopsy material potentially available from more than one BCC, although until recently mutational analysis was difficult on formalin-fixed material and required fresh tissue. With next-generation sequencing mosaic mutations may be found more frequently. As such mosaicism may still explain at least part of the difference between those with and without PTCH1 pathogenic variants. It is possible that some other sporadic patients may have fulfilled GS criteria by chance due to excess sun exposure, although the need for at least two major criteria makes this unlikely. It is possible that our techniques have failed to identify a few patients with PTCH1 inactivation, although this is unlikely given that RNA analysis was also performed. It is therefore likely that further as yet unidentified gene(s), likely within the hedgehog signalling pathway, accounts for most of the remaining unexplained cases.22
Although we were only able to identify nine patients with SUFU pathogenic variants, the rates of medulloblastoma and meningioma were significantly higher than those for individuals with PTCH1 pathogenic variants. A recent study of somatic mutations in meningiomas found that 5 of 775 contained a somatic SUFU mutation, but none contained a pathogenic PTCH1 mutation.27
A study of 131 childhood medulloblastoma cases identified germline SUFU variants in eight cases.10 Variants were identified in all 3 individuals with medulloblastoma with extensive nodularity, 4 of 20 with desmoplastic/nodular medulloblastomas and 1 of 108 with other subtypes. The study had already excluded four previously reported familial SUFU patients. The authors concluded that germline SUFU mutations (12 of 142; 8.5%) were more common than PTCH1 mutations (3 of 142; 2%) as a cause of childhood medulloblastoma, although they had only assessed PTCH1 through features of GS. A more recent study of 133 childhood medulloblastoma cases found germline SUFU mutations in 6 of 133 (4.5%) compared with 2 of 133 (1.5%) with a PTCH1 mutation.28 In contrast, somatic PTCH1 mutations are a more common cause of childhood medulloblastoma than SUFU.28 29 Indeed in their study of 133 SHH-related medulloblastomas, Kool et al 28 found more PTCH1 mutations (60 cases) than in SMO (19 cases) or SUFU (10 cases).28 Definite constitutional disease-causing truncating variants in SUFU are exceptionally rare: none are present on the ExAC database (http://exac.broadinstitute.org/) of over 60 000 individuals. A frameshift variant (p.Trp465LeufsTer6) was seen in 41 alleles; however, it was seen in homozygous form once and occurs in the last exon, meaning that it is likely to escape nonsense-mediated decay and is unlikely to be pathogenic. In contrast, eight PTCH1 truncating variants were present on the ExAC database. If one considers the 3–4-fold higher frequency of germline SUFU mutations in a series of medulloblastoma and a possible 8-fold higher frequency of germline PTCH1 in the general population, then one would expect a 24–32-fold higher incidence of medulloblastoma in SUFU mutation carriers. Even taking into account the 1 in 15 000 birth incidence estimate for GS, this would mean a 12–16-fold higher risk of medulloblastoma, which is consistent with the difference between a 2% risk in individuals with PTCH1 variants and 33% in individuals with SUFU variants in the present report. Therefore, while there is still some doubt over the true risk of medulloblastoma in SUFU-associated GS,18 it is likely to be many times higher than the risk for PTCH1-associated GS. Knowledge of a germline SUFU or PTCH1 mutation in a child with medulloblastoma is extremely important as SUFU patients are resistant to smoothened (SMO) inhibition.28
Although other reports have queried whether SUFU mutation carriers show clear features of GS,10 we have shown that all nine individuals in the present report met the diagnostic criteria, with the presence of key features such as skeletal anomalies (57%), ovarian fibromas (43%) and falx calcification (100%). It is also not clear whether the carrier parents of a SUFU mutation found in the French report had full assessment including a skeletal survey for GS.29 Although the risk of BCC from a SUFU mutation may be lower than that from a PTCH1 mutation, seven of nine (78%) patients with SUFU variants in the current study had developed BCCs,18 and two (22%) had developed more than 20 BCCs, including one individual (n=45 BCCs) who had not undergone radiotherapy. There are also reports of SUFU pathogenic variants in patients with GS,19 30 31 and an individual with hereditary infundibulocystic BCC has also been reported with a splicing mutation in SUFU.32 Meningiomas also appear more commonly in individuals with SUFU variants, although both in the current series had undergone radiotherapy for medulloblastoma. Nonetheless, a germline SUFU mutation has been shown to be the causative mutation in a family with familial meningioma,33 as well as one of the previously reported GS patients with a meningioma.30 31 To date, jaw keratocysts do not appear to be a feature of SUFU-associated GS with a highly significant absence compared with PTCH1-associated GS (62.7%) in the present study (p=0.0004).
In addition to the phenotypic differences between PTCH1-associated and SUFU-associated GS, we have shown there are also phenotypic correlations between different PTCH1 germline mutation types, with missense variants causing an apparently milder phenotype than truncating variants, with fewer BCCs, later age at diagnosis and fewer skeletal anomalies. Indeed, while not significantly reduced, all other GS features were less frequent. A number of other inherited tumour syndromes show correlations with missense mutations, including von Hippel-Lindau disease,34 SMARCB1-associated schwannomatosis35 and NF2.36 37 In von Hippel-Lindau, missense variants are associated with later onset and lower risk of retinal angiomas and renal cell carcinoma, but an increased risk of phaeochromocytoma.34 In schwannomatosis and NF2, missense variants cause a milder phenotype with later onset, and longer life expectancy for NF2.37 In addition, SMARCB1 missense variants are not associated with development of rhabdoid tumours unlike the majority of truncating mutations and large rearrangements.35 It is likely that many missense mutations although deleterious may retain some function in the protein product and thus represent hypomorphic variants.38
There are some limitations to the current study. Not all patients underwent a skeletal survey, and if more had done so further correlations could have been identified. We cannot be certain that all the missense variants are disease-causing, although they all segregated with disease in ascertained families or were shown to have occurred de novo. Although a missense change could have been a chance association, the frequencies of these variants in sporadic and inherited GS identified a number of clear cases that were identical to those with other PTCH1 mutations and quite different to the ratio in cases with no identified mutation. We have not adjusted for multiple testing within the same family statistically. Nonetheless, even though the missense variants did not have significance below an adjusted p value of 0.01, the fact that frequencies were below other PTCH1 mutations for all features means it is unlikely that these are chance findings. It is also consistent with our clinical impression that these patients have a milder phenotype. Finally, we did not screen PTCH2 in our cohort, so it is possible that PTCH2 mutations may account for a small subset of patients in whom no pathogenic variant was identified and who may have their own phenotypic characteristics.
In summary, the present report has identified a number of clear genotype–phenotype correlations that predict the presence of a germline PTCH1 or SUFU pathogenic variant. The relatively small numbers of patients with each of the different classes of PTCH1 pathogenic variant mean that more of these correlations may emerge in the future as they have for larger cohorts of patients with von Hippel-Lindau syndrome34 and NF2.37
References
Footnotes
Contributors DGE planned the study, drafted the manuscript and is responsible for the overall content of the study. MJS carried out in silico analysis and created table 2. DO, MJS, DR, EA, WGN and JTL collated and analysed the data. All authors reviewed, edited and approved the final manuscript.
Competing interests None declared.
Patient consent Not obtained.
Provenance and peer review Not commissioned; externally peer reviewed.