Tumor Mapping in 2 Large Multigenerational Families With CYLD Mutations: Title and subTitle BreakImplications for Disease Management and Tumor InductionTumor Mapping in Families With CYLD Mutations FREE

Heterozygous mutations within the CYLD gene locus have been identified as the cause of 3 clinically distinct dermatologic phenotypes: familial cylindromatosis (OMIM 132700), Brooke-Spiegler syndrome (OMIM 605041), and multiple familial trichoepitheliomas (OMIM 601606).^{1 - 3} These conditions are unified by a predisposition to inherited skin appendage tumors, the diagnostic hallmark being the presence of cylindromas. Although rare in the population, these disorders are associated with a high level of morbidity, which can greatly influence the quality of life of the affected individual. In particular, these highly disfiguring tumors often affect the face and scalp,⁴ which in some cases culminates in entire scalp removal.

Cylindromas are purported to arise from hair follicle stem cells within the bulge region of the hair follicle,⁵ a model supported by the numerous tumors seen on hair-bearing sites. Because the wildtype CYLD allele is lost in 70% of tumors, it has been postulated that CYLD has a classic tumor suppressor role in cylindroma formation.³ Ultraviolet radiation has been suggested as the major predisposing factor for loss of heterozygosity and tumor initiation⁶ on the basis of the presence of face and scalp tumors. However, careful clinical mapping of the distribution of the tumors does not exist to support this hypothesis.

Four CYLD knockout mouse models have been derived, which may help inform clinical phenotypes associated with germline CYLD mutations. The phenotypes seen in CYLD-null (CYLD^null) mice include an increased sensitivity to cutaneous squamous papilloma formation,⁶ immunologic defects involving T-cell maturation⁷ and B-cell responses,⁸ an increased predisposition to inducible colitis and colorectal tumor formation,⁹ and a protective effect against lethal Streptococcus pneumoniae infections.¹⁰

The CYLD locus encodes a ubiquitin hydrolase.¹¹ This has been implicated in the negative regulation of cell proliferation through a direct role in both the nuclear factor κB and c-JUN pathways.^{11 - 12} Recently, studies have identified a group of drugs including aspirin that can inhibit the nuclear factor κB pathway downstream of CYLD's putative point of action.¹³ Excitingly, this opens up new avenues for chemoprevention and potential nonsurgical approaches to the treatment of tumors in these patients.

Previous work (Table 1) investigating clinical phenotypes across multiple families has shown little phenotype-genotype correlation with regard to the position of a germline CYLD mutation and the tumor types seen in each family.¹ We were interested in examining clinical phenotypes, tumor distribution, and the burden of disease associated with the diagnosis of germline CYLD mutations in 2 large multigenerational families. This information is a prerequisite for future accurate genetic counseling for individuals both within and outside of these families.

Appropriate approval from the local ethical review board was obtained and written informed consent was obtained from patients.

Two large families with known germline mutations³ within the CYLD locus were identified. Family A had a frameshift mutation (c.2460delC), resulting in a premature termination codon, whereas family B had a splice site mutation (c.2469 + 1G>A). In total, 26 individuals determined by a dermatologist (N.R.) or geneticist (J.B.) to be affected were surveyed, of whom 18 were clinically examined by the same dermatologist (N.R.). Eight unaffected relatives—5 from family A and 3 from family B—provided blood samples to exclude the presence of a nonpenetrant mutation.

A search was conducted in PubMed for all publications with the following search terms: CYLD, cylindroma, Brooke-Spiegler syndrome, familial cylindromatosis, and multiple familial trichoepitheliomas. The phenotypes seen in CYLD knockout–associated models and all clinical phenotypic data were collated into a clinical questionnaire. The questionnaire aimed to gather clinical data on patient signs and symptoms, disease severity, hormonal exposure in females, associated diseases, and response to existing aspirin or nonsteroidal anti-inflammatory drug therapy.

Because of the extensive geographic spread of the families, a template was designed, and affected members were asked to self-report the exact locations of tumors. However, because most patients lived in the northeast of England, detailed examination of 18 patients was performed to ensure corroboration. Tumor mapping done in the clinic correlated well with the template data on the torso and face. Detailed location on the scalp was accurate when there were fewer than approximately 10 tumors; accurate resolution was not possible when confluent tumors were present.

Eight affected patients who were not examined for geographic reasons were interviewed by telephone. These patients self-reported lesion locations on a template and supplemented this subjective description with confirmatory digital photographic images. Histologic reports were reviewed to confirm the nature of the tumors where available. The distribution of tumors on each patient was ultimately marked on a template and a composite image was created by overlaying the templates according to sex by means of image processing software (Adobe Photoshop).

Genomic mutation analysis for the known familial CYLD germline mutations was undertaken in affected and unaffected individuals within both families. This was performed by the National Health Service molecular diagnostic laboratory based in the Institute of Human Genetics, Newcastle Upon Tyne.

The pedigrees of the families are shown in Figure 1. They both demonstrate an autosomal dominant pattern of inheritance across 7 generations in family A, with 34 of 133 individuals being reported as affected, and 5 generations in family B, with 16 of 77 individuals being reported as affected. In total, 50 of 210 members of both families were reported as clinically affected, 27 female and 23 male. This number is significantly less than would be expected with an autosomal dominant inheritance pattern and could be accounted for by either nonpenetrance of the mutation or a milder presentation of the condition, which was underreported especially in the earlier generations. In particular, female IV-1 and male V-3 in family A, obligate carriers by family history for the CYLD mutation, were reported as unaffected. Neither patient was alive to confirm clinical status.

To address the possibility of nonpenetrance within these families, we undertook CYLD mutation analysis in all available living affected individuals and contacted nonaffected individuals to offer them testing within both families. Nine members of family A, in whom a diagnosis of Brooke-Spiegler syndrome and familial cylindromatosis had been made, were shown to have the familial mutation by sequencing. Five unaffected members (age range, 33-64 years) were shown not to have inherited the mutation. Family B was similarly examined, with 3 affected members having a familial cylindromatosis phenotype found to have the familial mutation and 3 unaffected members (age range, 40-54 years) shown not to have the mutation.

Although numbers are small, we found no indication for nonpenetrance of either CYLD mutation. The unaffected obligate individuals present in this family most likely indicate a variation in CYLD mutation expression with family A.

In our population, 5 female patients and 1 male patient developed severe scalp disease that necessitated entire scalp removal and repair with split-thickness skin grafts. Their mean age at the time was 55 years (range, 50-60 years). Patients who had undergone this procedure were all noted to have multiple tumors on the trunk and genital areas. Other markers of severity included the use of wigs preoperatively (6 of 26 patients [23%]) and painful tumors (12 of 23 [52%]; 3 did not respond to this question).

Although there is the suggestion of clustering within the pedigree of individuals with a severe tumor burden, relatives of these patients were often noted to have only mild disease. These patients had a few tumors that were apparent only on close inspection of the scalp. Conversely, 1 patient in family A (V-14), who had only 3 tumors on the face, fathered 4 children who each presented with more than 30 tumors on the face. In addition, unaffected obligate carrier IV-1 had 11 affected progeny, 2 of whom required total scalp removal.

The clinical and tumor data of 26 affected individuals are summarized in Table 2. The average age at onset of tumor formation was 16 years (range, 8-30 years), consistent with tumor initiation occurring after adrenarche. In both families, the most common site for the presenting tumor was the scalp. The clinically diagnosed tumor phenotypes were supported by histologic findings in 10 of 26 patients.

Family A presented with classic scalp cylindromas, eccrine spiradenomas, trichoepitheliomas, and milia. This wide variation in tumor type and clinical presentation in family A supported a clinical diagnosis of Brooke-Spiegler syndrome in most patients. However, some individuals with exclusively facial trichoepitheliomas, if seen in an isolated setting, would have been diagnosed as having multiple familial trichoepitheliomas. In contrast, family B presented almost exclusively with cylindromas, supporting a diagnosis of familial cylindromatosis in most of these individuals.

The detailed tumor distribution maps comprising a total of 14 female and 12 male patients from both families were combined to produce composite maps (Figure 2). The tumors extended beyond the head and neck, with the trunk being involved in 18 patients (69%). The tumor distribution was V-shaped on the trunk, with tumors being more concentrated in the midline. Noticeably, tumors were highly concentrated on the hair-bearing pubic and genital areas, with a relative absence on other areas of the body such as the lower legs and axillary skin. Tumors were not seen on hairless areas of the body such as the palms and soles.

In addition, although not depicted on the tumor maps, patients with advanced scalp tumors (family A: VI-1; family B: IV-1, IV-2, and IV-3) showed a higher density of tumors over areas that are predisposed to androgenic alopecia²⁹ (Figure 3D). Although in these patients tumors did occur on the occipital and temporal scalp, they were smaller and not confluent, suggesting this to be a true predisposition rather than one that was biased because of balding. Furthermore, this distribution was also seen in female patients. Interestingly, 1 woman with severe disease reported an increase in the number of tumors during pregnancy and also when menopausal hormone therapy was started. As this had been previously noted anecdotally,³⁰ the remaining 13 affected females were directly questioned. No relationship was seen between number of pregnancies, hormonal contraception, menopausal hormone therapy, or duration of menarche-menopause in other patients.

Pain and disfigurement were commonly reported in individuals from both families as major signs and symptoms. Lesion-associated pain was a major complaint in more than 50% of affected individuals and was attributed by the patients to histologically confirmed cylindromas as well as spiradenomas. Although eccrine spiradenomas could be distinguished clinically because they were usually painful on light pressure and had a deep blue hue (Figure 3A), some painful cylindromas were also noted to have blue regions that correlated histologically with areas of spiradenomatous differentiation (Figure 4). This finding of regions within the same tumor showing spiradenomatous and cylindromatous patterns has been previously reported.³¹

Apart from pain, tumors on the back that were compressed during sitting or sleeping were susceptible to pressure necrosis, which resulted in problems due to ulceration and chronic discharge. Tumors in the pubic area were found in 11 (42%) of the affected individuals and resulted in sexual dysfunction (Figure 3B). Seven (27%) of the affected individuals were noted to have numerous confluent cylindromas in the ear canal, resulting in a conductive deafness when occlusion of the ear canal occurred (Figure 3C).

Taken together, the effect on quality of life caused by repeated surgery, scalp removal, painful tumors, deafness, and sexual dysfunction appears to be underreported in the literature and a cause of major unaddressed morbidity within families.

In keeping with a previously reported association,³² 1 individual was also diagnosed as having bilateral parotid gland tumors and underwent bilateral parotid surgical excision (family B, III-5). Unfortunately, we were unable to obtain the histology report on these tumors.

Because of the immunodysregulation observed in CYLD^null mouse models, affected patients were also questioned as to symptoms associated with either an immunodeficiency or autoimmune phenotype. The results showed no evidence of a predisposition to recurrent infections, autoimmune disease, colitis, infertility, or noncutaneous cancers. One patient reported vitiligo; other diseases reported included ischemic heart disease and osteoarthritis.

This is, to our knowledge, the most comprehensive study of the clinical features associated with germline heterozygous mutations within the CYLD gene locus.

In addition to the well-recognized morbidity associated with face and scalp cylindromas in these families, the tumor predisposition associated with this condition extends beyond the face and scalp, affecting both the trunk and genital areas. These less well-recognized sites of tumor formation can greatly affect the quality of life experienced by these patients. Unless actively sought, these complications may be underdiagnosed and not treated.

In our patient cohort, we found only a slight female preponderance of the disease. However, the impact of this disease appears to be more severe in females, with 5 women, compared with only 1 man, undergoing entire scalp removal. Although clinical photographs support the underlying clinical severity that warranted this procedure, it is also feasible that the psychological impact and social disability have a greater effect on females.

Basal cell adenomas and adenocarcinomas of the parotid glands and minor salivary glands have been reported in association with this disorder.³² In our families only 1 such case was reported, suggesting that this is an uncommon event in these families.

None of our patients developed any of the clinical features associated with the CYLD^null mouse models.^{7 - 10 ,33} Conversely, none of the CYLD^null mice developed cutaneous cylindromas. This phenotypic difference may result from the germline molecular heterozygosity found in the patients compared with the homozygote status of the CYLD^null mouse models and would suggest either that loss of the second CYLD allele in humans is a rare event outside of the hair follicle or that somatic loss of the second CYLD allele in cells from other tissues may be nonviable. However, this explanation does not account for the observation that there is overrepresentation of mutations within the 3′ end of the CYLD gene in all published human cohorts ascertained on their predisposition to cutaneous tumors.

Thus, an intriguing alternative explanation for the phenotypic difference between CYLD^null mice and the patients is that the human mutations represent hypomorphic alleles, which could potentially result in the production of a truncated protein with the potential for tissue-specific dominant negative pathogenic interactions. Evidence underpinning this theoretical possibility comes from recent murine models in which truncating mutations (CYLD^truncated) were engineered into the mCYLD locus. In direct contrast to the viable phenotype seen in the CYLD^null models,^{7 - 10 ,33} the animals in the CYLD^truncated mouse models died shortly after birth.^{34 - 35} However, in humans, mechanistic data are still limited. In cylindroma tissue, a truncated form of the CYLD protein has not been visualized on immunohistochemistry despite the antibody recognizing the N-terminus of the CYLD protein.⁶ In addition, the only heterozygous full CYLD gene deletion reported to date is in a 14-year-old girl with a large heterozygote chromosome deletion encompassing the CYLD locus.³⁶ Although the preliminary report does not indicate the presence of cutaneous tumors, she is currently too young for this to be excluded as a possible future complication. Further work is necessary before precise mechanisms of the pathogenicity of the mutated CYLD locus can be formulated.

The pluripotent hair follicle stem cell as the probable cell of origin of cylindromas⁶ is supported by our clinical data. The occurrence of multiple, different, rare³⁷ skin appendage tumors in these patients suggests that tumors may arise from a cell that is not fully committed to a lineage of differentiation. Furthermore, sites recognized to have an absence of hair follicles, such as the palms and soles, are spared. Further evidence of a pluripotent cell of origin comes from direct histologic examination of excised tumors showing evidence of a mixed differentiation lineage, with features of cylindroma and eccrine spiradenoma seen within the same tumor (Figure 4). As the hair follicle stem cell compartment becomes better defined,³⁸ it is to be hoped that the cell of origin of cylindromas will become clearer.

One key event in familial cancer tumorigenesis is loss of the wild-type allele in a previously heterozygous cell (loss of heterozygosity). Because of the high incidence of tumor formation on the face and scalp in patients with a CYLD mutation, it has been postulated that UV radiation–associated DNA damage may be the main initiating factor in the development of cylindromas.^{33 ,39} Our data support this hypothesis: all patients had a tumor on the face or scalp, and the greatest number of tumors was at these sites. In contrast, however, we demonstrate a high incidence of genital and trunk cylindromas and spiradenomas in more than half of the affected patients. Presence of tumors at lifelong light-protected sites such as the pubic skin indicates that other mechanisms independent of UV radiation must play a role in tumor initiation.

Further supportive data come from a study of patients with nevoid basal cell carcinoma syndrome, in which the cutaneous tumors were also thought to arise from hair follicles. That study⁴⁰ showed that patients with nevoid basal cell carcinoma syndrome (who carry PTCH mutations) have a disproportionately high proportion (59%-65%) of tumors on the trunk compared with the general population (9%-12%), who present with basal cell carcinoma. There was no correlation to sun exposure behavior (although numbers were small), and tumors were seen at genital sites as well as in highly pigmented skin types. Taken together, the distribution of skin tumors resulting from germline mutations in 2 known tumor suppressor genes suggests that alternative, non–UV radiation–mediated mechanisms of tumor induction are pertinent.

One feature of human hair that may demonstrate such mechanisms is the differential sensitivity of follicles at different body sites to hormonal stimulation. Puberty highlights follicles that change from fine vellus hair to coarse hair in response to hormonal stimulation in some areas such as the beard area, axilla, and pubic region. Another follicular response seen is acne, in which hormonally driven proliferation at the opening of the hair follicle at certain sites results in comedones, inflammatory pustules, and papules. The detailed mapping of tumor distribution in our patient cohort demonstrates an increased incidence of tumors in areas associated with hormonally stimulated hair follicles such as the pubic hair and upper trunk. The distribution on the torso correlates well with the distribution of hair follicles affected by acne at puberty. The converse of this, a relative absence of tumors on unstimulated sites such as distal limbs, palms, and mucous membranes, was also seen. Further evidence that the tumors are more likely to arise from hair follicles that are hormonally stimulated comes from the observation that the onset of tumors occurs after adrenarche¹⁵ and from the larger size and confluency of scalp tumors in areas predisposed to androgenic alopecia (male-pattern balding is thought to reflect follicles that preferentially senesce after androgen stimulation⁴¹ ). Finally, recent preliminary data in mouse follicles have shown that hair follicle cycling is dependent on CYLD.⁴² Although the precise mechanisms remain undetermined, the clinical data are supportive of hormonally sensitive hair follicles being vulnerable to loss of heterozygosity at the CYLD locus and tumor induction.

The data we present herein confirm that, even within one family, mutations within the CYLD locus give rise to a wide variation of clinical phenotypes. Fifty mutations have now been identified in CYLD, with the majority found toward the carboxyl terminus of the protein, the position of the catalytic residues of ubiquitin hydrolase.^{2 ,39} Clustering in this location suggests that the loss of deubiquitinating activity is important for the development of a cutaneous phenotype. The poor genotype-phenotype correlation with regard to tumor histologic findings and the wide variation in clinical expression that we demonstrate in our families has been noted by other authors.^{1 ,18 ,43} In contrast, there have been families in China with up to 4 generations affected^{16 ,19} with predominantly multiple familial trichoepitheliomas, whereas the same mutation in other populations has resulted in heterogeneous phenotypes, highlighting the possibility of population-specific modifier genes.

This wide variation in clinical phenotype means that genetic counseling and prognostication in families with known CYLD mutations is not straightforward. The presence of 2 obligate carriers in our pedigrees who were reported as unaffected by their relatives, but who had severely affected children, underlines the importance of offering all “at-risk” individuals within a family predictive testing if available or examination by a specialist with a knowledge of the condition.

A review of the literature places the oldest patient to develop a first tumor at 42 years.⁴⁴ Taken with our data from unaffected patients of the pedigree who underwent sequencing, this suggests that unaffected individuals in families who reach this age are unlikely to be carriers of a mutation.⁴⁵ Because most couples have children before this age, predictive testing for a known familial mutation rather than clinical examination of the “at-risk” individual would be required if reassurance concerning a future or ongoing pregnancy were requested. Our results also indicate the impact of the disease on the lives of future generations of affected individuals cannot be gauged by the clinical severity of previously affected individuals.

On the basis of the variation seen in our European families, we suggest that the clinical diagnosis of Brooke-Spiegler syndrome, multiple familial trichoepitheliomas, or familial cylindromatosis has no prognostic, diagnostic, or descriptive value and should be abandoned. Individuals found to have a germline CYLD mutation should be given the diagnosis of “CYLD cutaneous syndrome” because the eventual clinical phenotype may be uncertain. This nomenclature will also avoid confusion with regard to possible genetic heterogeneity.

The need for nonsurgical approaches in the treatment of these families is emphasized by the repeated surgical procedures that individuals in both families underwent, the frequencies often being restrained only by available clinical resources.

Our current understanding of CYLD's function as a negative regulator of the nuclear factor κB pathway¹¹ highlighted a role for aspirin in the treatment of these patients.¹³ However, the clinical efficacy of aspirin in 1 trial of topical aspirin was limited, with only 2 of 12 tumors showing complete remission after 24 weeks.⁴⁶

However, this preliminary work does suggest that further studies are required to analyze the dose, timing, and efficacy of nuclear factor κB–modifying drugs, perhaps at an earlier stage of tumorigenesis or prophylactically in these patients.

To summarize, wide variation in both clinical severity and tumor type was present in 2 large multigenerational families with CYLD cutaneous syndrome. The histologic heterogeneity of the tumors adds weight to the argument that a hair follicle stem cell is the cell of origin. Significantly, the distribution of tumors suggests that hormonally sensitive hair follicles may be predisposed to tumor formation. Further understanding of the role of CYLD in regulating pathways such as nuclear factor κB may help in the development of nonsurgical treatments of this devastating condition.

Correspondence: Neil Rajan, MBBS, MRCP, Institute of Human Genetics, University of Newcastle Upon Tyne, Newcastle Upon Tyne NE1 3BZ, England (drneilrajan@googlemail.com).

Accepted for Publication: April 9, 2009.

Author Contributions: Drs Rajan and Trainer had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Rajan, Langtry, Chapman, Burn, and Trainer. Acquisition of data: Rajan, Roberts, Chapman, Burn, and Trainer. Analysis and interpretation of data: Rajan, Ashworth, Burn, and Trainer. Drafting of the manuscript: Rajan, Langtry, Ashworth, Chapman, and Trainer. Critical revision of the manuscript for important intellectual content: Rajan, Langtry, Ashworth, Roberts, Burn, and Trainer. Obtained funding: Rajan, Burn, and Trainer. Administrative, technical, and material support: Roberts, Chapman, and Trainer. Study supervision: Langtry, Ashworth, Burn, and Trainer.

Financial Disclosure: None reported.

Funding/ Support: This study was supported in part by the North East Skin Research Fund and Cancer Research UK.

Additional Contributions: We are indebted to the families who made this study possible. David Bourn, PhD (Institute of Human Genetics, University of Newcastle Upon Tyne), provided technical help in the work.