Multiple self-healing squamous epithelioma (MSSE): rare variants in an adjacent region of chromosome 9q22.3 to known TGFBR1 mutations suggest a digenic or multilocus etiology.

TO THE EDITOR Multiple self-healing squamous epithelioma (MSSE; OMIM132800), also known as Ferguson-Smith disease, is a rare, inherited skin cancer syndrome characterized by multiple invasive keratoacanthoma-like skin tumors that regress spontaneously leaving pitted scars (Ferguson-Smith, 1934; Goudie et al., 1993). The majority of the first-identified affected families have a shared Scottish ancestry, but MSSE has also been described in families of non-Scottish origin (D'Alessandro et al., 2007). Recently, the causative gene for MSSE was identified as TGFBR1, encoding a transmembrane serine/threonine kinase receptor involved in TGF-β signaling (Goudie et al., 2011). TGFBR1 mutations in MSSE are functionally null, implying a tumor-suppressor action in this disease, whereas missense mutations in the same gene can lead to Marfan syndrome–related disorders (Loeys et al., 2005; Goudie et al., 2011). TGFBR1 had previously been excluded as an MSSE candidate because the gene lies outside the original shared at-risk haplotype (SRH) in the Scottish families (Bose et al., 2006). However, using large-scale sequencing technology, Goudie et al. (2011) identified TGFBR1 germline mutations in a total of 18 MSSE families including 12 Scottish families. Of nine Scottish families with the SRH, seven families shared the same TGFBR1 mutation (p.G52R); importantly however, two families had different causative mutations (p.N45S and p.R414X). This suggests that unidentified rare variants associated with MSSE pathogenesis may exist in the non-TGFBR1 SRH region that has been conserved in nine families. The fact that this relatively large conserved haplotype had not undergone recombination over many generations in these nine affected Scottish MSSE families suggested that manifestation of the disease may require both the causative TGFBR1 mutation and additional variants located within the SRH region. To search for these SRH variants, we used the Agilent SureSelect Target Enrichment system (Santa Clara, CA), followed by sequencing on the Illumina (San Diego, CA) GAII platform to sequence all of the target genes between the markers D9S197 and D9S1809 on 9q22-31 (D'Alessandro et al., 2007). Thirteen families, including seven Scottish families with the SRH, were available (Supplementary Table S1 online). Family numbers hereafter correspond to the study by Goudie et al. (2011). We performed targeted sequencing in five families with the SRH and five families without the SRH, along with six Centre d’Etude du Polymorphisme Humain (CEPH) controls. Coding regions of the 15 then-known genes were sequenced previously (D'Alessandro et al., 2007); however, based on NCBI36 (hg18), there are now 29 known genes in the targeted region (Supplementary Table S5 online), encompassing ~2.2-Mb in total, and this entire genomic region of these genes was sequenced. After filtering out common variants included in dbSNP and shared with CEPH controls, we identified nine noncoding variants shared by all five families with the SRH. None of the nine rare variants were detected in MSSE families without the SRH. These nine variants are located in the intronic regions of five different genes: FAM120A, PHF2, C9orf3, FANCC, and PTCH1 (Table 1). Sanger dideoxy sequencing confirmed these nine variants in all five families plus two additional families (Supplementary Table S1, Supplementary Figure S1 online). Thus, all nine variants were conserved over this ~2.2-Mb region in seven Scottish families with the SRH, although TGFBR1 mutations were different in families 2 and 18 (Figure 1a). In 231 unrelated healthy controls, including 118 Scottish individuals, the minor allele frequency of the nine variants was rare, ranging from 0 to 0.022, and the association of these variants with the MSSE phenotype was highly significant (Table 1). This suggests that these nine variants are MSSE-associated and segregated in individuals with this rare skin malignancy condition. Interestingly, these variants are located at either end of the ~2.2-Mb target region, leaving an ~1.4-Mb central region where 24 genes are densely located (Figure 1a). Further analysis of MSSE families lacking the SRH identified distinct MSSE-associated rare variants in two Scottish families (Supplementary Tables S2 and S3 online). In family 17, we identified eight distinct variants that were detected in four affected family members (Supplementary Table S3 online). Interestingly, these eight variants were all clustered in a ~1.4-Mb central region of the SRH that excluded the nine MSSE-associated variants discussed above (Figure 1a). This family harbored the most complex TGFBR1 mutation (c.1059_1062delACTGinsCAATAA) that was not observed in other families. All of these variants are noncoding and not frequently found in 162 healthy Scottish controls tested. Three of the nine variants found were located in the intronic regions of the PTCH1 gene, of which germline loss-of-function mutations are responsible for nevoid basal cell carcinoma syndrome (Hahn et al., 1996; Johnson et al., 1996). Our previous study using mouse models suggested that a gain-of-function polymorphic variant in the mouse Ptch1 gene conferred susceptibility to squamous cell carcinoma (SCC) development (Wakabayashi et al., 2007). We considered the possibility that these rare PTCH1 variants may have a role in predisposition to SCC development in MSSE patients. We performed a computational analysis of the possible functional significance of all variants using the program AliBaba2.1.(AliBaba2.1; http://www.gene-regulation.com/pub/programs/alibaba2/index.html). This analysis identified the 97309311G>C PTCH1 variant as having possible functional significance. An electromobility shift assay was performed on the major 97309311G>C variant, which is reported to bind to multiple transcription factors, including SP1 and PU.1 (http://genome.ucsc.edu/ENCODE/). Although the major allele (97309311G) could bind to SP1 and PU.1 transcription factors in nuclear extracts from HaCaT-immortalized human keratinocyte cells, the minor allele (97309311C) showed complete disruption of this binding, indicating that the MSSE-associated rare variant perturbs normal interaction between this binding site and transcription factors (Figure 1b). In our study, we identified a rare MSSE-associated haplotype comprising nine variants, spanning a 2.2-Mb region, lying 1.5 Mb proximal to the causative TGFBR1 gene. Several of the genes in this region have been implicated in cancer development, and regulatory polymorphisms may affect tumor susceptibility (Sinha et al., 2008). The coinheritance of these variants in the majority of Scottish families with MSSE suggests that these variants, or other as yet unknown linked variants (or their combinations), may affect the expression of the skin cancer phenotype induced by the mutations in TGFBR1. Their locations at opposite ends of the conserved haplotype may suggest that they modify chromatin loop structure to influence the expression of genes within the region. Alternatively, the variants may directly affect the expression of relevant genes, particularly the PTCH1 gene, which is implicated in susceptibility to different forms of skin cancer and carries three intronic variants, one of which affects transcription factor binding. Functional analysis of the possible roles of these and other variants in determining the MSSE phenotype, e.g., by gene expression or chromatin immunoprecipitation sequencing analysis to assess differential transcription factor binding, will require further studies using keratinocytes or cultured primary tumor cells from MSSE-affected patients. Finally, it is possible that, as loss-of-function germline mutations in TGFBR1 are very rare, the conserved haplotype may act as a modifier to increase survival of individuals carrying strong mutations in this potent developmental regulator. This mechanism is supported by the identification of unlinked variant genetic modifiers that are preferentially inherited in mice haploinsufficient for Tgfb1 (Benzinou et al, 2012). The authors state no conflict of interest. This work was supported by Program Project Grant 5P01AR050440 from the US National Institute Of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health. AB is supported by grants from the National Cancer Institute Mouse Models of Human Cancer Consortium 2U01 CA08422-06. We thank Malcolm Dunlop for providing normal control samples used in the study. SUPPLEMENTARY MATERIAL Supplementary material is linked to the online version of the paper