Molecular alterations in basal cell carcinoma subtypes

A number of genes have been implicated in the pathogenesis of BCC in addition to the Hedgehog pathway, which is known to drive the initiation of this tumour. We performed in-depth analysis of 13 BCC-related genes (CSMD1, CSMD2, DPH3 promoter, PTCH1, SMO, GLI1, NOTCH1, NOTCH2, TP53, ITIH2, DPP10, STEAP4, TERT promoter) in 57 BCC lesions (26 superficial and 31 nodular) from 55 patients and their corresponding blood samples. PTCH1 and TP53 mutations were found in 71.9% and 45.6% of BCCs, respectively. A high mutation rate was also detected in CSMD1 (63.2%), NOTCH1 (43.8%) and DPP10 (35.1%), and frequent non-coding mutations were identified in TERT (57.9%) and DPH3 promoter (49.1%). CSMD1 mutations significantly co-occurred with TP53 changes (p = 0.002). A significant association was observed between the superficial type of BCC and PTCH1 (p = 0.018) and NOTCH1 (p = 0.020) mutations. In addition, PTCH1 mutations were significantly associated with intermittent sun exposure (p = 0.046) and the occurrence of single lesions (p = 0.021), while NOTCH1 mutations were more frequent in BCCs located on the trunk compared to the head/neck and extremities (p = 0.001). In conclusion, we provide further insights into the molecular alterations underlying the tumorigenic mechanism of superficial and nodular BCCs with a view towards novel rationale-based therapeutic strategies.

These investigational studies revealed new potential BCC driver genes, although their contribution to the genetic network underlying tumorigenesis and tumour evolution is not yet completely explained.
We examined the molecular alterations across 13 genes, selected on the basis of their potential role in the pathogenesis of superficial and nodular tumour subtypes, in order to provide further insights into the molecular sub-classification of BCC lesions.
Focusing on the analysis of single mutated gene according to the specific BCC subtypes, PTCH1 and NOTCH1 mutations were found significantly associated with superficial BCCs (p = 0.018 and p = 0.020, respectively). In details, PTCH1 variants were 1.6 times (OR = 5.537, 95% CI = 1.367-22.43) and NOTCH1 mutations 2.0 times more frequent (OR = 4.457, 95% CI = 1.304-15.24) in superficial than in nodular BCCs. www.nature.com/scientificreports/ The Principal Component Analysis (PCA) multivariate approach confirmed a significant association between PTCH1 mutations and the superficial BCC subtypes that were recognized as genetically similar group for PTCH1 mutations in a separate cluster of the PCA diagram (Fig. 3).
The analysis of the mutational status according to patients and tumour characteristics revealed that PTCH1 mutations were significantly associated with intermittent sun exposure (p = 0.046), and with the occurrence of single BCC lesions (p = 0.021), and NOTCH1 mutations were more frequent in BCCs arising on the trunk compared to the head/neck and extremities (p = 0.001).

Discussion
Our in-depth analysis was focused on a panel of 13 genes potentially associated to BCC tumorigenesis. We showed a high prevalence of Hh pathway mutations and a high rate of mutations in CSMD1, NOTCH1 and DPP10 genes, and in TERT and DPH3 promoter regions. Interestingly, NOTCH1 and PTCH1 mutations were significantly more frequent in superficial than in nodular BCCs, and CSMD1 mutations occurred along with TP53 changes. PTCH1 alterations were significantly associated with intermittent sun exposure and with the development of a single BCC, while NOTCH1 mutations with location on the trunk.
In line with previous studies 14-17, 25, 26 , 98.2% of BCCs showed at least one alteration among the analysed genes. PTCH1 and TP53 mutations were found in 71.9% and 45.6% of BCCs, respectively, being UV-fingerprint mutations (C > T and CC > TT transitions) the most common. Moreover, multiple PTCH1 and TP53 mutations were found in 28.0% and 17.6% of BCC lesions, respectively. We identified eight PTCH1 splice site mutations including the c.1347 + 1G > A and c.1216-2A > T, which have been recently recognized as novel PTCH1 variants 27 , probably pathogenic according to Associations for Clinical Genetic Science and American College of Medical Genetics and Genomics criteria 28,29 . Notably, a significant association was found between PTCH1 mutations and the superficial BCC type that was further confirmed by the PCA multivariate statistical approach. This is in line with evidences that topical imiquimod, which has been shown to negatively regulate Hh signalling 30 , is a successful treatment for superficial BCC. Moreover, a clinical trial is investigating the efficacy of topical patidegib, a novel topical Hh inhibitor, to decrease the number of surgically eligible BCC lesions in patients with multiple BCCs (NCT04155190) 31 . We found no mutations in SMO gene in our BCC samples, in contrast with the 10-20% frequency rate previously described [13][14][15][16][17] . Notably, this discrepancy might reflect clinical differences in the patient population: in our study only treatment-naïve patients have been included in the analysis while previous study showed a 2.2-fold higher frequency of SMO mutations in vismodegib-resistant compared to treatment-naïve sporadic BCCs (P = 5 × 10 − 2, Fisher's exact test) 14 .
In addition to the established BCC-associated genes, we identified high frequency of mutations in CSMD1, NOTCH1, DPP10, TERT promoter and DPH3 promoter genes. CSMD1 gene, encoding an inhibitor of the complement system, is believed to act as a tumour suppressor gene whose functions seems to be inactivated in many cancers [32][33][34][35] . Consistently with previous data 14, 15 , we detected CSMD1 somatic mutations in 63.2% of BCCs, thus representing the second most prevalent mutated genes after PTCH1. In addition, 38.6% of BCCs harboured more than one CSMD1 alterations. It is conceivable that, under the selective pressure during cancer development, multiple CSMD1 variants reside in separate subclones within the same tumour population. Notably, we found a significant association of CSMD1 alterations with TP53 mutation rate, which was previously described www.nature.com/scientificreports/ in mucosal head and neck squamous cell carcinoma 33 . The increased cellular proliferation driven by TP53 loss, might provide CSMD1 mutant clone to sufficiently expand. NOTCH signalling pathway in physiological condition plays a critical role in the regulation of cell differentiation, self-renewal and homeostasis primarily controlling the interplay between adjacent cells 36 , while in cancer can function as either an oncogene or tumour suppressor gene depending on the cell type and context 37,38 . Shi et al. 39 demonstrated that NOTCH pathway activity is suppressed in BCCs highlighting its tumour suppressor function in human epithelial malignancies. Moreover, NOTCH1 loss-of-function mutations were found to specifically promote tumour persistence in sporadic BCCs suggesting therapeutic restoring of the NOTCH tumour-suppressor function as a potential approach to eradicate persistent tumour cells 40 . Previous reports found NOTCH1 loss-of-function alterations (missense, truncating or loss of heterozygosity) in 30-50% of BCCs 14,15,25 . In our study, NOTCH1 mutations were found in 43.8% of BCC lesions and were mainly inactivating alterations, which support the tumour-suppressor role of NOTCH1 in BCC tumorigenesis. Interestingly, we found that NOTCH1 mutations were 2-times more frequent in superficial BCCs than in nodular subtypes, suggesting that tumorigenic pathways may differ across BCC subtypes.
The DPP10 gene encodes a trans-membrane protein, which regulates potassium channels activity involved in cell proliferation and apoptosis 41 . We found mutations of DPP10 gene in 35.1% of BCCs, while a previous study on 12 BCCs showed mutations in 75% of cases 15 . Such difference in frequency rate might be due to the sample size or different subtypes of BCC lesions analysed.
Noncoding somatic mutations of TERT promoter are emerging to have a crucial pathogenetic role in BCCs. TERT non-coding mutations were previously identified in 39-74% of BCCs and considered to contribute to telomeres length maintenance in cancer 26,[42][43][44][45][46][47][48] . In addition, TERT promoter mutations in BCCs were associated with shorter telomere and increased transcription of the telomerase reverse transcriptase subunit 13,26 . It is thought that the genome instability caused by critically short telomeres promotes telomerase upregulation, thus sustaining cell proliferation and tumorigenesis 49 . In the present study, TERT promoter mutations were found in 57.9% of tumours although they were mutually exclusive.
We also identified DPH3 promoter mutations in 49.1% of BCCs, being the -121 C > T and -122 C > T transitions the most frequent ones. Mutations in the DPH3 promoter were previously described in 42% of BCCs. 42 In a recent study, 73/191 (38.2%) BCCs were found to carry DPH3 promoter mutations that were significantly more frequent in BCC patients with a clinical history of cutaneous neoplasms 24 . However, the effect of DPH3 promoter mutations on transcription of adjacent or distant genes remains currently unclear 42 . The exclusive identification of mutations at dipirimidinic sites into DPH3 and TERT promoter further highlights the role of UV-induced DNA damage in BCC tumorigenesis.
Additional in vivo studies evaluating the function of the novel variants (56%), of which 22.2% with uncertain significance, are needed to clarify their implications in BCC tumours.
In conclusions, our study provides further insights into the molecular alterations underlying the tumorigenic mechanism of superficial and nodular BCCs showing that additional genes and pathways beyond PTCH1-axis might contribute to BCC development and progression.

Material and methods
Patients and tumour samples. Sporadic BCC tumour tissues and matched blood samples were collected at the Institute of Dermatology, Catholic University-Fondazione Policlinico Universitario A. Gemelli-IRCCS, (Rome, Italy) and at the Dermatologic Clinic of the University of L' Aquila (L' Aquila, Italy) from January 2015 to July 2017 with complete medical records. Examples of superficial and nodular BCC lesion included in our study are reported in Supplementary Fig. S1. Informed consent was obtained from all patients after study approval by the local ethical committee (IRB number:15272/14, Ethical Committee Fondazione Policlinico Universitario A. Gemelli-IRCCS) and the research was performed in accordance with relevant guidelines and regulations stated in the Declaration of Helsinki. Patients and tumours characteristics are illustrated in Table 1. Only superficial and nodular types of BCCs, histologically reviewed by a single histopathologist, were included in the study.
During surgical excision, a 4-mm intra-tumoral punch biopsy specimen was obtained and stored in RNA later solution at -20 °C. DNA was extracted from fresh-frozen tumour samples using Qiagen All Prep-DNA/RNA/ miRNA extraction kit (Qiagen, Hilden, Germany) after tissues homogenization in a Precellys24 homogeniser (Bertin instruments, Montigny-le-Bretonneux, France). DNA from whole blood was purified using Qiagen QIAamp Blood Midi Kit (Hilden, Germany) and quantified by Qubit dsDNA HS Assay Kit (ThermoFisher Scientific, Waltham, MA USA) on Qubit 2.0 Fluorometer instrument (Invitrogen, Carlsbad, CA, USA). Genomic DNA at least 10 ng/μL concentrated was subjected to library preparation. www.nature.com/scientificreports/ Browser 18 . Raw data were then processed for somatic variant calling and annotation by following the VarDict workflow in paired samples (tumour/normal) analysis mode. The allele frequency and mapping quality thresholds were set at 0.05 and 30, respectively. Only the variants recognized as "StrongSomatic" were selected. To reduce the false positive rate during somatic mutation calling, several steps were used in order to retain only the high-quality variants with a minimum depth of total coverage ≥ 300 reads and each variant coverage ≥ 20 reads. Filtered variants were functionally annotated using Annovar software version 2018 Apr 16 19 .
Prediction tools analysis. All  Statistical analysis. The relationship between mutations and clinic-pathological features was evaluated using logistic regression analysis with estimation of OR and 95% CI. Semi-quantitative data were analysed by means with Student's t test or by medians with Mann-Whitney test. P values < 0.05 were considered statistically significant. Computations were performed using the R v3.6.2 statistical package. Mutual exclusivity analysis was performed by using the cBioPortal for Cancer Genomics tool 20 that identify patterns of mutual exclusivity or cooccurrence through a Fisher's exact test 21 . The PCA (Principal Component Analysis) multivariate analysis was performed in order to identify genetic variations among superficial and nodular BCC subgroups. PCA diagram was generated with the prcomp R function and plotted with ggplot2 23 .
Ethics declarations. The study was approved by the local ethical committee (IRB number: 15272/14, Ethical Committee Fondazione Policlinico Universitario A. Gemelli-IRCCS) and the research was performed in accordance with relevant guidelines and regulations stated in the Declaration of Helsinki. The patients in this manuscript have given written informed consent to publication of their case details.