Loss-of-function mutations in Notch receptors in cutaneous and lung squamous cell carcinoma

Identification of Notch Mutations.

Three additional cSCCs were sequenced on the whole-exome level, in addition to the eight originally reported (14). For these 11 samples, ∼1,300 mutations per genome were identified; >85% were made up of G>A transitions consistent with UV damage. Nine samples showed at least one NOTCH1 or NOTCH2 mutation, including two samples, cSCCs P8 and P10, derived from solid organ transplantation patients exposed to long-term immune suppression (Fig. 1 and Table 1; ref. 14). Six of these mutations were duplicated truncations with loss of the wild-type allele. One additional primary tumor and 14 cSCC cell lines were Sanger sequenced for 30/34 exons of NOTCH1 and 30/34 exons of NOTCH2, revealing 15 additional missense mutations and yielding an overall nonsynonymous mutation prevalence of NOTCH1 or NOTCH2, per sample, of 19/26 (∼75%). Thirty-three of 43 total missense or nonsense mutations in Notch genes resulted from a G>A transition. By contrast, exome or transcriptome sequencing of five primary basal cell cancers (BCCs) and directed Sanger sequencing of three BCCs did not identify missense mutations in NOTCH1 or NOTCH2 (Table S1). No association with clinical subtype or TP53 status was noted (Table S2).

Table 1.

Identified amino acid substitution mutations in Notch receptors and pathway genes in primary and immortalized cSCCs

Open in a new tab

Shaded regions were not assessed (cell lines were Sanger sequenced for 30/34 exons each of NOTCH1 and NOTCH2 only). *denotes stop codons; bold/italicized changes are homozygous. RBJP, MAML1-3, and JAG1 mutations were not identified in any samples (data not shown).

Recent exome-sequencing data from 40 lung SCCs acquired by The Cancer Genome Atlas (TCGA) project were also examined. Three missense mutations and one splice site mutation were identified in NOTCH1, as well as one missense and one nonsense mutation in NOTCH2—a combined mutation frequency of 5/40 samples or 12.5% (Table S1). Sanger sequencing for NOTCH1 and NOTCH2 in lung SCC cell lines SW900 and HCC95, esophageal SCC lines TT and TE10, and lung adenocarcinoma lines H549 and A549 revealed a single heterozygous nonsense NOTCH1 mutation in TE10.

Analysis of Unique Notch1 Loss-of-Function Mutants.

Amino acid (missense) changes were analyzed by PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/) for potential structural effects. Eighteen of 27 mutations were identified as “probably damaging” with the remainder being labeled “probably benign” (Table S3). Besides overt loss-of-function frameshift and nonsense mutations, NOTCH1 point mutations involved the extracellular EGF-like repeats, the juxtamembrane heterodimerization domain, and the intracellular RAM domain. To exclude the possibility that the observed changes are nonpathogenic passenger events, we tested the functional consequence of mutations involving each of these domains. An EGF-like repeat mutation (D469G) and a unique heterodimerization domain mutation (R1594Q) interfered with ligand-mediated activation of Notch1 (Fig. 2A). D469, a residue within EGF repeat 12 (Fig. 2B), contributed to binding of a Ca²⁺ ion (27), which fixed the orientation of EGF repeat 11 with respect to repeat 12 (28). Prior genetic studies have shown that EGF repeats 11–12 are necessary and sufficient for ligand binding by Drosophila Notch (29), and biochemical studies have implicated D469G in binding of recombinant Notch1 EGF repeats 11–14 to the human Delta-like-1 ligand (30).

P1770S is a mutation in the Notch1 RAM domain, a region that binds the downstream transcription factor RBPJ and thereby contributes to the assembly of Notch transcription complexes (31–33). The P1770S substitution greatly diminished signaling when scored in ΔEGFΔLNR, a truncated form of Notch1 that signals in a ligand-independent fashion (Fig. 3A). This mutation also prevented formation of stable ICN1/RBPJ complexes on DNA (Fig. 3B). Thus, P1770S appears to interfere with Notch signaling at the level of transcription complex assembly.

Sample Acquisition.

Tumor and matched normal tissue samples were obtained as part of an established skin cancer study protocol, with all subjects providing informed consent according to procedures approved by the University of California, San Francisco Committee on Human Research. Samples were either immediately frozen in liquid nitrogen or deposited in RNAlater preservative (Qiagen). Diagnosis of either cSCC or BCC was confirmed for all tumors through histological examination of a standard biopsy specimen by a board-certified dermatopathologist. Patient information and genomic profiling for lung SCCs analyzed here were obtained from sequencing completed by TCGA and deposited in the Database of Genotypes and Phenotypes. Further details are provided in SI Materials and Methods.

Isolation and culture details, as well as detailed characterization of cSCC lines SCCRDEB2-4, SCCT1-3, and -8 and SCCIC1, have been described (51). Two additional lines sequenced here, SCCIC8 and SCCIC12, were established from female immunocompetent patients aged 51 and 87, respectively. The tumor sites were, respectively, buttock (poorly differentiated spindle cell) and left calf (moderately to poorly differentiated SCC).

Sequencing.

For exome sequencing of 11 cSCCs and matching normal tissue, ∼40 megabases of coding region were selected from each genomic DNA sample by using oligonucleotide-based hybrid capture and sequenced by using the Illumina sequencing-by-synthesis platform. Three primary BCCs were similarly sequenced on the whole-exome level, and two additional samples were sequenced for transcribed sequences only. Standard methods for alignment, PCR duplicate removal, recalibration of base scores, and mutation calling were applied (detailed description is provided in SI Materials and Methods).

Three primary BCCs and 1 additional primary cSCC and matching tissue, as well as 14 cSCC cell lines, were capillary PCR-sequenced (Sanger) for NOTCH1 and NOTCH2 mutations. Thirty of 34 exons of NOTCH1 and 30/34 exons of NOTCH2 were reliably PCR amplified. Coverage of coding sequence was assessed by computing sequenced bases per exon with Phred-scaled quality scores, Q, ≥20. For the targeted exons of NOTCH1, a minimum of 64% exons met this threshold with a median of 93%; for NOTCH2, a minimum of 83% of exons met threshold with a median of 96%.

For cSCC cell lines, matching normal lines were not available; therefore, some variants discovered in these samples are likely germ-line in origin. To minimize inclusion of germ-line variants in the final results, all variants exhibiting overlap with markers from the dbSNP build 131 database (52) were filtered out, and three amino acid substitutions also present in other mammalian organisms were excluded.

Site-Directed Mutagenesis.

QuikChange site-directed mutagenesis was performed according to the manufacturer's instructions (Stratagene), and mutagenized cDNA sequences were confirmed by resequencing.

Reporter Gene Assays.

Notch1 reporter gene assays were as described (37, 53). Briefly, to assess ligand-mediated Notch1 activation, pcDNA3 plasmids encoding Notch1–Gal4 DNA binding domain fusion receptors (100 ng) were transfected into U2OS cells along with Gal4–firefly luciferase and human thymidine kinase Renilla luciferase reporter genes. After 24 h, transfected cells were split onto control NIH 3T3 cell feeders or feeders expressing the Notch ligand Jagged2 in the presence or absence of the gamma-secretase inhibitor compound E (1 μM). After an additional 24 h, cells were harvested, and dual luciferase assays were carried out. Results were normalized to the internal Renilla luciferase control and expressed relative to empty vector control, which was arbitrarily set to a value of 1. To assess effects of mutations on ligand-independent Notch1 activity, U2OS cells were cotransfected with the vector pcDNA3 encoding ΔEGFΔLNR, a form of Notch1 lacking the EGF and LNR repeats that is subject to constitutive ADAM metalloprotease and gamma-secretase cleavages, along with an artificial RBPJ–firefly luciferase and the internal control Renilla luciferase reporter genes, as described (54). Normalized luciferase activities were measured 48 h after transfection as described above.

Electrophoretic Mobility Shift Assays.

Recombinant RBPJ and ICN1 polypeptides were expressed and purified to homogeneity from Escherichia coli as described (55). Electrophoretic mobility shift assays were performed in nondenaturing polyacrylamide gels as described (55).