Glioblastoma (n=39), colorectal carcinoma (n=39), and head and neck squamous cell carcinoma (HNSCC, n=60) samples were obtained at the time of surgery at Memorial Sloan-Kettering Cancer Center and University of California, Los Angeles, and snap frozen in liquid nitrogen. Matched normal tissue or peripheral blood was also obtained for each patient. All patients provided informed consent as part of institutional review board-approved protocols at MSKCC and UCLA. After pathologist confirmation of histology, source DNA was extracted. The glioblastoma patients had not received prior treatment with temozolomide82,83
. Samples were fingerprinting with a 44 single nucleotide polymorphism panel on the Sequenom platform to confirm that tumor and normal DNA were correctly matched84
Copy number analysis, expression analysis, and bioinformatics
Copy number data were analyzed using array CGH data from the Tumorscape dataset39
. Copy number alterations were assigned using GISTIC, significance determined with false discovery rate-adjusted q values85
, and data manually visualized with the Integrative Genomics Viewer.
FAT1 copy number was assessed in frozen GBM samples included in mutational analysis, and an additional 3 samples for which there was insufficient material for sequencing. Copy number was determined using genomic quantitative PCR (Taqman, Applied Biosystems) in triplicate. Pre-validated primers for FAT1
were obtained from Applied Biosystems (accession numbers HS00869981_cn, HS00703603_cn, HS01357303_cn). Reference human genomic DNA (Roche Applied Science, #11691112001) was used as reference and RNase P (Life Technologies, #4403328) as a diploid control86
. Inferred copy number of <0.3 was considered a homozygous deletion.
Gene expression analysis in cell lines was performed with the Human Genome U133A 2.0 microarray (Affymetrix). CEL files were imported into Partek Genomics Suite software (Partek, Inc) and normalized using RMA quartile normalization and log probe summarization. Differentially expressed genes were identified by assembling a list of genes that were differentially expressed across all 3 cell lines, comparing scrambled siRNA with each of two FAT1 siRNAs, at p<.05 (analysis of variance). This generated a list of 1539 differentially expressed genes, corresponding to 2031 probes, after excluding six probes (0.3%) with inconsistent directionality between the 2 siRNAs (Supplementary Table 5
). Enriched pathways were identified in the differentially expressed genes, in Ingenuity Pathway Analysis (Ingenuity Systems) and the Biocarta, KEGG and Reactome modules87
. Ingenuity incorporates directional fold change, and a threshold of log2
fold change >1.2 was incorporated. For Biocarta, KEGG and Reactome analyses, the conservative EASE score, a modified Fisher exact test, was used for statistical analyses. The EASE score conservatively estimates the upper bound of probability using a delete-1 observation jackknife, and is therefore higher than standard p values.88
Gene expression data in 404 GBM and 590 ovarian cancer samples were obtained from The Cancer Genome Atlas data portal 5,60
. Data were imported into Partek, RMA-normalized and log-transformed. Samples were categorized dichotomously as low or normal/high FAT1 expressors based on probe ID 201579_at(FAT1), with low expressors defined as the lowest normalized quartile. ANOVA followed by false discovery correction was used to identify genes that were differentially expressed (FDR q<0.05; absolute log2
fold change > 1.2) between groups. FAT1, as the classifier, was not included in the gene list. Differentially expressed genes were imported into Ingenuity Pathway Analysis to assess over-representation of functional categories.
Clinical and microarray data in the 590 ovarian cancer samples, and in 297 glioma samples in the NCI Rembrandt dataset were used to identify low and normal/high FAT1 expressing tumors, with matched clinical outcome data. Low FAT1 expressors were defined as samples in the lowest normalized expression quartile. Cox multivariable regression analysis was performed to compare survival of patients with low FAT1 expressing tumors, compared to other patients, controlling for other clinicopathologic variables significant on univariate analysis (histology, performance status and ethnicity in glioma patients). Survival data were expressed as the hazard ratio for death in the low FAT1 expressing group. Mutation and copy number data were available for the ovarian tumors, allowing survival analysis of tumors with FAT1 mutation or homozygous deletion (GISTIC score of -2). These analysis was performed using IBM SPSS 19 and the MSKCC Computational Biology Cancer Genomics Portal.
The functional consequences of FAT1 mutations were predicted using Polyphen-2 version 2.2.2, an algorithm that uses protein sequence data, structural data, and multiple alignments of vertebrate genomes as detailed in Supplementary Table 341
PCR amplification and sequencing
Standard Sanger methodology was used for sequencing. Exonic regions for the genes (NCBI Human Genome Build 36.1) were broken into amplicons of maximum 1000bp, and specific primers with M13 tails were designed using Primer3 (Supplementary Table 9
). PCR reactions were carried out in 384-well plates in a Duncan DT24 thermal cycler with 10 ng of whole-genome amplified DNA (REPLIg, Qiagen) as a template, using a touchdown PCR protocol with KAPA Fast HotStart (Kapa Biosystems). The touchdown PCR method consisted of: 95 °C for 5 min; 3 cycles 95 °C for 30 s, 64 °C for 15 s, 72 °C for 30 s; 3 cycles 95 °C for 30 s, 62 °C for 15 s, 72 °C for 30 s; 3 cycles 95 °C for 30 s, 60 °C for 15 s, 72 °C for 30 s; 37 cycles 95 °C for 30 s, 58 °C for 15 s, 72 °C for 30 s; 70 °C for 5 min. Templates were purified using AMPure, and sequenced bidirectionally with M13 forward/reverse primers and the Big Dye Terminator Kit v.3.1 (Applied Biosystems) at Agencourt Biosciences. Dye terminators were removed using CleanSEQ (Agencourt Biosciences), and sequence reactions were run on ABI PRISM 3730xl (Applied Biosystems).
Passing reads were assembled against the gene reference sequence, using command line Consed 16.089
. Assemblies were passed to Polyphred 6.02b90
, which generated a list of putative candidate mutations, and to Polyscan 3.091
, which generated a second list of putative mutations. Outputs were merged, mutation calls normalized to ‘+ ’ genomic coordinates and annotated using the genomic mutation consequence calculator92
. A Postgres database was used to annotate each mutation call (assembly position, coverage and methods supporting mutation call). To reduce false positives, only point mutations supported by at least one bidirectional read pair and at least one sample mutation called by Polyphred were considered, and only the putative mutations annotated as nonsynonymous, within 11bp of an exon boundary, or with conservation score >0.699, were included. Single nucleotide polymorphisms were identified and filtered out using dbSNP (NCBI) and referencing sequencing data from matched normal DNA. All putative mutations were manually reviewed, and indels were included in the candidate list if found to hit an exon. All putative mutations were confirmed by a second PCR and sequencing reaction in tumor and matching normal DNA, to confirm all mutations were somatic.
Cloning and site-directed mutagenesis
Expression of FAT1 was accomplished by cloning a gene comprised of the N-terminus, first 2 cadherin repeats, all 5 EGF repeats, the transmembrane region and the cytoplasmic tail, into the vector pcDNA 3.1. A Flag tag was added to the C-terminus. This was named FAT1_Trunc. Mutations identified in cancer were engineered into the constructs using QuikChange II XL (Stratagene). We created a non-adhesive FAT1 chimeric construct, fusing the IL2 receptor extracellular and transmembrane regions to the intracytoplasmic tail of FAT1, annotating this IL2R-FAT1_IC. IL2R protein was cloned as an additional control. Constructs were verified by Sanger sequencing.
Cell culture, soft agar assay, growth curve, colony formation assay
Cell lines were obtained from American Type Culture Collection and cultured using the recommended media (Invitrogen) + 10% FBS (Invitrogen) and penicillin/streptomycin at 37 °C in 5% CO2. Specifically, Dulbecco Modified Eagle Medium was used to culture SNB19, HS683, U87, U251, IHA, and 293T; Roswell Park Memorial Institute 1640 medium, SF295; F-12 with proline, Chinese Hamster Ovary cells. Cancer cell line DNA was extracted using the Qiagen Gentra Puregene Cell Kit.
Functional assays were performed in cell lines as specified. Assays involving overexpression of FAT1 plasmids, were performed in cell lines with low/intermediate endogenous FAT1 expression: colony formation assays were performed in SF295, SNB19 and U87 glioma cells; Subsequent assays of cell growth were performed in SF295, SNB19 and HS683 cell lines, due to more efficient plasmid transfection in HS683. FAT1 knockdown experiments were performed in cell lines with intermediate/high endogenous FAT1 expression: U87, HS683, SF295, and IHA. Experiments requiring co-transfection of 2 or more plasmids and/or siRNAs, were performed in cell lines with high transfection efficiency. Specifically, SF295 cells were chosen for experiments involving synchronous knockdown of endogenous FAT1, and expression of FAT1_Trunc, due to high transfection efficiency and intermediate levels of FAT1 expression. 293T cells were chosen for luciferase reporter assays involving co-transfection of FAT1, β-catenin, and reporter plasmids, because of transfection efficiency and low baseline level of catenin-related transcription. Chinese Hamster Ovary cells were chosen for co-transfection of FAT1 and β-catenin plasmids, because of transfection efficiency in similar co-transfection experiments with other cadherin proteins32
. U251 cells were chosen for assays involving co-transfection of siRNAs targeting FAT1 and β-catenin; and U251 and SF295, for luciferase reporter assays requiring co-transfection of plasmids and siRNAs, because of transfection efficiency and intermediate/high endogenous FAT1 expression. Transfection was performed using Lipofectamine (Invitrogen) and media changed at 6 hours. Stable clones were selected using G418. Cells in soft agarose assays were quantified and measured using ImageJ software (Research Services Branch, National Institutes of Health). Growth curve assays were performed in triplicate, and quantified using the Vi-Cell XR Cell Viability Analyzer (Beckman Coulter), or in real time, in quadruplicate, with the xCELLigence System (Roche Applied Science), which detects as few as 100 cells/well. xCELLigence plates were seeded with 5000 and 10000 cells per well, and growth reported as Cell Index, a dimensionless, relative measure of impedance reflecting viable, adherent cells, with a consistent, logarithmic relationship to cell number.
For the microarray experiment, FAT1 was knocked down using 2 siRNAs (see below, “siRNA knockdown of FAT1”) in duplicate, in IHA, U87 and U251 cells, and RNA extracted after 48 hours with Trizol (Invitrogen). RNA quality was determined with the Agilent 2100 Bioanalyzer.
Immunoprecipitation and immunohistochemistry
For the endogenous immunoprecipitation assay, cell lysates were incubated with antibodies (40 μl of anti-beta Catenin, BD Transduction Laboratories; 35 ul of anti-FAT1, Sigma; or 3.3 ul mouse IgG as negative control, Invitrogen) and precipitated by using protein A- Sepharose beads that had been blocked with 3% powdered milk. For Flag-tagged transfected plasmids, EZview Red ANTI-FLAG M2 Affinity Gel (Sigma) was used. Beads were washed four times with lysis buffer and then mixed with 2× Laemmli sample buffer.
Immunohistochemistry was performed in IHA and U251 cells. Cells cultured on chamber slides (Thermo Fisher Scientific) were fixed by 3.7% paraformaldehyde (PFA) at room temperature for 10 minutes. Slides were blocked with 1% BSA in PBST (PBS + 0.3% Triton X-100) and then incubated with anti-β-catenin (1:200, mouse IgG, BD Biosciences), anti-FLAG (1:500, Rabbit, Sigma-Aldrich) antibodies overnight at 4°C. After primary antibody incubation, slides were washed with PBST three times following incubation with secondary antibodies (1: 500) at room temperature for 1 hour. After wash, slides were dehydrated and mounted with DAPI-containing Prolong Gold antifade mounting fluid (Invitrogen). Images were acquired on Leica TCS SP2 confocal microscopy, and nuclear staining pixel intensity measured in ImageJ.
SiRNA knockdown and plasmid transfection
siRNAs were obtained from Qiagen and Ambion. β-catenin siRNAs were obtained from Santa Cruz Biotechnology and Dharmacon (Supplementary Table 9
). siRNAs were transfected in antibiotic-free media using Lipofectamine RNAiMAX (Invitrogen), media changed at 18-24 hours, and cells harvested at 48 hours. Cells undergoing both FAT1 knockdown and FAT1_Trunc transfection were first transfected with siRNAs, then with plasmids after 18 hours, in serum and antibiotic-free media, and harvested after 48 hours.
Cells were trypsinized, fixed and stained using the standard propidium iodide method 48 h after transfection. Cell cycle analysis was performed on stained cells using a MoFlo cell sorter (Cytomation).
β-catenin-mediated transcription (TOPFLASH) assay
For FAT1 expression, 293T cells were transfected using FuGene (Roche) with β-catenin, FAT constructs, or control expression constructs, the β-galactosidasegene, along with either TCF wildtype (Topflash) or mutated control (Fopflash) luciferase reporter plasmids. For FAT1 knockdown, siRNAs were transfected along with the β-galactosidase gene and Topflash or Fopflash plasmids, in IHA, U251 and SF295 cells, which express endogenous FAT1. Cells were harvested in luciferase assay buffer. Luciferase and β-galactosidase activities were measured (Luciferase assay system, Promega; Aurora Gal-XE chemiluminescent β-Galactosidase reporter, MP Biomedicals) on a Microplate luminometer (Turner Biosystems). Data were normalized by sample-specific β-galactosidase activity, and expressed as Topflash/Fopflash ratios. Experiments were performed in quintupilicate.
Mouse xenograft studies and immunohistochemistry
1 × 106 stably FAT1-transfected SNB19 cells suspended in 50% Matrigel were injected into the flanks of severe combined immunodeficiency mice. Growth was measured with calipers. 26 mice were injected and 8-16 tumors were assessed for each of 5 conditions (empty pCDNA 3.1 vector, FAT1_Trunc, and mutants FAT1 Pro4309Ala, FAT1 Ala4419Ser, FAT1 Thr4511Ile). Mice were sacrificed and tumors harvested at 19 weeks. Immunohistochemistry for Ki-67 was performed (Vector Laboratories #VP-K451). Photomicrographs were taken with a Nikon Digital Sight DS-Fi1 camera on a Nikon Eclipse TE200-E microscope.
Fluorescence-based cell-cell adhesion assay
U87 and U251 cells, which express endogenous FAT1, were transfected with FAT1 siRNA as described above. SNB19 cells, which do not express endogenous FAT1, were transfected with pcDNA 3.1 empty vector, FAT1_Trunc, IL2R, and IL2R-FAT1_IC as described above. Forty-eight hours post transfection, 1 × 1010 cells/mL were resuspended in serum-free media + 5uM calcein AM and incubated at 37°C for 30 minutes. Cells were washed twice with serum-free media and 1 × 105 cells were added to microplate wells containing confluent (unlabeled) cells. Calcein-labeled cells were allowed to adhere for 45 minutes at 37°C. Nonadherent calcein-labeled cells were washed away with media, and PBS added to each well. Fluorescence was measured at an absorbance of 494 nm and emission of 517 nm using a SpectraMax M5 Multilabel Microplate Reader (Molecular Devices). Images were taken using a Nikkon Eclipse TE2000E microscope (Nikon), NIS Elements AR 3.2 software and Photometrics CoolSnap HQ2 camera.
Antibodies used included FLAG (Sigma #F7425), FAT1 (Sigma #hpa023882), β-catenin (BD #610154 and Cell Signaling #9587s), actin (Sigma #A2066), c-myc (Cell Signaling #9402s), cyclin D (Cell Signaling #2922), TCF8 (Cell Signaling #3396s), ITF2 (Cell Signaling #2569), Claudin (Cell Signaling #4933), Id2 (Santa Cruz #sc-489), EGFR D38B1 (Cell Signaling #8839), GLAST (Miltenyi Biotec #130-095-821), GFAP (BD Biosciences #561483), and Ki-67 (Vector #VP-K451).
Two-tailed t-test, one-way ANOVA with posthoc Tukey comparisons, chi-squared, Wilcoxon, log-rank and Cox regression analyses were performed in GraphPad Prism or SPSS 19 software, with a priori level of alpha <.05.