The contribution of somatic mutations to metastasis of colorectal cancers is currently unknown. To find mutations involved in the colorectal cancer metastatic process, we performed deep mutational analysis of 676 genes in 107 stages II to IV primary colorectal cancer, of which half had metastasized. The mutation prevalence in the ephrin (EPH) family of tyrosine kinase receptors was 10-fold higher in primary tumors of metastatic colorectal than in nonmetastatic cases and preferentially occurred in stage III and IV tumors. Mutational analyses in situ confirmed expression of mutant EPH receptors. To enable functional studies of EPHB1 mutations, we demonstrated that DLD-1 colorectal cancer cells expressing EPHB1 form aggregates upon coculture with ephrin B1 expressing cells. When mutations in the fibronectin type III and kinase domains of EPHB1 were compared with wild-type EPHB1 in DLD-1 colorectal cancer cells, they decreased ephrin B1–induced compartmentalization. These observations provide a mechanistic link between EPHB receptor mutations and metastasis in colorectal cancer. Cancer Res; 77(7); 1730–40. ©2017 AACR.

Colorectal cancer develops due to a series of well-characterized mutations that arise in a particular order and affect both oncogenes and tumor suppressor genes (TSG; ref. 1). The progression of colorectal cancer has been extensively studied and characterized on a genetic level, due to the availability of biopsies from various stages in disease progression, that is, from benign polyps to advanced carcinomas (2). However, while much is known about frequent mutations causing colorectal cancer, the genetic basis of metastasis is unclear (3). Stratifying colorectal cancer patients into those at a higher risk of developing metastatic disease is of great clinical importance, as the decision to give adjuvant therapy following resection of the primary tumor is difficult (4). Many patients who may never develop metastatic disease receive chemotherapy resulting in unnecessary side effects, and a number of patients who would benefit from elimination of micrometastases following radical resection remain unidentified. Discovery of mutations that can predict development of metastatic disease would help to reduce the number of patients who are overtreated, and could help to increase the overall survival for higher risk patients.

Although several features of cancer genomes have been associated with the dissemination of colorectal cancer, specific gene mutations related to metastatic processes have yet to be confirmed. For example, loss of 1p36 has consistently been associated with metastatic disease (5, 6) and mutations in the tumor suppressor FBXW7 did not occur with distant metastases (7, 8). If metastasis-causing mutations arise late in tumor development, they may be subclonal and therefore present at a low frequency in the primary tumor (9). For this reason, and as the vast majority of genetic events known to contribute to colorectal cancer development are base-level somatic mutations, we applied targeted deep sequencing to determine the mutational spectrum of pathways and systems associated with colorectal cancer development. Our aim was to analyze tumor samples from metastatic and nonmetastatic patients in order to identify less frequently mutated candidate cancer genes with high sensitivity to gain insight into the mutational spectrum of these tumors and discover novel mechanisms of metastasis.

Study design

We aimed to collect 20 to 25 cases each of stages II and III colorectal cancers with and without distant metastases, as well as 25 stage IV cases (6, 10). Fresh frozen tissue and whole blood samples from 112 colorectal cancer patients [224 matched tumor/normal (T/N) paired samples] were collected (Table 1). Sequencing libraries prepared from gDNA extracted from all patients (tumor and matched normal tissue or blood) was sequenced on an Illumina HiSeq platform and a somatic mutation analysis was performed. All data were included if the T/N pairs were correctly matched, a read depth >250-fold was obtained for >90% of the sequencing region of interest and the tumor was >40%.

Table 1.

Summary of patients and sequencing

Instability phenotypeCIN (n = 83)MSI (n = 24)
Distant metastasesNoYes, during follow-upYes, at diagnosisNoYes, during follow-upYes, at diagnosis
Stage II III II III IV II III II III IV 
Cases 15 16 20 25 10 
Postoperative adjuvant therapy 11 — — 
Median no. of mutations per sample (range) 10 (2–13) 9.5 (6–16) 9 (3–17) 9 (1–36) 8 (3–33) 83 (7–122) 69.5 (10–161) 81.5 (61–102) 87 (82–92) 128.5 (87–170) 
Point mutations 111 146 55 181 221 330 530 99 112 169 
InDels 21 12 10 25 30 201 228 64 62 88 
Instability phenotypeCIN (n = 83)MSI (n = 24)
Distant metastasesNoYes, during follow-upYes, at diagnosisNoYes, during follow-upYes, at diagnosis
Stage II III II III IV II III II III IV 
Cases 15 16 20 25 10 
Postoperative adjuvant therapy 11 — — 
Median no. of mutations per sample (range) 10 (2–13) 9.5 (6–16) 9 (3–17) 9 (1–36) 8 (3–33) 83 (7–122) 69.5 (10–161) 81.5 (61–102) 87 (82–92) 128.5 (87–170) 
Point mutations 111 146 55 181 221 330 530 99 112 169 
InDels 21 12 10 25 30 201 228 64 62 88 

NOTE: The protein coding regions of 676 genes in 107 T/N pairs, representing 100 colon cancer and 7 rectal cancer cases, were enriched and sequenced by Illumina sequencing to >1,000-fold average sequence depth followed by mutational analysis. Tumors were considered MSI if ≥1 of five Bethesda markers showed instability.

Patient samples

DNA was extracted from 196 frozen tissue samples (3 × 10-μm-thick sections) on a liquid handling workstation (Tecan Evo 150 MCA LiHa RoMa; ref. 11). DNA from 28 EDTA whole blood samples was extracted using a QIAamp DNA Blood Midi Kit (Qiagen). DNA was quantified using the Qubit HS dsDNA Kit (Invitrogen by Life Technologies). For the validation cohort of metastatic colorectal cancer samples used for BMPR2 mutation detection, DNA from 19 microsatellite unstable (MSI) FFPE tumor sections was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen).

Tumor tissue purity, determination of genomic stability, and T/N matching

Affymetrix SNP 6.0 microarrays were used to assess tumor purity and to confirm that samples had ≥40% TCC as initially assessed histologically. Microsatellite instability status was determined using MSI Analysis System, version 1.2 (ProMega) with 6 ng genomic DNA and analysis of five mononucleotide repeat Bethesda markers (BAT25, BAT26, NR-21, NR-24, and MONO-27) on a 3130xl genetic analyzer (Applied Biosystems). T/N pair matching was performed using an MLGA based genotyping method described in ref. 12.

Target enrichment by HaloPlex and Illumina sequencing

Genomic DNA (225 ng) from tumor and normal tissue or blood was used in HaloPlex target enrichment (Agilent) for the enrichment of the coding regions of 676 genes. The quality and molarity of the sequencing libraries was assessed on a Bioanalyzer instrument using a High Sensitivity DNA Kit (Agilent). The enriched and barcoded targets were then deep sequenced on a next-generation sequencing HiSeq platform (Illumina). The fraction coverage of the region of interest was required to be greater than 90% and the mean read depth >600-fold. Samples not passing these criteria on the first attempt were resequenced and sequencing data from both runs was merged. Samples not exceeding >600-fold mean read depth thereafter were accepted if the mean read depth was greater than 250-fold (three T/N pairs).

Statistical analyses

Illumina sequencing adaptors were removed by cutAdapt version 0.9.5 (13) and the trimmed reads were subsequently aligned to the reference genome (hg19, March 2009 assembly) using MosaikAligner version 2.1.33 allowing for a maximum of 5% mismatches in a read. A software for somatic mutation analysis in deep sequence datasets, ConfIdent (Adlerteg and colleagues; manuscript in preparation), was used for somatic mutation detection. We assessed which genes were significantly mutated in CIN and MSI tumors using a combination of MutSigCV (14), number of nonsynonymous mutations per Mb and the nonsynonymous to synonymous ratio (NS: S). We used the following cut-offs for CIN tumors: Q-value of <0.1 for MutSigCV output, >1 mutation per Mb, and >2:1 NS: S ratio. Differences in the average number of mutations and in mutation prevalence between metastatic and nonmetastatic groups were assessed using Welch t test and Fisher exact test. The Bonferroni method was used to correct for multiple comparisons.

In situ mutation analysis with padlock probes

Fresh frozen 4 μm sections of colon tumors with EPH receptor mutations were mounted on Superfrost Plus slides (ThermoFisher Scientific). Because of limited sample availability, a total of 15 EPH receptor mutations in 12 tumors were assessed by in situ mutation detection (two tumor sections contained two separate sections, called 1 and 2). Probes specific for EPH receptor mutations were designed with one general and one specific detection oligo sequence (Supplementary Table S1). Probe performance was first tested in T47D and U251 cell lines with known ephrin receptor mRNA expression levels (Supplementary Table S2). The in situ detection method was performed as outlined in Grundberg and colleagues (15). Briefly, tissue sections and cells were fixed in 3.7% PFA for 45 or 20 minutes, respectively. Next, mRNA was reverse transcribed using specific LNA primers (Supplementary Table S3) amplifying wild-type and mutated regions of the tumor-specific EPH receptor mutations. Single-stranded cDNA was created through Rnase H cleavage and padlock probes were hybridized and ligated, followed by rolling circle amplification. Rolling circle products (RCP) were identified through fluorophore-labeled detection oligos (Supplementary Table S4) and the nuclei was stained with 4',6-diamidino-2-phenylindole (DAPI). The tissues and cells were then imaged with an automated Zeiss Axioplan II epifluorescence microscope (Zeiss) using a Z-stack of 0.49 μm × 6 and a tile overlap of 10%. Images were orthogonally processed and tiles were stitched together using ZEN software (Zeiss). The colon sections contained features that were highly autofluorescent; therefore, each RCP was dual-labeled (one general stain and one specific stain) to minimize false-positive signals. Moreover, for each RCP all fluorophore intensities were measured and a quality score was applied to call the correct signal (16). A quality score from 0.5 to 0.8 was used depending on the autofluorescence in each tumor section. Images were analyzed in Cellprofiler v.2.1.1 calling ImageJ plugins (Broad Institute, MA) and specific signals were called using a Matlab script (v.8.5.1, Mathworks). All tumor sections were then stained for hematoxylin and eosin (Sigma), imaged and aligned to the DAPI (nuclei) stain of the fluorescent images.

Cell lines and cell culture

Parental DLD-1 (CCL-221) cells were purchased from ATCC, in 2010. All the DLD-1 cell lines used in this study were authenticated by short tandem repeat (STR) profiling from the ATCC cell line authentication service in October 2016. DLD-1 cells ectopically expressed EPHB1-GFP and its four mutant versions were generated by lentiviral transductions. DLD-1 cells ectopically expressing GFP, RFP, EPHB2-GFP, and ephrin B1-RFP were from ref. 17. All cell lines were maintained in DMEM (Invitrogen) medium supplemented with 10% FBS and 1% penicillin–streptomycin (Invitrogen) at 37°C in 5% CO2.

Generation of stable ectopic expressing cell lines of wild-type and mutant EPHB1

Lentiviral particles were purchased from Labomics. The day before transduction, 50,000 cells were plated in each well of a 24-well plate. Viruses were diluted in 250 μL of normal growth medium with 7.5 mg/mL Sequa-Brene per well. The plating medium was removed and 250 μL of diluted virus was added to each well. After 24 hours of incubation at 37°C, virus containing media were replaced with fresh medium. After 48 hours of incubation, transduced cells were selected with Puromycin (1 μg/mL) for two to four passages to remove any Puromycin-resistant cells in the cell pool. Expression levels of EPHB1 and its four different mutated transcripts were determined by qPCR.

In vitro compartmentalization experiments

Compartmentalization experiments were performed as described in ref. 17. Briefly, DLD-1 cells expressing GFP and RFP were mixed in suspension at a 1:3 ratio and plated at a density of 130,000 cells/cm2 on coverslips coated with 2 mg/cm2 laminin and incubated at 37°C in 5% CO2. Culture medium was changed every 24 hours. After 48 hours, the coverslips were fixed in 4% paraformaldehyde and mounted in Fluoromount G with DAPI (SouthernBiotech). This experiment was performed twice.

Confocal image analysis and quantitation of GFP clusters by ImageJ software

Slides from compartmentalization experiments were subjected to confocal (LSM 700) image analysis. Images were acquired from five random fields with 20× objective and two confocal planes in z-axis from two experiments. Cell sorting was quantified by counting the number of cells present in each GFP-positive cluster of approximately 10 representative fields at two different confocal planes in the z-axis and from two experimental repeats by creating an ImageJ macro. Images throughout the paper show the basal plane.

Ethical approval

This study was approved by the Regional Ethical Review Board of Uppsala (2007/116).

Biosafety declaration

The Swedish work environment authority approved the work with genetically modified and replication-deficient lentiviral particles (Arbetsmiljöverket ID 202100-2932 v72). All the experiments with GMO lentiviral particles were conducted under Biosafety Level 2.

Samples and mutation frequency

Tumor and normal tissues from 112 patients exhibiting stages II, III, and IV colorectal cancer, of which approximately half (58 patients) had metastasized either at diagnosis or during follow-up were analyzed (Table 1). Five T/N pairs were excluded due to contamination of the normal sample with tumor tissue or a failed sequencing library preparation. The coding regions and adjacent splice sites of 676 genes (Supplementary Table S5) were enriched in the remaining samples (Supplementary Table S6) using HaloPlex (Agilent) and sequenced. Genes were chosen on a pathway-oriented basis, including members of canonical colorectal cancer pathways such as Wnt, Ras-MAPK, PI3K, p53, and TGFβ as well as families and processes with a putative role in colorectal cancer. The mean read depth in regions of interest was 1,063-fold (range 256–3516) in tumor samples and 1,103-fold (range 258–14,956) in the normal samples. A conservative mutational analysis tool (ConfIdent, see Supplementary Methods) was applied to call somatic mutations at base positions covered by ≥30 reads in both tumor and patient-matched normal sample. We identified 3,392 somatic mutations, of which 696 were synonymous single nucleotide variations (SNV), 1,941 nonsynonymous SNVs, 717 insertion/deletion polymorphisms (InDels), and 38 splice site mutations in the genes of interest (Supplementary Table S7). The chromosomally unstable (CIN) tumors had an average of 9.9 nonsynonymous somatic mutations (SNV, InDels, and splicing mutations) per patient, whereas the MSI tumors had 75.4 such mutations in the 676 genes of interest (Fig. 1A; Supplementary Table S7). We then assessed the genes with the highest nonsynonymous mutation density in both CIN and MSI tumors to confirm expected mutation patterns. In CIN tumors, these were APC, TP53, KRAS, SMAD4, PIK3CA, and BRAF. In contrast, the genes with the highest nonsynonymous mutation density in MSI tumors were ACVR2A, SETD1B, TGFBR2, BMPR2, BRAF, and TCF7L2 (Fig. 1B and C). The recurrent frameshift InDels in SETD1B and BMPR2 were validated by Sanger sequencing (Supplementary Materials and Methods). Using a separate cohort of 19 stage IV MSI-high colorectal cancers, we sequenced the region of BMPR2 containing this mutation by Sanger sequencing and found that 31% (6/19) of tumors had this mutation. Using a luciferase reporter assay to measure BMP pathway activity using 293T cell lines overexpressing either the mutant or wild-type BMPR2 protein, we observed a downregulation of activity when overexpressing the mutant form of BMPR2, although this difference was not significant (Supplementary Methods and Supplementary Fig. S1).

Figure 1.

Deep targeted mutational analyses of metastatic and nonmetastatic primary colorectal cancers. A, Number of mutations per sample in 676 genes in colorectal cancer pathways and systems. Red bars, MSI-high tumors; blue bars, MSI-low tumors; empty bars, CIN tumors. Black bars, patients metastatic at diagnosis; orange bars, patients that developed metastases; empty bars, patients that did not develop metastases. B and C, Genes with the highest mutation prevalence in CIN tumors (B) and MSI tumors (C). The percentage of tumors with mutation is indicated. Black bars, missense mutations; blue bars, frameshift InDels; red bars, nonsense mutations; grey bars, splicing mutations; empty bars, silent mutations.

Figure 1.

Deep targeted mutational analyses of metastatic and nonmetastatic primary colorectal cancers. A, Number of mutations per sample in 676 genes in colorectal cancer pathways and systems. Red bars, MSI-high tumors; blue bars, MSI-low tumors; empty bars, CIN tumors. Black bars, patients metastatic at diagnosis; orange bars, patients that developed metastases; empty bars, patients that did not develop metastases. B and C, Genes with the highest mutation prevalence in CIN tumors (B) and MSI tumors (C). The percentage of tumors with mutation is indicated. Black bars, missense mutations; blue bars, frameshift InDels; red bars, nonsense mutations; grey bars, splicing mutations; empty bars, silent mutations.

Close modal

Enrichment of Eph receptor mutations in metastatic tumors

The aim of the study was to investigate if alterations in a particular pathway or gene group could predict the metastasis of CIN tumors. The average number of mutations per sample did not differ between the metastatic and nonmetastatic groups in CIN tumors (10 vs. 9.4, P = 0.71, Welch two-sample t test). We did not find any significant differences in mutation prevalence between the metastatic and nonmetastatic samples in any of the pathways or large gene families included in the study (Supplementary Table S8). To assess enrichment of biological themes in the metastatic samples compared with the nonmetastatic, we compared the 50 genes with the highest number of nonsynonymous mutations per Mb in each group. Genes found only in the metastatic group were analyzed using the DAVID functional annotation tool (18), revealing an enrichment of the Ephrin receptor family (Supplementary Methods and Supplementary Table S9). The mutation prevalence of EPH receptor genes in CIN patients with metastases (18/52 of stage IV and stage II and III tumors that later developed distant metastases) versus those that did not develop metastases (2/31 patients) was significantly different (P = 0.0032, Fisher exact test; Supplementary Table S10). There was no difference in sequencing read depth in any of the EPH receptor genes between the metastatic and nonmetastatic groups (Supplementary Table S11) and 21/24 mutations were independently validated by Sanger sequencing (Supplementary Table S10). The three mutations that could not be validated had a variant allele ratio (VAR) 0.11 to 0.13, which is at the limit of detection for Sanger sequencing. Sixty-seven percent (16/24) of mutations were located in a functional domain, with six mutations in a kinase domain. Six Eph receptor mutations had a low VAR of 0.1 to 0.15 despite being found in tumors with 50% to 75% tumor cell content (TCC). Mutations in EPH receptors seemed to predominantly affect stage III and IV cancers (22/24 mutations). Together, this suggests that Eph mutations in colorectal cancers are late events or part of subclones that become increasingly more prevalent due to selection pressure as the disease progresses. As there is a high level of sequence similarity between the Eph receptors, we used the prediction tool Consurf to assess the degree of conservation at the affected amino acids (19). We found that the affected amino acids with the highest conservation scores (>7) were generally those with the most damaging mutations as predicted by the effect of specific point mutations using PredictProtein (20). Seventy-nine percent (19/24) of these alterations were predicted to have damaging effects using the PolyPhen-2 prediction tool (Supplementary Tables S10 and S12; ref. 21). Of the two mutations in nonmetastatic samples, one was predicted to be benign whereas the other was predicted to have a probably damaging consequence. In total, there were 6/24 (25%) positions in a predicted functional residue, which is greater than the fraction of functional residues found in the EPH receptors (15%), indicating that they are selected for. Randomly selecting 14 genes (the same number of Eph receptor genes) from the panel 100,000 times and comparing the number of significant differences in mutation rate between nonmetastatic and metastatic CIN patients indicated that EPH receptors are involved primarily in the metastatic groups (see Supplementary Methods). EPHB1, EPHA5, and EPHA6 were the 1st, 3rd, and 7th most frequently found in groups of genes showing significant differences. Other genes frequently found in significant groups had a lower mutation density than any of the EPH receptor genes (Supplementary Fig. S2). Comparing the number of mutations in EPH receptor genes in the two groups indicated that these alterations might have clinical relevance in predicting the development of metastatic disease (Fig. 2A). The mutations found were distributed across the coding regions of both EPHA and EPHB-type receptors (Fig 2B and C). Importantly, of the two patients with EPH receptor mutations that did not develop metastatic disease, one had adjuvant therapy after radical resection of the tumor, which may have eradicated metastatic cells remaining after surgery.

Figure 2.

Mutations in protein-coding regions of Ephrin receptor tyrosine kinases are associated with metastasis of colorectal cancer. A, The number of nonsynonymous mutations in Eph receptor genes in the two groups by disease stage is shown. Red bars, metastatic stage IV patients; black bars, metastatic stage III patients; gray bars, metastatic stage II patients; empty bars, stage III patients that did not develop metastases. B and C, Schematic representation of the distributions of mutations in EPHA genes (B) and EPHB genes (C). Symbols on top of the gene ideogram represent mutations in genes found in this study (no border) and in nonhypermutated and CIN samples (black border) in refs. 7, 24, 49, and 50. All mutations are exonic and nonsynonymous.

Figure 2.

Mutations in protein-coding regions of Ephrin receptor tyrosine kinases are associated with metastasis of colorectal cancer. A, The number of nonsynonymous mutations in Eph receptor genes in the two groups by disease stage is shown. Red bars, metastatic stage IV patients; black bars, metastatic stage III patients; gray bars, metastatic stage II patients; empty bars, stage III patients that did not develop metastases. B and C, Schematic representation of the distributions of mutations in EPHA genes (B) and EPHB genes (C). Symbols on top of the gene ideogram represent mutations in genes found in this study (no border) and in nonhypermutated and CIN samples (black border) in refs. 7, 24, 49, and 50. All mutations are exonic and nonsynonymous.

Close modal

Validation of mutant transcript expression in tumors by in situ mutational analyses

EPH receptor mutations in tissue samples from 12 patients were evaluated by in situ mutation detection using padlock probes. This revealed both the expression of these receptors in the tissue as well as validating the mutations present on an RNA transcript level (Supplementary Fig. S3; Table S13). There was no localization of any of the mutations in the tissue; rather signals that were detected were spread throughout the section. For the majority of mutations, the mutant: wild type ratio was greater at the transcript level than at the genomic level, revealing a preferential expression of the mutant transcript. A number of EPH receptors showed a low overall expression level; namely, EPHA5 and EPHA8 (Supplementary Fig. S3A) and EPHA6 (Supplementary Fig. S3H). Mutations in EPHB1 (C268T) and EPHA3 (A2399G) had a lower mutant: wild type ratio at the transcript level compared with the corresponding VAR at the genomic level (Supplementary Table S13), indicating that these variants may be silenced at the transcript level. The mutations found in EPH receptors may have various different roles in disease progression and are therefore seen both up and downregulated in these tumors.

In vitro compartmentalization assay to study EPHB1 mutations

EphB receptor signaling is of particular importance in the colonic crypt, where expression of Eph receptors and ephrin ligands control the correct positioning of cells in the intestine due to repulsive mechanisms (22). We therefore sought to establish a model system with the potential to compare mutant phenotypes across the EPHB receptor family. It is known that Ephrin-B1 activation of EphB2 and EphB3 receptors induces sorting and compartmentalization of colorectal cancer cells in vitro (17). However, this phenotype has not previously been demonstrated for EphB1. We first assessed whether ephrin B1–expressing cells induce in vitro compartmentalization of EPHB1 expressing colorectal cancer cells using the coculture system described in ref. 17 complemented by DLD-1 cell lines engineered to express wild-type or mutant EPHB1 tagged with GFP (Supplementary Fig. S4). Continuous cell contact-mediated EPHB1 and ephrin B1 bidirectional activation (23) was achieved by coculturing these two cell populations (Fig. 3A–E). Intermingling of GFP and RFP populations was virtually absent in EPHB1-GFP and EphB2-GFP cocultures with ephrin B1-RFP (Fig. 3A; Supplementary Fig. S5A). Conversely, GFP and RFP cells were completely mixed and scattered to a similar extent in the case of EPHB1-GFP (Fig. 3F–J), EphB2-GFP, or GFP (Supplementary Fig. S5A) expressing cells cocultured with RFP-labeled control cells. Quantification of cell distribution demonstrated the presence of large homogenous GFP clusters (>50 cells) of EPHB1-GFP (Fig. 3K) and EphB2-GFP (Supplementary Fig. S5A and S5B) expressing cells in the presence of ephrin B1-RFP cells compared with the cell distribution using RFP expressing cell alone. Thus, DLD-1 cells expressing EphB1 display a similar clustering phenotype as EphB2 expressing cells (17) when exposed to ephrin B1. Notably, the cell sorting phenotype of EPHB1-GFP cells was comparatively weaker than EphB2-GFP cells in the presence of ephrin B1-RFP cells (Fig. 3A and K; Supplementary Fig. S5A and S5B).

Figure 3.

EPHB1 activity induces cell sorting and compartmentalization of colorectal cancer cells, and point mutations in the fibronectin and tyrosine kinase domains reduced these effects in vitro. A–J,In vitro compartmentalization assay. Representative confocal images of DLD-1 cells expressing ephrin B1-RFP ligand (A–E) or RFP alone (F–J) cocultured with DLD-1 cells expressing EPHB1-GFP alone (A and F), four different mutated versions of EPHB1, (G251A)-GFP (B and G), (C268T)-GFP (C and H), (C1051T)-GFP (D and I), and (G1882T)-GFP (E and J). Images were captured 48 hours after plating. Arrows point to examples of large, homogeneous GFP+ cell clusters indicative of cell sorting and compartmentalization. K, Quantitative results of the compartmentalization experiments. Cell distribution was quantified by counting the percentage of GFP+ cells forming clusters of different sizes. In cocultures of ephrin B1 ligand with EPHB1 and its two mutated versions, (G251A) and (C268T), a high percentage of GFP+ cells were distributed into large homogeneous clusters (>50), whereas this was significantly reduced in coculture of ephrin B1 ligand with mutated versions (C1051T) and (G1882T) where the majority of GFP+ cells form small groups of fewer than ten cells. The experiments were performed twice and quantitation was based on these two experimental repeats. Error bars, SD (n = 10 random fields). Statistical analysis was performed by Student t test, where **P < 0.01.

Figure 3.

EPHB1 activity induces cell sorting and compartmentalization of colorectal cancer cells, and point mutations in the fibronectin and tyrosine kinase domains reduced these effects in vitro. A–J,In vitro compartmentalization assay. Representative confocal images of DLD-1 cells expressing ephrin B1-RFP ligand (A–E) or RFP alone (F–J) cocultured with DLD-1 cells expressing EPHB1-GFP alone (A and F), four different mutated versions of EPHB1, (G251A)-GFP (B and G), (C268T)-GFP (C and H), (C1051T)-GFP (D and I), and (G1882T)-GFP (E and J). Images were captured 48 hours after plating. Arrows point to examples of large, homogeneous GFP+ cell clusters indicative of cell sorting and compartmentalization. K, Quantitative results of the compartmentalization experiments. Cell distribution was quantified by counting the percentage of GFP+ cells forming clusters of different sizes. In cocultures of ephrin B1 ligand with EPHB1 and its two mutated versions, (G251A) and (C268T), a high percentage of GFP+ cells were distributed into large homogeneous clusters (>50), whereas this was significantly reduced in coculture of ephrin B1 ligand with mutated versions (C1051T) and (G1882T) where the majority of GFP+ cells form small groups of fewer than ten cells. The experiments were performed twice and quantitation was based on these two experimental repeats. Error bars, SD (n = 10 random fields). Statistical analysis was performed by Student t test, where **P < 0.01.

Close modal

Somatic mutations in the fibronectin type III domain and tyrosine kinase domain compromised the compartmentalization of EphB1-expressing cells

We found 7 somatic, nonsynonymous EPHB1 mutations in CIN colorectal cancer cases that developed metastatic disease. Of these, four were located in functional domains of EPHB1 and therefore more likely to be of relevance to protein function (Supplementary Table S10). These four mutations were also predicted to be damaging using PolyPhen-2 (Supplementary Table S12) and in situ mutation detection revealed expression of three of the mutations in the Ephrin binding and tyrosine kinase domains (Fig. 4). We therefore engineered DLD-1 colorectal cancer cell lines each expressing one of the four different EPHB1 mutants G251A, C268T, C1051T, and G1882T tagged with GFP and performed compartmentalization assays by coculturing with RFP-tagged ephrin B1-RFP ligand expressing cells (Fig 3B–E) or RFP alone (Fig 3G–J; ref. 17). The presence of the desired mutation was confirmed by Sanger sequencing of the constructs (Supplementary Table S14; Supplementary Fig. S6A–S6D). There were no deleterious effects on the growth of DLD-1 cells ectopically expressing mutated EPHB1 (Supplementary Fig. S6E) or ephrin B1-RFP (Supplementary Fig. S6F). Cells expressing EphB1 C1051T and G1882T showed abrogated compartmentalization capacity when cocultured with ephrin B1 ligand-expressing cells (Fig. 3D, E, and K), whereas the mutations G251A and C268T formed large clusters (>50 cells) to a similar extent as wild type EPHB1–ephrin B1 ligand interaction (Fig. 3B, C, and K). The fraction of GFP+ cells expressing EPHB1 with C1051T and G1882T mutations present in large homogenous clusters of >50 cells was more than two-fold lower (P = 0.0085 and 0.0066, respectively) than the fractions of wild-type EPHB1 and G251A and C268T mutants in such clusters (Fig. 3K). Notably, the number of small clusters (≤10 cells) was more than two-fold higher in these two mutants as compared with normal EPHB1. Taken together, the C1051T and G1882T EPHB1 mutations yield protein products with impaired compartmentalization ability as compared with wild-type EPHB1.

Figure 4.

Schematic representation of EPHB1 including the four nonsynonymous mutations studied in the compartmentalization experiments and in situ detection of wild-type and mutant transcripts in patient samples. The number of signals in the tissue represents the expression level of the transcript in the section.

Figure 4.

Schematic representation of EPHB1 including the four nonsynonymous mutations studied in the compartmentalization experiments and in situ detection of wild-type and mutant transcripts in patient samples. The number of signals in the tissue represents the expression level of the transcript in the section.

Close modal

Somatic mutations in the exomes of colorectal cancers have previously been identified in a limited sample set by Sanger sequencing and hybridization capture enrichment coupled to Illumina sequencing to 20-fold coverage (7, 24). Although the number of samples analyzed limited the first approach, the second was limited in coverage by the sequence depth. Other groups have performed targeted deep sequencing of colorectal cancer with panels of genes implicated in the disease, showing a high degree of concordance between primary and metastatic lesions from the same patient (25). Here, we have performed pathway-oriented targeted deep sequencing for the further characterization of colorectal cancer genomes. In contrast to other targeted sequencing studies, we included genes with both known and putative roles in colorectal cancer development and progression to uncover novel mechanisms for metastatic disease development. We sequenced these genes to approximately 1,000-fold coverage in normal and tumor samples to validate our approach to mutation detection, confirm known patterns of mutations in colorectal cancer and extend the compendium of potential cancer genes in these pathways. The expected frequencies and types of mutations were observed in known colorectal cancer genes such as APC, KRAS, and TP53, confirming the sensitivity and specificity of mutation detection (7). Interestingly, we uncovered recurring mutations in repeat sequences in BMPR2 and SETD1B in MSI tumors. The frameshift mutation N583fs in the A (7) microsatellite repeat sequence of BMPR2 observed in 56% of MSI tumors has been reported in both gastric cancer and colorectal cancer (frequency of 5–13%; refs. 26, 27). Functional studies including those reported here revealed an impaired expression of BMPR2 in MSI colorectal cancers (28) and downregulation of the BMP pathway (Supplementary Fig. S1), indicating a potential role of this mutation in colorectal cancer. In addition, it was recently reported that 35% of Lynch syndrome colorectal cancers had the same mutation seen in this study (29). Similarly, the frameshift InDel seen here in 72% of MSI tumors in SETD1B (H8fs) has been reported as a confirmed somatic mutation in colorectal cancer, albeit at a mutation rate of ∼6% (7). The mutation prevalence of TCF7L2, thought to have a tumor suppressor role in colorectal cancer (30, 31), was also higher than expected in MSI tumors (48% vs. 27% in hypermutated tumors; ref. 7). We hypothesize that the higher mutation rate in microsatellites shown here compared with similar studies is due to improved sensitivity of our mutation calling in these repeat sequences, due to the ability of ConfIdent to filter out artefactual InDels in the corresponding normal sample that are due to enzyme slippage introduced by the sequencing technique.

In this study, we aimed to assess mutational differences between patients that developed metastases and those that did not. It has been proposed that FBXW7 mutations may be protective for the development of metastatic disease (7). Although we did not find a statistically significant correlation between mutations in FBXW7 and the development of metastatic disease (P = 0.23, Fisher exact test), 79% (11 out of 14) of mutations in FBXW7 occurred in patients that did not develop metastases. We next examined mutations in Ephrin receptor tyrosine kinases as we noted an enrichment of this gene family in tumors giving rise to metastasis. A major challenge in understanding Eph receptor mutations observed in cancers is the combinatorial nature and complexity of Ephrin signaling. In the intestine, TCF and β-catenin inversely control the expression of EphB genes, and the expression of EphB genes is required for the stabilization of the correct positioning of epithelial cells along the intestinal crypt (22). Eph receptors have previously been associated with metastatic disease development due to their role in tumor growth, invasiveness, angiogenesis, and metastasis in vivo (32). However, no mutational evidence has yet been presented to explain this association. The Ephrin receptors have a complex role in tumor progression and have been found both up- and downregulated in several cancer types. In colorectal cancer, EphA1-3, EphA8, and EphB4 have upregulated expression, whereas EphA6, EphA7, EphB1, and EphB2 have downregulated expression (33). A tumor suppressor role for Ephrin B receptors has been suggested based on animal models and transcriptional downregulation in human colorectal cancers (34). EphB2, EphB3, and EphB4 were silenced in metastatic intestinal tumors (35–37) and EphB2 expression and silencing of EphB2 was inversely correlated with patient survival in colorectal cancer (35, 38).

Here, EPHB1, EPHA5, and EPHA6 were the most frequently mutated Eph receptors, accounting for 63% of Eph mutations. EPHA5 has been associated with metastases in breast cancer, where promoter hypermethylation results in downregulation of expression (39, 40). Although EphA6 expression was found strongly downregulated in colorectal cancer compared with normal colon, mutations have not previously been reported in CIN tumors (41). Reduced EPHB1 expression in colon cancer is associated with poor differentiation and increased invasive capacity (42). In our in situ mutation detection assay, we demonstrate that the expression of wild type EPHA6 and EPHB1 is generally reduced compared with the mutant transcript, indicating preferential expression of mutant transcripts in a number of samples (Supplementary Fig. S3; Supplementary Table S13). In addition, four nonsynonymous mutations (Supplementary Table S10; Fig. 3) in protein coding functional domains of EPHB1 were assessed using an in vitro compartmentalization assay. To our knowledge, this is the first time the interaction of EPHB1 with ephrin B1 ligand in vitro has been demonstrated. As EphB receptors play an important role in controlling the position of different cell types in the crypt–villus axis of the epithelium (22), we hypothesized that dysregulation of EphB receptor activity by somatic mutations could be linked to metastases by abrogating restrictive forces maintained between cells under normal conditions. It was found that the mutations C1051T and G1882T in the fibronectin and tyrosine kinase domains respectively, abrogated the cell sorting and compartmentalization capacity more than two-fold when compared with wild-type EPHB1 (Fig. 3K). In contrast, mutations in the ephrin-binding domain (G251A and C268T) had no such effect (Fig. 3K). The lack of phenotypes of these mutations can either mean that they are passenger mutations or that the compartmentalization assay is a suboptimal tool to study their phenotype. Interestingly, in situ mutation detection revealed that the mutant allele ratio for C268T was decreased compared with the VAR found by DNA sequencing, indicating that expression of this mutation may be downregulated at a transcript level. The opposite was true for the G251A and G1882T mutations (Supplementary Table S13). The fibronectin domain is responsible for the interactions of many extracellular matrix proteins such as integrins (43, 44) and plays an important role in metastasis and invasion (45). Bidirectional downstream signaling of EPHB1 is transduced by phosphorylation of the tyrosine kinase domain upon binding with ephrin B1 (46, 47). Therefore, mutations in these domains may disrupt normal function of the protein and enhance tumor progression as well as metastatic capacity of cells bearing these mutations by reducing the repulsive interactions between cells. To better understand which aspects of Eph signaling the compartmentalization assay measures and due to the fact that kinase independent functions of Eph receptor signaling were found to be mediated by PI3K (48), we performed the assay with and without kinase inhibitors to EPH kinase, Abl and PI3K (Supplementary Fig. S7). We found that a strong compartmentalization phenotype was dependent on Eph kinase activity but not Abl or PI3K activity, indicating that kinase independent phenotypes may not be captured effectively by this assay and the use of alternative assays could be beneficial to convey the function of these mutations.

The pattern of mutations seen in this study and others (7, 25, 49, 50) suggest a metastasis suppressor role for Eph genes as the mutations are widely distributed over the coding region, both inside and outside functional domains (Fig. 2B and C), and some mutations result in loss of function (51–53). Stimulating Eph forward signaling by activating Eph receptors using soluble ephrin ligands has a tumor suppressive function by promoting contact-dependent growth inhibition and reducing cell motility and invasive capacity (32). Therefore, it is reasonable to hypothesize that mutations abrogating forward signaling through Eph receptors are tumor suppressive. However, the mutations seen here do not appear to have the characteristics of classical tumor suppressors (requiring inactivation of both alleles), perhaps due to a dominant negative effect, heterodimerization effects and/or crosstalk with other signaling pathways (32, 52). In fact, it has been shown that depending on the ratio of mutant:wild type Eph receptor expressed in a cell, inactivating mutations of one Eph receptor can exert a dominant negative effect on a different wild type receptor (54). Interestingly, three patients had mutations in more than one Eph receptor gene, suggesting that in some cases, several of these genes work together to suppress tumorigenesis, and the cumulative effect of several mutations is required to promote metastasis (52). Concurrent expression of >1 Eph receptor mutation is shown in Supplementary Fig. S3 for two of these patients (panels A and J). The finding that some of the mutations found in EPHB1 may contribute to an increased invasive capacity of cancers with Eph receptor mutations is novel, and is potentially of great clinical importance to identify patients who require close monitoring to detect recurrence and to stratify patients that would benefit most from adjuvant treatments.

No potential conflicts of interest were disclosed.

Conception and design: L. Moens, C. Cortina, M. Sundström, J. Botling, E. Batlle, B. Glimelius, M. Nilsson, T. Sjöblom

Development of methodology: S. Kundu, T. Adlerteg, C. Cortina, A. Moustakas, E. Batlle, B. Glimelius, M. Nilsson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Mathot, S. Kundu, J. Svedlund, V. Rendo, C. Bellomo, M. Sundström, P. Micke, J. Botling, A. Moustakas, H. Birgisson, B. Glimelius, M. Nilsson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Mathot, S. Kundu, V. Ljungström, J. Svedlund, L. Moens, T. Adlerteg, E. Falk-Sörqvist, C. Bellomo, M. Mayrhofer, A. Isaksson, A. Moustakas, B. Glimelius, M. Nilsson

Writing, review, and/or revision of the manuscript: L. Mathot, S. Kundu, V. Ljungström, J. Svedlund, T. Adlerteg, E. Falk-Sörqvist, V. Rendo, C. Bellomo, M. Sundström, J. Botling, A. Isaksson, A. Moustakas, H. Birgisson, B. Glimelius, M. Nilsson, T. Sjöblom

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Adlerteg, M. Nilsson

Study supervision: B. Glimelius, M. Nilsson, T. Sjöblom

We thank Uppsala Genome Center, Jeremy Adler, and Simin Tahmasebpoor for expert technical assistance. Imaging was performed with support of the Science for Life Lab BioVis Platform, Uppsala, Sweden.

This study was supported by the research grants awarded to T. Sjöblom from the Swedish Cancer Foundation (2006/2154, 2007/775, and 2012/834), the Uppsala-Umeå Comprehensive Cancer Consortium (U-CAN), the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 601939 (MERIT) and the Swedish Foundation for Strategic Research (F06-0050), and by grants awarded to M Nilsson from VINNOVA (Companion diagnostic initiative) and the Innovative Medicines Initiative (IMI) Joint Undertaking under grant agreement no. 115234 (OncoTrack).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Vogelstein
B
,
Papadopoulos
N
,
Velculescu
VE
,
Zhou
S
,
Diaz
LA
 Jr
,
Kinzler
KW
. 
Cancer genome landscapes
.
Science
2013
;
339
:
1546
58
.
2.
Muto
T
,
Bussey
HJ
,
Morson
BC
. 
The evolution of cancer of the colon and rectum
.
Cancer
1975
;
36
:
2251
70
.
3.
Jones
S
,
Chen
WD
,
Parmigiani
G
,
Diehl
F
,
Beerenwinkel
N
,
Antal
T
, et al
Comparative lesion sequencing provides insights into tumor evolution
.
Proc Natl Acad Sci U S A
2008
;
105
:
4283
8
.
4.
Schmoll
HJ
,
Van Cutsem
E
,
Stein
A
,
Valentini
V
,
Glimelius
B
,
Haustermans
K
, et al
ESMO Consensus Guidelines for management of patients with colon and rectal cancer. a personalized approach to clinical decision making
.
Ann Oncol
2012
;
23
:
2479
516
.
5.
Ogunbiyi
OA
,
Goodfellow
PJ
,
Gagliardi
G
,
Swanson
PE
,
Birnbaum
EH
,
Fleshman
JW
, et al
Prognostic value of chromosome 1p allelic loss in colon cancer
.
Gastroenterology
1997
;
113
:
761
6
.
6.
Mayrhofer
M
,
Kultima
HG
,
Birgisson
H
,
Sundstrom
M
,
Mathot
L
,
Edlund
K
, et al
1p36 deletion is a marker for tumour dissemination in microsatellite stable stage II-III colon cancer
.
BMC Cancer
2014
;
14
:
872
.
7.
The Cancer Genome Atlas Network
. 
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
8.
Tan
Y
,
Sangfelt
O
,
Spruck
C
. 
The Fbxw7/hCdc4 tumor suppressor in human cancer
.
Cancer Lett
2008
;
271
:
1
12
.
9.
Yachida
S
,
Jones
S
,
Bozic
I
,
Antal
T
,
Leary
R
,
Fu
B
, et al
Distant metastasis occurs late during the genetic evolution of pancreatic cancer
.
Nature
2010
;
467
:
1114
7
.
10.
Birgisson
H
,
Edlund
K
,
Wallin
U
,
Pahlman
L
,
Kultima
HG
,
Mayrhofer
M
, et al
Microsatellite instability and mutations in BRAF and KRAS are significant predictors of disseminated disease in colon cancer
.
BMC Cancer
2015
;
15
:
125
.
11.
Mathot
L
,
Wallin
M
,
Sjoblom
T
. 
Automated serial extraction of DNA and RNA from biobanked tissue specimens
.
BMC Biotechnol
2013
;
13
:
66
.
12.
Mathot
L
,
Falk-Sorqvist
E
,
Moens
L
,
Allen
M
,
Sjoblom
T
,
Nilsson
M
. 
Automated genotyping of biobank samples by multiplex amplification of insertion/deletion polymorphisms
.
PLoS One
2012
;
7
:
e52750
.
13.
Martin
M
. 
Cutadapt removes adapter sequences from high-throughput sequencing reads
.
EMBnet J
2011
;
17
:
10
2
.
14.
Lawrence
MS
,
Stojanov
P
,
Polak
P
,
Kryukov
GV
,
Cibulskis
K
,
Sivachenko
A
, et al
Mutational heterogeneity in cancer and the search for new cancer-associated genes
.
Nature
2013
;
499
:
214
8
.
15.
Grundberg
I
,
Kiflemariam
S
,
Mignardi
M
,
Imgenberg-Kreuz
J
,
Edlund
K
,
Micke
P
, et al
In situ mutation detection and visualization of intratumor heterogeneity for cancer research and diagnostics
.
Oncotarget
2013
;
4
:
2407
18
.
16.
Ke
R
,
Mignardi
M
,
Pacureanu
A
,
Svedlund
J
,
Botling
J
,
Wahlby
C
, et al
In situ sequencing for RNA analysis in preserved tissue and cells
.
Nat Methods
2013
;
10
:
857
60
.
17.
Cortina
C
,
Palomo-Ponce
S
,
Iglesias
M
,
Fernández-Masip
JL
,
Vivancos
A
,
Whissell
G
, et al
EphB-ephrin-B interactions suppress colorectal cancer progression by compartmentalizing tumor cells
.
Nat Genet
2007
;
39
:
1376
83
.
18.
Huang
da W
,
Sherman
BT
,
Lempicki
RA
. 
Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists
.
Nucleic Acids Res
2009
;
37
:
1
13
.
19.
Ashkenazy
H
,
Erez
E
,
Martz
E
,
Pupko
T
,
Ben-Tal
N
. 
ConSurf 2010: calculating evolutionary conservation in sequence and structure of proteins and nucleic acids
.
Nucleic Acids Res
2010
;
38
:
W529
33
.
20.
Yachdav
G
,
Kloppmann
E
,
Kajan
L
,
Hecht
M
,
Goldberg
T
,
Hamp
T
, et al
PredictProtein–an open resource for online prediction of protein structural and functional features
.
Nucleic Acids Res
2014
;
42
:
W337
43
.
21.
Adzhubei
IA
,
Schmidt
S
,
Peshkin
L
,
Ramensky
VE
,
Gerasimova
A
,
Bork
P
, et al
A method and server for predicting damaging missense mutations
.
Nat Methods
2010
;
7
:
248
9
.
22.
Batlle
E
,
Henderson
JT
,
Beghtel
H
,
van den Born
MM
,
Sancho
E
,
Huls
G
, et al
Beta-catenin and TCF mediate cell positioning in the intestinal epithelium by controlling the expression of EphB/ephrinB
.
Cell
2002
;
111
:
251
63
.
23.
Pasquale
EB
. 
Eph-ephrin bidirectional signaling in physiology and disease
.
Cell
2008
;
133
:
38
52
.
24.
Sjöblom
T
,
Jones
S
,
Wood
LD
,
Parsons
DW
,
Lin
J
,
Barber
TD
, et al
The consensus coding sequences of human breast and colorectal cancers
.
Science (New York, NY
2006
;
314
:
268
74
.
25.
Brannon
AR
,
Vakiani
E
,
Sylvester
BE
,
Scott
SN
,
McDermott
G
,
Shah
RH
, et al
Comparative sequencing analysis reveals high genomic concordance between matched primary and metastatic colorectal cancer lesions
.
Genome Biol
2014
;
15
:
454
.
26.
Wang
K
,
Kan
J
,
Yuen
ST
,
Shi
ST
,
Chu
KM
,
Law
S
, et al
Exome sequencing identifies frequent mutation of ARID1A in molecular subtypes of gastric cancer
.
Nat Genet
2011
;
43
:
1219
23
.
27.
Seshagiri
S
,
Stawiski
EW
,
Durinck
S
,
Modrusan
Z
,
Storm
EE
,
Conboy
CB
, et al
Recurrent R-spondin fusions in colon cancer
.
Nature
2012
;
488
:
660
4
.
28.
Kodach
LL
,
Wiercinska
E
,
de Miranda
NF
,
Bleuming
SA
,
Musler
AR
,
Peppelenbosch
MP
, et al
The bone morphogenetic protein pathway is inactivated in the majority of sporadic colorectal cancers
.
Gastroenterology
2008
;
134
:
1332
41
.
29.
Pinheiro
M
,
Pinto
C
,
Peixoto
A
,
Veiga
I
,
Lopes
P
,
Henrique
R
, et al
Target gene mutational pattern in Lynch syndrome colorectal carcinomas according to tumour location and germline mutation
.
Br J Cancer
2015
;
113
:
686
92
.
30.
Angus-Hill
ML
,
Elbert
KM
,
Hidalgo
J
,
Capecchi
MR
. 
T-cell factor 4 functions as a tumor suppressor whose disruption modulates colon cell proliferation and tumorigenesis
.
Proc Natl Acad Sci U S A
2011
;
108
:
4914
9
.
31.
Bass
AJ
,
Lawrence
MS
,
Brace
LE
,
Ramos
AH
,
Drier
Y
,
Cibulskis
K
, et al
Genomic sequencing of colorectal adenocarcinomas identifies a recurrent VTI1A-TCF7L2 fusion
.
Nat Genet
2011
;
43
:
964
8
.
32.
Pasquale
EB
. 
Eph receptors and ephrins in cancer: bidirectional signalling and beyond
.
Nat Rev Cancer
2010
;
10
:
165
80
.
33.
Xi
HQ
,
Wu
XS
,
Wei
B
,
Chen
L
. 
Eph receptors and ephrins as targets for cancer therapy
.
J Cell Mol Med
2012
;
16
:
2894
909
.
34.
Clevers
H
,
Batlle
E
. 
EphB/EphrinB receptors and Wnt signaling in colorectal cancer
.
Cancer Res
2006
;
66
:
2
5
.
35.
Jubb
AM
,
Zhong
F
,
Bheddah
S
,
Grabsch
HI
,
Frantz
GD
,
Mueller
W
, et al
EphB2 is a prognostic factor in colorectal cancer
.
Clin Cancer Res
2005
;
11
:
5181
7
.
36.
Batlle
E
,
Bacani
J
,
Begthel
H
,
Jonkheer
S
,
Gregorieff
A
,
van de Born
M
, et al
EphB receptor activity suppresses colorectal cancer progression
.
Nature
2005
;
435
:
1126
30
.
37.
Guo
DL
,
Zhang
J
,
Yuen
ST
,
Tsui
WY
,
Chan
AS
,
Ho
C
, et al
Reduced expression of EphB2 that parallels invasion and metastasis in colorectal tumours
.
Carcinogenesis
2006
;
27
:
454
64
.
38.
Lugli
A
,
Spichtin
H
,
Maurer
R
,
Mirlacher
M
,
Kiefer
J
,
Huusko
P
, et al
EphB2 expression across 138 human tumor types in a tissue microarray: high levels of expression in gastrointestinal cancers
.
Clin Cancer Res
2005
;
11
:
6450
8
.
39.
Fu
DY
,
Wang
ZM
,
Wang
BL
,
Chen
L
,
Yang
WT
,
Shen
ZZ
, et al
Frequent epigenetic inactivation of the receptor tyrosine kinase EphA5 by promoter methylation in human breast cancer
.
Hum Pathol
2010
;
41
:
48
58
.
40.
Kober
P
,
Bujko
M
,
Oledzki
J
,
Tysarowski
A
,
Siedlecki
JA
. 
Methyl-CpG binding column-based identification of nine genes hypermethylated in colorectal cancer
.
Mol Carcinog
2011
;
50
:
846
56
.
41.
Hafner
C
,
Schmitz
G
,
Meyer
S
,
Bataille
F
,
Hau
P
,
Langmann
T
, et al
Differential gene expression of Eph receptors and ephrins in benign human tissues and cancers
.
Clin Chem
2004
;
50
:
490
9
.
42.
Sheng
Z
,
Wang
J
,
Dong
Y
,
Ma
H
,
Zhou
H
,
Sugimura
H
, et al
EphB1 is underexpressed in poorly differentiated colorectal cancers
.
Pathobiology
2008
;
75
:
274
80
.
43.
Johansson
S
,
Svineng
G
,
Wennerberg
K
,
Armulik
A
,
Lohikangas
L
. 
Fibronectin-integrin interactions
.
Front Biosci
1997
;
2
:
d126
46
.
44.
Mao
Y
,
Schwarzbauer
JE
. 
Fibronectin fibrillogenesis, a cell-mediated matrix assembly process
.
Matrix Biol
2005
;
24
:
389
99
.
45.
Akiyama
SK
,
Olden
K
,
Yamada
KM
. 
Fibronectin and integrins in invasion and metastasis
.
Cancer Metastasis Rev
1995
;
14
:
173
89
.
46.
Davis
S
,
Gale
NW
,
Aldrich
TH
,
Maisonpierre
PC
,
Lhotak
V
,
Pawson
T
, et al
Ligands for EPH-related receptor tyrosine kinases that require membrane attachment or clustering for activity
.
Science
1994
;
266
:
816
9
.
47.
Stein
E
,
Lane
AA
,
Cerretti
DP
,
Schoecklmann
HO
,
Schroff
AD
,
Van Etten
RL
, et al
Eph receptors discriminate specific ligand oligomers to determine alternative signaling complexes, attachment, and assembly responses
.
Genes Dev
1998
;
12
:
667
78
.
48.
Genander
M
,
Halford
MM
,
Xu
NJ
,
Eriksson
M
,
Yu
Z
,
Qiu
Z
, et al
Dissociation of EphB2 signaling pathways mediating progenitor cell proliferation and tumor suppression
.
Cell
2009
;
139
:
679
92
.
49.
Bardelli
A
,
Parsons
DW
,
Silliman
N
,
Ptak
J
,
Szabo
S
,
Saha
S
, et al
Mutational analysis of the tyrosine kinome in colorectal cancers
.
Science
2003
;
300
:
949
.
50.
Wood
LD
,
Parsons
DW
,
Jones
S
,
Lin
J
,
Sjoblom
T
,
Leary
RJ
, et al
The genomic landscapes of human breast and colorectal cancers
.
Science
2007
;
318
:
1108
13
.
51.
Faoro
L
,
Singleton
PA
,
Cervantes
GM
,
Lennon
FE
,
Choong
NW
,
Kanteti
R
, et al
EphA2 mutation in lung squamous cell carcinoma promotes increased cell survival, cell invasion, focal adhesions, and mammalian target of rapamycin activation
.
J Biol Chem
2010
;
285
:
18575
85
.
52.
Lisabeth
EM
,
Fernandez
C
,
Pasquale
EB
. 
Cancer somatic mutations disrupt functions of the EphA3 receptor tyrosine kinase through multiple mechanisms
.
Biochemistry
2012
;
51
:
1464
75
.
53.
Huusko
P
,
Ponciano-Jackson
D
,
Wolf
M
,
Kiefer
JA
,
Azorsa
DO
,
Tuzmen
S
, et al
Nonsense-mediated decay microarray analysis identifies mutations of EPHB2 in human prostate cancer
.
Nat Genet
2004
;
36
:
979
83
.
54.
Janes
PW
,
Griesshaber
B
,
Atapattu
L
,
Nievergall
E
,
Hii
LL
,
Mensinga
A
, et al
Eph receptor function is modulated by heterooligomerization of A and B type Eph receptors
.
J Cell Biol
2011
;
195
:
1033
45
.

Supplementary data