Preferential Induction of EphB4 over EphB2 and Its Implication in Colorectal Cancer Progression

The erythropoietin-producing hepatoma (Eph) family of receptor tyrosine kinases and their membrane-bound ligands, the ephrins, have been implicated in regulating cell adhesion and migration during development by mediating cell-to-cell signaling events (1–3). These receptors, however, carry unique functions depending on the organ or cell lineage. EphB4 and EphrinB2, for example, have restricted expression in the venous and arterial compartments, respectively (4, 5). Unlike any other member of the family, they are critically required for blood vessel maturation, such that knockout of either gene results in embryonic lethality (6).

Certain EphB receptors are Wnt/β-catenin/T-cell factor 4 (TCF4) regulated (7). The Wnt/β-catenin pathway initiates a signaling cascade critical in both normal development and the initiation and progression of cancer (8–10). Wnt is a secreted glycoprotein that binds to receptors belonging to the Frizzled family leading to activation of disheveled or DSH. DSH induces dephosphorylation of β-catenin and releases it from its inactive complex with APC. Free β-catenin translocates to the nucleus and complexes with members of the T-cell factor (TCF)/lymphoid enhancer factor family of transcription factors. TCF/β-catenin recruits the transcriptional coactivator cyclic AMP-responsive element binding protein–binding protein (CBP), or its closely related homologue p300, and other components of the basal transcription machinery to initiate transcription of target genes. In the absence of Wnt activation, phosphorylated β-catenin is ubiquitinated in the cytoplasm and degraded in proteasomes. Colon cancers frequently harbor genetic defects that inhibit β-catenin degradation, leading to nuclear β-catenin translocation and activation of target genes. APC mutations are found in >80% whereas mutations in β-catenin NH₂ terminus are present in ∼5% of colorectal cancers (11, 12). Interestingly, whereas EphB2 is expressed in the progenitor cells located at the bases of normal intestinal crypts and controls cell compartmentalization along the crypt axis, most human colorectal cancers lose expression of EphB2 at the adenoma-carcinoma transition (13). Further, reduction of EphB2 activity accelerates tumorigenesis in the colon and rectum of APC^Min/+ mice (13). Reduced expression of EphB2 is associated with increased invasiveness and metastases (14) and poor survival (15). In contrast, EphB4 analysis in fresh frozen colorectal cancer and normal adjacent tissue shows increased mRNA and protein levels (16, 17). However, these findings were not reproduced in formalin-fixed immunohistochemical analysis (13, 18).

Reagents, tissues, and cell lines. Human tissues were harvested at surgery from Institutional Review Board–approved, consenting patients. All cell lines were obtained from American Type Culture Collection and cultured under standard conditions. Details of reagents and expression vectors are provided in Supplementary data.

In vitro experiments. Immunoblotting (20 μg cell lysates), immunoprecipitation (100 μg cell lysates), and immunostaining were done following standard protocols detailed elsewhere (19). The short hairpin RNA (shRNA) expression vector was constructed according to Dickins and colleagues (20). In parallel experiments, chemically transfectable vectors were generated by deleting long terminal repeat sequences and self-ligation.

Animal experiments. All animal experiments were done under Institutional Animal Care and Use Committee–approved protocols and adhered to NIH guidelines. For xenograft experiments, appropriate numbers of cells were implanted in the flanks of 10- to 12-week-old, male BALB/c athymic mice. Tumor size was measured every other day, and volume estimated as 0.52 × a × b², where a and b are the largest and smallest lengths of the palpable tumor, respectively. To study metastases, cells were injected in spleens of 10- to 12-week-old, male BALB/c athymic mice through a flank incision. After 1 min, splenectomy was done and mice were recovered. To study tet-regulation, one set of mice was fed 0.2 mg/mL doxycycline in 0.5% sucrose solution as drinking water in light-proof bottles, replaced every 4 d.

We studied protein expression by immunoblotting and immunofluorescence using two different monoclonal antibodies highly sensitive and specific to EphB4 (Supplementary Fig. S1). Anti-EphB4 #V131 antibody recognizes native human EphB4 but not any other member of EphB family (Supplementary Fig. S1A). Preincubation of the antibody with the antigenic peptide used to generate the antibody (blocking peptide) completely abolishes immunofluorescence signal, confirming antibody specificity (Supplementary Fig. S1B). Anti-EphB4 #V265 antibody recognizes denatured human EphB4 on immunoblotting under reducing conditions with high sensitivity (Supplementary Fig. S1C) and weakly recognizes mouse EphB4 but not other Eph receptors (Supplementary Fig. S1D). We evaluated EphB4 and EphB2 expression in 102 fresh frozen colorectal tumor specimens and matched adjacent normal tissues. Tissue arrays composed of 7-μm sections of tumor and adjacent normal tissues were analyzed by immunofluorescence to ensure identical processing conditions. A single observer (J.S.S.) graded the staining as reduced, no change, or increased expression. EphB4 and EphB2 band intensity on immunoblotting was compared between tumor and normal tissue normalized to β-actin. There was excellent correlation between immunoblotting and immunofluorescence (Spearman's r = 0.3, P = 0.004 for EphB4 and r = 0.4, P = 0.0001 for EphB2). EphB4 expression was minimal to none in all normal colon samples studied, whereas 67 of 90 (73%) primary colorectal cancer specimens expressed significantly higher levels of EphB4 compared with adjacent normal mucosa by immunofluorescence (Fig. 1A; Supplementary Fig. S2A and B) and immunoblotting (Fig. 1B; Supplementary Fig. S2B). EphB4 is specifically expressed on the membrane of tumor cells, and not in the surrounding stroma (Fig. 1A), and moderately expressed in normal and tumor vasculature (data not shown). In contrast to EphB4, there was consistent expression of EphB2 by immunoblotting and immunofluorescence in all normal colon tissues. In 58 of 90 (64%) primary tumors, there was a significant decline of EphB2 expression both by immunofluorescence (Fig. 1A; Supplementary Fig. S2A) and immunoblotting (Fig. 1B). EphB4 expression in tumors correlated inversely with EphB2 expression (Spearman r = −0.3, P = 0.02). Among the 12 liver metastases, adjacent normal liver did not express EphB4 or EphB2, whereas all colorectal metastases expressed EphB4 but not EphB2 (Fig. 1C and D; Supplementary Fig. S2C). Similarly, lymph node metastasis expressed EphB4 but not EphB2 (Supplementary Fig. S2D). Preliminary quantitative reverse transcription-PCR analysis in five primary tumors, done as detailed before (19), revealed a median 3.8-fold higher level of EphB4 mRNA copies in tumor tissue compared with adjacent normal colon (Supplementary Fig. S2E). The intensity of EphB4 expression on immunoblotting (Fig. 1E) and immunofluorescence (Fig. 1F) correlated with increasing tumor stage and grade, assessed from H&E staining by a single pathologist (D.H.) unaware of the expression data, whereas the opposite was true for EphB2. Thus, EphB4 and EphB2 are differentially regulated in colon cancer with induction of EphB4 and decline in EphB2 levels as disease progresses from early to late stage. In metastatic cancer, EphB2 is completely lost and EphB4 is consistently expressed at high levels. Our finding of EphB2 silencing in colon cancers is in agreement with the previous reports of Batlle and colleagues (13) and Guo and colleagues (14). However, we show differential regulation and overexpression of EphB4 in colon cancers, in contrast to the report by Batlle and colleagues. Some of these differences could arise from the antibodies used and variations in tissue processing. In their study, Batlle and colleagues studied EphB4 expression by staining formalin-fixed paraffin-embedded sections using a polyclonal anti-EphB4 antibody. We studied fresh frozen tissues and used highly sensitive and specific monoclonal antibodies to study expression in two complementary ways (i.e., immunoblotting and immunofluorescence), with excellent correlation between the two techniques. Furthermore, we observed nearly uniform membrane staining of EphB4 in all sections, allowing us to compare the intensity of staining rather than the number of positive cells as studied by Batlle and colleagues We also compared expression in tumors to matched normal colon samples, a logical internal control for baseline expression. We have used the American Joint Committee on Cancer tumor-node-metastasis staging to compare expression between various tumor stages, in contrast to the Dukes' staging used in Batlle's study. Lastly, Batlle studied colorectal cancers from the Netherlands, and although unlikely, we cannot rule out a biological difference in the cancers in the United States.

To study the regulation of EphB4 following mutations in the β-catenin pathway, we evaluated tumor-free (“normal”) colon mucosa, polyp, cancer, and an omental metastatic deposit from a patient with APC mutation and familial adenomatous polyposis (FAP; Fig. 2A and B). Unlike normal colon in patients without APC mutation, the APC mutant colon expresses detectable levels of EphB4 (compare normal colon in Fig. 2A to normal in Fig. 1A and Supplementary Fig. S2A and B). Expression of EphB4 was higher in polyp and furthermore in primary and metastatic cancer (Fig. 2A and B). EphB2, on the other hand, whereas expressed in the normal colon and adenomatous polyp, was undetectable in primary tumor and metastatic lesion. In addition, sporadic primary and metastatic colorectal tumor samples express high levels of β-catenin in the cytoplasm and nucleus by immunofluorescence (Supplementary Fig. S1F), confirming that the Wnt/β-catenin signaling cascade is indeed active in these tumors. To validate up-regulation of EphB4 following APC mutation, we evaluated normal colon and colon adenomas harvested from APC^Min/+ mouse (APC min) and normal colon from age- and background-matched wild-type mouse (Fig. 2C). Normal colon of APC min mouse expressed EphB4, unlike wild-type colon, and there was a further 2- to 4-fold increase in expression in the polyps. EphB2 expression was also induced in APC^Min/+ colon compared with wild-type colon, and there was no appreciable change in EphB2 expression in adenomas. To study the mechanisms that regulate EphB4 expression, we first examined seven colorectal cancer cell lines and one normal colonic epithelial line in vitro (Fig. 2D). Six of seven cancer cell lines express EphB4, whereas normal colonic epithelial line does not express EphB4. EphB2 is expressed by normal colonic epithelium and six of seven cancer cell lines. EphB4 and EphB2 are inversely expressed in some of the cell lines, especially HT29 and COLO-205. HT29 has high levels of EphB4 and low levels of EphB2. Fluorescence-activated cell sorting (FACS) analysis of HT29 cells shows that EphB4 is expressed on the membrane of the vast majority of cells, and about 28% cells also express the cognate ligand EphrinB2 (Supplementary Fig. S3A). EphB4 can be phosphorylated in a dose- and time-dependent manner by clustered EphrinB2-Fc, but not Fc fragment alone, confirming that the receptor is functionally active (Supplementary Fig. S3B). Under serum-free conditions, EphB4 is unphosphorylated, whereas, when cultured in the presence of serum, EphB4 is phosphorylated even in the absence of exogenous EphrinB2 stimulation (Supplementary Fig. S3C). When endogenous EphrinB2 expression is abolished using siRNA, a significant proportion of EphB4 still remains phosphorylated, suggesting ligand-independent phosphorylation in HT29 cells. Reintroduction of wild-type APC, but not empty vector, into HT29 cells that harbor APC mutation results in a significant decrease in EphB4 expression (Fig. 2E), confirming that EphB4 expression is regulated by the Wnt/β-catenin/APC pathway.

Wnt/β-catenin signaling promotes self-renewal in a variety of tissue stem cells, including neuronal and hematopoietic stem cells. However, activation of the Wnt/β-catenin pathway has been shown to either promote or inhibit differentiation (8, 21, 22). Recently, based on a series of investigations (8, 21–23), Kahn has proposed that β-catenin/CBP–driven transcription is critical for proliferation with maintenance of the undifferentiated state, whereas a switch to β-catenin/p300–mediated gene expression is an essential first step in initiating normal cellular differentiation. Aberrant regulation of the balance between these two related transcriptional programs may be associated with a wide array of diseases, including cancer. We thus hypothesized that the differential expression of EphB2 and EphB4 during colorectal cancer progression may result from differential usage of coactivators in the Wnt/β-catenin pathway. Treatment with siRNA directed against CBP down-regulates expression of EphB4, but not EphB2 (Fig. 3A and B). In contrast, siRNA directed against p300 down-regulates EphB2, but not EphB4, whereas GFP-siRNA has no effect on expression of either protein. Therefore, EphB2 is a TCF/β-catenin–regulated gene that is dependent on the coactivator p300, whereas EphB4 expression relies on the coactivator CBP. With disease progression, the Wnt/β-catenin signaling cascade may become more dependent on the usage of the coactivator CBP resulting in expression of CBP/β-catenin/TCF4–mediated gene products [e.g., survivin (8), S100A4 (21), and EphB4] at the expense of the p300/β-catenin/TCF4 interaction–dependent gene products (e.g., EphB2). To study differential coactivator usage, CBP or p300 was immunoprecipitated from total protein extracted from primary colorectal cancer samples, and β-catenin association was assessed by immunoblotting (Fig. 3C and D). In the patient with FAP (case #22, Fig. 3C), the tumor-free colon and adenomatous polyp showed equal association of β-catenin with both CBP and p300. However, in primary tumor and metastatic deposit, a significantly higher amount of β-catenin was associated with CBP compared with p300. Similarly, in sporadic colorectal tumors (Fig. 3D), β-catenin was preferentially associated with CBP compared with p300. This difference may be secondary to preferential utilization of CBP as transcriptional coactivator or due to a difference in relative amounts of CBP and p300. In any event, progression from tumor-free colon mucosa to cancer and metastases shows increased association of nuclear β-catenin with CBP relative to p300, which may contribute to increased expression of EphB4 and down-regulation of EphB2.

Induction of EphB4 at the expense of EphB2 led us to speculate that EphB4 may play a tumor-promoting role in colorectal cancer, in keeping with its role in mammary tissue, where overexpression of EphB4 in the context of neuT accelerates the development of mammary tumors and their metastases (24). We addressed the biological role of EphB4 in colorectal cancer by knockdown of receptor expression using specific siRNA. Transfection of EphB4-specific siRNA results in a dose-dependent (Fig. 4A,, middle) and time-dependent (Supplementary Fig. S4A) reduction in EphB4 expression and cell number in HT29 cells (Fig. 4A,, top). Similar results were seen with three other active siRNAs targeting other regions of the EphB4 mRNA (data not shown), confirming that cell death is unlikely to result from an off-target siRNA effect. Mutations in three bases in the siRNA (siRNAΔ) within the first eight critical nucleotides fail to knock down EphB4 or reduce cell numbers. Knockdown of the exclusive ligand for EphB4, EphrinB2 (Fig. 4A,, bottom), however, has no effect on cell number, suggesting that the prosurvival function of EphB4 is ligand independent. Treatment of HT29 cells with EphB4 siRNA reduces surface expression of EphB4, as expected (Supplementary Fig. S4B). In addition, siRNA directed against human EphB4 has no effect on murine EphB4 expression (Supplementary Fig. S4C). To further confirm the target-specific effect of EphB4 siRNA, we introduced murine EphB4 in HT29 cells, which abrogated reduction in cell number resulting from human EphB4-specific siRNA (Supplementary Fig. S4D). In addition, EphB4-specific siRNA has no effect on the number of COLO-205 cells that express minimal EphB4 (Fig. 4B) or SW620 cells that do not express EphB4 (Supplementary Fig. S4E). In contrast, EphB2 siRNA has no effect on cell number in COLO-205 cells that express high levels of EphB2 (Fig. 4B), indicating that EphB4, but not EphB2, provides survival advantage in colon cancer. Target specificity of the siRNA used was confirmed by immunoblotting (Supplementary Fig. S4F).

We also tested the activity of endogenously expressed siRNA. We generated expression cassettes that produce shRNAs, which are processed into EphB4-specific siRNAs, and stably transfected them into HT29 cells under control of the tet-repressor (Supplementary Fig. S5A; ref. 20), hereon identified as the FF4 cell line. The clone expressing the tet-repressor gene alone was designated HT29-mock. Withdrawal of doxycycline from FF4 cell culture results in a dose-dependent increase in shRNA production (Supplementary Fig. S5B), a dose-dependent (Fig. 4C,, top) and time-dependent (Fig. 4C,, bottom) decrease in EphB4 protein expression, and a dose-dependent decrease in cell number (Fig. 4D). HT29-parent cells, HT29-mock cells, and FF4 cells were implanted in mice, and tumor growth was assessed with and without doxycycline ingestion (Fig. 5A). Unlike other tumors, FF4 tumors in mice that were not fed doxycycline show an 84% reduction in growth in vivo. Tumors harvested after 28 days of doxycycline withdrawal lack expression of EphB4 (Fig. 5A,, top right and left). In addition, there are large areas of necrosis, reduction in Ki-67 expression, increase in terminal deoxyribonucleotidyl transferase–mediated dUTP nick end labeling (TUNEL)–positive apoptotic cells, and reduction in CD31-positive tumor microvasculature (Fig. 5A , right). Reduction in tumor microvasculature suggests that EphB4 knockdown also has an antiangiogenic effect in addition to the direct effect on tumor cells. As expected, doxycycline withdrawal in mice implanted with FF4 and HT29-mock cells in either flank showed selective inhibition of FF4 tumors with no effect on HT29-mock tumors (Supplementary Fig. S6).

Regulation of cell migration and invasion by EphB4 was tested in vitro using EphB4 siRNA. Significant impairment in the ability of tumor cells to repopulate a cell-free area in confluent cell culture (Supplementary Fig. S7A) and invasion of Matrigel-coated membranes in response to epidermal growth factor is observed (Supplementary Fig. S7B). Under these experimental conditions, no significant cell death was observed, although early apoptotic changes may play a role in reduced tumor cell migration. To study tumor metastases in vivo, we injected FF4 or HT29 cells in spleens of mice, and livers were examined after 4 weeks. One group of mice bearing FF4 cells was deprived of oral doxycycline. Nearly 80% of HT29- and FF4/doxycycline–treated mice had liver tumors, whereas none of the mice injected with FF4 and not receiving doxycycline developed liver metastasis (Fig. 5B). Thus, EphB4 enhances tumor cell migration, invasion, and metastasis.

Deregulation of apoptosis is a frequent mechanism by which tumor cells gain a survival advantage. Therefore, we studied apoptosis in EphB4-depleted cells. Transfection of HT29 cells with EphB4-specific siRNA leads to cytoplasmic accumulation of nucleosomes (Supplementary Fig. S8A) and increase in caspase-8 activity (Supplementary Fig. S8B). To further study these events, we introduced varying doses of control and EphB4 siRNA along with increasing doses of tumor necrosis factor-related apoptosis-inducing ligand (TRAIL; R&D Systems) into HT29 cells for 24 h (Fig. 6A). Parent HT29 cells are relatively resistant to TRAIL-induced cell death. EphB4 knockdown increases the sensitivity of tumor cells to TRAIL-induced apoptosis, suggesting that EphB4 interrupts the TRAIL/death receptor pathway. EphB4 knockdown had no effect on the expression levels of TRAIL, DR4, DR5, or decoy receptors (Supplementary Fig. S8C). Decrease in cell numbers after conditional expression of EphB4 siRNA in FF4 cells is inhibited by TRAIL-neutralizing antibodies (Fig. 6B), indicating that EphB4 knockdown induces cell death in vitro by sensitizing cells to endogenous TRAIL. Furthermore, knockdown of TRAIL with siRNA also rendered cells resistant to cell death after induction of EphB4 siRNA (Supplementary Fig. S9). We next tested if forced expression of EphB4 would induce resistance to TRAIL. COLO-205 cells that have low basal levels of EphB4 were transfected with full-length EphB4 expression vector or empty vector (Fig. 6C,, top left) and treated with various concentrations of TRAIL. Whereas parent cells and vector only–transfected cells are highly sensitive to TRAIL, forced expression of EphB4 renders cells resistant to TRAIL-induced apoptosis (Fig. 6C,, bottom left). Conversely, forced expression of EphB3 in COLO-205 cells does not alter their sensitivity to TRAIL (Fig. 6C , right). EphrinB2 knockdown data had suggested that forward signaling downstream of EphB4 may not play an important role in providing survival signals to tumor cells. We therefore generated a mutant EphB4 protein, wherein the cytoplasmic tail was replaced by GFP (EphB4-eGFP; Supplementary Fig. S10A). EphB4-eGFP is unable to forward signal while retaining its ability to bind EphrinB2 and induce reverse signaling (Supplementary Fig. S10B). Expression and membrane localization of EphB4-eGFP in 293 cells (293-EphB4-eGFP) were confirmed (Supplementary Fig. S10C and D). Exposure of 293 cells transfected with wild-type EphB4 (293-EphB4) to 2 μg/mL clustered EphrinB2-Fc results in phosphorylation of overexpressed EphB4 (Supplementary Fig. S10E, top). However, 293-EphB4-eGFP cells show decreased phosphorylation of EphB4, thus showing a dominant negative function for the mutant protein. Both 293-EphB4 and 293-EphB4-eGFP can bind EphrinB2-AP (Supplementary Fig. S10E, bottom) and induce reverse signaling.

COLO-205 cells transfected with mutant EphB4-eGFP expression vector are also resistant to TRAIL-induced cell death (Fig. 6C,, left), indicating that the extracellular domain of EphB4 is sufficient to protect tumor cells from TRAIL. SW480 cells express intermediate levels of EphB4 and are moderately sensitive to TRAIL. Knockdown of EphB4 in these cells increases their sensitivity to TRAIL-induced apoptosis (Supplementary Fig. S11A). Further, forced expression of full-length EphB4 or mutant EphB4-eGFP renders these cells resistant to TRAIL (Supplementary Fig. S11B). Wild-type COLO-205 cells are also sensitive to TRAIL in vivo. COLO-205 cells (parent) were implanted in BALB/c athymic nude mice (Fig. 6D,, top), and after tumors were established, mice treated with exogenous TRAIL show a 47% reduction in tumor volumes at 3 weeks, showing sensitivity to exogenous TRAIL. Treatment with TRAIL neutralizing antibody results in a 1.7-fold increase in tumor volume compared with no treatment or isotype IgG (P = 0.004), showing sensitivity to endogenous TRAIL. We then generated COLO-205 cells stably expressing mutant EphB4-eGFP protein (EphB4-eGFP). When implanted in mice, these cells result in 1.9-fold larger tumors compared with parent cells at 3 weeks (Fig. 6D , bottom). Treatment of mice bearing COLO-205-EphB4-eGFP tumors with TRAIL has no effect on tumor volumes, indicating that signaling-deficient EphB4 is sufficient to confer resistance to TRAIL in vivo.

In summary, whereas both EphB2 and EphB4 are expressed in early-stage colon cancer, expression diverges as disease advances, such that there is progressive loss of EphB2 and increase in EphB4 levels. In the normal colorectal epithelium, EphB2 is expressed at high levels in the region of precursor stem cells. Benign polyps express high levels of EphB2 and show induction of EphB4. With disease progression toward overt colorectal cancer, early and differentiated tumors continue to express EphB2 while EphB4 levels increase further. Poorly differentiated and higher-stage tumors show loss of EphB2 and high levels of EphB4. Such a differential expression has not previously been shown and is surprising considering that both genes are targets of the Wnt/β-catenin pathway, which is induced in the vast majority of colorectal cancers. We found that as disease progresses, the transcriptional coactivator CBP is preferentially used over p300 by the Wnt pathway, which may represent one of the mechanisms through which expression of EphB4 is induced while EphB2 is down-regulated. This differential expression provides survival benefit to tumor cells via two distinct mechanisms. First, EphB4 on tumor cells, unlike EphB2 or EphB3, provides a direct and cell-autonomous effect that protects tumor cells from cell death, specifically endogenous TRAIL. It remains to be determined how EphB4, in particular its extracellular domain, confers a survival phenotype and interferes with the TRAIL pathway. Second, EphB4 knockdown in tumors also has an antiangiogenic effect, thereby inhibiting tumor growth. This is not surprising, based on the critical requirement of EphB4, but not of EphB2 or EphB3, for vessel maturation during mouse embryogenesis (25). EphB4 on tumor cells interacts with EphrinB2 on tumor vasculature, promoting tumor angiogenesis and growth. EphB4 expression also correlates with stage and grade of colorectal cancer. Further work is required to determine the potential prognostic significance of this expression. Thus, EphB4 is preferentially overexpressed in colorectal cancer, and targeting EphB4 has the potential to modulate both tumor angiogenesis and tumor cell viability.

No potential conflicts of interest were disclosed.

Grant support: National Cancer Institute grant RO1CA79218 (P.S. Gill) and Tobacco-Related Disease Research Program grant 14DT-0125 (S.R. Kumar). M. Kahn wishes to thank the Jeannik M. Littlefield-AACR Grant in Metastatic Colon Cancer Research.

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.

We thank Lee Ellis, M.D. (M.D. Anderson Cancer Center, Houston, TX) for providing tumor samples, and Hemant Bid, Loubna Hassanieh, Pramod Sutrave, and LiXin Yang for excellent technical assistance.