Nuclear Export of Ubiquitinated Proteins Determines the Sensitivity of Colorectal Cancer to Proteasome Inhibitor

Colorectal cancer is one of the most common malignancies with estimated more than 1.2 million new cases each year worldwide. Noteworthy, its incidence rate has been increasing in many developing areas over the last several decades despite the downward trend in developed countries (1). As the majority of patients are diagnosed at advanced stages, and the relapse of the tumor that has become resistant to the 5-fluorouracil- and oxaliplatin-based combination chemotherapy, the 5-year survival rate for patients with colorectal cancer remains quite low (2). Thus, early diagnosis through targeted screening and finding new targets and approaches for the treatment are sorely needed to improve the prognosis of individuals with colorectal cancer.

It has been shown that alterations of the ubiquitin–proteasome system are critical for cancer initiation and development, often through enhanced degradation of tumor suppressors and reduced degradation of oncoproteins (3, 4). Genomic analyses also revealed that mutations of ubiquitination-related proteins, including APC, p53, Fbw7, and Smad4, are the characteristic changes of colorectal cancer (5). Proteasome inhibitors such as bortezomib are able to induce growth arrest and apoptosis in colorectal cancer cells in vitro (6). Bortezomib has also been approved for the treatment of multiple myeloma and mantle cell lymphoma by FDA (7, 8). However, bortezomib or newer generation of proteasome inhibitors had minimal antitumor activity in patients with advanced colorectal cancer or other solid tumors (9–11). These results prompted significant efforts to combine proteasome inhibitors with other antitumor strategies, including conventional chemotherapy, radiation, and other targeted therapy. At present, the promise of proteasome inhibitors in the treatment of solid tumors has yet to be realized.

Precisely controlled transportation of protein across nuclear membrane is critical for proper growth, death, and differentiation of eukaryotic cells (12). It has been shown that chromosome region maintenance 1 (CRM-1) recognizes nuclear export signal (NES) of target proteins and mediates the nuclear export of many tumor suppressor proteins (TSP), such as p53, FOXO, RB1, and CDKN1A (13, 14). CRM-1 is upregulated in a variety of cancers, and responsible for aberrant cytoplasmic localization and inactivation of tumor suppressors. Furthermore, specific CRM-1 inhibitor KPT330 (selinexor) has broad antitumor activity in various tumors and is being actively explored as a novel cancer therapeutic agent (15, 16). FDA has designated selinexor orphan drug status for certain types of leukemia and lymphoma (17, 18).

In this study, we examined the nuclear protein exportation upon proteasome inhibitor and KPT330 exposure, proposing possible hypothesis of chemoresistant mechanism of proteasome inhibitors. The synergistic effect and its molecular mechanism of combining bortezomib and KPT330 on colorectal cancer were investigated in vitro and in patient-derived xenografts (PDX) models. The intent of these studies was to provide a rationale for the combination therapy using inhibitors for proteasome and nuclear export in the treatment of colorectal cancer.

The human cell lines HeLa, SW480, SW620, HCT116, and RKO were obtained from ATCC in September 2014, which were cultured in DMEM (Hyclone) supplemented with 10% FBS (Hyclone) and 1% penicillin/streptomycin, and were maintained at 37°C in an incubator under an atmosphere containing 5% CO₂. The cell lines were obtained directly from ATCC that performs cell line characterizations or authentication by the short tandem repeat profiling and passaged in our laboratory for fewer than 6 months after receipt.

Cell viability assays were conducted using the CCK8 method. Briefly, cancer cells were suspended and seeded on 96-well plates (10³ cells per well) in 100-μL culture medium. Twenty-four hours later, the cells were treated with the indicated drugs with another 100 μL culture medium for an additional 72 hours. Ten microliters of CCK8 reagent (Dojindo) was added to each well, and the plates were incubated at 37°C for another 2 hours. The absorbance was measured with a spectrophotometer at 450 nm. Two types of chemotherapeutic drugs, bortezomib (Selleck) and KPT330 (Selleck,), were used in our study. The results of the combined treatment were analyzed according to the isobolographic method of Chou and Talalay (19) using the Calcusyn software program. The resulting combination index (CI) was used as a quantitative measure of the degree of interaction between the two drugs. CI > 1 indicates additivity, and CI < 1 indicates synergism.

Colorectal cancer cells were cultured for 24 hours prior to drug treatment. Cells were then treated with 5 nmol/L bortezomib and 100 nmol/L KPT330 for another 24 hours, and then fixed with 4% paraformaldehyde. After blocking with 5% BSA for 1 hour, slides were incubated overnight with anti-p53 Ab (Santa Cruz Biotechnology) and anti-ubiquitinated proteins (Santa Cruz Biotechnology). Cells were then washed and incubated with fluorescence conjugated goat anti-rabbit IgG for 1 hour. Slides were analyzed using Leica Fluorescence Inversion Microscope System.

Western blotting was performed as described previously (20). The following specific antibodies were used for analysis: anti-cleaved PARP1, anti-Bax, anti-p21, anti-lamin A/C, and anti-tublin antibodies were purchased from Cell Signaling Technology. Anti-actin, anti-p53, anti-ubiquitinated proteins, anti-Mdm2, and anti-Ki67 antibodies were purchased from Santa Cruz Biotechnology.

For quantification of apoptosis, the Pharmingen Annexin V Apoptosis Detection Kit (BD Biosciences) was used according to the manufacturer's instructions. Apoptosis was further assessed by the measurement of caspase-3 and -7 activity using a luminometric caspase-Glo-3/7 assay (Promega) according to the manufacturer's protocol.

Cells were seeded in 6-well plates and allowed to attach overnight. After treating with bortezomib (5 nmol/L) and KPT330 (100 nmol/L) for 48 hours, cells were fixed with 75% ethanol, stained with RNase-containing propidium iodide (PI), and analyzed by flow cytometry after 20 minutes incubation.

Nude mice (4–6 weeks old, male) were used as an in vivo mouse model. All mouse procedures were approved by the Animal Care and Use Committees of Xinhua Hospital (Shanghai, P.R. China). Mice were inoculated subcutaneously in both flanks with HCT116 cells (2 × 10⁶), and were randomly divided into four groups. The groups were treated with vehicle (control), bortezomib (1 mg/kg, intraperitoneal administration, twice a week), KPT330 (10 mg/kg, oral administration, twice a week), or a combination of bortezomib and KPT330. Tumors were measured twice a week with a caliper. Their volumes were calculated as follows: 0.5 × length × width². After 18 days of treatment, the tumors were removed from euthanized mice, photographed, and paraffin imbedded. Tumor inhibition rate = (control group volume − treatment group volume)/control group volume × 100%. Tumor inhibition rate >30% was considered as sensitive treatment, tumor inhibition rate <30% was considered as resistant treatment, according to Response Evaluation Criteria in Solid Tumors (RECIST) criteria (21).

Tumor tissue specimens from freshly resected colon were washed and cut into 2 to 3 mm³ pieces in antibiotic-containing PBS medium. Under anesthesia with pentobarbital, one tumor piece was implanted subcutaneously by a small incision in one side of axilla into 4- to 6-week-old male nude mice. Tumors were harvested when they reached a size of 1,500 mm³ (Px1 xenografts). Xenografts from Px1 mice were divided into small pieces and then implanted again subcutaneously as described above to obtain Px2 xenografts. This process was further repeated and the experiments were performed on xenografts Px3.

Slide sections of tumor specimens were baked at 60°C for 1 hour, deparaffinized, and rehydrated with xylene and ethanol. After antigen retrieval with microwave heating, endogenous peroxidase activity was blocked with 3% hydrogen peroxide. Nonspecific staining was minimized by incubation in 5% FBS. Slides were then incubated with the primary antibodies at 4°C overnight. After washing and incubating with secondary antibodies at room temperature for 1 hour, specific staining was visualized using the Horseradish Peroxidase Color Development Kit (Beyotime). Photomicrographs were taken using an Olympus microscope (Olympus). Expression index = % of positive cells × staining intensity (1+ 2+ or 3+).

Statistical analyses were performed using SPSS 13.0 software. The paired, two-tailed Student t test was used to determine the significance between two groups. P < 0.05 was regarded as the threshold value for statistical significance.

Proteasome inhibition offers an effective strategy to kill tumor cells, and proteasome inhibitors such as bortezomib have been approved to treat multiple myelomas and lymphoma. However, bortezomib or newer generation of proteasome inhibitors had minimal antitumor activity in patients with advanced colorectal cancer or other solid tumors. To explore the potential chemoresistant mechanisms to proteasome inhibitors, we examined the distribution of ubiquitinated proteins in proteasome inhibitor MG132-treated HeLa cells. As revealed by immunofluorescence staining, exposure to MG132 increased ubiquitinated proteins in cells, most notably in cytoplasm, whereas the predominant staining was found in the nuclei in the presence of CRM1 inhibitor LMB (Fig. 1A). Nuclear and cytosolic ubiquitinated proteins were also examined by immunoblotting after the cells were treated with the inhibitors. As shown in Fig. 1B, MG132 treatment dramatically increased ubiquitinated proteins in both nuclear and cytoplasmic fractions (Fig. 1B, lanes 5–8 vs. 1–4). When compared with cells treated with MG132 alone, a significantly higher increase of ubiquitinated proteins in the nucleus, as revealed by a higher nuclear/cytoplasmic ratio, was detected in cells simultaneously treated with MG132 and LMB (Fig 1B). We then investigated the distribution of ubiquitinated protein in colorectal cancer cells HCT116 and RKO exposed to proteasome inhibitor bortezomib and CRM1 inhibitor KPT330. Similar to the finding in HeLa cells, ubiquitinated proteins were exported from nuclei upon bortezomib treatment, and combination of bortezomib and KPT330 led to increased ubiquitinated protein in the nuclei (Fig. 1C and D). These results indicated that proteasome inhibition promotes the export of ubiquitinated protein from nuclei. It is likely that the NESs in these proteins are responsible for proteasome inhibition–induced nuclear export.

To assess whether the export of nuclear proteins is related to the cytotoxic action of bortezomib, colorectal cancer cells HCT116, RKO, SW480, and SW620 cells were cultured in the presence of different concentrations of bortezomib or KPT330 for 72 hours. The IC₅₀ of different cells indicated that HCT116 and RKO were moderately more resistant to bortezomib compared with SW480 and SW620 (22 and 99 nmol/L vs. 5 and 9 nmol/L; Fig. 2A), whereas all of them were relatively insensitive to KPT330 (303, 1,790, 1,079, and 2,345 nmol/L, respectively; Fig. 2A). When colorectal cancer cells were treated concurrently with bortezomib and KPT330 at the indicated concentrations, a markedly greater inhibition of proliferation was observed in all the cells (Fig. 2B and C, left). Noteworthy, CI value from isobologram analysis (19) revealed that there is a synergistic effect between bortezomib and KPT330 in HCT116 (p53 wild-type) and RKO (p53 wild-type) cells with CI < 1 (Fig. 2B, right), but not in SW480 (p53 mutant) and SW620 (p53 mutant) cells (Fig. 2C, right). Under microscope, there were also profound morphologic alterations when the cells were exposed to bortezomib in the presence of KPT330 (Fig. 2D). To further explore whether the order of drug treatment affects the synergistic action, HCT116 and RKO cells were exposed to bortezomib and KPT330 at different orders. As shown in Supplementary Fig. S1, there were no significant differences in the synergistic action regardless of the order. Thus, bortezomib and KPT330 were given concurrently in the following studies. Furthermore, combination of bortezomib and KPT330 had significantly increased inhibition of colony formation on HCT116 cells than either drug alone (Fig. 2E). Colony numbers were counted and shown in Fig. 2F. These results suggest that the synergistic effects depend on the function of p53. To further confirm that other 20S proteasome inhibitor carfilzomib exert the similar synergistic effect with KPT330, HCT116, and SW480 cells were treated with carfilzomib and KPT330 concurrently. Similar to bortezomib, the cell viability assay and CI values revealed that combinational treatment have synergistic effect on HCT116 rather than SW480 cells (Supplementary Fig. S2A and S2B).

To minimize the possibility that the effect of KPT330 is a result of its action on cellular processes other than nuclear exportation, we knocked down CRM1 in HCT116 cells using two specific siRNAs. The efficiency of specific siRNA on CRM1 level was confirmed by real-time PCR analysis (Fig. 2G). As shown in Fig. 2H, similar to KPT330 treatment, CRM1 knockdown significantly enhanced the cytotoxic effect of bortezomib. Taken together, these results demonstrated that inhibition of nuclear exportation synergistically enhanced the cell killing activity of proteasome inhibition in p53^+/+ colon cancer cells HCT116 and RKO.

To further understand the cytotoxic effects of bortezomib and KPT330 on HCT116 and RKO cells, Annexin V staining and caspase-3/7 activities were assessed using the flow cytometric analysis and luminometric caspase-Glo-3/7 Assay Kit. Although the relative low doses of each drug alone induced a moderate increase of Annexin V staining and caspase-3/7 activation, their combination markedly augmented both the staining and the activation (Fig. 3A and B). The increased Annexin V staining was effectively inhibited by the caspase inhibitor Z-VAD-FMK (Fig. 3C). Moreover, the combination of bortezomib and KPT330 also led to more PARP1 cleavage in HCT116 and RKO cells, indicating synergetic apoptotic effect of the two agents (Fig. 3D). The effects of bortezomib and KPT330 on cell cycle were further examined. Although the compounds alone did not significantly change cell cycles, HCT116 and RKO cells exposed to both bortezomib and KPT330 underwent a G₂–M cell-cycle arrest (Fig. 3E), which is a hallmark of p53-mediated cell-cycle block (22).

These findings prompted us to examine the action of KPT330 and bortezomib on tumor xenografts derived from HCT116 cells in nude mice. When the xenografts became palpable (10–20 mm³), tumor-bearing nude mice were randomly assigned to receive vehicle, bortezomib, KPT330, or both (n = 10 per group). They showed tolerance to treatment and maintained normal activities. Regular twice a week measurements found no marked changes in body weight. After 18 days of treatment, the mice were euthanized to dissect the tumors. The combination of the drugs significantly enhanced tumor growth inhibition compared with vehicle (83.8% tumor reduction, P < 0.01), bortezomib (45.7% tumor reduction, P < 0.01), or KPT330 (55.2% tumor reduction, P < 0.01) by the final day of treatment (Fig. 4A). To further confirm the synergistic effect, late-stage tumors, with the volume reaching approximately 200 mm³ before treatment, were also examined (n = 6 per group). Similar to the results in Fig. 4A, the combinational treatment group exerted more significant therapeutic action compared with other groups in late-stage tumors (Fig. 4B), indicating the synergistic effect of boretezomib and KPT330 both existed in early- and late-stage tumors in terms of volume. The expression of p53, Ki67, and DNA fragmentation in the tumors were then evaluated by IHC and TUNEL assay (Fig. 4C). Compared with these from the vehicle group, bortezomib treatment decreased the ratio of nuclear/cytoplasmic p53 in the tumor (0.57 vs. 0.73, P < 0.01), whereas KPT330 and bortezomib combination significantly increased the ratio (2.52 vs. 0.73, P < 0.01) compared with control group (Fig. 4D). Interestingly, the combination of bortezomib and KPT330 also significantly increased the level of p53 in tumors. Furthermore, compared with vehicle or either drug alone, the combination treatment resulted in significantly reduced level of Ki67 and increased DNA fragmentation, indicating that these tumors had the decreased cell growth and likely p53-mediated apoptosis (Fig. 4D).

The level of p53 in cells is mainly controlled through ubiquitination and proteasomal degradation. Association of p53 expression with the cytotoxic effect of KPT330 and bortezomib prompted us to examine its subcellular localization. In colorectal cancer cells, accumulated p53 upon bortezomib treatment was mostly in cytoplasm, whereas cotreatment with KPT330 resulted in a predominant nuclear staining of p53 (Fig. 5A; Supplementary Fig. S3). To further confirm the result, cellular fractionation was performed to examine p53 in nuclei and cytosol by immunoblotting. As shown in Fig. 5B and Supplementary Fig. S4A, the results were consistent with the findings from immunofluorescence and indicated that bortezomib induced nuclear export of p53, whereas it could be effectively blocked by KPT330.

It is conceivable that nuclear retention of p53 might reduce its proteasomal degradation and activate the transcription of its target genes. We therefore analyzed the level of p53 and the expression of its targeted genes p21, Bax, and Mdm2 by Western blotting. The results showed that KPT330 further increased bortezomib-induced p53 accumulation, and the expression of Mdm2, p21, and Bax (Fig. 5C and D; Supplementary Fig. S4B and S4C). These changes likely account for the synergistic apoptosis and G₂–M arrest when HCT116 and RKO cells were treated with bortezomib and KPT330. To further determine the pivotal role of p53 in the process, we used siRNA to knockdown its expression, which was confirmed by real-time PCR analysis (Fig. 5E; Supplementary Fig. S4D). As shown in Fig. 5F and Supplementary Fig. S4E, the synergistic cytotoxic effect of bortezomib and KPT330 was markedly attenuated by p53 knockdown in HCT116 and RKO cells. It is not clear at this moment whether the lack of complete reversal is due to experimental limitations or involvement of additional factors in the synergistic action. To further confirm whether p53 plays a critical role in the synergistic effect of carfilzomib, nuclear distribution of p53 was investigated in HCT116 cells exposed to carfilzomib and KPT330. Similar to the results in Fig. 5A, there was significant nuclear export of p53 upon carfilzomib exposure, which was inhibited by KPT330 treatment (Supplementary Fig. S2C). When the expression of p53 was knocked down, the synergistic cytotoxicity was largely reversed (Supplementary Fig. S2D).

To further evaluate the potential therapeutic effects of bortezomib and KPT330, patient-derived primary human colorectal cancer xenografts (PDX model) were established and used in our study. We transplanted primary tumor tissues from various colorectal cancer patients into nude mice. The clinical pathologic features and p53 status were shown in Table 1. After the initial xenografts were established, they were reimplanted into a panel of nude mice to expand the colony (Supplementary Fig. S5A). Hematoxylin and eosin (H&E)-stained tumor sections (Model CRC0008) indicated that the initial xenografts and their passages were histologically similar to the original tumors (Supplementary Fig. S5B). We then monitored the growth of four PDXs with bortezomib and KPT330 treatment (Fig. 6A–D, n = 6 per group). Compared with tumors from the control group, bortezomib or KPT330 alone led to moderate tumor regression by 9.6% to 36.6%, whereas their cotreatment inhibited tumor growth more effectively by 60.9% to 76.7% in CRC0008 (p53 wild-type), CRC0014 (p53 wild-type), and CRC0005 (p53 wild-type; Fig. 6A–C). However, Model CRC0006 (R248G, loss-of-function p53 mutation) was the least responsive to cotreatment with 24% tumor regression (Fig. 6D).

Then tumor sections of Model CRC008 were examined by IHC (Fig. 6E). Similar to the findings in HCT116 xenograft, the ratio of nuclear/cytoplasmic p53 protein was decreased in bortezomib group compared with the control group (0.49 vs. 0.98, P < 0.01), whereas the combination of bortezomib and KPT330 treatment increased the ratio significantly (2.36 vs. 0.98, P < 0.01; Fig. 6F). Furthermore, the combination treatment markedly reduced Ki67 expression detected by IHC, and increased apoptosis detected by the TUNEL assay (Fig. 6F).

To further assess the efficiency of bortezomib and KPT330 combination treatment, additional primary tumor-derived xenografts from 5 colorectal cancer patients were tested (Supplementary Fig. S5C). Three of the xenografts were sensitive to bortezomib and KPT330 treatment, and two were relatively not sensitive. Among the sensitive xenografts, CRC3496 and CRC3547 contained wild-type p53, whereas CRC3405 harbored the C176Y p53 mutation. Interestingly, it has been shown that p53 with C176Y mutation was transcriptionally active in a number of experimental systems (23). For the two xenografts that were insensitive to the combination treatment, CRC6227 had the loss-of-function p53 mutantion (G245V; refs. 23–25), and CRC3612 contained a wild-type p53. As shown in Table 1, taken together, six cotreatment–responsive models possessed functional p53, two nonresponsive models harbored loss-of-function p53 mutation, whereas only one nonresponsive model (CRC3612) harbored wild-type p53.

For the purpose of comparison, 5 PDXs (CRC3496, CRC3547, CRC3405, CRC3612, CRC6227) were also treated with various therapeutics, including cetuximab, bevacizumab, cisplatin, regorafenib, sorafenib, doxorubicin, olaparib, gefitinib, everolimus, and imatinib. The tumor reduction rates of all 13 drugs were produced (Supplementary Fig. S5C). In the three PDXs (CRC3496, CRC3547, CRC3405) that are sensitive to bortezomib and KPT330, cotreatment was the first or second most effective therapy compared with other therapeutics. Notably, the two PDX not sensitive to the combination treatment (CRC6227 and CRC3612) were also resistant to cisplatin and doxorubicin, two drugs that act on DNA and are known to induce p53 activation (26, 27). Thus, it is conceivable that the tumor CRC3612 has defect in response to p53. Taken together, these findings provide a rational basis for the clinical use of this combination for the treatment of CRC patients with wild-type p53.

Despite extensive investigations and clinical trials, development of resistance to chemotherapy remains a major challenge for the treatment of colorectal cancer (28). In the effort to explore the mechanisms of the resistance and find novel strategies and targets to improve the prognosis of colorectal cancer patients, we found that proteasome inhibition induced export of ubiquitinated nuclear proteins in colorectal cancer cells, which might represent a mechanism of chemoresistance. It has been found that CRM1, the transport protein responsible for nuclear export of many major tumor suppressors and growth regulators, is upregulated in many tumors (29). Small molecules targeting CRM1 have been developed, and FDA has designated one of the inhibitor selinexor (KPT330) orphan drug status for certain types of leukemia and lymphoma (17, 18). We demonstrated in this study that inhibition of nuclear export sensitized colorectal cancer cells to the cytotoxic action of proteasome inhibitor, which led to G₂–M cell-cycle block and apoptosis.

Tumor suppressor p53 functions as a critical guardian of genome. In response to genotoxic stimuli, upregulated p53 induced G₂–M cell-cycle arrest and apoptosis (30, 31). As cellular level of p53 is mainly controlled through ubiquitination-mediated proteasomal degradation, proteasome inhibitors are known to accumulate p53 in cells (32, 33). We found in the study that proteasome inhibition induced nuclear export of p53, and bortezomib and KPT330 have synergistic cytotoxic action on colorectal cancer cell lines with wild-type p53 in vitro and in nude mice, but not the cells with mutated p53. In the sensitive cells, the combination treatment led to further increased nuclear p53 and expression of target genes. Furthermore, knockdown of p53 largely abolished the synergistic action. These results indicated that KPT330 and bortezomib together increase nuclear p53, which in turn initiates the apoptosis program in tumor cells. It is worth noting that both proteasome and CRM1 inhibitors have profound effects on many cellular processes and may kill tumor cells through a variety of mechanisms. In our study, bortezomib and KPT330 can effectively kill colorectal cancer cells containing mutated p53 (SW480 and SW620). However, their cytotoxic actions are not synergistic in these cells, suggesting the two drugs induced cell death though different pathways in these cells.

Although tumor cell line–derived xenografts have been used for decades in assessing cytotoxic action against tumor cells, they are limited in many aspects to mimic human tumors, including reduced intratumoral heterogeneity, lack of stromal cells, and modest diversity of molecular subtypes (34). Patient-derived xenografts in mice, which largely avoided these limitations and provided more accurate depictions of human tumors, have become a “gold standard” for evaluating antitumor chemotherapeutics (34, 35). It has been shown that the effect of drugs on PDXs from colorectal tumors correlated with clinical outcome (35). We generated primary tumor-derived xenografts from 9 patients. The data indicated that the combination of bortezomib and KPT330 was more effective than either drug alone or other therapeutics in inhibiting tumor growth, and the six PDXs that response to the combination therapy all contained functional p53, supporting its further clinical trial. Noteworthy, we also tested a number of therapeutics on these PDXs. The two PDXs that did not response to bortezomib and KPT330 also failed to be inhibited by the treatment of cisplatin and doxorubicin, which exerted cytotoxic effect by acting on DNA and inducing p53. (26, 27). It is conceivable that these two tumors have defects in p53 signaling pathway.

In summary, our preclinical data suggest that colorectal cancer cells could exert self-protective function upon proteasome inhibition through nuclear export of ubiquitinated proteins, including p53. CRM1 inhibitor KPT330 synergistically sensitizes colorectal cancer cells to bortezomib treatment in vitro and in vivo, through inhibiting nuclear export and restoring functions of p53. Taken together, these findings provide a rational basis for the clinical use of this combination for the treatment of colorectal cancer patients with wild-type p53.

No potential conflicts of interest were disclosed.

Conception and design: T. Wu, W. Chen, Y. Zhong, S. Fang, X. Ouyang, L. Cui, Y. Yang

Development of methodology: T. Wu, W. Chen, Y. Zhong, S. Fang, C.-Y. Liu, G. Wang, Y.-Y. Huang, X. Ouyang, H.Q.X. Li, L. Cui, Y. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Wu, W. Chen, Y. Zhong, X. Hou, G. Wang, T. Yu, Y.-Y. Huang, X. Ouyang, H.Q.X. Li, Y. Yang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Wu, W. Chen, Y. Zhong, X. Hou, C.-Y. Liu, G. Wang, T. Yu, X. Ouyang, Y. Yang

Writing, review, and/or revision of the manuscript: T. Wu, X. Hou, S. Fang, C.-Y. Liu, T. Yu, X. Ouyang, L. Cui, Y. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Chen, Y. Zhong, S. Fang, X. Ouyang, H.Q.X. Li, Y. Yang

Study supervision: S. Fang, C.-Y. Liu, X. Ouyang, H.Q.X. Li, L. Cui, Y. Yang

We acknowledge Jumei Lou for collecting and providing us the patient-derived colorectal cancer samples for PDX model establishment.

L. Cui is supported by the National Natural Science Foundation of China (grant no. 81372636). Y. Yang is supported by the National Natural Science Foundation of China (grant no. 81302089).

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.