Abstract
Purpose: Cancer stem–like cells have been well accepted to be involved in recurrence and metastasis of cancers, but the prognostic potential of biomarkers integrating with metastasis and cancer stem–like cells for colorectal cancer is unclear.
Experimental Design: We identified three proteins, CLIC4, ERp29, and Smac/DIABLO, from metastatic cancer stem–like cells of colorectal cancer and verified the proteins' role in metastatic behaviors. The proteins were detected by IHC in colorectal cancer tumors and matched colonic mucosa from patients with colorectal cancer who underwent radical surgery in the training cohort. The associations between proteins expression levels and five-year disease-specific survival (DSS) were evaluated to predict the survival probability in the training cohort of 421 cases and the validation cohort of 228 cases.
Results: A three-protein panel including CLIC4, ERp29, and Smac/DIABLO, which was generated from multivariate analysis by excluding clinicopathologic characteristics from the training cohort, distinguished patients with colorectal cancer into very low-, low-, middle-, and high-risk groups with significant differences in five-year DSS probability (88.6%, 63.3%, 30.4%, 11.4%; P < 0.001). The panel is independent from tumor–node–metastasis staging system and histologic grading to predict prognosis, and also enables classification of validation cohort into four risk stratifications (five-year DSS probability is 98.2%, 80.2%, 25.6%, and 2.7%; P < 0.001).
Conclusions: CLIC4, ERp29, and Smac/DIABLO integrated into a novel panel based on cancer stem–like cells in association with metastasis stratify the prognostic risks of colorectal cancer. Prediction of risks with molecular markers will benefit clinicians to make decisions of individual management with postoperative colorectal cancer patients. Clin Cancer Res; 20(14); 3809–17. ©2014 AACR.
Translational Relevance
Colorectal cancer is the third leading cause of death of malignancies in China and worldwide, especially in the developed countries. The major importance of this work is that we have identified a novel panel of three proteins associated with metastatic behaviors of cancer stem–like cells of colorectal cancer. The three-protein signature with IHC examination of primary tumors to predict prognosis is independent from and better than tumor–node–metastasis staging system and it stratifies patients with colorectal cancer into four grade risks of prognosis after radical surgery. In the very-low risk group, the probability of patients surviving more than 5 years was 88.6% in the training cohort and 98.2% in the validation cohort, respectively. Therefore, these findings could provide a tool for clinicians to make decisions of individual therapy and for identifying patients with colorectal cancer in low-, middle-, and high-risk groups that will benefit from adjuvant chemotherapy following surgical resection.
Introduction
Colorectal cancer is the third most common type of cancer and killed more than 600,000 people every year mainly owing to metastasis despite aggressive surgery, postoperative chemotherapy, and biotherapy worldwide (1, 2). In the past decade, gene expression signature and proteins panel provided prognostic value and set up prediction of local recurrence, distant metastases, and chemotherapy-associated assessments in several solid tumors (3–7). KRAS mutation screening allows for the optimal incorporation of anti-EGFR monoclonal antibody therapy to inhibit metastasis of colorectal cancer (8), and personalization of colorectal cancer screening would be important for treatment and family members surveillance just as biomarkers in Lynch syndrome (9, 10). Recent evidence has demonstrated that cancer stem cells (CSC) play a decisive role in metastases and resistance to radiochemotherapy (11–14). Therefore, identification of novel targets by IHC examination of surgical samples is urgent to predict outcome of patients with colorectal cancer and improve the clinical management.
The tumor–node–metastasis (TNM) staging is the most effective standard for prediction of outcome of patients with colorectal cancer after surgical resection (15). Postoperative adjuvant chemotherapy has been shown to improve the survival in particular in stage II/III colorectal cancer patients (16–18). Optimal use of adjuvant chemotherapy is a challenging issue as there are few established clinical criteria to separate patients with good prognosis from poor prognosis. However, molecular markers with independent prognostic value have demonstrated more accurate prediction of prognosis than TNM staging alone in patients with colorectal cancer (19). Thus, CSCs markers associated with metastasis might be another factor to identify the patients with poor prognosis who may need to receive adjuvant chemotherapy for potential benefits.
However, the precise molecules of CSCs involved in the prognostic value remain unclear. Previous studies have identified CD133 as an effective biomarker for colorectal stem cells (20–22), but both CD133+ and CD133− metastatic colorectal cancer cells initiate tumor development (23). There is not a consensus for the “best marker” to identify particular tumor-initiation stem cells; EpCAM, CD44, CD166, and ALDH1 identify colorectal cancer stem cells (22, 24–26), meanwhile, CD44, CD133, and ALDH1 are used to identify breast CSCs (27–30). Well-accepted concept of CSCs indicates that distant metastases could be resulted from migrating CSCs (31), but there are no biomarkers associated with metastasis derived from migrating CSCs to predict the prognoses of colorectal cancer. Therefore, metastatic colorectal cancer stem cells may be capable of tumorigenesis and metastases resulting from their biomarkers correlating with poor prognosis.
In this study, we sought to explore molecular markers associated with CSCs to stratify the prognostic risks of colorectal cancer, three proteins, chloride intracellular channel 4 (CLIC4), endoplasmic reticulum protein 29 (ERp29), and second mitochondria-derived activator of caspases (Smac/DIABLO) were identified from cancer stem–like cells with metastasis of SW480 by proteomic analyses. Colorectal cancer stem–like cells were identified from CD133+ single cell-derived progenies (SCP) of SW480 with different metastatic potentials, respectively, as described in our previous research (ref. 32; metastatic SCP17 and nonmetastatic SCP40). The possible roles of the three proteins relevant to metastasis and stemness were confirmed in vitro and in vivo. CLIC4, ERp29, and Smac have been studied in renal, ovarian, bladder prostate, and breast cancer (33–37), but their prediction value in colorectal cancer has not been known yet. For analysis of clinical tissues in 649 colorectal cancer cases (421 cases of training cohort vs. 228 cases of validating cohort), we described here that a novel panel of three proteins stratified patients with colorectal cancer into four grade risks of prognosis after radical surgery.
Patients and Methods
Additional data are available in the Supplementary Materials and Methods section.
Results
Additional data are available in the Supplementary Results section (Supplementary Fig. S1–S7).
Patient characteristics
Patients' median age at surgery was 57.45 years (range, 19–86 years), 54.2% of the patients were in stage I/II, and 45.8% of them were in stage III/IV. The median survival time was 56.05 months (range, 1–128 months) in this study, with a median follow-up of 62 months (range, 6–128 months) for all the cases, and a median disease-specific survival (DSS) of 61 months. All the deaths of patients with colorectal cancer were confirmed by recurrences or metastases and their associated complications. Table 1 listed the clinical and pathologic characteristics of the patients with colorectal cancer in the training and validation cohorts. The incidence of advanced colorectal cancer patients might be attributable to different socioeconomic circumstances (e.g., lifestyle, food, and medical cost) between the two cohorts in China (Training cohort: Pearl Triangle area of Guandong Province, a coastal developed district of China; Validation cohort: Chongqing, southwest area of China, a less developed district). For instance, the less developed district (north area of Guangdong Province) had a higher percentage of advanced colorectal cancer patients than the well-developed district (Pearl Triangle area of Guangdong Province; ref. 38).
Clinicopathologic characteristics of patients with colorectal cancer in the training and validation cohorts
Obtaining three proteins and validating their roles in metastasis
Isolating and identifying of cancer stem–like cells in association with different metastatic potentials are described in the Supplemental Data (Supplementary Fig. S1–S3). 2D-gel portraits revealed 27 protein spots among Sph17 (the tumor sphere of metastatic cancer stem–like cells), Sph40 (the tumor sphere of nonmetastatic cancer stem–like cells), and SW480 cells. We focused on the 27 protein spots that showed significant variations more than twice among Sph17, Sph40, and SW480. The proteins name and their possible function are shown in Supplementary Table S1. These proteins were correlated with metastasis and underwent gene ontology using an online analysis (http://www.geneontology.org/; Supplementary Fig. S4). Finally, we confirmed three proteins as candidates for further validation and functional studies. ERp29, CLIC4, and Smac were highly expressed in Sph17, as shown in the enlarged images (Fig. 1A). Meanwhile, their parental cells showed low levels of ERp29, CLIC4, and Smac expression (Fig. 1B). The real-time RT-PCR quantitative analysis of genes abundance was consistent with the intensity of the proteins observed on the 2D-gel electrophoresis images (Fig. 1C). Western blotting assay displayed stronger ERp29, CLIC4, and Smac in fresh tumors than in their paired normal mucosa tissues in eight cases of colorectal cancer (Fig. 1D). Confocal microscopic observations revealed that the corresponding proteins expressions were consistent with the 2D-gel results (Fig. 1E–G). Thus, these data indicate that ERp29, CLIC4, and Smac are a potential signature that endows metastatic potentials derived from the cancer stem–like cells of colorectal cancer (Supplementary Fig. S5 and S6).
Screening out and identifying four proteins associated with the metastasis of cancer stem–like cells. (A), three proteins (CLIC4, ERp29, and Smac) enlarged 2D-gel images with different expression among metastatic cancer stem–like cells (Sph17), nonmetastatic cancer stem–like cells (Sph40), and their parental cells SW480. B, Western blotting displayed expression of the three proteins in Sph17, Sph40, and their parental cells [SW480, single cell-derived progenies (SCP17, and SCP40)]. C, real-time PCR amplification and quantitative analysis revealed that the abundance of CLIC4, ERp29, and Smac were consistent with Western blotting results. D, Western blotting of CLIC4, ERp29, and Smac in fresh tumor samples (T1-T8) and paired normal mucosa (N1-N8) in eight cases of colorectal cancer. Fluorescent images of CLIC4 (E), ERp29 (F), and Smac (G) in Sph17, Sph40, and SW480.
Optimizing the components associated with unfavorable prognosis
CLIC4, ERp29, and Smac were stained with a percentage of immunoreactivity in cancer cells of 421 patients in the training cohort at 67.2%, 64.6%, and 68.5%, and the percentages of their matched mucosa were 2.1%, 2.4%, and 5.5%, respectively. Representative images of immunostaining with positivity and intensity of the three proteins in tumor cells and normal colonic glands were shown in Supplementary Fig. S7. Only a few glandular cells in some of the cases were positive for CLIC4, ERp29, and Smac in the crypts of glands, which indicates the location of intestinal stem cells (39, 40). The expression of these three proteins in cancer cells with colorectal cancer was considered to be related to cancer stem–like cells distribution and prone to higher expression in cancer cells than normal mucosa (P < 0.001).
In quantitative analysis, ROC curve was generated for the sensitivity and specificity as a predictor of death after obtaining the accumulated points of each biomarker in every case of the training cohort. Figure 2A illustrates the plot with the maximum AUC for each protein of its accumulated points and the cutoff values for dichotomizing each predictor. AUC for CLIC4, ERp29, and Smac is more than 0.7 with significance of dichotomizing (P < .001, respectively). Then, the cutoff values are set with respect to low values for favorable prognosis, and they are also used to dichotomize the validation cohort.
(A), ROC curve generates the maximum AUC for assessment of sensitivity and specificity, and provides the cutoff values of dichotomizing CLIC4, ERp29, and Smac to predict prognosis of colorectal cancer survival status. (B), integrating the three proteins (CLIC4, ERp29, and Smac) into a panel, ROC curve generates the maximum AUC to a value of 0.808. Kaplan–Meier survival analysis estimates for very-low-, low-, middle- and high-risk (corresponding to the first, second, third and fourth line both in the graph C and D) patients with colorectal cancer as defined by the three-protein (CLIC4, ERp29, and Smac) panel of IHC scoring system. DSS curves of evaluated patients in training (C) and validation (D) cohorts. Log-rank test used to calculate P values.
Cox proportional hazards regression modeling was analyzed to reveal the univariate associations of clinicopathological characteristics, including age, sex, tumor location, histopathological grade, TNM stage, and lymph node involvement, and the expression of the three proteins with 5-year DSS in all 421 patients with colorectal cancer from the training cohort as well (Table 2). It showed that lymph node metastasis, advanced TNM stage (stage III and IV), and poor-differentiated tumors (grade 3) correlate to poor prognosis (P < 0.05). Patients' age, sex, and tumor location did not influence on the outcomes of colorectal cancer. CLIC4, ERp29, and Smac were associated with the unfavorable prognosis of colorectal cancer, respectively.
Patient characteristics, immunomarker expression, and univariate HRs for associations with 5-year DSS among patients from training cohort
Multivariate association of the three proteins and clinicopathologic features was performed using Cox proportional hazards regression analysis again, which optimized the components associated with 5-year DSS of the patients in the training cohort. The factors with P significance value less than 0.05 by univariate association analysis were enrolled into multivariate association analysis. Table 3 indicates that CLIC4, ERp29, and Smac correlate to the patients outcome according to the multivariate association analysis (P < 0.001), although histopathology grades, tumor stages, and lymph nodes involvement had a HR more than 1.0, and without statistical significances (P > 0.05). It seems that CLIC4, ERp29, and Smac play a prominent role and more than TNM staging for prognostic prediction. Thus, according to the multivariate association analysis in optimizing the prognostic factors, histopathology grades, tumor stages, and lymph nodes involvement should be ruled out in the subsequent analysis, so that we can make an effort to scrutinize the influence of survival assessment. The three biomarkers were optimized to be a panel in correlation analysis with colorectal cancer prognosis.
Multivariate HRs for associations with 5-year DSS among patients from training cohort
Three-protein panel classifies prognostic risks grade
By integrating the three proteins into a panel, another ROC curve depicted the maximal sensitivity and specificity to generate the AUC of 0.808 (Fig. 2B, P < 0.001). As predicted prognostic survival of the training cohort, the probability of the patients surviving for more than 5 years is 88.6% at the subgroup of very-low risk, the low-risk subgroup is 63.3%, the middle-risk subgroup is 30.4%, and the high-risk subgroup decreases to 11.4% (P < 0.001, Table 4). By setting the predicted very low-risk subgroup as reference, the low-, middle-, and high-risk subgroups were endowed with powerful prediction of 5-year DSS ratio at HRs of 4.56 [95% confidence interval (CI), 2.40–8.66], 10.96 (95% CI, 6.01–20.01), and 20.51 (95% CI, 11.05–38.05), respectively. Therefore, we can designate a given patient with colorectal cancer into a corresponding group as the detection points of the three-protein panel as 0 (very-low risk), 1 (low risk), 2 (middle risk), and 3 (high risk) after IHC assessment in surgical sample.
Accumulative points of IHC by dichotomizing cutoff values to predict the risks of patients with colorectal cancer in the training cohort
Kaplan–Meier curves estimate a tendency of survival ratio with the three-protein panel expression classifiers for very low-, low-, middle-, and high-risk patients with colorectal cancer defined as multivariate associations in the training cohort (Fig. 2C), and the resembling curves confirm the coincidence in the validation cohort (Fig. 2D). Overall comparisons by log-rank test showed statistical significance among the four stratifications of risks both in the training and validation cohorts (P < 0.001).
Discussion
Recurrence and metastasis mainly account for the cause of colorectal cancer death (2). CSCs theory proposes an opinion to interpret tumor metastasis with two underlying requirements: (i) cancer cells that reside at target organs or circulation should be capable of self-renewal enough to establish secondary foci and (ii) in which is the subtype of cancer cells with migration capacities (12, 14, 41, 42). Therefore, we sought to develop a novel panel combining with self-renewal and metastatic characteristics to predict prognosis with clinically heterogeneous outcomes even with the same clinical stage. Our results indicate that the three-protein panel can better categorize patients with colorectal cancer into very low-, low-, middle-, and high-risk groups with large differences in 5-year DSS rates. This panel is better than and independent of tumor stage, lymph node involvement, histologic grade, and patient characteristics.
Because of metastatic cancer stem–like cells with inherited molecules for dissemination, it is important that prognostic biomarkers of metastasis can identify patients with good prognosis who may not require further adjuvant treatment. However, until now, the molecules of metastatic cancer stem–like cells with colorectal cancer remained unclear. Here, we have identified a novel proteins panel that provides new tools for making optimal clinical decisions enabling clinicians to identify very low-risk patients with colorectal cancer for mild treatment, and they might not need to receive adjuvant therapy. In contrast, low-, middle-, and high-risk patients with colorectal cancer, as classified by this panel, may benefit from adjuvant chemotherapy and/or molecular target therapy. These patients might be also required to receive intensive surveillance of metastasis and recurrence. Thus, the three-protein panel provides clinicians with a valid and reliable tool for prediction of colorectal cancer prognosis.
Many other prognostic signatures based on gene expression profiles and immunobiomarkers have previously been demonstrated to distinguish prognosis between favorable and unfavorable subgroups for solid tumors (3–7, 43, 44). But there has been no available molecular signature mainly involving in CSCs for prediction of colorectal cancer. Three to four immunobiomarkers are applicable to clinical practice and cost-effective laboratory examination avoiding uncertain reproducibility of complicated molecular biology methods with mRNA management, especially in developing countries. Therefore, the three-protein panel in this study, which has been verified to be related with metastatic behavior for IHC examination, might be more readily adaptable to clinical assessments.
In addition, the published data indicated that chemoresistance and dissemination of cancer stem–like cells were associated with epithelial-to-mesenchymal transition (EMT) properties (45–47). In this study, the metastatic cancer stem–like cells also expressed stronger mesenchymal proteins with regard to EMT than the nonmetastatic cancer stem–like cells. High expressions of CLIC4, ERp29, and Smac were the innate factors contributing to the aggressive abilities of the metastatic cancer stem–like cells of colorectal cancer, because up or downregulation of CLIC4, ERp29, or Smac in the metastatic cancer stem–like cells or the nonmetastatic cancer stem–like cells demonstrated reciprocal invasiveness and metastasis in vitro and in vivo. By optimizing components with biomarkers and clinical pathologic characteristics in univariate and multivariate association analysis, the three proteins were an effective panel to predict colorectal cancer prognosis rather than TNM staging and histologic grades for both of the training and validation cohorts.
In conclusion, this study reveals that CLIC4, ERp29, and Smac coupling with metastatic phenotype for IHC examination can accurately distinguish patients with colorectal cancer with substantially different clinical outcomes dealing with personalized therapies. However, poly-molecule–targeted treatments are necessary to explore antimetastatic therapies and will provide important clues aimed at cancer stem–like cells in the future.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: Y.-J. Deng, G.-Q. Li, X.-W. Bian, W.-G. Fang, Y.-Q. Ding
Development of methodology: Y.-J. Deng, N. Tang, C. Liu, J.-Y. Zhang, Y.-L. Peng, L.-L. Ma, G.-Q. Li, Q. Jiang, C.-T. Hu, Y.-Q. Ding
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y.-J. Deng, N. Tang, C. Liu, J.-Y. Zhang, Y.-L. Peng, L.-L. Ma, G.-Q. Li, Q. Jiang, C.-T. Hu, Y.-Z. Liang, W.-G. Fang, Y.-Q. Ding
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y.-J. Deng, N. Tang, C. Liu, J.-Y. Zhang, S.-L. An, Y.-L. Peng, L.-L. Ma, G.-Q. Li, Q. Jiang, C.-T. Hu, X.-W. Bian, Y.-Q. Ding
Writing, review, and/or revision of the manuscript: Y.-J. Deng, N. Tang, C. Liu, J.-Y. Zhang, Y.-L. Peng, L.-L. Ma, Q. Jiang, C.-T. Hu, X.-W. Bian, Y.-Q. Ding
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.-J. Deng, N. Tang, Y.-N. Wang, W.-G. Fang, Y.-Q. Ding
Study: Y.-J. Deng, N. Tang, G.-Q. Li, X.-W. Bian, W.-G. Fang, Y.-Q. Ding
Grant Support
This work was supported by the National Basic Research Program of China (973 Program, No. 2010CB529402 and 2010CB529403), the State Key Program of the National Natural Science Foundation of China (U1201226), and the National Nature Science Foundation of China (Grants 81172381, 81372584, 81090422, 81071735, and 81201970), and also funded by the Science and Technology Innovation Foundation of Guangdong Higher Education (CXZD1016), and Key Program of National Natural Science Foundation of Guangdong Province, China (2010B031500012).
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.
Acknowledgments
The authors thank Dr. Bharath Ramachanadran (Faculty in Department of Pathology, Southern Medical University) for correcting the article.
Footnotes
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
- Received July 9, 2013.
- Revision received April 5, 2014.
- Accepted May 3, 2014.
- ©2014 American Association for Cancer Research.
References
Volume 20, Issue 14
