A total of 298 cases of primary colorectal cancer were surgically resected. All tissues were collected from the Chinese PLA General Hospital according to the approved guidelines of the Chinese PLA General Hospital’s Institutional Review Board. In addition, three colorectal cancer cell lines (LoVo, HCT116, and SW620) were included in this study. All colorectal cancer cell lines were purchased from the Chinese Academy of Medical Science (CAMS) and Peking Union Medical College (PUMC), which were previously established from primary colorectal cancer. These cell lines were maintained in 90% RPMI 1640 (Invitrogen, CA, USA) supplemented with 10% fetal bovine serum. For the analysis of inducible expression of RAI2 by retinoic acid, colorectal cancer cell lines were split to low density (25% confluence) 12 h before treatment. In the detection of retinoic acid inducible effect on RAI2, Cells were treated with retinoic acid (RA) (Sigma, St. Louis, MO) at different concentrations of 0, 1, 5, 10μM in the growth medium. After that, the protein was extracted for Western blot detection of RAI2 expression.

Total RNA was extracted using Trizol reagent; cDNA was synthesized according to the manufacturer’s instructions (Invitrogen, CA, USA). Actin was used as control. The RAI2 PCR primer sequences were as follows: 5′-ATGGCAATGCCACCTACGTCATG-3′ (forward) and 5′- GATACGGCACCAAGAGGGTGGC -3′ (reverse); CtBP2 PCR primer sequences were as follows: 5′- CAAGGCCTTTGGATTCAGCGTC -3′ (forward) and 5′- AGGGCTTGTGCTAAGGCTTTCT -3′ (reverse); the Actin PCR primer sequences were as follows: 5′- CTCCATCCTGGCCTCGCTGT -3′ (forward) and 5′- GCTGTCACCTTCACCGTTCC-3′ (reverse). Each experiment was repeated three times.

IHC staining was performed on 4-μm thick serial sections derived from formaldehyde-fixed paraffin blocks. Rabbit polyclonal antibody against RAI2 (Abcam, ST. Louis, MO) was diluted at 1:30, Rabbit anti-phospho-β-catenin (Thr41) (BIOSS, Beijing, China) was diluted at 1:1200, Rabbit anti- achaete-scute complex homolog 2 (ASCL2) (BIOSS, Beijing, China) was diluted at 1:200 and Rabbit anti-Leucine-rich repeat-containing G-protein coupled receptor 5 (LGR5) (Abcam, ST. Louis, MO) was diluted at 1:200. IHC was performed and the staining was evaluated as described previously (24), and samples were divided into 2 groups: high RAI2 expression group (score>3) and low RAI2 expression group (score ≤ 3).

The expression vectors for RAI2 were subcloned into Plenti6 lentivirus expression vector, and RAI2 expression lentiviral or empty vectors were packaged using ViraPower™ Lentiviral Packaging Mix (Invitrogen, CA, USA) to infect LoVo and HCT116 cells to establish stable expression cells. The infected cells were selected by blasticidin (Invitrogen, CA, USA) at a concentration of 10 μg/ml. Lipofectamine 3000 (Invitrogen, CA, USA) was used for plasmid transfection. For transient transfection, the RAI2 CDS region was cloned into the pcDNA3.1 (+) plasmid (Era Biotech, Shanghai, China). To generate a RAI2 protein incapable of CtBP2 binding, we constructed the mutated RAI2 expression vector (RAI2-M), in which the binding sites, two conserved sequences (ALDLS) in RAI2 protein (amino acids 316–320 and 342–346 in human sequence), were mutated to ³¹⁶ ALDAA ³²⁰ and ³⁴² ALDAA ³⁴⁶ (23). All constructs were confirmed by sequencing.

Four shRNA molecules were designed to target all four transcripts of RAI2 and constructed into pGPU6/GFP/Neo vector (GenePharma, Shanghai, China). These shRNAs were then transfected into SW620 cells according to the manufacturer’s instructions. The target sequences in the RAI2 gene were as follows: shRNA-1 (5′-GCTGTGCTCCAGAATTTGTTT-3′), shRNA-2 (5′-GCCACACGGTCATTAAGATGG-3′), shRNA-3 (5′-GGGAAGAGTCCATGGGAAATG-3′), and shRNA-4 (5′-GAAATACACATGCTCCCAATC-3′). The most effective construct, shRNA-2, was selected used in the follow-up experiment.

Three cell lines were used in this experiment, including LoVo, HCT116 and SW620. And pcDNA3.1(+) expression plasmids containing either the wild-type RAI2 (labeled with RAI2) or mutated RAI2 cDNA sequence (labeled with RAI2-M) were employed in this experiment, with pcDNA3.1(+) empty vector as control (labeled with vector). LoVo or HCT116 cells were transfected with 4 μg of above RAI2/RAI2-M/vector expression plasmids using Lipofectamine 3000 (Invitrogen, CA, USA). The following experiment was performed using Pierce Co-Immunoprecipitation(Co-IP) Kit (#26149,Thermo Scientific™, MA, USA) according to the manufacture’s instruction. IgG group was used as a negative control. Extracted cell lysates before adding antibodies (input) were also used as a control for target proteins, Actin, or GAPDH detection.

Protein preparation and Western blot were performed as described previously (24). The antibodies for Western blot analysis were as follows: rabbit anti-RAI2 (#97857, Cell signaling technology, MA, USA, 1:1000), rabbit anti-phospho-β-catenin (ab27798, Abcam, Cambridge, UK, 1:500), rabbit anti-CtBP2 (ab128871, Abcam, Cambridge, UK, 1:10000), rabbit anti-β-catenin (ab32572, Abcam, Cambridge, UK, 1:5000), rabbit anti-CyclinD1 (ab134175, Abcam, Cambridge, UK, 1:10000), rabbit anti-c-Myc (ab39688, Abcam, Cambridge, UK, 1:500), rabbit anti ASCL2 (TA313480, ORIGENE, Rockville, US, 1:1000), rabbit anti LGR5 (ab199335, Abcam, Cambridge, UK, 1:1000), Anti-SOX2 (ab137385, Abcam, Cambridge, UK, 1:1000), Anti-CD133 [EPR20980-104] (ab216323, Abcam, Cambridge, UK, 1:1000), Anti-APC [EP701Y] (ab40778, Abcam, Cambridge, UK, 1:2000), and Anti-Oct4 antibody [GT486] (ab184665, Abcam, Cambridge, UK, 1:1000). And rabbit anti-GAPDH (ab181602, Abcam, Cambridge, UK, 1:10000) or anti-Actin (Cell Signaling Technology, Danvers, USA, 1:1000) was used as a control. Each experiment was repeated three times.

LoVo and HCT116 cells were seeded at 5 × 10³ cells/well in 96-well culture plates 24 h before transfection. To examine transcriptional activity driven by TCF/LEF, the above cells were transfected with 20 ng/well Topflash/Fopflash reporter vector (TCF/LEF–responsive reporter) and 2 ng/well pRL-TK. And 20ng/well pCI-neo-β-catenin-wt expressing vector, which expresses wild-type β-catenin was used to activate the reporter gene. Then, 200 ng/well of RAI2 was transfected into cells together with the Topflash/Fopflash reporter vector, pRL-TK control vector and pCI-neo-β-catenin-wt was used to evaluate the regulative effect of RAI2 on the Wnt signaling pathway. Forty-eight hours after transfection, relative luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega, Madison, USA) according to the manufacturer’s protocol. In addition, LoVo and HCT116 cells, with or without RAI2 transfected, were treated with 20 mM LiCl (Sigma, St. Louis, MO) for 36hs before luciferase activity detection, and another two groups of LoVo and HCT116 cells, with or without RAI2 transfected, were treated with 10 µM XAV939 (MedChemExpress, Shanghai, China) for 36hs before luciferase activity detection. For each experiment, the luciferase reporter assay was performed three times.

Cells were grown on glass culture slides (BD Biosciences, San Jose, USA) and fixed with 4% cold methanol at -20°C for 10 min. Subsequently, cells were blocked with 10% goat serum for 1 h and incubated with primary antibodies against β-catenin (ab32572, Abcam, Cambridge, UK, 1:250) or RAI2 (PA5-62305 Invitrogen, CA, USA, 1:500) at 4°C for 1 h and then incubated with fluorescent labeled secondary antibodies for 1h at room temperature. After being counterstained with DAPI (Invitrogen, CA, USA), the slide was observed under a confocal microscope (Zeiss).

LoVo and HCT116 cells were analyzed by fluorescence-activating cell sorting (FACS; BD bioscience). The cells were harvested using 0.05% trypsin (Gibco, CA, USA), washed twice with DPBS supplemented with 5% FBS and resuspended in DPBS supplemented with 10% FBS with mouse anti-human CD133/1 antibody (Miltenyi Biotec, Bergisch Gladbach, Germany) for 30 min at 4°C. The cells were washed twice with pre-cooled DPBS and centrifuged at 1200 rpm at 4°C and incubated in DPBS supplemented with 10% FBS with goat anti-mouse Alexa Fluor^® 488 for 30 min at 4°C in dark. After washing with DPBS twice, cells were sorted by FACS. CD133-negative and CD133-positive CRC cells were collected for further experiments, and they were cultured in RPMI 1640 supplemented with 10% FBS with 1% of penicillin/streptomycin.

In total, 1×10³ cells were seeded in 6-well ultra-low cluster plates (Corning, NY, USA) for 10 days. Spheres were cultured in RPMI 1640 serum-free medium (Invitrogen, CA, USA) supplemented with 2% B27 (Invitrogen, CA, USA), 20 ng/ml EGF, 20 ng/ml bFGF, 0.4% BSA, and 5 μg/ml insulin (Sigma, Beijing, China).

LoVo/HCT116 cells were treated with a concentration gradient of anticancer drugs before and following re-expression of RAI2. The anticancer drugs, oxaliplatin (L−OHP) (Sigma, Beijing, China) and 5−fluorouracil (5−FU) (Princeton, NJ, USA) were tested in six dilutions respectively. The concentrations of L−OHP for treatment were 0.00, 1.00, 2.00, 4.00, 8.00, 16.00μM in LoVo/HCT116 cells and 0.00, 2.00, 4.00, 8.00, 16.00, 32.00μM in SW620 cells. The concentration of 5−FU for treatment was 0.00, 6.25, 12.50, 25.00, 50.00, 100.00μM in LoVo/HCT116/SW620 cells. The cells were seeded at a density of 3x10⁴ cells/well on 96−well plates in RPMI−1640 medium, each well was repeated in triplicate. The following incubation overnight at 37˚C, the medium growth was replaced by a fresh medium containing the test drug. All the cells were treated for 48 h. At the end of the treatment, cell viability was measured by the MTT assay (KeyGEN Biotech, Nanjing, China). Absorbance was measured on a microplate reader (Thermo Multiskan MK3, MA, USA) at a wavelength of 490 nm. IC50 was defined as the concentration of L-OHP/5-FU for 50% inhibition of cell growth. Each experiment was repeated three times.

The RNA sequencing (RNA-Seq) data for RAI2 and CtBP2 gene expression in the CRC dataset were downloaded from The Cancer Genome Atlas (TCGA) (http://xena.ucsc.edu/). Statistical analysis was performed using SPSS 17.0 software. Chi-square or Fisher’s exact tests were used to evaluate the relationship between RAI2 expression status and clinical pathological characteristics. The two-tailed independent samples t-test was applied to determine the statistical significance of the differences between the two experimental groups. Spearman rank correlation test was used to determine the correlation between two groups of ordinal data. Survival rates were calculated by the Kaplan-Meier method, and the differences in survival curves were evaluated using the log-rank test. Cox proportional hazards models were fit to determine independent associations of RAI2's low expression with the 5-year OS and 5-year relapse-free survival (RFS) outcomes. Two-sided tests were used to determine the significance, and P < 0.05 was considered to be statistically significant.

RAI2 was initially identified as a retinoic acid-induced gene (21). Firstly, we confirmed the induction effect of retinoic acid (RA) on RAI2 in LoVo and HCT116, in which basal RAI2 expression levels were low before RA treatment ( Supplementary Figure 1 ). The primary RAI2 amino acid sequence has two highly orthologically conserved sequences (ALDLS) in the internal region (amino acids 316–320 and 342–346 in human sequence) of the RAI2 protein, and It was reported that RAI2 can interact with CtBP2 by ALDLS domain in breast cancer (23). In this study, to confirm the interaction between RAI2 and CtBP2 in colorectal cancer, we also employed the double mutant RAI2 with both sites were mutated to 316 ALDAA 320 and 342 ALDAA 346. The pull-down experiments demonstrated that when both sites were mutated, there was a marked decrease in coprecipitated CtBP2 in LoVo cells. To assess whether endogenous CtBP2 and RAI2 interact in CRC cells, we performed the same experiment in SW620 cells, in which basal RAI2 expression levels were high, and confirmed the interaction between RAI2 and CtBP2 ( Figure 1A ). Interestingly, in the input experiment group, we found that the expression level of CtBP2 in LoVo cells transfected with RAI2 expression vector was less than that of empty vector- or RAI2-M plasmid-transfected LoVo cells. In addition, by searching The Cancer Genome Atlas (TCGA) (http://xena.ucsc.edu/) databases and analyzing RNA sequencing (RNA-Seq) data in 434 cases of CRC samples, we found that the expression of CtBP2 was negatively correlated with the expression of RAI2 (p<0.0001) ( Figure 1B ). We further detected the expression of CtBP2 in LoVo and HCT116 cells with or without RAI2 restored by RT-PCR and found that the expression of CtBP2 was significantly decreased in LoVo and HCT116 cells with RAI2 re-expressed. And in SW620 cells, the expression of CtBP2 was significantly increased after RAI2 was knocked down ( Figure 1C ). The above results suggested that besides interacting with CtBP2, RAI2 can also down-regulate the expression of CtBP2 in colon cancer cells.

Aberrant activation of Wnt/β-catenin signaling contributes to carcinogenesis and progression of colorectal cancer. CtBPs(CtBP1/2) was determined to be transcriptional co-repressor and oncogene in human cancer (26). CtBPs could promote the activation of Wnt/β-catenin by interacting with mutated APC or directly binding to TCF/LEF (16). To clarify the effect of RAI2 on Wnt signaling, we detected the luciferase activity of Wnt/β-catenin signaling before and after RAI2 transfection in colorectal cancer cells LoVo and HCT116, respectively. As shown in Figure 2A , the increased activity induced by wild-type β-catenin was significantly suppressed by RAI2 (P=0.0021 in LoVo, P=0.0004 in HCT116). In addition, LiCl and XAV939, the agonist and antagonist of Wnt/β-catenin signaling respectively, were employed to clarify the role of RAI2 in Wnt signaling. As expected, the agonist LiCl significantly enhanced the activity of Wnt signaling induced by wild-type β-catenin with RAI2 transfected (P-value was 0.0007 and 0.0006 respectively in LoVo and HCT116 cells), While, the antagonist XAV939 significantly inhibited Wnt/β-catenin signaling activity in both LoVo (P=0.0027) and HCT116 (P=0.0004) cells, and the inhibitory effect was similar or even more significant in RAI2-transfected CRC cells (P-value was 0.0021 and 0.0005 respectively in LoVo and HCT116 cells) ( Figure 2A ). We detected the effect of RAI2 on the expression of CtBP2, β-catenin, phosphorylated β-catenin, and the Wnt target genes, c-Myc, CyclinD1, ASCL2, and LGR5, by western blot. Significant reduction of CtBP2 and increase of phosphorylated β-catenin (p-β-catenin) was found in RAI2-transfected LoVo and HCT116 cells, with reduced expression of c-Myc, CyclinD1, ASCL2, and LGR5, the Wnt target genes. Besides, we found that the effects of RAI2 on expression of p-β-catenin, c-Myc, CyclinD1, ASCL2, and LGR5 could be eliminated by LiCl addition in RAI2-tranfected LoVo and HCT116 cells. While, no significant change of CtBP2 and β-catenin expression was found in RAI2-transfected LoVo and HCT116 cells with or without LiCl added. In contrast, increased expression of CtBP2 and reduced phosphorylated β-catenin (p-β-catenin) were found in SW620 cells with RAI2 knocked down (shRAI2-transfected SW620 cells) compared with control group (shNC-transfected SW620 cells), with upregulation of c-Myc, CyclinD1, ASCL2, and LGR5. While, except CtBP2, no significant change of above other proteins was found in RAI2-knocked down SW620 cells with XAV939 addition compared with shNC-transfected SW620 cells, which revealed that the effect of RAI2 on p-β-catenin and target genes of Wnt signaling was reversed by XAV939 ( Figure 2B ). The above results suggested that RAI2 served as an antagonist of Wnt/β-catenin signaling, while, the regulation of CtBP2 by RAI2 was not through Wnt signaling and RAI2 may inhibit Wnt/β-catenin by interacting with or down-regulating CtBP2, resulting in increased degradation of β-catenin. Furthermore, the localization of β-catenin in CRC cells with or without RAI2 restored was investigated by cell immunofluorescence assay, and Mutant type RAI2 (RAI2-M), in which ALDLS sequences were destroyed, was then employed in our study. Firstly, we confirmed the expression and localization of RAI2 in RAI2/RAI2-M-transfected LoVo cells, and found RAI2 expression both in cell cytoplasm and nucleus ( Figure 3A , left panel). In LoVo cells without RAI2 restoration, β-catenin was mainly localized at the cell nucleus, which represents the activation of Wnt signaling. While, in LoVo cells with RAI2 re-expressed, β-catenin was mainly localized at cytoplasm. However, both in RAI2-M-transfected and vector control LoVo cells, β-catenin was localized at the cell nucleus ( Figure 3A , right panel). Thus, we proposed that RAI2, not RAI2-M, inhibits the translocation of β-catenin from the cytoplasm to nucleus. consistently, Western blot detection found that the effect of RAI2 on phosphorylation of β-catenin, expression of CtBP2 and Wnt target genes c-Myc, CyclinD1, ASCL2, and LGR5 were lost when ALDLS sequences of RAI2 was destroyed. Thus, we found no significant change of the above proteins in LoVo/HCT116 cells transfected with RAI2-M compared with vector control ( Figure 3B ).

In addition, we further detected the effect of RAI2 on the interaction between CtBP2 and APC by Co-IP and found that APC was pulled down by CtBP2 antibody in vector control LoVo cells with mutant type APC, which was eliminated in RAI2-transfected LoVo cells. In contrast, APC was not pulled down by CtBP2 antibody in all three groups of HCT116 cells that have wild-type APC with vector, RAI2, or RAI2-M transfected ( Figure 3C ). It revealed that the inhibition of Wnt signaling by RAI2 in CRC with mutant APC may through destruction of interaction between CtBP2 and mutant APC, while in CRC with wild-type APC, the inhibition of Wnt signaling by RAI2 may involve a different mechanism, such as transcriptional down-regulation of CtBP2, an indirect approach.

These results suggested that the up-regulation of phosphorylated β-catenin depends on conserved sequences ALDLS of RAI2, by which RAI2 interacts with CtBP2. In addition, we can see the suppression of CtBP2 by RAI2, which was not seen in RAI2-M-transfected CRC cells. This result suggested that RAI2 may inhibit Wnt signaling by promoting the phosphorylation and subsequent degradation of β-catenin, and the effect of RAI2 on Wnt signaling may depend on the interaction between RAI2 and CtBP2 or the direct down-regulation of CtBP2 expression by RAI2.

We have reported the reduced expression of RAI2 regulated by promoter region methylation in colorectal cancer in our previous study. However, by searching The Cancer Genome Atlas (TCGA) (http://xena.ucsc.edu/) databases, the RAI2 expression data obtained by RNA sequencing (RNA-Seq) showed no association between RAI2 mRNA expression and 5-year overall survival (OS)/5-year relapse-free survival (RFS) (24). In this study, we detected the expression of RAI2 in 298 cases of colorectal cancer patients by IHC. Low expression of RAI2 was found in 33.89% (101/298) of the colorectal cancer samples, and the expression of RAI2 was positively correlated with phosphorylated β-catenin (r=0.8866, P<0.0001), which is further validation of the inhibitory effect of RAI2 on Wnt/β-catenin signaling. And we further detected the expression of ASCL2 and LGR5, two main Wnt target genes in CRC, in above 298 cases of CRC patients, and found that expression of RAI2 was negatively correlated with that of ASCL2 (r=-0.1674, P=0.0037) and LGR5 (r=-0.1580, P=0.0063) ( Figure 4A ). Though no association was found between RAI2 expression and clinical characteristics such as gender, TNM stage, lymph node metastasis, age, differentiation, tumor location, and size (P >0.05) ( Table 1 ), Kaplan-Meier analysis indicated that low expression of RAI2 was significantly associated with poor 5-year RFS (P = 0.0029) and 5-year OS (P = 0.0102) ( Figure 4B ). According to univariate analysis, RAI2 expression and TNM stage were associated with both poor 5-year OS (all P < 0.05) and 5-year RFS (all P < 0.05, Table 2 ). In addition, age was associated with poor 5-year OS (P < 0.01), but not 5-year RFS (P >0.05). By multivariate analysis, RAI2 expression and TNM stage were associated with both poor 5-year OS and 5-year RFS (all P < 0.05, Table 2 ). Age was associated with poor 5-year OS (P < 0.01, Table 2 ). These results suggest that RAI2 expression was an independent prognostic marker for poor 5-year OS (P = 0.027, Table 2 ) and 5-year RFS (P = 0.022, Table 2 ).

Current findings support that the Wnt/β-catenin pathway plays a crucial role in the maintenance of stem cell-like properties and chemo-resistance in colorectal cancer (27). CD133 is a representative stem cell maker (28). In this study, we isolated CD133+ CRC cells (LoVo and HCT116) by flow cytometry ( Figure 5A , left panel). Tumor sphere formation assay showed that CD133+ cells with RAI2 re-expressed formed fewer and smaller spheres than the control group (P=0.0285 in LoVo, P=0.0048 in HCT116) ( Figure 5A , right panel), suggesting that RAI2 inhibited the self-renewal ability of CRC cells. while we found no significant difference in CD133+ cells with RAI2-M transfected compared to vector control group cells (P>0.5 in both LoVo and HCT116) ( Figure 5B ). Western blot assay showed that CD133+ LoVo and HCT116 cells with RAI2 re-expressed exhibited lower levels of stem cell makers like CD133, SOX2, and Oct4 than control LoVo and HCT116 cells ( Figure 5C ), while we found no significant difference between RAI2-M-transfected CD133+ cells and vector control group cells ( Figure 5D ). The above results revealed that re-expression of RAI2 inhibited the stem cell-like properties of CRC cells.

Chemoresistance is a hallmark of cancer stem cells (CSCs) (29). Thus, we detected the effect of RAI2 on chemosensitivity of colorectal cancer cells to oxaliplatin (L-OHP) and fluorouracil (5-FU) by MTT. We raised its level by RAI2 transfection in LoVo/HCT116 cells. To confirm the importance of the ALDLS sequence in RAI2 protein, we also performed the chemosensitivity assay in RAI2-M-transfected CRC cells, with the vector-transfected CRC cells as the control group. As shown in Figure 5E , The IC50 of 5-FU was decreased in RAI2-transfected LoVo cells (30.20 ± 6.09 μM versus 16.31 ± 2.89 μM, P=0.0234) and RAI2-transfected HCT116 cells (20.89 ± 4.23 μM versus 7.76 ± 1.66 μM, P=0.0075), while, we found no significant difference between RAI2-M-transfected CRC cells and vector control group in both LoVo and HCT116 (P>0.05). We gain similar results in L-OHP-treated cells, the IC50 of L-OHP was 7.90 ± 2.21μM versus 3.27 ± 0.37μM in LoVo cells with empty vetor and RAI2 transfected (P=0.0233), and 4.52 ± 1.25 μM versus 2.09 ± 0.56 μM (P=0.0373) in HCT116 cells with empty vetor and RAI2 transfected, while no significant difference was found in RAI2-M-transfected CRC cells controlled with vector group in both LoVo and HCT116 cells (P>0.05) ( Figure 5F ). On the contrary, reduced chemosensitivity was found in SW620 cells with RAI2 knocked down. The IC50 of 5-FU was 12.11 ± 1.90 μM versus 24.41 ± 6.51 μM (P=0.0349) in shNC- and shRAI2-transfected SW620 cells, and the IC50 of L-OHP was 4.76 ± 0.77 μM versus 8.74 ± 1.49 μM (P=0.0147) in shNC- and shRAI2-transfected SW620 cells, while, no significant difference was found between CRC cells with both shRAI2 and RAI2-M transfected and CRC cells with only shRAI2 transfected in both LoVo and HCT116 (P>0.05) ( Figure 5G ). The above results suggested that RAI2 increased the chemosensitivity of colorectal cancer cells to L-OHP and 5-FU. RAI2 may serve as a potential biomarker of chemosensitivity in colorectal cancer.