RNA sequencing (RNA-seq) profiles, mutation data, and clinical data for CRC patients were obtained from The Cancer Genome Atlas (TCGA), comprising 51 normal and 650 tumor samples. Two external CRC datasets and corresponding clinicopathological data were also sourced from the Gene Expression Omnibus (GEO) as validation cohorts (GSE17536 and GSE39582). Patients with less than one month of follow-up information were excluded. A total of 14 malignant tumor features were included for the study of CRC patient prognosis. Supplementary Table 1 integrates the cancer-associated functional status of all gene sets derived from the CancerSEA database [19].

Pathway enrichment analysis was conducted on a total of 1575 genes linked to cancer. Heatmaps were used to visualize the enrichment pathways, and interactive network diagrams were created based on the Metascape database. Next, single nucleotide variation (SNV) and copy number variation (CNV) in these 1575 genes were analyzed using the Gene Set Cancer Analysis (GSCA) database. In addition, genes with a high mutation frequency were visualized utilizing the “maftools” R package (v4.1.3). To investigate underlying molecular mechanisms, differentially expressed genes (DEGs) between high- and low-risk patient groups were analyzed using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG).

Cancer functional status-related genes (CFSRGs) with differential expression between para-cancerous and CRC tissues were identified utilizing the “limma” package (v3.19) (False Discovery Rate 0.05 and Fold-Change $\ge$2). Subsequently, the prognostic significance of these differentially expressed CFSRGs was evaluated by univariate Cox regression analysis. To resolve the paradox of multicollinearity, the least absolute shrinkage and selection operator (LASSO) algorithm and multivariate Cox regression analysis were employed to develop a prognostic panel based on the CFSRGs. The risk score for each CRC patient was then calculated using a formula that takes into account the expression and coefficient of CFSRGs. Patients were subsequently categorized as either low- or high-risk (better or worse outcome, respectively) according to the median score. Prognostic value was assessed by Kaplan-Meier (KM) analysis and time-dependent receiver-operating characteristic (ROC) analysis.

Two independent GEO cohorts (nGSE17536, n = 177; nGSE39582, n = 573) were employed to validate the clinical value of the CFSRG panel in CRC patients. The R package was utilized to filter and standardize the chip data before evaluating prognostic performance. KM and area under the curve (AUC) analyses were subsequently performed for each cohort.

Associations between the risk score and different clinical parameters were evaluated, including age, sex, and tumor stage. Independent prognostic markers for CRC were identified by multivariate Cox regression analysis. The accuracy of the CFSRG-based prognostic panel for predicting the outcome of CRC patients was subsequently evaluated using a nomogram, and tested using a calibration curve. ROC curves were used to assess its predictive power on the training, test, and combined cohorts. Differences in clinical characteristics between high- and low-risk groups were visualized using heatmaps and KM survival curves.

Seven algorithms (QUANTISEQ, TIMER, EPIC, XCELL, MCPCOUNTER, CIBERSORT-ABS, and CIBERSORT) were employed to study the associations between immune cell infiltration and risk scores. Infiltrating immune and stromal cells are known to play vital roles in tumorigenesis. Based on gene expression values, we computed the immune, stromal, and tumor purity scores of CRC patients by estimating the ratio of stromal and immune cells [20]. Single sample Gene Set Enrichment Analysis (ssGSEA) was then used to calculate the proportion of immune cells and the immune response in different samples within a given data set. Gene Set Enrichment Analysis (GSEA) specifically explores the major signaling pathways in patient samples using the “GSVA” R package (v4.1.3).

Since somatic mutations are strongly associated with cancer, the mutation status of the top 15 genes in patients classified as high- or low-risk was initially assessed by employing the “maftools” package (v4.1.3). Subsequently, the R software packages “survival” (v3.6.4) and “survminer” (v0.4.9) were employed to ascertain the optimal TMB threshold and to perform KM survival analysis in combination with the risk scores. Subsequently, CRC patients were classified into distinct immune subtypes, including wound healing (C1), Interferon gamma Antibody (IFN-$\gamma$) dominant (C2), inflammatory (C3), lymphocyte-depleted (C4), immunologically quiet (C5), and Transforming growth factor beta (TGF-$\beta$)-dominant (C6) [21]. These subtypes were then examined in relation to the risk scores. The Tumor Immune Dysfunction and Exclusion (TIDE) score for the different risk groups was compared using the Wilcoxon test. Finally, low- and high-risk groups were compared in terms of their expression of immune checkpoints.

The “Prophetic” R package (v4.1.3) was applied according to cancer drug sensitivity genomics (GDSC) to predict how CRC patients respond to common chemotherapy drugs. The Wilcoxon test was performed to compare the half-maximal inhibitory concentration (IC${}_{50}$) between the two risk groups.

CRC cell lines (SW480, HCT-116, LOVO) and a normal colon epithelial cell line (NCM460) were purchased from Wuhan Procell Life Science and Technology. Dulbecco’s Modified Eagle Medium (DMEM; Gibco, Thermo Fisher Scientific, Waltham, MA, USA) containing 10% fetal bovine serum (FBS; Gibco, Thermo Fisher Scientific, Waltham, MA, USA) was used to culture the cells in a humidified incubator at 37 °C with 5% CO${}_2$ atmosphere. All cell lines were validated by short tandem repeat (STR) profiling and tested negative for mycoplasma.

The proliferation status of CRC cells was quantified with a Cell Counting Kit-8 (CCK-8; Gibco, Thermo Fisher Scientific, Waltham, MA, USA). Transfected LOVO and SW480 cells were seeded into 96-well plates (1 $\times$ 10${}^3$ cells per well) and cultured under normal conditions. Cell proliferation was assessed after 24, 48, and 72 h by adding 10 µL of CCK-8 solution and incubating for 4 h. A microplate reader was employed to measure absorbance at a wavelength of 450 nm, which was then used to estimate cell proliferation.

Transfected cells were cultured in 6-well plates until confluence was reached. An incision along a straight line on the bottom of the culture dish was performed with a 200 µL pipette tip. Floating cells were removed and closure of the wound was measured at 0 h and 48 h.

Non-matrigel-coated Transwells were employed to evaluate the migration ability of cells. Serum-free culture medium was introduced into the upper chamber of Transwells containing transiently transfected cells. Complete medium containing 20% FBS was then introduced into the culture dish. Following a 24 h incubation period, the upper chamber cells were removed, whereas the lower layer cells were fixed, stained, and counted. The invasion assay used Transwells coated with Matrigel, and all other steps were the same as for the cell migration experiment.

RNA-plus reagent (Thermo Fisher Scientific, Waltham, MA, USA) was utilized to isolate total RNA from cells. qRT-PCR was conducted according to the protocol provided with the ReverTra Ace qPCR RT Kit (Toyobo, Shanghai, China). Complementary DNA (cDNA) was synthesized using oligonucleotide primers. The expression of STC2 was evaluated by qRT-PCR using 2 $\times$ S6 Universal SYBR qPCR Mix (EnzyArtisan, Shanghai, China), with the sequence of each primer shown in Supplementary Table 2.

Tissue samples were obtained from CRC patients who underwent surgery at the Fourth Affiliated Hospital, Harbin Medical University. Sample collection for this study was approved by the Research Ethics Committee of the Fourth Affiliated Hospital, Harbin Medical University (No. 2023-LLSC-04). All patients have given informed written consent.

Tissue slices were dewaxed and then subjected to a 30 minute treatment at 100 °C with 10 mM sodium citrate buffer (Yeasen Biotechnology, Shanghai, China) (pH 6.0) to unmask antigens. The sections were then blocked with normal serum and subsequently incubated overnight at 4 °C with primary antibody against STC2 (1:100, Proteintech, Wuhan, China). They were then washed and incubated with a secondary antibody (Zsgb-Bio, Beijing, China). Signal amplification was performed using SignalUp™ Immunostaining Enhancer as directed by the manufacturer (Beyotime, Shanghai, China). Images showing positive staining (brown color) were recorded with an optical microscope (Nikon, Tokyo Metropolis, Japan).

siRNA-mediated silencing of STC2 was achieved by transfecting cells with synthetic siRNA duplex molecules for 24–96 h. The siRNA sequences are listed in Supplementary Table 2. The negative control was a siRNA with no sequence homology to mammalian STC2 (GenePharma, Shanghai, China).

Statistical analyses were conducted using R 4.1.3 (https://www.r-project.org/, Boston, MA, USA), GraphPad Prism 9.4 (v9.4.0, GraphPad Software Inc., San Diego, CA, USA), and SPSS 26.0 (R26.0.0.0, IBM-SPSS Statistics, Chicago, IL, USA). KM survival analysis and log-rank tests were performed to compare the survival probability between different risk groups. Spearman’s rank correlation test was used to examine correlations between CFSRGs and the abundance of immune cells. A two-sided p-value of 0.05 was used as the criterion for statistical significance, unless otherwise stated.

The molecular mechanisms involving 1575 CFSRGs were studied using a heatmap to display the top 20 enriched signaling pathways. The CFSRGs participate mainly in cell proliferation (GO:0008283), DNA repair (R-HSA-73894), cell cycle (R-HSA-1640170), and epithelial cell differentiation (GO:0030855) (Supplementary Fig. 1A). Interactive network maps revealed connections between various pathways, such as DNA repair being closely linked to the cell cycle (Supplementary Fig. 1B). Missense mutations were shown to be the most common type of somatic mutation in CRC patients from the TCGA database (Supplementary Fig. 1C). Single-nucleotide polymorphism (SNP) variants also occurred more frequently than deletions (DEL) or insertions (INS), with C $>$ T being the major subclass of SNP (Supplementary Fig. 1D,E). Histograms were used to show the number and classification of mutations per sample (Supplementary Fig. 1F,G). Lastly, we evaluated specific mutated genes in CRC samples. The 10 most frequently mutated genes are shown in Supplementary Fig. 1H, with the highest frequency being 75% for Adenomatous Polyposis Coli (APC).

The workflow diagram for this study is shown in Fig. 1. Overall, 2632 DEGs were identified between adjacent normal tissues and CRC, of which 1361 were upregulated and 1271 were downregulated (Fig. 2A and Supplementary Table 3). Of these 2632 DEGs, 360 were CFSRGs (Fig. 2B). Ten CFSRGs with substantial prognostic value were identified by univariate analysis, of which 6 were associated with better patient outcome (EPHB2, CDC25C, CCNF, EREG, CXCL11, and MMP3) and 4 with worse outcome (GPM6A, CCBE1, TIMP1, and STC2) (Fig. 2C). Hub CFSRGs were identified utilizing LASSO-Cox regression analysis, with a graph used to show the correlation between the log ($\lambda$) and the partial likelihood deviation. Subsequently, an optimal panel was selected as the final prognostic model for CRC patients (Fig. 2D,E). The risk score for each CRC patient was calculated as follows: Risk Score = EPHB2*(–0.3667) + CCNF*(–0.3583) + EREG*(–0.1336) + TIMP1*(0.4388) + CXCL11*(–0.1602) + STC2*(0.3784) + MMP3*(–0.1270). A bar graph showing the coefficient for each CFSRG included in construction of the model is shown in Fig. 2F.

Fig. 1.

Workflow diagram for this study. CRC, Colorectal cancer; TCGA, The Cancer Genome Atlas; GEO, Gene Expression Omnibus; FC, Fold Change; STC2, Stanniocalcin 2; qRT-PCR, Real-Time Reverse Transcription Polymerase Chain Reaction; KM, Kaplan-Meier; ROC, Receiver operating characteristic analysis; AUC, Area Under Curve; CCK-8, Cell Counting Kit-8.

Fig. 2.

Differential expression analysis and construction of a prognostic panel. (A) Differentially expressed genes between normal and colorectal cancer (CRC) tissues are displayed by the Volcano plot. (B) Venn diagram showing the differentially expressed cancer function status-related genes (CFSRGs). (C) Univariate Cox regression analysis of differential CFSRGs linked to prognosis, with p 0.05 indicating statistical significance. (D) Specific $\lambda$ selection via 10-fold cross-validation. (E) The value of the $\lambda$ parameter represents the coefficients of the chosen features. The $\lambda$ value and independent variable coefficients are respectively represented along the abscissa and ordinate. (F) The bar graph represents the coefficients of the seven hub CFSRGs.

The expression and risk coefficients for each CFSRG were integrated to create a prognostic signature. Additionally, the median risk score was used to categorize each CRC patient into low- and high-risk groups, corresponding to better and worse outcomes, respectively. The TCGA dataset was evenly divided into two groups, namely the training and test cohorts. Fig. 3A–F shows the scatter distributions for the training, test, and overall cohorts in terms of survival time and survival status, as well as correlations with the risk score. As the risk score increased, a marked reduction was observed in patient survival. In all study cohorts, KM analysis revealed unfavorable survival for high-risk patients (Fig. 3G–I). High-risk patients also had worse progression-free survival (PFS) compared to low-risk patients (Fig. 3J–L). Time-dependent ROC curves were generated for the training, test, and overall cohorts. For the training cohort, the corresponding AUCs at 1-, 3-, and 5-years were 0.742, 0.762, and 0.785, respectively (Fig. 4A), for the test cohort they were 0.686, 0.703, and 0.662, (Fig. 4B), and for the overall cohort they were 0.715, 0.733, and 0.728 (Fig. 4C).

Fig. 3.

Evaluation of the risk panel in the The Cancer Genome Atlas (TCGA) cohort. (A–C) Risk score and survival status distribution across the training, test, and overall cohorts. (D–F) The survival time for different samples is represented by a scatter plot corresponding to their risk scores. (G–I) Kaplan-Meier (KM) curves that compare the overall survival (OS) of the training, test, and overall cohorts between groups with high and low risk. (J–L) KM curves comparing disease-free survival (DFS) of the training, test, and overall cohorts between groups with high and low risk.

Fig. 4.

Evaluation of independent prognostic value and construction of nomograms. Time-dependent receiver operating characteristic (ROC) curves for the prediction of 1-, 3- and 5-year OS in the training (A), test (B), and overall cohorts (C). (D) Clinicopathological parameters and risk scores are subjected to univariate Cox analysis. (E) Clinicopathological parameters and risk scores are subjected to multivariate Cox analysis. (F) Area under the curve (AUC) comparison between the risk score and other clinicopathological features. (G) Nomogram calibration curves were utilized to predict OS at one, three, and five years. (H) Survival probability predictions at 1, 3, and 5 years for CRC patients using a nomogram that combines clinicopathological parameters with risk scores. * p 0.05, *** p 0.001.

Additional analyses were performed to assess the clinical relevance of the risk score for CRC patients. The risk score, TNM stage, and age all showed significant correlations with patient survival in univariate analysis (p 0.001; Fig. 4D). Comprehensive clinicopathological data for the TCGA cohort is displayed in Supplementary Table 4. Multivariate analysis showed that addition of other variables to the model did not reduce the independent prognostic value of the risk score (Fig. 4E). Moreover, the panel outperformed other clinical parameters for the prediction of CRC patient outcome according to the AUC results (Fig. 4F). In the next step, multiple clinical parameters were incorporated and nomograms were developed to enhance the accuracy of prognosis. This involved calculating individual scores for each variable and utilizing the risk score as a prognostic indicator. The observed and projected values showed a satisfactory correlation on the calibration curves, as depicted in Fig. 4G,H.

Risk score heatmaps that incorporated clinical variables and the various T, M, and tumor stages were generated for the two risk groups (Supplementary Fig. 2A). With all clinical features, KM analysis showed that high-risk patients have reduced survival compared to low-risk patients. This included age ($>$65 years, Supplementary Fig. 2B), gender (female and male, Supplementary Fig. 2C,D), M0 (Supplementary Fig. 2E), N (N0 and N1–3, Supplementary Fig. 2F,G), T (T3–T4, Supplementary Fig. 2H) and stage (stages I–II and stages III–IV, Supplementary Fig. 2I,J). Results from previous analyses were largely consistent with these findings.

Two independent and external GEO cohorts (nGSE17536, n = 177; nGSE39582, n = 573) were used to test the predictive ability of the 7-CFSRG panel for CRC patient survival. The median risk score for each patient group was positively associated with survival (Fig. 5A,D), with high-risk groups showing shorter survival times (Fig. 5B,E). Moreover, the 1-, 3- and 5-year AUC values for the GSE17536 cohort were 0.682, 0.650, and 0.598, respectively, and 0.669, 0.651, and 0.605 for the GSE39582 cohort (Fig. 5C,F). These results indicate the CFSRG panel is a robust and accurate predictor of CRC patient outcome.

Fig. 5.

Risk score validation in the Gene Expression Omnibus (GEO) cohorts. (A) Scatter plot illustrating the risk scores, survival status, and survival time distribution in the GSE17536 cohort. (B) KM curves for the OS in the GSE17536 cohort for high- and low-risk patients. (C) Prediction of 1-, 3-, and 5-year OS in the GSE17536 cohort using time-dependent ROC curves. (D) Scatter plot illustrating the risk scores, survival status, and survival time distribution in the GSE39582 cohort. (E) KM curves for the OS in the GSE39582 cohort for high- and low-risk patients. (F) Predicting 1-, 3-, and 5-year OS in the GSE39582 cohort using time-dependent ROC curves. A higher AUC value signifies a more robust predictive capability exhibited by the signature.

The composition and proportion of immune and stromal cells in the TME are major determinants of tumor progression. Gene expression levels provide the basis for the MCP-counter and xCell algorithms, while the other five algorithms are based on deconvolution [22, 23]. A total of 22 cell types were therefore assessed using 7 algorithms: XCELL, CIBERSORT, QUANTISEQ, EPIC, MCPCOUNTER, CIBERSORT-ABS, and TIMER (Fig. 6A). Analysis of gene expression in immune and stromal cells using the ESTIMATE algorithm allowed calculation of the corresponding immune and stromal scores. Notably, the stromal scores for high-risk patients were significantly higher than low-risk patients (p = 0.0026, Fig. 6B). However, the two risk groups could not be distinguished by the ESTIMATE and immune scores (Fig. 6C,D). Moreover, significant activation of the APC_co_stimulation, inflammation-promoting, and MHC_class I pathways was observed (Fig. 6E,G). We next compared 22 immune cell types between patients with and without CRC. Significantly more macrophages were found in CRC patients, but there were significantly less Treg, Th1, Th2, pDC, and NK cells. These results indicate the occurrence of an immunosuppressive TME in high-risk patients (Fig. 6F).

Fig. 6.

Analysis of CFSRG-based risk scores and infiltrating immune cells. (A) The following tools were used to examine the link between the risk score and infiltrating immune cells: TIMER, CIBERSORT, quanTIseq, EPIC, MCP counter, CIBERSORT-ABS, and xCell. (B–D) Immune, stromal, and estimation scores for both the high- and low-risk populations are displayed in a box plot. (E) Heatmap showing the risk scores and enrichment degree of 13 immune-related functions in the two risk groups. (F) The two risk groups’ risk scores for 16 different kinds of immune cells are presented in a box plot. (G) Box plots illustrate the risk scores and enrichment degrees of 13 immune-associated functions in the two risk groups. * p 0.05, ** p 0.01, *** p 0.001.

The 7-CFSRG panel and potentially related biological mechanisms were further investigated by GSEA. High-risk patients showed a remarkable enrichment for focal adhesion and ECM-receptor interactions, indicating strong tumor invasive properties (Supplementary Fig. 3A). Multiple pathways associated with cell death were enriched in low-risk patients, as well as the cell cycle, DNA replication, and peroxisomes (Supplementary Fig. 3B). Next, we examined the 15 genes with the highest frequency of mutations. A waterfall plot showed that CSMD3, RYR2, KRAS, LRP1B, FAT4, MUC16, ZFHX4, TTN, and TP53 were more often mutated in high-risk patients, whereas SYNE1, PIK3CA, OBSCN, FBXW7, and DNAH5 were less frequently mutated compared to low-risk patients. The frequency of APC mutation did not differ between the two groups (Supplementary Fig. 3C,D). The TMB and risk scores were combined to assess the prognostic significance. High-risk scores and high TMB were found to correlate with a worse outcome for CRC patients (Supplementary Fig. 3E). Additionally, the STEMNESS index showed a significant negative association with the risk score, indicating increased stem cell activity and tumor dedifferentiation in high-risk CRC patients (Supplementary Fig. 3F).

Despite advances in treatment methods, CRC patients still have poor prognosis due to tumor heterogeneity [24]. CRC patients from the TCGA cohort exhibited four distinct immune subtypes. Most high-risk patients had C3 or C4 immune subtypes, whereas the most frequent immune subtype in the low-risk group was C2. Additionally, significant differences in immune subtypes were found between the two risk groups (p = 0.002, Fig. 7A). TIDE may help to improve the selection of patients for ICB therapy [25]. Many patients classified as TIDE responders were observed to have a poor clinical outcome (Fig. 7B). The high-risk CRC group scored higher in TIDE, Dysfunction, and Exclusion (Fig. 7C), suggesting that it may be resistant to ICB therapy. Expression of the immune checkpoint genes CD200R1, ICOS, IDO1, TMIGD2, TNFRSF9, CD80, BTLA, IDO2, HHLA2, CTLA4, and CD274 was shown to be downregulated in high-risk patients, whereas VTCN1, TNFSF4, CD70, and TNFSF9 were highly expressed (Fig. 7D). Targeting of these immune checkpoint genes might therefore improve the efficacy of immunotherapy in high-risk populations.

Fig. 7.

Prediction of immune response and risk score. (A) An immunologic subtype bar chart is shown for the two different risk populations. (B) Histogram showing responses to Tumor Immune Dysfunction and Exclusion (TIDE) among CRC patients. (C) Violin plots represent differences in TIDE, dysfunction, and exclusion scores in the two different risk groups. (D) Fifteen immune checkpoint genes are differentially expressed between high- and low-risk groups. * p 0.05, ** p 0.01, *** p 0.001.

To enhance the therapeutic effect of neoadjuvant chemotherapy in CRC patients, we tested the ability of the 7-CFSRG panel to identify drugs that may be more effective in treating CRC. The IC${}_{50}$ value of each small molecule was calculated using the GDSC database. A wide range of compounds were tested for their potential utility in treating CRC patients. Because BI-2536, CAL-101, FR-180204, Lisitinib, LY317615, Pyrimethamine, Ruxolitinib, and ZSTK474 all showed higher IC${}_{50}$ values in high-risk patients, they may potentially be of more benefit to low-risk patients (Fig. 8A–H,K–R). Conversely, lower IC${}_{50}$ values for YM155 and PHA-665752 were observed in high-risk patients (Fig. 8I,J,S,T), suggesting they could be more sensitive to these drugs. Hence, our risk score could be useful for individualizing chemotherapy in CRC patients.

Fig. 8.

Analysis of CRC patient risk scores to predict small molecule compounds. Scatter plots show the association of half-maximal inhibitory concentration (IC${}_{50}$) values with risk scores for different drugs. (A) BI-2536, (B) CAL-101, (C) FR-180204, (D) Lisitinib, (E) LY317615, (F) ZSTK474, (G) Pyrimethamine, (H) Ruxolitinib, (I) YM155 and (J) PHA-665752. Box plots show IC${}_{50}$ values for potentially active compounds in low- and high-risk CRC patients. (K) BI-2536, (L) CAL-101, (M) FR-180204, (N) Lisitinib, (O) LY317615, (P) ZSTK474, (Q) Pyrimethamine, (R) Ruxolitinib, (S) YM155 and (T) PHA-665752.

The TCGA results showed that all 7 hub CFSRGs were significantly upregulated in CRC tissues relative to normal tissues. Consistent with this, their expression was also higher in CRC cell lines compared to the normal cell line (Supplementary Fig. 4). Specifically, high expression of CXCL11, CCNF, EREG, EPHB2, and MMP3 correlated with better prognosis, whereas upregulation of STC2 and TIMP1 was associated with shorter survival time (Supplementary Fig. 4A–G). ROC analysis showed that STC2 (0.943) had a higher AUC value compared to TIMP1 (0.667) (Supplementary Fig. 4H,I). Furthermore, the high expression of STC2 was confirmed by qRT-PCR (Supplementary Fig. 4J). STC2 was therefore selected for subsequent cellular phenotyping experiments. Analysis of several GEO cohorts also confirmed that STC2 was upregulated in CRC tissues and was associated with worse prognosis (Fig. 9A,B). STC2 expression was further examined by IHC in 24 pairs of in-house clinical samples and found to be enhanced in CRC tissues compared with normal paraneoplastic tissues (Fig. 9C). Results in Supplementary Tables 5,6 show a strong correlation between STC2 expression and the pathologic characteristics of CRC patients. Thus, STC2 may serve as a biomarker of prognosis in CRC patients.

Fig. 9.

SCT2 expression is upregulated in CRC tissues and is correlated with poor outcomes. (A) Box plots showing significant up-regulation of STC2 in CRC tissues from the 6 GEO cohorts. Comparison between groups using the Wilcoxon test, with p 0.05 representing statistical significance. (B) KM survival curves show that CRC patients with high STC2 expression have worse survival. (C) Immunohistochemical detection of STC2 expression in three representative images. Scale bar = 125 µm.

Cox analysis of the TCGA and GEO datasets showed that STC2 expression has significant prognostic value in CRC patients (Fig. 10A). STC2 expression was also positively correlated with pathological stage (Fig. 10B,C), and was highly expressed in three CRC cell lines (LOVO, SW480, and HCT116) compared to normal intestinal epithelial cells (NCM460) (Fig. 10D). In subsequent experiments, si-STC2-interfering plasmid and negative plasmid were transfected into SW480 and LOVO cells. qRT-PCR results showed that mRNA expression of STC2 was significantly decreased following transfection (Fig. 10E). Moreover, transfection with si-STC2 considerably decreased the proliferation rate of SW480 and LOVO cells, as determined by the CCK-8 assay (Fig. 10F,G). In addition, wound healing experiments showed that inhibition of STC2 effectively suppressed CRC cell migration compared to the controls (Fig. 10H,I). Transwell experiments also revealed that SW480 and LOVO cells had reduced migration and invasion ability following inhibition of STC2 (Fig. 10J,K). Based on these findings, we surmise that targeting of STC2 may be a novel approach for the treatment of CRC.

Fig. 10.

Inhibition of STC2 suppresses the malignant properties of CRC cells. (A) The hazard ratio for STC2 expression in CRC patients from TCGA and GEO datasets was evaluated via Cox regression analysis. (B) Comparison of different pathological stages of CRC from the TCGA based on the SCT2 z-score. Comparative analyses between groups via the Kruskal-Wallis test, with p 0.05 representing statistical significance. (C) Comparison of different pathological stages of CRC from GSE28722 based on their SCT2 z-score. Comparative analyses between groups via ANOVA, with p 0.05 representing statistical significance. (D) STC2 expression in epithelial and CRC cells was determined by quantitative real-time polymerase chain reaction (qRT-PCR). (E) STC2 knockdown efficiency is shown in the bar graph. (F,G) SW480 and LOVO cell proliferative rate was assessed by Cell Counting Kit-8 (CCK-8) assay in three independent experiments. (H,I) Cell migration by SW480 and LOVO cells was assessed with the wound healing assay (original image, 200$\times$, scale bar = 200 µm). (J,K) A Transwell assay was conducted to determine cell migration and invasive ability (original image, 200$\times$, scale bar = 200 µm). * p 0.05, ** p 0.01, *** p 0.001.