Identification of ROS and KEAP1‐related genes and verified targets of α‐hederin induce cell death for CRC

In this study, we analyzed and verified differentially expressed genes (DEGs) in ROS and KEAP1 crosstalk in oncogenic signatures using GEO data sets (GSE4107 and GSE41328). Multiple pathway enrichment analyses were finished based on DEGs. The genetic signature for colorectal adenocarcinoma (COAD) was identified by using the Cox regression analysis. Kaplan–Meier survival and receiver operating characteristic curve analysis were used to explore the prognosis value of specific genes in COAD. The potential immune signatures and drug sensitivity prediction were also analyzed. Promising small‐molecule agents were identified and predicted targets of α‐hederin in SuperPred were validated by molecular docking. Also, expression levels of genes and Western blot analysis were conducted. In total, 48 genes were identified as DEGs, and the hub genes such as COL1A1, CXCL12, COL1A2, FN1, CAV1, TIMP3, and IGFBP7 were identified. The ROS and KEAP1‐associated gene signatures comprised of hub key genes were developed for predicting the prognosis and evaluating the immune cell responses and immune infiltration in COAD. α‐hederin, a potential anti‐colorectal cancer (CRC) agent, was found to enhance the sensitivity of HCT116 cells, regulate CAV1 and COL1A1, and decrease KEAP1, Nrf2, and HO‐1 expression significantly. KEAP1‐related genes could be an essential mediator of ROS in CRC, and KEAP1‐associated genes were effective in predicting prognosis and evaluating individualized CRC treatment. Therefore, α‐hederin may be an effective chemosensitizer for CRC treatments in clinical settings.

providing cytoprotection (Mirzaei et al., 2021).These findings highlight the potential of targeting ROS and KEAP1 crosstalk in CRC.
As a phytochemical compound which derived from plants of the Hedera species, α-hederin has gained attention for its potential therapeutic properties.α-hederin has been studied for its various biological effects, particularly in the context of cancer research (Cao et al., 2022).Previous studies have shown that α-hederin inhibits expression of GPX2 and GSH synthetase (GSS), triggering oxidative stress events such as ROS accumulation and GSH depletion in A549 and PC9 cells (Sun et al., 2019).Since α-hederin exhibits promising antiproliferative and antitumor activities in different types of cancer cells, in this study, we investigated the effect and mechanism of the potent small-molecule inhibitor α-hederin in regulating Nrf2 or KEAP1 in CRC cells.
In this research, we identified potential ROS and KEAP1-related oncogenes in CRC.Interestingly, we constructed and validated KEAP1related genes prognostic signature among GSE4107 and GSE41328 in CRC.Furthermore, we evaluated the precision and sensitivity of the constructed model using Kaplan-Meier (KM) and receiver operating characteristic (ROC) analysis.Also, based on the prognostic analysis, we explored changes of immune cell infiltration, and assessed effects of critical genes on the prognosis of CRC patients.We discussed possible genetic signatures and drug sensitivity prediction, along with molecular docking of small-molecule drugs.Additionally, we used SuperPred, a webserver for predicting the targets of compounds based on Anatomical Therapeutic Chemical (ATC) codes.The identification of ATC codes and targets of small molecules aids in the drug development process.It suggested that α-hederin can regulate the expression of ROS-dependent proteins involved in KEAP1-mediated NRF2 activity in HCT116 CRC cells.Hence, we aimed to identify ROS and KEAP1-related genes and assess their impact on CRC patient survival through bioinformatics and statistical analysis.

| Data processing
Two expression profile microarray data sets (GSE41328 and GSE4107) were selected and downloaded from the GEO database for analysis.GSE41328 is an expression profile based on the five colorectal adenocarcinomas (COADs) and matched normal colonic tissues were analyzed with Affymetrix HG-U133-Plus-2.0microarrays (Lin et al., 2006).Two labs independently generated microarray data with the same array platform on the same biological samples.GSE4107 is an expression profile based on the RNA extracted from colonic mucosa of healthy controls (10 samples) and patients (12 samples) were analyzed using GeneChip U133-Plus 2.0 Array (Hong et al., 2007).Patients and controls were age-(50 or less), ethnicity-(Chinese) and tissue-matched.T-test, hierarchical clustering, mean fold-change and principal component analysis were used to identify genes that differentiate between patients and controls.
These were subsequently verified by real-time polymerase chain reaction (PCR) technology.Additionally, RNA sequencing (RNA-seq) data and clinicopathological information of 220 CRC patients were obtained from The Cancer Genome Atlas program (TCGA; https:// portal.gdc.cancer.gov/)by querying KEAP1 in cBioPortal for Samples with mutation and CNA data (https://www.cbioportal.org/).

| Analysis of microarray data sets
According to the progression of CRC, in the subsequent analysis, GEO2R tool was used to analyze the two expression data sets and to identify differentially expressed genes (DEGs).GEO2R is an interactive web tool which allows users to compare different groups of samples in a GEO series to screen genes that can be applied to various technologies such as microarrays, RNA-seq, quantitative PCR, and protein technologies.DEGs were identified and determined based on the following criteria: |fold change (FC)| >1 and both the p value and false discovery rate (FDR) <0.05.The identified DEGs were selected for subsequent studies.Furthermore, a comprehensive list of ROS and KEAP1-related CRC genes were acquired from the GeneCards database (https://www.genecards.org/), and a total of 3751 ROS-related genes and 3008 KEAP1related genes were identified.Then, the Venn diagram tool (http:// bioinformatics.psb.ugent.be/webtools/Venn/)was used to compare and analyze the results of the intersection analysis.Based on the intersection of data sets, the final DEGs were obtained.In this study, we selected DEGs according to the intersection of at least two expression profile data sets to avoid the disadvantages of a single data set and then integrated the results for further biological function analysis.

| Enriched pathway analysis
By performing the Venn diagram, differentially expressed ROS and KEAP1-related genes were identified for further gene ontology (GO) function enrichment analyses including cellular component (CC), biological process (BP) and molecular function (MF) levels) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses.The FDR < 0.01 was considered statistically significant.

| Correlation of ROS and KEAP1-related DEGs
To identify the overlap between the expressed ROS and KEAP1related genes, the expression matrix of the DEGs were subjected to intersection analysis using a Venn diagram.Furthermore, the Pearson correlation test was employed to evaluate the interactions between the ROS and KEAP1-related DEGs at the mRNA level, which was finished using the OECloud tools (https://cloud.oebiotech.com).

| Integration of protein-protein interaction network and module analysis
The protein-protein interaction (PPI) network was drawn by using the STRING tool (http://version10.string-db.org/).The Cytoscape software is a useful tool for visualizing and analyzing molecular interaction networks and pathways.In this study, we utilized Cytoscape to visualize and analyze the interactions among the 48 candidate DEGs associated with ROS and KEAP1.In the PPI network, interactions with a combined score >0.4 were considered significant.Furthermore, the Molecular Complex Detection (MCODE) was used for module screening in Cytoscape, with criteria of scores >3 and nodes >4.

| Correlations between hub genes and clinical features in CRC
Hub genes were identified by overlapping data sets from Pearson correlation analysis of ROS and KEAP1-related DEGs, as well as the top genes in the PPI network using the Venn diagram.To assess the significance of these hub genes in CRC, the correlation between KEAP1 and the hub genes was evaluated using Spearman's correlation coefficient in both normal and tumor samples.The "Pathological Stage Plot" module in GEPIA2 was used to examine the relationship between key oncogene expression and CRC pathological stages.Additionally, univariable Cox regression analysis was performed on TCGA database data to analyze the connections between key oncogenes and OS in CRC patients.

| Establishment and validation of hub genes for prognosis model
The risk score for each CRC patient was calculated, and patients were then classified into high-risk or low-risk groups based on the median risk score.The prognostic signature of the ROS and KEAP1-related genes were evaluated using the KM analysis between two groups.ROC curves were plotted using the "ROC plotter" tool to evaluate the predictive performance of the risk score model.Furthermore, univariate and multivariate Cox regression analyses were completed to assess the prognostic value of the risk score in predicting CRC patient outcomes.

| Correlation of immune infiltrates and hub genes in CRC
To investigate the impact of hub genes on the TME, we assessed the relationships between hub genes and the infiltration levels of immune cell in TCGA-COAD by using the infiltration scores of six kinds of immune cells (B cell, CD4 + T cell, CD8 + T cell, neutrophil, macrophage, and dendritic cell) obtained from the TIMER database based on the TCGA data.

| Tumor infiltration levels of somatic copy number alterations module of hub genes in COAD
The somatic copy number alterations (SCNA) module helps us explore tumor infiltration levels in tumors with distinct SCNA for hub genes.The GISTIC module identifies significant genomic regions with amplification or deletion across samples and assigns a G-score based on amplitude and frequency of occurrence.The SCNAs were determined using GISTIC 2.0 and categorized into deep deletion (−2), arm-level deletion (−1), diploid/normal (0), arm-level gain (1), and high amplification (2).We investigated the correlation between amplification of hub genes and coefficients in COAD associated with tumor infiltration levels, utilizing data from TCGA and heatmap analysis to visualize hub gene expression patterns.

| Target predicition of α-hederin
The ATC code and target prediction of α-hederin were performed in SuperPred (https://prediction.charite.de/subpages/target_result.php).The ATC is established based on machine learning and trained on Morgan fingerprints.Query compounds are evaluated and scored using the ATC.

| Molecules interaction study
Molecular docking is a widely utilized approach for predicting the binding conformation of a ligand to its target protein.In this study, molecular docking analysis was conducted using Schrödinger (Maestro 12.8 software).Schrödinger is designed for developing state-of-the-art chemical simulation software for use in pharmaceutical, biotechnology, and materials science research.
Maestro provides many viewing options to accommodate the varied needs of different applications.NFE2L2 and KEAP1 were predicted targets of α-hederin in SuperPred, with a probability of 53.8% and 51%, respectively, and model accuracies of 96% and 82%.Based on these findings, two crystal structures corresponding to the human KEAP1 bound to NFE2L2 and COL1A1 receptor were searched and selected from Protein Data Bank (PDB) database: The crystal structures of KEAP1 bound to NFE2L2 (PDB ID = 7K2F) and COL1A1 (PDB ID = 2LLP).The interactions between the ligand and target proteins were evaluated using Schrödinger (Maestro 12.8 software) to assess the forces involved.The binding sites for small molecules were identified based on the evolutionary conservation of the binding-site amino acids prepared for the inverse molecular docking.With the reduction of the docking space and by solely focusing on the protein binding sites, the time and complexity of the inverse molecular docking were also shortened.The binding site for a specific protein was determined with the use of the binding site similarity of all PDB entries with co-crystallized ligands of the same protein.The procedure is described in detail by Konc et al. (2021) and Štular et al. (2016).

| Cell culture
The HCT116 and Caco-2 cell lines were sourced from the ATCC.CRC cell lines were cultured in DMEM supplemented with 10% FBS and 50 U/mL penicillin and 50 µg/mL streptomycin (Gibco) in a 37°C humidified incubator containing 5% CO 2 .Cell condition was detected and kept well before all experiments.

| Detection of cell viability
The cytotoxicity of α-hederin was assessed using the Cell Counting Kit-8 (CCK-8) assay (Biotechwell).Cells were seeded into the 96-well plate with 100 µL medium per well.Subsequently, CCK-8 solution (10 µL) was added to each well for incubation (37°C for 4 h).Samples were measured at 450 nm using a spectrophotometer.

| Apoptosis detection
The apoptosis was detected using an Annexin V-FITC apoptosis kit (Beyotime).Briefly, CRC cells were seeded in six-well plates and reached 80%-90% confluence.After applying different treatments, the cells were harvested and washed twice with ice-cold PBS.Subsequently, cells were resuspended in 400 µL of binding buffer combined with 5 µL Annexin V and 5 µL of propidium iodide (PI) in the dark at 37°C for 30 min.Cell apoptosis was detected using the flow cytometry and analyzed using Flowjo V10 software.

| Detection of ROS
The levels of ROS were measured using DCFH-DA (Beyotime) according to the manufacturer's protocol.Briefly, CRC cells were treated with α-hederin or not.Next, cells were labeled with 20 µM DCFH-DA and incubated at 37°C for 30 min in the dark.Following the incubation, the fluorescence intensity of DCF was measured using flow cytometry (BD Biosciences).

| Quantitative real-time polymerase chain reaction
TRIzol reagent (Invitrogen) was used to extract total RNA of cells.
According to the manufacturer's protocol, reverse transcription was completed.Quantitative real-time polymerase chain reaction was conducted using SYBR Green qPCR Master Mix (None ROX) and the DragonLab PCR system.The primer sequences were as follows: Nrf2:

| Western blot analysis
Cell lysates were prepared by collecting and lysing cells on ice for 30 min with RIPA buffer (Beyotime).Equal amounts of cell lysates were separated by SDS-PAGE gel electrophoresis and transferred onto PVDF transfer membranes for antibody blotting.Membranes were blocked with 5% nonfat dry milk for 1 h at room temperature (RT) and incubated overnight at 4°C with primary antibodies against KEAP1, Nrf2, GPX4, HO-1, BGN, and COL1A1.
After incubation, the membranes were washed and incubated with an HRP-conjugated secondary antibody for 1 h at RT. Target bands were detected using an ECL reagent.

| Statistical analysis
The mRNA expression levels between HCT116 cell lines were compared using Student's t test.The relationship between the expression of key oncogenes and OS in CRC patients was examined using univariable Cox regression analysis.The correlations between KEAP1 and hub genes were analyzed using the Spearman correlation coefficient.For the analysis of DEGs, GO, and KEGG, the FDR was adjusted using the Benjamini-Hochberg procedure to control for

| Identification and comparison and analysis of DEGs
To investigate the activation of ROS-KEAP1 in cancer progression, we identified the DEGs in GSE41328 and GSE4107.As shown in Figure 1a,b, 1744 genes were identified as DEGs in GSE4107, and 1309 genes were identified as DEGs in GSE4107.The intersected genes in four genes sets were shown by Venn diagram, and 48 genes were finally identified as key genes (Figure 1c).
By performing the KEGG analysis using the 48 key genes, it showed that cancer-related pathways such as "the pathways in cancer," "proteoglycans in cancer," and "PI3K-Akt signaling pathway" were significantly enriched (Figure 1d,e).

| Pearson correlation of DEGs in the PPI network and identification of hub genes
Significant positive or negative correlations between the 48 key genes were constructed and 20 top-correlated genes were found (Figure 3a,b).Furthermore, the interactions of 48 genes ranked by degree were identified.as key genes (Figure 3c,d).Furthermore, 17 top genes (MCODE Score >1.5, Clustered Node) were identified in the PPI network and shown in the Table 1.Then, the Venn diagram of 20 top-correlated genes in Pearson correlation and 17 top genes in the PPI network were constructed in Cytoscape, and 7 overlapping genes (COL1A1, COL1A2, CXCL12, CAV1, FN1, TIMP3, and IGFBP7) were identified as hub genes (Figure 3e).The expression heatmap of seven hub genes is shown in Figure 3f.

| Development and validation of hub genes' prognostic signature for COAD
Next, expressions of seven hub genes in COAD and normal tissues were analyzed in the TCGA-COAD.It suggested that COL1A1 and COL1A2 were significantly increased while CXCL12 and CAV1 were significantly decreased in tumor compared to normal tissues (Figure 4a).Additionally, analyses based on tumor stages revealed differences in gene expressions for COL1A2 and FN1 (Figure 4b).KM survival curves further demonstrated that high-risk patients with COL1A1 and COL1A2 had lower OS compared to low-risk patients (Figure 4c).In the training cohort, the area under the ROC curves (AUCs) for the OS of patients.The "ROC plotter" tool was used to validate the predictive value of hub genes.The AUCs of COL1A1 for treatment with 5-Fu, capecitabine, and fluoropyrimidines monotherapy were 0.61, 0.70, and 0.72, respectively (Figure 4d).The AUC of COL1A2 for treatment with Bevacizumab was 0.63, while the AUC of CAV1 for treatment with fluoropyrimidines monotherapy was 0.63, and the AUC of CXCL12 for treatment with Bevacizumab was 0.62.expression in COAD samples compared to normal samples (Figure 4e).Moreover, distinct connections were observed between the significant differed expression of COL1A1 and COL1A2 with different stages.On the other hand, both CXCL12 and CAV1 exhibited significantly lower expression in COAD samples compared to normal samples, and there were no significant differences in their expression based on different stages (Figure 4e).However, high expression of COL1A1 and COL1A2 was observed in Adenocarcinoma and Mucinous adenocarcinoma compared to normal tissues (Figure 4f).In contrast, CXCL12 and CAV1 expression showed a decrease in Adenocarcinoma and Mucinous adenocarcinoma in a histological subtype-dependent manner.

| Correlation between the prognosis-related genes and the TME
The TISCH database was used to analyze the expression of key hub genes in TME-related cells.In immune cells (CD4Tconv, Tprolif, DC) and epithelial cells, COL1A1 expression was low to moderate (Figure 5a).Fibroblasts/Myofibroblasts showed the highest expression of COL1A1 and COL1A2, while DC, Endothelial, and Malignant cells had lower to moderate COL1A2 expression.Additionally, CXCL12 and CAV1 had the highest expression in Endothelial cells, Fibroblasts, and Myofibroblasts.
Fibroblasts, Leydia cells, and Endothelial cells were the most abundant cell types analyzed in the HPA database.Furthermore, CXCL12 and COL1A1 had higher infiltration in TME-related cells compared to CAV1.Furthermore, the distributions and expressions of COL1A1, COL1A2, CXCL12, and CAV1 in different cells at the single-cell level were analyzed using the HPA database (Figure 7b-i).The infiltrations of COL1A1 (Figure 5b,c) and COL1A2 (Figure 5d,e) in TME-related cells were higher compared to CXCL12 (Figure 5f,g) and CAV1 (Figure 5h,i), consistent with the findings presented in

| Correlation between immune infiltrates and hub genes in COAD
To investigate the impact of hub genes on the TIM, we examined the correlations between the expression of COL1A1, COL1A2, CXCL12, and CAV1 and the infiltration levels of immune cells in TCGA-COAD.We observed strong associations between hub genes and immune cell infiltrations in COAD according to the TIMER database (Figure 6a-d).The accompanying linear regression analysis demonstrated that high expression of COL1A1, COL1A2, CXCL12, and CAV1 were associated with increased levels of immune cell infiltration.Notably, CD4 + T cells, macrophages, and dendritic cells exhibited the most significant coefficients in relation to COL1A1, COL1A2, CXCL12, and CAV1 in COAD.
Furthermore, infiltration levels of immune cells in different copy number alternations (CNAs) were compared as well, which suggested that CNAs of hub genes could influence the infiltration levels of immune cells in COAD (Figure 6e).

| Target predicition and molecular interaction analysis
Predicting ATC codes or targets of α-hederin is valuable for drug development.NFE2L2 and KEAP1 were predicted targets of α-hederin in SuperPred, with a probability of 53.8% and 51%, respectively, and model accuracies of 96% and 82% (Table 2).We generated an interaction network between α-hederin and predicted targets using Cytoscape (Figure 7a).The chemical structure of α-hederin is illustrated in Figure 7b.
To study the binding interactions between α-hederin and the core molecular targets (KEAP1-NFE2L2 and COL1A1), we used AutoDock Maestro 12.8 software for molecular docking analysis.The PDB structures of KEAP1-NFE2L2 and COL1A1 were employed for representing the targets in the docking analysis (Figure 7c,d).
Furthermore, we visualized the molecular docking results by displaying small-molecule drug binding (Figures 7e,f).For instance, α-hederin potentially binds to KEAP1-NFE2L2, forming hydrogen bonds with GLY-527, GLY-528, GLY-530, and ARC 482 near the active site.This may contribute to its biological functions.To evaluate α-hederin's docking scores against the active sites of COL1A1, we used the Glide module in Schrodinger suite software.The docking score of α-hederin with COL1A1 was −3.39, indicating a favorable interaction.The 2D and 3D interaction diagrams of α-hederin in the active site of COL1A1 showed the formation of two hydrogen bonds with CLN B:20 and CLY C:19 (Figure 7f).

| In vitro validation of predicted targets of α-hederin in CRC cells
3.9.1 | α-hederin inhibits cell growth and promotes cell death in HCT116 cells To assess the cytotoxic effect of α-hederin on CRC cell lines, we treated HCT116 CRC cell lines with different concentrations of α-hederin or Oxaliplatin for varying durations (12, 24, 36, and 48 h) and evaluated cell viability using the CCK-8 assay.Our findings demonstrated that α-hederin exerted a significant concentration-and time-dependent reduction in the viability of HCT116 CRC cell lines compared to Oxaliplatin or untreated cells (Figure 8a,b).In addition, the IC 50 value for α-hederin in HCT116 cells was 16.22 μM (Figure 8c).Thus, α-hederin demonstrated a considerably superior cytotoxic effect on HCT116 cells.
Effects of α-hederin on CRC apoptosis were detected.Results revealed that α-hederin treatment for 24 or 48 h induced apoptosis.
Interestingly, after 12 h of treatment with α-hederin in HCT116 cells, a fraction of the cells accumulated in the Annexin V−/PI+ region (Figure 8d).This observation suggests that α-hederin may induce a distinct form of cell death within 12 h, different from apoptosis.
Additional cell death assays were conducted to investigate further.
Results showed that α-hederin-induced cell death within 48 h is mainly apoptotic, indicating that apoptosis could become more prominent with longer treatments (Figure 8e).

| α-hederin induces ROS generation by mediating KEAP1\Nrf2-HO-1 activation in CRC cells
Our findings demonstrated that α-hederin induced rapid and concentration-dependent production of ROS in HCT116 cells, as compared to Oxaliplatin or untreated cells (Figure 9a,b).GSH, an essential antioxidant involved in regulating redox balance (Cao et al., 2022).By investigating the mechanism underlying α-hederin-induced ROS generation, results revealed a decrease in GSH levels upon treatment with α-hederin in HCT116 cells (Figure 9c).To understand the mechanism of α-hederin-induced cell death, we performed RNA-sequencing analysis.We found that α-hederin induced oxidative stress, leading to ROS activation.Specifically, we examined Nrf2, an upstream regulator of ROS-responsive genes.In HCT116 cells treated with α-hederin, we observed a significant concentration-dependent decrease in gene expression of Nrf2 and HO-1, compared to cells treated with Oxaliplatin (Figure 9d).
Protein level analysis after 24 h of α-hederin treatment also showed reduced levels of Nrf2 and HO-1 in HCT116 cells (Figure 9e).These results suggest that α-hederin inhibits the Nrf2-HO-1 signaling pathway, contributing to cell death in HCT116 cells.
Moreover, Western blot analysis revealed that α-hederin treatment resulted in a decrease in KEAP1\Nrf2 activation within 24 h in HCT116 cells (Figure 9f).T A B L E 2 The predicted targets of α-hederin in SuperPred.To investigate the involvement of CAV1 and COL1A1 in α-hederininduced cell death, we treated HCT116 cells with α-hederin and examined their expression levels.Western blot analysis revealed that α-hederin treatment led to reduced activation of CAV1 and COL1A1 in HCT116 cells within 24 h (Figure 9g).This suggests that α-hederin may influence the expression of CAV1 and COL1A1, potentially contributing to its ability to induce cell death in HCT116 cells.

| DISCUSSION
This study aimed to evaluate the clinical significance of ROS and KEAP1-related genes in CRC development, prognosis, and drug selection for clinical individualized treatment.We examined the interactions between hub genes of ROS and KEAP1-related genes and immune cell infiltration levels in TCGA-COAD to determine their participation in TME regulation.Results showed that COL1A1, COL1A2, CXCL12, and CAV1 expression were associated with increased immune cell infiltration, particularly CD4 + T cells, macrophages, and dendritic cells.These four genes showed the highest significant coefficients among the eight hub genes examined.
TME, particularly immune microenvironment, plays a crucial role in CRC recurrence and metastasis (Mei et al., 2021;Xiong et al., 2019), while immune cell infiltration has been implicated in CRC progression, metastasis, and immunosuppression (Picard et al., 2020).Our study investigated the correlation between TME, immune cells, and ROS and KEAP1-related hub genes.Using the TISCH and TIMER database, we assessed immune cell infiltration levels in CRC patients.We observed a higher proportion of CD4 + T cells, macrophages, and dendritic cells in the high-risk group.Studies have indicated that increased macrophage infiltration in the TME is associated with poor prognosis in CRC patients (Ma et al., 2022;Tokunaga et al., 2020).
Fibroblasts, the dominant stromal cell type in the TME of CRC, have been linked to poor prognosis in CRC patients due to high expression of fibroblast-specific stromal genes (Khaliq et al., 2022).
Cancer-associated fibroblasts (CAFs) produce ECM components for the tumor and contribute to its development and metastasis through ECM remodeling (Sandberg et al., 2019;Stadler et al., 2021).
Furthermore, CAFs participate in immunosuppression formation and immune cells recruitment in the TME via secretion of cytokines and chemokines such as CXCL12 (Song et al., 2021).Related metaanalysis of CXCL12 also verified that CXCL12 can predict the prognosis of various cancer types including CRC (Samarendra et al., 2017;Zou et al., 2020).Moreover, newly studies also show  distinct differences of TIM between primary and metastatic CRC (Lal et al., 2022;Sorrentino et al., 2021).These results support that CXCL12 and COL1A1 connected closely to the TIM in CRC.
CAV1 has been identified as a potential biomarker for CRC metastasis (Lin et al., 2022;Subbarayan et al., 2022) and is significantly associated with CRC with liver metastasis (CRCLM) in CRC, as well as worse survival rates (Dai et al., 2023;Ng et al., 2022).
CAV1 expression is also linked to CRC migration and invasion, and its upregulation in HCT116 cell lines was found to promote cell proliferation and invasion, (Ng et al., 2022) while inhibition of CAV1 in CRC cell lines significantly reduced these processes.
Additionally, increased COL1A1 expression in CRC has been shown to promote serosal invasion, lymph and hematogenous metastases, highlighting its potential as a therapeutic target (Manoochehri et al., 2022;Zhang et al., 2018).Therefore, our findings suggest that CAV1 and COL1A1 may serve as novel therapeutic targets for CRC.
Studies have hinted that the natural phytochemicals show strong potential in anti-CRC treatments.In the meantime, studies have tried to find effective phytochemicals or derivatives targeting the Nrf2/ KEAP1 axis, which plays a crucial role in tumor development (Deshmukh et al., 2017;Fourquet et al., 2010;Yamamoto et al., 2008).

2. 9 |
Connections among the prognosis-related genes and the tumor microenvironment Tumor microenvironment (TME) plays a key role in the initiation and progression of tumors.Tumor Immune Single-cell Hub 2 (TISCH2) is a scRNA-seq database focusing on TME.It offers detailed cell-type annotation at the single-cell level, facilitating the study of TME in various cancer types.Therefore, we utilized multiple data sets from the TISCH2 database to investigate the expression patterns of COL1A1, COL1A2, CXCL12, and CAV1 in TME-related cells.Additionally, the HPA database was used to analyze the single-cell expression profiles of ROS and KEAP1-related hub genes across 10 distinct cell types.

F
I G U R E 1 Differential expression analysis and comparison of the DEGs in ROS-KEAP1-related genes and two GEO data sets (GSE4107 and GSE41328).(a) The volcano plot of the expression of GSE4107.(b) The volcano plot of the expression of GSE41328, comparison of the expression of the DEGs.(c) At the intersection of the Venn diagram, 48 differentially expressed ROS-KEAP1-related genes were finally identified.(d, e) KEGG analysis of ROS and KEAP1-related DEGs.DEG, differentially expressed gene; KEGG, Kyoto Encyclopedia of Genes and Genomes; ROS, reactive oxygen species.F I G U R E 2 (See caption on next page).multiple testing.A significance level of p < .05 was considered statistically significant.
To elucidate the underlying mechanism by which hub genes impact the prognosis of COAD patients, we utilized the UALCAN online database to analyze 374 COAD samples and 41 normal samples from TCGA.Both COL1A1 and COL1A2 showed significantly higher F I G U R E 2 Functional enrichment analysis of up-or downregulated DEGs in ROS and KEAP1-related genes for GO and KEGG analysis.(a) Heatmap of DEGs in ROS-KEAP1-related genes and two GEO data sets (GSE4107 and GSE41328).A total of 17 upregulated and 15 downregulated DEGs were identified.Gene expressions and clustering analysis were shown in heatmap.(b) Analysis of the GO Classification of up-or downregulated DEGs.(c, d) Analysis of the GO enrichment terms of up-or downregulated DEGs.(e, f) The KEGG Pathway Classification analysis of up-or downregulated differentially expressed genes.(g, h) Analysis of the up-or downregulated DEGs ranked by degree were identified in the KEGG enrichment.DEG, differentially expressed gene; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; ROS, reactive oxygen species.

Figure 5a .
Figure 5a.These results provide further evidence supporting the close association between COL1A1, COL1A2, and the tumor immune microenvironment (TIM) in CRC.

F
I G U R E 3 Pearson correlation test and the protein-protein interaction (PPI) networks of the hub genes ranked by degree were identified between ROS and KEAP1-related DEGs.(a, b) Pearson correlation test between ROS and KEAP1-related DEGs.(c, d) The PPI network.The interaction relationship of the top genes for candidate genes.(e) At the intersection of the Venn diagram, seven hub genes in ROS-KEAP1related differentially expressed genes were finally identified.(f) Heatmap of hub genes in ROS-KEAP1-related DEGs and two GEO data sets (GSE4107 and GSE41328).DEG, differentially expressed gene; ROS, reactive oxygen species.F I G U R E 4 (See caption on next page).

F
I G U R E 4 Construction and validation of the ROS and KEAP1-related hub gene prognostic signature.(a) The expression profiles of ROS and KEAP1-related hub gene in the validation cohort and entire cohort of TCGA-COAD.(b) The correlation between expression levels of hub genes and the pathological stages in TCGA-COAD.(c) Predictive ability of the prognostic signature in the TCGA-COAD cohort and KM survival curves of OS for COAD patients.(d) ROC curves (AUCs) for COL1A1, COL1A2, CXCL12, and CAV1 for COAD patients' chemotherapy or monotherapy survival in the TCGA-COAD cohort and validation cohort.(e) Correlations among the glycolysis-and lactate-related genes and gender and stage.(f) The ROS and KEAP1-related hub gene expression in COAD based on histological subtypes of Adenocarcinoma and Mucinous adenocarcinoma.AUC, area under the curve; COAD, colorectal adenocarcinoma; KM, Kaplan-Meier; OS, overall survival; ROC, receiver operating characteristic; TCGA, The Cancer Genome Atlas.*p < .05,**p < .01,and ***p < .001.F I G U R E 5 Correlation between the prognostic-related genes and the TME.(a) Correlation analysis between COL1A1, COL1A2, CXCL12, and CAV1 in CRC and the TME using the TISCH database.(b, c) The distribution of COL1A1 in different cell types in the HPA database.(d, e) Cell types and distribution of COL1A2 in the HPA database.(f, g) The distribution of CXCL12 in different cell types in the HPA database.(h, i) The cell types and their distribution of CAV1 in the HPA database.CRC, colorectal cancer; TME, tumor microenvironment.

F
I G U R E 6 Immune score of hub genes and immune cells' infiltrates CNV in COAD.(a-d) Correlation analysis of COL1A1, COL1A2, CXCL12, and CAV1 with the infiltration levels of immune cells using the TIMER database.(e) Analysis of SCNAs revealed a difference in COL1A1, COL1A2, CXCL12, and CAV1 expression at the dendritic cell level among these groups.Immune subsets and copy number status were shown in COAD.The infiltration level for each SCNA category is compared with the normal using a two-sided Wilcoxon rank-sum test.COAD, colorectal adenocarcinoma.
. Although there have been reports on the regulation of Nrf2 or KEAP1 and the detection of Cul3dependent E3 ubiquitin ligase activity in cancer cells using selective phytochemicals and small-molecule inhibitors, further validation in clinical research and deeper mechanistic investigations are still lacking.Therefore, the potential anti-CRC mechanism of α-hederin was explored in our study.Results suggested that α-hederin could effectively inhibit CRC cells growth.Moreover, significant upregulation of KEAP1 expression was observed in HCT116 cells, which was notably downregulated upon exposure to α-hederin.Moreover, we discovered that α-hederin exerts an anticancer effect by targeting the Nrf2/KEAP1 axis in CRC.Importantly, α-hederin enhances the sensitivity of HCT116 cells and regulates CAV1 and COL1A1, which are key oncogenes providing novel targets for individualized treatment of CRC.Some limitations in the present study are worth noting.First, functional enrichment validation by fundamental experiments could be finished in the next step.Second, more validations about connections between α-hederin and verified hub genes in our study are needed.Third, related clinical applications or validations of α-hederin in CRC patients will increase the translational value.In summary, by using multiple CRC data sets, our study successfully identified key genes (such as CAV1, COL1A1, and COL1A2) in CRC development, which may be translated into new targets for CRC treatments.In the meantime, by performing related verifications, we found that α-hederin could serve as a potential anti-CRC agent for clinical CRC treatments, and possible connections between α-hederin and verified hub genes in this study were found as well.Furthermore, this study provides novel insights into oncogenic mechanisms driven by the KEAP1-related signaling pathway.It also seeks to identify substrate proteins deubiquitinated by Nrf2/KEAP1, with the ultimate goal of realizing the potential of KEAP1 as a drug target in CRC.