Colorectal Cancer Liver Metastasis‐Associated Ferroptosis‐Related Genes Modulate Lipid Peroxidation in Colorectal Cancer Cells

ABSTRACT

Distant metastasis, predominantly to the liver, remains the leading cause of death in colorectal cancer (CRC), yet biomarkers that capture metastatic competence remain limited. Ferroptosis is an iron‐dependent, lipid peroxidation‐driven form of regulated cell death that can restrain tumor progression, but whether primary CRC from patients with liver metastasis shows ferroptosis‐resistance‐related features remains incompletely understood. In a small exploratory set of T‐stage‐matched primary CRC tumors with or without liver metastasis, we quantified glutathione redox and lipid peroxidation‐related readouts and assessed SLC7A11 and GPX4 expression. We integrated GSE62321 transcriptomic profiles with a FerrDb ferroptosis gene set, evaluated prognosis in TCGA‐COAD/READ, and performed genetic knockdown, MDA assays, C11‐BODIPY lipid ROS staining, ferrostatin‐1 rescue assays, and Transwell assays in CRC cell models. Primary tumors from patients with liver metastasis showed a more reduced redox profile and increased expression of core ferroptosis‐suppressive proteins, consistent with enhanced ferroptosis resistance potential but not direct evidence of lower in vivo ferroptotic cell death. Integrative discovery highlighted fatty acid binding protein 4 (FABP4), α‐synuclein (SNCA), and discoidin domain receptor 2 (DDR2) as CRC‐LM‐associated ferroptosis‐related candidates. High expression of each gene was associated with unfavorable disease‐free survival. In CRC cell models, including the lymph‐node‐metastasis‐derived SW620 line and additional validation lines, silencing FABP4, SNCA, or DDR2 increased bulk MDA and/or C11‐BODIPY‐detected lipid ROS, altered ferroptosis susceptibility, and suppressed migratory and/or invasive phenotypes. Ferrostatin‐1 partially rescued knockdown‐induced viability loss, lipid ROS accumulation, and migratory/invasive defects, supporting involvement of ferroptosis‐associated lipid peroxidation while not excluding broader stress‐response mechanisms. FABP4, SNCA, and DDR2 are CRC‐LM‐associated ferroptosis‐related candidates that modulate lipid peroxidation, ferroptosis susceptibility, and migratory/invasive phenotypes in CRC cell models, warranting further validation in viability‐controlled and liver metastasis‐specific models.

Abbreviations

CCK‐8

Cell Counting Kit‐8

CPTAC

Clinical Proteomic Tumor Analysis Consortium

CRC

colorectal cancer

CRC‐LM

colorectal cancer liver metastasis

CRC‐nLM

colorectal cancer non‐liver metastasis

DDR2

discoidin domain receptor 2

DEGs

differentially expressed genes

DFS

disease‐free survival

FABP4

fatty acid binding protein 4

FBS

fetal bovine serum

GEO

Gene Expression Omnibus

GEPIA

Gene Expression Profiling Interactive Analysis

GO

Gene Ontology

GSH

glutathione

GSH/GSSG

glutathione to oxidized glutathione ratio

GSSG

oxidized glutathione

IHC

immunohistochemistry

KEGG

Kyoto Encyclopedia of Genes and Genomes

MDA

malondialdehyde

PBS

phosphate‐buffered saline

ROS

reactive oxygen species

SNCA

α‐synuclein

TCGA

The Cancer Genome Atlas

WB

western blot

1. Introduction

Colorectal cancer (CRC) remains a leading cause of cancer morbidity and mortality worldwide [1, 2]. Distant spread, predominantly manifesting as liver metastasis, serves as the principal determinant of long‐term prognosis, yet clinical tools for identifying patients at high risk remain imperfect [3, 4]. Beyond conventional clinicopathological factors, there is a clear unmet need for molecular readouts from the primary tumor that can (i) capture metastatic competence at an early stage and (ii) reveal actionable vulnerabilities that may be exploited to prevent or treat distant metastasis of CRC.

Ferroptosis is a distinct form of regulated cell death driven by iron‐dependent lipid peroxidation and collapse of redox homeostasis. Because metastatic dissemination imposes intense oxidative and metabolic stress, ferroptosis has been proposed as a key barrier that limits metastatic outgrowth [5, 6]. Conversely, tumor cells that successfully metastasize are expected to acquire mechanisms that restrain lipid peroxidation, preserve glutathione metabolism, or upregulate antioxidant defenses such as the system x _c ⁻/GPX4 axis. However, it remains unclear whether ferroptosis resistance is already “pre‐wired” in the primary CRC lesion of patients who present with liver metastasis, and which genes constitute this program.

A clinically anchored ferroptosis‐resistance‐related signature in primary CRC may have translational value for risk stratification and for identifying potential vulnerabilities. Therefore, this study aimed to identify genes associated with colorectal cancer liver metastasis (CRC‐LM) risk and to test whether they modulate ferroptosis susceptibility and motility‐related phenotypes in CRC cell models. We first profiled glutathione redox status, lipid peroxidation‐related readouts, and core ferroptosis‐suppressive proteins in a small exploratory set of T‐stage‐matched primary CRC specimens from patients with or without liver metastasis. Guided by these clinical observations, we integrated a public transcriptomic cohort [7] containing normal mucosa, primary CRC, and liver metastases with a curated ferroptosis gene set to nominate metastasis‐associated ferroptosis regulators [8].

2. Materials and Methods

2.1. Human Subjects

This study was conducted on patients with CRC recruited from Qilu Hospital of Shandong University. Primary colorectal tumor tissues were obtained from patients with or without liver metastases. The two cohorts were matched for T stage; however, because of limited specimen availability, complete matching for N stage and differentiation status was not feasible. The detailed clinical characteristics of the patients are presented in Tables S1 and S2. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Qilu Hospital of Shandong University (KYLL‐2026‐01‐029).

2.2. Cell Culture

Human CRC cell lines SW620, HCT116, and HT29 were obtained from Cyagen Biosciences. SW620 is derived from a lymph‐node metastasis of CRC and was used in this study as a metastasis‐derived CRC model. HCT116 and HT29 cells were included as additional CRC cell lines for validation of selected findings. All cell‐line‐based experiments were therefore interpreted as assays of ferroptosis susceptibility and metastatic CRC cell behaviors, including migration and invasion. All cells were maintained in DMEM (Gibco, 11965092) supplemented with 10% (v/v) fetal bovine serum (FBS, Gibco, 10099141) at 37°C in a humidified atmosphere containing 5% CO₂.

For pharmacological ferroptosis induction experiments, SW620 cells were treated with 2 μM RSL3 (MCE, HY‐100218A) or 5 μM erastin (MCE, HY‐15763) for 48 h before downstream analyses. For the Transwell experiments, cells were first exposed to RSL3 or erastin for 48 h, then trypsinized and reseeded at equal cell numbers into Transwell chambers; the same concentrations of RSL3 or erastin were maintained in both the upper and lower chambers during the subsequent migration (12 h) or invasion (24 h) assay. Thus, the total exposure time was 60 h for migration assays and 72 h for invasion assays. For gene‐knockdown rescue experiments, 5 μM ferrostatin‐1 (MCE, HY‐100579) was added 24 h after transfection. For CCK‐8 and C11‐BODIPY assays, cells were analyzed after 24 h of ferrostatin‐1 treatment; for Transwell rescue assays, cells were exposed to ferrostatin‐1 before seeding and ferrostatin‐1 was maintained in both chambers during the 12‐h migration or 24‐h invasion period. RSL3, erastin, and ferrostatin‐1 were dissolved in DMSO, and an equal volume of DMSO was used as the vehicle control.

2.3. Glutathione to Oxidized Glutathione Ratio (GSH/GSSG Ratio) Measurement

GSH/GSSG ratio was quantified using the GSH/GSSG Ratio Detection Assay Kit (Beyotime, S0053), following the manufacturer's protocol. For tissue samples, homogenates were prepared from 50 mg of tumor tissue; for cell samples, cells from one well of a 6‐well plate were collected for homogenization. Concentrations were determined in μM. For tissue comparisons, relative GSH and GSSG levels were normalized to the mean value of the CRC‐nLM group, which was set as 1. The GSH/GSSG ratio was calculated from the measured GSH and GSSG concentrations.

2.4. Malondialdehyde (MDA) Measurement

MDA concentration was quantified using the MDA Assay Kit (Beyotime, S0131), following the manufacturer's protocol. Values were normalized to total protein concentration. Tissue MDA levels were expressed as nmol/mg tumor tissue protein, and cellular MDA levels were expressed as nmol/mg protein.

2.5. Western Blot (WB)

WB was performed according to our standard protocol [9]. Briefly, proteins were separated using commercial gels (Epizyme, PG212/PG214), transferred to PVDF membranes (Millipore, IPVH00010), and probed with primary antibodies (see Table S3) overnight at 4°C, followed by incubation with HRP‐conjugated secondary antibodies. Signals were detected by ECL (Proteintech, PK10002) and quantified using ImageJ software.

2.6. Immunohistochemistry (IHC)

IHC experiments were performed on tissues following our previously established methods [9]. Specifically, after deparaffinization, rehydration, antigen retrieval, and blocking of endogenous peroxidase activity, paraffin sections were stained using an immunohistochemistry kit (ZSGB‐BIO, PV9000). The antibodies and concentrations used are detailed in Table S3. The Immunoreactive Score was assessed based on previous studies [10] and evaluated independently by two researchers, with consensus reached on any discrepancies.

2.7. Cell Transfection

To silence fatty acid binding protein 4 (FABP4) and α‐synuclein (SNCA), cells were transfected with target‐specific or nontargeting siRNA (siCtrl) using Lipomaster 3000 (Vazyme, TL301). Discoidin domain receptor 2 (DDR2) knockdown was performed using shRNA plasmids (or a nontargeting shRNA plasmid as shCtrl), which were also transfected with Lipomaster 3000. Protein was extracted 72 h after transfection for validation of knockdown efficiency by WB. All siRNA and shRNA sequences are provided in Table S4.

2.8. Cell Viability Assay

Cell viability was assessed using the Cell Counting Kit‐8 (CCK‐8) assay for the gene‐knockdown/ferrostatin‐1 rescue experiments. Cells were seeded into 96‐well plates at a density of 2 × 10³ cells per well and cultured overnight to allow attachment. Transfection was then performed as described above, ferrostatin‐1 was added 24 h after transfection, and CCK‐8 analysis was performed after 24 h of ferrostatin‐1 treatment using a commercial kit (Sparkjade, CT0001) according to the manufacturer's protocol. No separate CCK‐8 panel was included for the prolonged RSL3/erastin Transwell exposure conditions.

2.9. BODIPY 581/591 C11 Staining

Lipid peroxidation was assessed using the BODIPY 581/591 C11 probe. Cells were seeded into 6‐well plates and cultured overnight to allow attachment. Transfection and ferrostatin‐1 treatment were then performed as described above. After the treatments, cells were incubated with 2 μM BODIPY 581/591 C11 (Beyotime, S0043S) in PBS for 30 min at 37°C in the dark. After incubation, cells were washed twice with PBS and then examined under a fluorescence microscope.

2.10. Transwell Assays

With reference to previous studies [11], Transwell plates with an 8 μm pore size (Falcon, 353097) were used to examine migratory and invasive abilities. For migration assays, cells (1 × 10⁵ cells per well) were suspended in FBS‐free medium and seeded into the upper chamber, while the lower chamber was filled with medium containing 10% FBS. The plates were incubated at 37°C for 12 h. For invasion assays, the upper chamber was pre‐coated with 50 μL of Matrigel (diluted 1:10; Corning, 356234), and the incubation time was extended to 24 h. Cells were then fixed with 4% paraformaldehyde, stained with crystal violet solution, and imaged after removal of cells remaining on the upper surface of the membrane. For gene‐knockdown rescue assays, cells were transfected and treated with ferrostatin‐1 as described above, then trypsinized and seeded at equal cell numbers; ferrostatin‐1 was maintained in both chambers throughout the assay. For inducer‐only assays, SW620 cells were pretreated with 2 μM RSL3 or 5 μM erastin for 48 h before seeding, and the same inducer concentration was maintained in both chambers during migration or invasion.

2.11. Microarray Data

The gene expression profiles of GSE62321 [7] were obtained from the publicly available Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). This data set utilized the GPL97 platform ([HG‐U133B] Affymetrix Human Genome U133B Array) and encompassed 57 expression matrices representing primary CRC, normal colon mucosa, and liver metastases.

2.12. Identification of Differentially Expressed Genes (DEGs)

Differential analysis was performed using the limma R package (version 3.40.6) to identify DEGs [12]. Prior to the analysis, we initially subjected the expression profiles to a log₂ transformation. Subsequently, a linear model was then fitted to the data using the lmFit function. Statistical significance was assessed with the empirical Bayes approach implemented in the eBayes function, which moderates the standard errors across genes to compute moderated t‐statistics, moderated F‐statistics, and log‐odds of differential expression. The identified DEGs were subsequently subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis.

2.13. Survival Prognosis Analysis

Survival analysis was performed based on RNA sequencing data from TCGA using the Gene Expression Profiling Interactive Analysis (GEPIA) online tool to evaluate the impact of key genes on disease‐free survival (DFS) in CRC patients, and survival curves were generated accordingly [13]. Furthermore, clinical data and gene expression profiles were obtained from the TCGA‐COAD and TCGA‐READ projects. A prognostic nomogram was constructed via Cox regression analysis, incorporating clinicopathological variables (survival time, age, sex, and stage) and the expression levels of three specific genes. This model was evaluated using a cohort of 613 samples, and its predictive accuracy was assessed with calibration curves.

2.14. Exploratory Analysis of Gene Expression and Estimated Immune Infiltration

Expression data for FABP4, DDR2, and SNCA were extracted from the TCGA Pan‐Cancer data set via the UCSC database (https://xenabrowser.net/). Expression values were log₂(x + 1) transformed and annotated with official gene symbols. Immune infiltration scores for B cells, CD4+ T cells, CD8+ T cells, neutrophils, macrophages, and dendritic cells were calculated using the TIMER algorithm implemented in the IOBR R package (version 0.99.9) [14].

2.15. Analysis of Gene Expression and Protein Levels

Using the “Gene_DE” module of TIMER2 (http://timer.cistrome.org/), we analyzed differential mRNA expression of FABP4, DDR2, and SNCA between tumor tissues and matched normal tissues across multiple cancer types from TCGA. In parallel, the UALCAN platform was used to compare protein expression levels of these genes in primary tumor versus normal tissues using CPTAC data sets (http://ualcan.path.uab.edu/analysis-prot.html). These pan‐cancer tumor‐versus‐normal comparisons were used only to describe broad dysregulation patterns.

2.16. Analysis of Genetic Alterations

Using the cBioPortal webserver (https://www.cbioportal.org/) [15, 16], we systematically analyzed the genetic alteration profiles—including mutation frequency and type—of FABP4, DDR2, and SNCA across the TCGA Pan‐Cancer Atlas data set through the “Cancer Types Summary” and “Mutations” modules.

2.17. Statistical Analysis

Data were analyzed using GraphPad Prism 10.1.2. Statistical methods included t‐tests, ANOVA, Dunnett's and Bonferroni's multiple comparisons tests. Error bars represent mean ± SD. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns nonsignificant.

3. Result

3.1. Primary CRC Tumors From Patients With Liver Metastasis Show a Reduced Redox State and Increased Expression of Ferroptosis‐Suppressive Proteins

To explore ferroptosis‐related redox and lipid peroxidation features in primary CRC tumors, we analyzed T3‐stage specimens from CRC patients with or without liver metastasis. Because the clinical cohorts were small (n = 5 per group for GSH/GSSG and MDA assays; n = 3 per group for WB and IHC analyses) and were not fully balanced for N stage and, in the biochemical cohort, differentiation status, these analyses were considered exploratory. In primary tumors from patients with liver metastasis (CRC‐LM group), GSH levels were comparable to those in tumors from patients without liver metastasis (CRC‐nLM group), whereas GSSG showed a trend toward lower levels (Figure 1A,B). Consequently, the GSH/GSSG ratio was higher in CRC‐LM group primary tumors than in CRC‐nLM tumors (Figure 1C). MDA levels were also lower in CRC‐LM group primary tumors (Figure 1D). These biochemical readouts suggest a more reducing redox environment and a lower lipid peroxidation burden in CRC‐LM primary tumors, but they do not directly measure the basal level of ferroptotic cell death in vivo. We next examined the expression of the core ferroptosis‐suppressive proteins SLC7A11 and GPX4 by WB. CRC‐LM group primary tumors showed a borderline increase in SLC7A11 and a marked increase in GPX4 protein expression relative to CRC‐nLM tumors (Figure 1E,F). Collectively, the higher GSH/GSSG ratio, lower MDA level, and increased expression of SLC7A11/GPX4 indicate that CRC‐LM primary tumors exhibit a more reduced redox state and enhanced expression of core ferroptosis‐suppressive machinery, consistent with increased ferroptosis resistance potential rather than a directly measured lower level of ferroptotic cell death. These exploratory clinical observations therefore motivated subsequent transcriptomic and functional analyses.

Figure 1.

Primary CRC tumors from patients with liver metastasis show a reduced redox state and increased expression of ferroptosis‐suppressive proteins. (A–D) Glutathione redox and lipid peroxidation‐related readouts in primary tumor tissues. (A) Relative GSH level. (B) Relative GSSG level. (C) GSH/GSSG ratio. (D) MDA level. For (A–D), n = 5 per group. (E) Representative WB images and (F) quantification of the ferroptosis‐suppressive proteins SLC7A11 and GPX4 in tumor tissues. For (E and F), n = 3 per group. These assays reflect redox status, lipid peroxidation, and ferroptosis‐suppressive capacity, rather than direct measurement of in vivo ferroptotic cell death. Data are presented as mean ± SD. Statistical significance was determined by two‐tailed Student's t‐test. Significance levels: *p < 0.05, **p < 0.01, ns nonsignificant.

3.2. Identification of Key Genes Associated With Ferroptosis and Metastasis

The GSE62321 data set comprises 18 normal colon tissues, 20 primary CRC tumors, and 19 liver metastases, among which 10 patients had complete tripartite matching, with pairwise matches including 15 normal‐primary, 13 primary‐metastasis, and 11 normal‐metastasis pairs. Differential expression analysis identified 968 DEGs between primary CRC and normal tissues (Figure 2A,B) and 1130 DEGs between liver metastases and normal tissues (Figure 2C,D). The intersection of these two DEG sets yielded 755 genes consistently associated with CRC progression and metastasis. To characterize the functional profiles of these 755 genes, GO and KEGG enrichment analyses were performed, with the results detailed in Figure S1.

Figure 2.

Identification of key genes associated with ferroptosis and metastasis. (A) Volcano plot and (B) heatmap of primary tumor versus normal intestinal tissues. (C) Volcano plot and (D) heatmap of normal intestinal tissues versus liver metastases. (E) Venn diagram for the identification of key genes.

Based on the observed more reduced redox state, a condition closely linked to ferroptosis resistance, in primary tumors of metastatic CRC (Figure 1), we sought to identify key genes that are concurrently associated with ferroptosis and CRC metastasis. Using the 564 ferroptosis‐related genes obtained from FerrDb, a Venn analysis was performed against the 755 CRC progression‐associated genes, revealing 17 overlapping candidates (Figure 2E). After excluding two long noncoding RNAs, we focused on 15 protein‐coding genes: AMN, ANO6, DDR2, FABP4, GPT2, HDDC3, LIFR, NEDD4L, PDK4, PLIN4, PRKAA2, PSAT1, SNCA, SOX2, and USP7, which are implicated in both ferroptosis regulation and CRC metastasis.

3.3. Survival Prognosis Analysis

To evaluate their prognostic relevance, we analyzed the association between the expression of the 15 candidate genes and DFS in CRC patients using the GEPIA database. High expression of FABP4, SNCA, and DDR2 was significantly correlated with shorter DFS (Figure 3A–C), whereas the other 12 genes showed no significant association (Figure S2). Consequently, FABP4, SNCA, and DDR2 were selected as key genes for further study.

Figure 3.

Survival prognosis analysis. DFS curves based on (A) FABP4, (B) SNCA, and (C) DDR2 expression in patients. (D) Nomogram for predicting overall survival based on the expression of FABP4, SNCA, and DDR2. (E) Calibration plot showing the predictive performance of the model, and (F) Kaplan–Meier survival curves stratified by the risk scores derived from the nomogram.

Next, we constructed a prognostic nomogram by integrating the expression levels of these three genes with clinical variables including age, gender, and tumor stage (Figure 3D). The nomogram demonstrated good predictive performance, with a C‐index of 0.735 (Figure 3E). Subsequently, patients were stratified into high‐risk and low‐risk groups based on the risk scores derived from the nomogram. Survival analysis confirmed a significant divergence in clinical outcomes between these two risk groups (Figure 3F).

3.4. Functional Validation of FABP4, SNCA, and DDR2 in Ferroptosis Resistance and Migratory/Invasive Phenotypes in CRC Cells

To further validate the clinical relevance of these key genes, IHC analysis was performed on primary CRC tissues. This analysis revealed elevated protein expression of FABP4 and DDR2 in CRC‐LM tissues compared to CRC‐nLM tissues. Although the increase in SNCA expression did not reach statistical significance (p = 0.1012), a consistent upward trend was observed (Figure 4).

Figure 4.

IHC analysis of primary CRC tumors with or without liver metastasis. (A, C, E) Representative IHC images and (B, D, F) corresponding quantitative analyses are shown for (A and B) FABP4, (C and D) SNCA, and (E and F) DDR2 in primary tumor tissues from CRC patients with or without liver metastasis. Scale bar, 20 μm. Data are presented as mean ± SD (n = 3 per group). Statistical significance was determined by two‐tailed Student's t‐test. Significance levels: *p < 0.05, ***p < 0.001.

To investigate whether these key genes modulate ferroptosis‐associated lipid peroxidation and motility‐related phenotypes in CRC cells, we used SW620 cells as the primary in vitro model. SW620 was selected because it is a metastasis‐derived CRC cell line with robust migratory and invasive capacity; however, it is derived from a lymph‐node metastasis rather than a liver metastasis. Therefore, SW620 was used here as a surrogate model for metastatic CRC cell behavior. To determine whether the observations were not restricted to SW620 cells, MDA changes were examined in HCT116 and HT29 cells, and ferroptosis susceptibility, C11‐BODIPY lipid ROS, and migration/invasion phenotypes were further evaluated in HCT116 cells. Efficient knockdown of FABP4 and SNCA was achieved using siRNA, while shRNA was employed for DDR2 because of the limited efficacy of siRNA targeting this gene (Figure 5A,C,E). In SW620 cells, FABP4 and SNCA knockdown reduced GPX4 protein expression, whereas DDR2 knockdown did not reduce GPX4 but markedly decreased SLC7A11 expression (Figure 5A,C,E). These results suggest that FABP4/SNCA and DDR2 may influence ferroptosis defense through partially distinct mechanisms rather than through a single shared pathway. FABP4 and DDR2 knockdown increased MDA levels in SW620, HCT116, and HT29 cells, whereas SNCA knockdown increased MDA levels in HCT116 and HT29 cells and showed a similar trend in SW620 cells (p = 0.0659) (Figure 5B,D,F). Because MDA is a bulk lipid peroxidation‐related endpoint that can also reflect broader oxidative stress, these findings were interpreted as preliminary evidence of increased lipid peroxidation and were further complemented by ferroptosis‐relevant rescue assays and C11‐BODIPY lipid ROS staining.

Figure 5.

Knockdown of FABP4, SNCA, or DDR2 alters ferroptosis‐defense protein expression and MDA levels. (A–F) Analysis of ferroptosis‐defense proteins and MDA levels following knockdown of FABP4, SNCA, or DDR2 in CRC cells. Protein levels of GPX4 and SLC7A11 following knockdown of (A) FABP4, (C) SNCA, or (E) DDR2 in SW620 cells. Corresponding MDA levels after knockdown of (B) FABP4, (D) SNCA, or (F) DDR2 in CRC cells. MDA was used as a bulk lipid peroxidation‐related endpoint; C11‐BODIPY lipid ROS assays are shown in Figure 6 and Figure S3. Data are presented as mean ± SD (n = 3). Statistical significance was determined by two‐tailed Student's t‐test. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns nonsignificant.

To assess whether knockdown‐induced viability loss and lipid peroxidation were related to ferroptosis rather than only a nonspecific stress response, we performed ferrostatin‐1 rescue assays together with C11‐BODIPY staining. In SW620 cells, knockdown of each of the three genes reduced cell viability, and this reduction was partially or substantially reversed by ferrostatin‐1 treatment (Figure 6A,C,E). Consistently, C11‐BODIPY staining showed increased oxidized lipid signal after FABP4, SNCA, or DDR2 knockdown, and this signal was suppressed by ferrostatin‐1 (Figure 6B,D,F). These findings were largely reproduced in HCT116 cells (Figure S3A,B,D,E,G,H). Together, these data provide ferroptosis‐relevant evidence that FABP4, SNCA, and DDR2 modulate ferroptosis susceptibility in CRC cells. However, because liproxstatin‐1 rescue, iron chelation, ACSL4‐dependence experiments, and ferrostatin‐1 reversal of knockdown‐induced MDA were not performed, we avoid concluding that these genes exclusively regulate ferroptosis independent of other stress responses.

Figure 6.

FABP4, SNCA, and DDR2 knockdown increases C11‐BODIPY‐detected lipid ROS and ferrostatin‐1‐rescuable viability loss in SW620 cells. (A, C, E) Cell viability assessed by CCK‐8 assay in SW620 cells with or without ferrostatin‐1 treatment following knockdown of FABP4, SNCA, or DDR2. One‐way ANOVA overall p < 0.0001 for each panel. Data are mean ± SD. Multiple comparisons by Bonferroni's test for the three indicated comparisons. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns nonsignificant. (B, D, F) Lipid ROS detected by BODIPY 581/591 C11 staining under the same conditions. Scale bar, 50 μm. These assays assess ferrostatin‐1‐rescuable viability loss and lipid ROS accumulation after gene knockdown.

Because migration and invasion are in vitro surrogates of metastatic cell behavior [17, 18, 19], we first used RSL3 and erastin as pharmacological ferroptosis inducers to provide supportive context. SW620 cells were exposed to 2 μM RSL3 or 5 μM erastin for 48 h, followed by Transwell seeding at equal cell numbers; the same compounds were maintained during the subsequent 12‐h migration or 24‐h invasion period. This treatment increased MDA levels and reduced Transwell migration/invasion (Figure 7A,B). We then focused on gene‐knockdown/ferrostatin‐1 rescue experiments, for which CCK‐8 viability rescue data are shown in Figure 6A,C,E and Figure S3A,D,G. Ferrostatin‐1 was added 24 h after transfection and maintained during the Transwell assays. FABP4 knockdown reduced SW620 cell migration and invasion, and this phenotype was partially rescued by ferrostatin‐1 (Figure 7C). SNCA knockdown similarly impaired migration and invasion, and ferrostatin‐1 largely restored these phenotypes (Figure 7D). DDR2 knockdown preferentially impaired invasion rather than migration, and ferrostatin‐1 partially rescued the invasion defect (Figure 7E). Comparable effects on migration and/or invasion were observed in HCT116 cells (Figure S3C,F,I). These findings suggest that ferroptosis‐associated lipid peroxidation contributes to the migratory and/or invasive phenotypes regulated by FABP4, SNCA, and DDR2. Moreover, because SW620 is not liver‐metastasis‐derived and Transwell assays do not model liver colonization, these data should be considered as evidence for metastatic CRC cell behavior rather than liver‐metastasis‐specific activity.

Figure 7.

Ferroptosis induction and ferrostatin‐1 rescue affect migratory and invasive phenotypes in SW620 cells. (A) MDA levels in SW620 cells treated with 2 μM RSL3, a GPX4 inhibitor, or 5 μM erastin, a system x _c ⁻ inhibitor, for 48 h. (B) Transwell assays performed after 48 h of RSL3 or erastin exposure; equal numbers of cells were reseeded and the same drug concentrations were maintained during the subsequent 12‐h migration or 24‐h invasion period. Because no separate viability panel was included for the exact total exposure durations used in these assays, these data should not be interpreted as migration‐ or invasion‐specific effects independent of viability. C‐E Transwell assays assessing migration and invasion after knockdown of (C) FABP4, (D) SNCA, or (E) DDR2, with or without ferrostatin‐1 treatment. For (C–E), ferrostatin‐1 was added after gene knockdown and maintained during the Transwell assays. The partial rescue by ferrostatin‐1 supports involvement of ferroptosis‐associated lipid peroxidation but does not establish ferroptosis as the sole mechanism. These assays evaluate in vitro migratory and invasive phenotypes and do not directly model liver‐specific metastatic colonization. Data are presented as mean ± SD (n = 3). Statistical significance was determined by one‐way ANOVA with Dunnett's multiple comparisons test. Significance levels: *p < 0.05, **p < 0.01.

3.5. Exploratory Immune‐Infiltration and Pan‐Cancer Analyses

To provide broader context without overextending the mechanistic conclusions, we condensed the immune and pan‐cancer analyses in the main text and present the detailed plots in Supplementary Data S1. In TCGA CRC, FABP4, SNCA, and DDR2 expression showed positive correlations with TIMER‐estimated immune infiltration scores for multiple cell types (Figure S4), and related correlation patterns were observed in selected tumor types in pan‐cancer analyses (Figure S5). However, these results cannot determine whether the gene expression originates from tumor cells or from stromal, adipocyte, endothelial, fibroblast, or immune compartments [20, 21, 22]. This point is particularly important for FABP4, which may be expressed by multiple nontumor cell types in the tumor microenvironment.

Pan‐cancer expression and alteration analyses showed heterogeneous, cancer‐type‐dependent patterns (Figures S6–S8). FABP4, SNCA, and DDR2 were downregulated in many tumor‐versus‐normal comparisons, with notable exceptions in specific tumor types. We therefore do not interpret high expression of these genes as uniformly tumor‐promoting across cancers. Instead, the apparent difference between the pan‐cancer tumor‐versus‐normal pattern and the higher expression observed in CRC‐LM primary tumors likely reflects distinct biological comparisons: tumor initiation and tissue composition in pan‐cancer analyses versus metastatic competence within CRC. Thus, these analyses are presented as exploratory context and were not used as primary support for the core mechanistic model.

4. Discussion

In this study, we started from human specimens and observed that primary CRC tumors from patients with liver metastasis exhibited a more reduced redox state and increased expression of the ferroptosis‐suppressive regulators SLC7A11 and GPX4, features consistent with enhanced ferroptosis resistance potential. Guided by this clinical observation, integrative transcriptomic screening and survival analyses nominated FABP4, SNCA, and DDR2 as candidates linking ferroptosis regulation with CRC progression and CRC‐LM‐associated risk. In CRC cell models, knockdown of these genes increased bulk MDA levels and C11‐BODIPY‐detected lipid ROS, reduced cell viability in a ferrostatin‐1‐rescuable manner, and suppressed migratory and/or invasive phenotypes with partial ferrostatin‐1 rescue. These data support a role for FABP4, SNCA, and DDR2 in modulating ferroptosis susceptibility and motility‐related phenotypes, but they do not exclude contributions from broader stress‐response pathways. Because SW620 is derived from a lymph‐node metastasis, and because Transwell assays do not recapitulate liver colonization, the cell‐line data should be interpreted as functional evidence for metastatic CRC cell behavior.

An important implication of our work is that redox and ferroptosis‐suppressive features may be detectable in primary tumors from patients who present with liver metastasis. The three identified genes appear to affect ferroptosis‐related phenotypes through distinct biological entry points. FABP4 is a fatty acid‐binding protein involved in lipid trafficking and metabolism [23]. Because ferroptosis is driven by peroxidation of susceptible membrane lipids, lipid handling is closely connected to ferroptosis sensitivity [24, 25]. In our CRC models, FABP4 knockdown decreased GPX4 expression, increased MDA and C11‐BODIPY‐detected oxidized lipid signal, and produced ferrostatin‐1‐rescuable viability and motility defects. These observations are consistent with an anti‐ferroptotic role for FABP4, while additional work such as lipidomic profiling and ACSL4‐dependent rescue or epistasis assays will be needed to determine whether FABP4 acts directly on ferroptosis pathways or indirectly through broader lipid metabolic stress. FABP4 is also expressed in stromal and immune compartments; therefore, defining the cellular source of FABP4 in metastasis‐associated primary tumors will be important for clinical translation.

SNCA is best known for its roles in neuronal physiology and neurodegenerative disease [26], and emerging evidence links SNCA to lipid homeostasis and oxidative stress [27, 28]. In our CRC models, SNCA knockdown reduced GPX4 expression, increased lipid peroxidation‐related readouts, and caused anti‐migratory/anti‐invasive phenotypes that were largely rescued by ferrostatin‐1. These findings suggest that ferroptosis‐associated lipid peroxidation is an important component of the SNCA‐regulated phenotype. Nevertheless, because SNCA can influence membrane composition and oxidative stress more broadly, future studies should define whether SNCA regulates ferroptosis through direct remodeling of peroxidizable membrane lipids, altered GPX4 stability or expression, or other stress‐response pathways.

DDR2 is a collagen‐activated receptor tyrosine kinase that transduces extracellular matrix (ECM) cues into intracellular signaling [29, 30]. DDR2 behaved differently from FABP4 and SNCA: DDR2 knockdown did not reduce GPX4 expression but markedly decreased SLC7A11, suggesting a possible link to the system x _c ⁻/cystine uptake‐glutathione axis or other antioxidant pathways. Functionally, DDR2 knockdown preferentially impaired invasion, and ferrostatin‐1 partially restored invasion. This pattern suggests that ferroptosis‐associated lipid peroxidation may contribute to DDR2‐dependent invasive behavior, but it also points to a broader ECM‐signaling and invasion program that is unlikely to be explained by ferroptosis alone.

Beyond tumor cell‐intrinsic effects, we retained the immune‐infiltration and pan‐cancer analyses only as exploratory context. FABP4, SNCA, and DDR2 expression correlated with TIMER‐estimated immune infiltration scores in TCGA CRC and in selected pan‐cancer settings [20, 21, 22]; however, these associations are derived from bulk RNA‐seq deconvolution and cannot determine whether expression is tumor cell‐intrinsic or derived from stromal, adipocyte, endothelial, fibroblast, or immune compartments. This distinction is especially important for FABP4. Accordingly, these data generate hypotheses about interactions between ferroptosis‐related programs and the tumor microenvironment, but they do not independently reinforce the cell‐intrinsic mechanistic model [31, 32, 33, 34].

The pan‐cancer expression patterns also require context‐specific interpretation. FABP4, SNCA, and DDR2 were downregulated in many tumor‐versus‐normal comparisons, with cancer‐type‐specific exceptions. This does not imply that high expression uniformly promotes tumorigenesis. Rather, tumor‐versus‐normal comparisons capture changes associated with tumor initiation, tissue composition, and compartmental abundance, whereas the CRC‐LM versus CRC‐nLM comparison addresses metastatic competence within CRC. Therefore, genes that are decreased during tumor initiation in some tissues may still be enriched in specific metastatic contexts or in particular tumor‐microenvironment compartments. Cell‐type‐resolved studies will be required to clarify the biological source and functional meaning of these patterns.

Collectively, our data nominate FABP4, SNCA, and DDR2 as CRC‐LM‐associated candidates with functional effects on ferroptosis‐associated lipid peroxidation, ferrostatin‐1‐rescuable viability loss, and migratory/invasive phenotypes in CRC cells. We therefore present them as candidate regulators of ferroptosis susceptibility and metastatic CRC cell behavior. The pharmacological RSL3/erastin Transwell findings are supportive only, because they may be partly confounded by treatment‐associated cytotoxicity, and the immune/pan‐cancer analyses provide hypothesis‐generating context rather than mechanistic proof. From a diagnostic standpoint, a small gene panel may be amenable to clinical implementation by immunohistochemistry or targeted transcript assays and may complement conventional staging in future validation cohorts. From a therapeutic standpoint, these genes represent biologically distinct nodes—lipid handling (FABP4), membrane/redox regulation (SNCA), and ECM‐sensing signaling (DDR2)—that warrant further investigation in CRC‐LM‐specific and viability‐controlled models.

5. Limitations

This study has several limitations. First, the clinical specimen analyses were performed in limited cohorts. Although the cases were matched for T stage, the CRC‐LM and CRC‐nLM groups were not fully balanced for N stage and, in the biochemical cohort, differentiation status. Second, the primary in vitro model used in this study was SW620, which is derived from a lymph‐node metastasis rather than a liver metastasis. Although selected findings were validated in HCT116 cells and MDA changes were also assessed in HT29 cells, we did not perform a complete paired SW480/SW620 comparison, nor did we use CRC‐LM‐derived organoids or in vivo liver metastasis models. Third, although we included C11‐BODIPY lipid ROS staining and ferrostatin‐1 rescue of viability and lipid peroxidation as ferroptosis‐relevant assays, we did not test liproxstatin‐1, iron chelation such as deferoxamine, ACSL4 dependence, or ferrostatin‐1 reversal of knockdown‐induced MDA. Fourth, we did not include a separate viability assay for the exact RSL3/erastin total exposure durations used in the inducer‐only Transwell experiments. Therefore, these inducer‐only Transwell results may reflect combined effects of ferroptosis induction, cytotoxicity, and motility changes, and should not be interpreted as migration‐ or invasion‐specific effects independent of viability. Future studies should incorporate multiple orthogonal ferroptosis‐specific readouts, lipidomic analysis, canonical pathway epistasis, single‐cell or spatial validation, viability‐controlled migration assays, and in vivo liver colonization or spontaneous metastasis models to determine whether these genes directly regulate liver‐specific metastatic outgrowth through ferroptosis‐dependent mechanisms.

6. Conclusion

Our study identifies FABP4, SNCA, and DDR2 as CRC‐LM‐associated ferroptosis‐related candidates that influence bulk lipid peroxidation, C11‐BODIPY‐detected lipid ROS, ferrostatin‐1‐rescuable viability loss, and migratory/invasive phenotypes in CRC cell models. These findings provide a rationale for further investigation of ferroptosis susceptibility as a potential vulnerability in metastatic CRC.

Author Contributions

Zheng Ge: data curation, formal analysis, investigation, visualization, writing – original draft. Wei Guo: conceptualization, methodology, project administration, validation, writing – review and editing. Jingxin Li: project administration, resources, supervision, writing – review and editing. Yanqing Wang: conceptualization, methodology, supervision, writing – review and editing. Kexin Wang: funding acquisition, resources, supervision, writing – review and editing.

Ethics Statement

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Qilu Hospital of Shandong University (KYLL‐2026‐01‐029).

Consent

All participants provided written informed consent for the use of their tissue specimens.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.