CLCF1 is associated with perineurial invasion and prognosis in colorectal cancer

Abstract

Perineural invasion (PNI) is implicated in tumor invasion, metastasis, cancer-related pain, and adverse clinical outcomes across various malignancies. Nonetheless, the core genes and molecular mechanisms underlying PNI in colorectal cancer (CRC) remain inadequately understood. The objective of our study was to identify key genes associated with PNI in CRC through bioinformatic analysis. Transcriptomic and clinical data for CRC were obtained from the Cancer Genome Atlas (TCGA) dataset. Using the limma package, we identified 2,446 differentially expressed genes (DEGs) between CRC samples with and without PNI. Additionally, 1,793 immune-related genes (IRGs) were obtained from the ImmPort database. The intersection of differentially expressed immune-related genes (DEIRGs) was subsequently employed for functional enrichment analysis and the development of risk models. A novel prognostic model related to PNI was successfully developed and validated, demonstrating its independent prognostic value in CRC. Cardiotrophin-like cytokine factor 1 (CLCF1) was identified as a key prognostic marker associated with PNI in CRC, with elevated expression observed in tumor samples. CLCF1 was shown to enhance the proliferation, invasion, and migration capabilities of CRC cells, suggesting its potential as a therapeutic target for CRC.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-15358-1.

Introduction

Colorectal cancer (CRC) is the third most common cancer globally and the second leading cause of cancer-related mortality¹. The 5-year survival rate for patients with advanced CRC is less than 30%, with recurrence and metastasis remaining the primary causes of death in these patients. Perineural invasion (PNI) is recognized as an independent risk factor for poor prognosis^2,3, being associated with an elevated risk of disease recurrence and lymph node metastasis in CRC patients^4,5. Consequently, it is imperative to investigate the underlying molecular mechanisms of PNI in CRC.

PNI is characterized by tumor cells encircling at least one-third of the nerve bundle’s circumference or infiltrating any layer of the nerve sheath⁶. Initially identified in head and neck tumors, PNI has also been documented in pancreatic cancer, cholangiocarcinoma, colorectal cancer, gastric cancer, prostate cancer, breast cancer, and thyroid cancer^7,8. The detection rate of PNI in CRC varies between 2.3% and 33%^9,10. PNI encompasses several critical stages: cancer cell survival, nerve homeostasis, inflammatory response, chemotaxis of cancer cells towards nerves, nerve regeneration, adhesion of cancer cells to the nerve sheath, and ultimately, nerve invasion⁷. Despite the growing recognition of this transfer process, the specific driving genes and molecular pathways remain to be elucidated.

Cardiotrophin-like cytokine factor 1 (CLCF1), a member of the interleukin-6 (IL-6) family, functions as a potent neurotrophic factor, B-cell stimulator, and neuroendocrine regulator of pituitary corticotropin function¹¹. The expression level of CLCF1 was up-regulated in wild-type mice with sciatic nerve injury, and this up-regulation promoted axon growth¹². Zhang Z et al. demonstrated that the miR-30a-5p/CLCF1 axis could regulate sorafenib resistance and glycolysis in hepatocellular carcinoma (HCC) cells¹³. Furthermore, Li Y et al. reported that miR-3065-3p enhanced the stemness and metastasis of CRC cells by targeting CRLF1¹⁴. However, no studies have investigated the impact of CLCF1 on the biological behavior of CRC.

In this study, we obtained the mRNA expression profiles and corresponding clinical data of CRC patients from a public database. Subsequently, we identified a prognostic signature using differentially expressed immune-related genes (DEIRGs) within the Cancer Genome Atlas (TCGA) cohort and validated the stability and reliability of this model through an internal validation cohort. Functional enrichment analysis was conducted to elucidate potential mechanisms underlying the identified signature. Our findings indicate that CLCF1 serves not only as a novel biomarker implicated in PNI of CRC but also influences the biological phenotypes of CRC cells, suggesting its potential utility in future cancer therapies.

Results

PNI is a prognostic factor in CRC

The baseline characteristics of the 198 CRC patients included in the TCGA database are as follows (Table S1). The flowchart of this study is shown in Fig. 1. PNI was identified in 26.26% of tumors (52 out of 198 patients) within the TCGA-CRC cohort (Fig. 2A). The prevalence of PNI was notably higher in tumors with advanced stages, being observed in 10.8% of stage I, 26.38% of stage II, 25% of stage III, and 52% of stage IV tumors. Additionally, PNI was associated with elevated serum carcinoembryonic antigen levels (CEA, Fig. 2B). Mortality rates among CRC patients with PNI were higher compared to those without PNI. Furthermore, patients exhibiting PNI demonstrated significantly poorer overall survival (OS) relative to these without PNI counterparts. These findings underscore that the presence of PNI in CRC is strongly correlated with aggressive tumor characteristics and adverse clinical outcomes (Fig. 2C). Patients with PNI had worse OS when compared to those without PNI (HR = 2.30, 95%CI 1.07–4.95, P = 0.03, Fig. 2D). Thus, we demonstrate that the presence of PNI in CRC is strongly associated with aggressive tumor behavior and poor clinical outcomes.

Fig. 1.

Flowchart of the present study. TCGA, The Cancer Genome Atlas; COAD, Colon adenocarcinoma; READ, Rectal adenocarcinoma; DEGs, Differentially expressed genes; DEIRGs, Differentially expressed immune-related genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; LASSO, Least absolute shrinkage and selection operator; ROC, Receiver operating characteristic.

Fig. 2.

PNI is a prognostic factor in CRC. (A) Prevalence of PNI in the TCGA-CRC cohort. (B) Prevalence of PNI in CRC stratified by tumor stage and CEA levels. (C) Mortality rate of CRC patients stratified by presence of PNI. (D) Kaplan-Meier plots of CRC patients stratified by presence of PNI. PNI, Perineural invasion; CRC, Colorectal cancer; CEA, Carcinoembryonic antigen; OS, Overall survival. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant.

Identification of PNI-related DEIRGs in CRC

We identified 2,446 differentially expressed genes (DEGs) between patients with and without PNI groups in CRC using the limma package (Table S2). A total of 1,793 immune-related genes (IRGs) were retrieved and compared with the list of DEGs, resulting in 186 overlapping DEIRGs (Fig. S1A). The expression profiles of these 186 genes are presented in Fig. S1B.

Enrichment analysis of DEIRGs in CRC

The functional enrichment analysis of the 186 DEIRGs was conducted. Gene Ontology (GO) analysis encompassed three categories: biological processes (BPs), molecular functions (MFs), cellular components (CCs). In terms of BP, these genes were predominantly implicated in cell activation, differentiation, and proliferation, immune processes, cytokine signaling pathways, inflammatory responses, and vasculature development (Fig. S1C). Within the CC category, these genes were enriched in membrane components, cell surfaces, secretory granules and vesicles, receptor complexes, external encapsulating structures, collagen-containing extracellular matrices, and cell-substrate junctions (Fig. S1D). Regarding MF, these genes were involved in signaling receptor binding, molecular transducer activity, receptor regulatory activity, cytokine activity and immune receptor activity (Fig. S1E). Additionally, 15 Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (www.kegg.jp/kegg/kegg1.html) associated with these DEGs were enriched and are illustrated in the bubble plot. Notably, the pathways involved include Cytokine-cytokine receptor interaction, MAPK signaling pathway, PI3K-Akt signaling pathway, and Ras signaling pathway (Fig. S1F).

Construction and evaluation of PNI-related risk model in CRC

From the 186 DEIRGs, 10 genes associated with survival were identified through univariate Cox analysis of OS (P < 0.05) (Fig. S2). A least absolute shrinkage and selection operator (LASSO) Cox regression analysis was then employed to compute a risk score based on the expression levels of these 10 DEIRGs (Fig. 3A). The model demonstrated optimal performance when 7 genes were incorporated (Fig. 3B). The patients were stratified into high- and low-risk cohorts based on the median risk score. Principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE) analysis indicated that the two patient groups were distinctly separated into two clusters (Fig. 3C-D). Distribution plots of risk scores, outcome statuses, and gene signature expression profiles are presented in Fig. 3E. The high-risk cohort exhibited a higher incidence of mortality compared to the low-risk cohort. Heatmap analyses revealed that in high-risk cases, the genes SLPI, CRABP2, IRF7, UNC93B1, CLCF1, and GDF9 were overexpressed, whereas NUDT6 expression was downregulated. Similarly, the Kaplan-Meier plot indicated a higher probability of survival in the high-risk group compared to the low-risk group (Fig. 3F). To evaluate the predictive efficacy of the risk score for OS, we employed the receiver operating characteristic curve (ROC), yielding an area under the ROC curve (AUC) of 0.73 at 1 year, 0.84 at 3 years, and 0.75 at 5 years (Fig. 3G).

Fig. 3.

Establishment of a prognostic gene signature using LASSO regression analysis. (A) LASSO coefficient profiles of 10 PNI-related potential prognostic genes. (B) Tenfold cross-validation for tuning parameter selection in the LASSO model. (C-D) PCA and t-SNE analysis of the TCGA cohort. (E) The distribution of the risk score, and the survival status of patients, heatmap of the 7 genes. (F) Kaplan-Meier curves for CRC patients in the TCGA cohort. (G), AUC of time-dependent ROC curves in the TCGA cohort. PCA, Principal component analysis; t-SNE, T-distributed stochastic neighbor embedding; AUC, Area under the ROC curve.

Validation of the prognostic model in internal test cohort

The risk score was calculated using the previously established formula within the internal validation cohort, subsequently categorizing patients into either high-risk or low-risk groups. In both internal validation cohorts, PCA and t-SNE analysis confirmed a robust difference in the distribution between the two subgroups (Fig. S3A-B, S4A-B). High-risk individuals exhibited a greater number of mortality events compared to low-risk individuals. The heatmaps revealed that in high-risk cases, SLPI, CRABP2, IRF7, UNC93B1, CLCF1 and GDF9 were overexpressed, whereas NUDT6 expression was downregulated (Fig. S3C, S4C). Additionally, patients in the high-risk group demonstrated a lower survival rate relative to those in the low-risk group (Fig. S3D). The AUC values for the risk score in predicting OS at 1, 3, and 5 years were 0.7, 0.85, and 0.88, respectively, within the internal test I cohort (Fig. S3E). Consistent with the results from the test I cohort, CRC patients with higher risk scores had a poorer prognosis than those with lower risk scores in the test II cohort (Fig. S4D). The AUCs for risk score in predicting OS were 0.75 at 1 year, 0.84 at 3 years, and 0.64 at 5 years (Fig. S4E). Our findings demonstrated that the seven-gene signature could effectively stratify patient prognosis.

Clinical correlation analysis of the prognostic model

Independent risk factors for OS included stage, T stage, N stage, M stage, lymphatic invasion (LI), PNI, and vascular invasion (VI, Fig. 4A). The risk score was also significantly associated with OS (HR = 3.77, 95% CI, 2.36–6.03, P < 0.001, Fig. 4A). After adjusting for other confounding variables, the seven-gene signature remained a robust prognostic factor, second only to metastasis (HR = 3.33, 95% CI, 1.934–5.773, P < 0.001, Fig. 4B). Violin plots illustrated that the calculated risk score was positively correlated with tumor infiltration depth, lymph node metastasis, distant metastasis, stage, LI, and VI (Fig. 4C-H).

Fig. 4.

Clinical features and prognostic value of the risk score. (A) Univariate Cox regression analyses in the TCGA cohort. (B) Multivariate Cox regression analyses in the TCGA cohort. (C-G) The risk score at different T stage, N stage, M stage, pathological stage, lymphatic invasion, venous invasion.

Tumor microenvironment analysis

Single sample Gene Set Enrichment Analysis (ssGSEA) was subsequently conducted to investigate the immune microenvironment across various risk score groups. A heat map was generated to illustrate the correlations among 28 immune infiltration cell types (Fig. 5A). Both differential and correlation analyses were employed to examine the immune infiltration characteristics between high-risk and low-risk groups. The integration of findings from these analyses identified six tumor-infiltrating immune cells associated with the risk signature. Among the six tumor-infiltrating immune cells (TICs), plasmacytoid dendritic cell, natural killer T cell, natural killer cell, regulatory T cell and T follicular helper cell were positively correlated with the risk score; Type 17 T helper cell was negatively correlated with the risk score (Fig. 5B-C). Furthermore, patients with Microsatellite Stable (MSS) status exhibited higher risk scores compared to those with Microsatellite Instability-Low (MSI-L) or Microsatellite Instability-High (MSI-H) status (Fig. 5D).

Fig. 5.

Correlation analysis between the risk scores and the immune microenvironment. (A) The ratio differentiation of 28 TICs between the high- and low-risk groups. (B) The correlations between the TICs and risk score. (C) 6 TICs identified from the intersection of correlation analysis and difference analysis. (D) Relationship between risk score and microsatellite instability. (E) The expression of immune check points in the two groups. TICs, Tumor-infiltrating immune cells.

Immune checkpoints (ICPs) are crucial molecules expressed by immune cells that modulate immune activation. In this study, we examined the correlation between CRC and 14 key ICPs. Our findings revealed that the expression levels of PD-L1, TIM3, CD47, CD80, CD86, CD155, and Galectin-9 were significantly elevated in the high-risk CRC group compared to the low-risk CRC group (Fig. 5E). These results suggest that abnormal immune infiltration and the heterogeneity of immune infiltration in CRC may serve as prognostic indicators and potential targets for immunotherapy, thereby holding substantial clinical implications.

Nomogram model for CRC patients

The nomogram was successfully constructed to quantify the prognostic risk assessment of CRC patients (Fig. 6A). The C-index for the nomogram was 0.81. Calibration plots demonstrated that the predicted 1-, 3-, and 5-year OS probabilities were es closely aligned with the ideal performance, as shown in Fig. 6B. Subsequently, decision curve analysis (DCA) was conducted to validate the predictive value of the nomogram. DCA revealed that the net benefit of our prognostic models surpassed that of the previous risk score model across a broad range of threshold probabilities (Fig. 6C).

Fig. 6.

Nomogram model for colorectal cancer in TCGA cohort. (A) Nomogram model to predict 1-, 3-, and 5-year survival for CRC. (B) Calibration graphs to evaluate the performance of this nomogram. (C) DCA curves of the nomogram. DCA, Decision curve analysis.

Expression and survival verification of the PNI-related genes in CRC

The expression levels of CRABP2 and CLCF1 were up-regulated, whereas NUDT6 was down-regulated in tumor tissue compared to adjacent normal tissue (Fig. 7A). The expression level of CLCF1 was found to be highest in tumor tissues exhibiting PNI, followed by tumor tissues without PNI, and was lowest in adjacent normal tissues. In contrast, the expression level of NUDT6 was lowest in samples lacking PNI (Fig. 7B). Kaplan-Meier survival analysis demonstrated that patients with higher levels of NUDT6 and lower levels of CLCF1 had a more favorable prognosis (Fig. 7C-D). Consequently, our study suggests that CLCF1 and NUDT6 may play significant roles in the perineural invasion of CRC. Additionally, we examined the expression levels of CLCF1 in four different CRC cell lines and normal human colon mucosal epithelial cells. The expression of CLCF1 was significantly elevated in HCT116, SW480, HT29 and SW620 cells, compared to NCM460 cells (Fig. 7E-F).

Fig. 7.

Expression analysis and Kaplan-Meier survival analysis of the seven PNI-related genes. (A) Expression analysis of the seven genes between tumor tissue and the adjacent normal tissue. (B) Expression analysis of the seven genes in with PNI tumor tissue, without PNI tumor tissue and the adjacent normal tissue. (C) Kaplan-Meier plots of OS of colorectal cancer patients stratified by CLCF1 expression. (D) Kaplan-Meier plots of OS of colorectal cancer patients stratified by NUDT6 expression. (E-F) CLCF1 mRNA and protein expression levels in CRC cells. Original western blot for Fig. 6F was presented in Supplementary Fig. 5.

CLCF1 knockdown inhibited proliferation of CRC cells

To further investigate the biological function of CLCF1, siRNA was employed to generate two independent CLCF1 knockdown cell lines (Si1# and Si2#) alongside control cells (NC, Fig. 8A-D). Following CLCF1 knockdown, the proliferation and colony-formation capacities of CRC cell lines were markedly reduced (Fig. 8E-G). Collectively, these findings suggest that CLCF1 may serve as a critical regulator of proliferation in CRC cell lines.

Fig. 8.

CLCF1 knockdown inhibits proliferation of colorectal cancer cells. (A-D) CLCF1 mRNA and protein expression levels of siRNA-mediated knockdown in HCT116 and SW480. (E-F) Time-dependent cell viability of CRC cells with CLCF1 knockdown measured by CCK-8 assay. (G) Representative images and quantitation of colony formation of CRC cells with CLCF1 knockdown cells of HCT116 and SW480. Original western blot for Fig. 7B and D were presented in Supplementary Fig. 5.

CLCF1 knockdown inhibited invasion and migration of CRC cells

Furthermore, the wound-healing rates, as well as the migratory and invasive abilities of HCT116 and SW480 cells, were significantly diminished upon CLCF1 knockdown. (Fig. 9). These findings underscore the significant role of CLCF1 as a contributor to metastasis in CRC cells.

Fig. 9.

CLCF1 knockdown inhibits migration and invasion of colorectal cancer cells. (A-B) Representative wound-healing images and quantitative analysis of wound-healing rates of HCT116 and SW480 cells with CLCF1 knockdown. (C-D) Representative images and quantitative analysis of stained migrated or invaded HCT116 and SW480 cells with CLCF1 knockdown. Scale bars: 200 μm.

Discussion

In this study, we demonstrated that presence of PNI is indicative of aggressive tumor behavior and poor clinical outcomes in CRC, corroborating the results of previous research. Then, we utilized public databases from the TCGA develop a seven-gene signature associated with PNI. Univariate and multivariate Cox regression analyses further confirmed that the signature-based risk score serves as an independent predictor of OS. Analysis of immune infiltration highlighted the crucial role that six tumor-infiltrating immune cells may play in supporting the seven-gene signature associated with CRC prognosis. Additionally, we developed a nomogram by incorporating variables such as age, gender, stage, N stage, M stage, and risk scores associated with this signature, thereby providing clinicians with a reliable and user-friendly tool for predicting the survival outcomes of CRC patients. In contrast to previous studies, our research is pioneering in its utilization of PNI-related genes. Furthermore, we conducted both expression and survival analyses to validate the selected seven genes, identifying CLCF1 as potentially playing a crucial role in the PNI of CRC. This study offers novel targets for the diagnosis and treatment of colorectal cancer.

The pathological mechanism of PNI is attributed to the interplay of neurotrophic factors, neurotransmitters, adhesion molecules, Schwann cells, and stromal cells^10,15,16. Recent investigations have demonstrated that neurotrophic factors, including NGF and BDNF, facilitate pancreatic nerve invasion^17–19. In the context of pancreatic cancer, substance P has been shown to enhance cancer cell proliferation, migration, and PNI through activation of the NK-1R/Akt/NF-κB signaling pathway²⁰. Furthermore, the adhesion molecule L1CAM activates STAT2, leading to the upregulation of matrix metalloproteinases MMP2 and MMP9, thereby promoting pancreatic cancer invasion²¹. Duchalais et al. identified that colorectal cancer tumor cells migrate along intestinal neurons via L1CAM and N-cadherin-mediated tumor-nerve adhesion. This mechanism may constitute the pathway for perineural invasion of colorectal cancer, given the extensive distribution of intestinal nerves in the colon and rectum²². Additionally, Schwann cells have been shown to facilitate nerve invasion in pancreatic and cervical cancers^23,24. Studies, including those by Deborde et al., have demonstrated that Schwann cells can disrupt the cohesion of tumor cell lines, transforming spherical cancer cell aggregates into chain-like structures. NCAM1 is significantly expressed at cell contact sites, facilitating the formation of projections and migration of tumor cells towards Schwann cells²⁵. The density of tumor-associated macrophages (TAMs) in the PNI group was higher than that in the without PNI group. An increased infiltration of TAMs in PNI was correlated with a poorer prognosis in patients with pancreatic ductal adenocarcinoma. Elevated Ach expression by nerves suppressed the expression of CCL5, subsequently impairing T cell recruitment and T cell-mediated anti-tumor immunity^26,27.

The ssGSEA analysis revealed that high-risk CRC patients exhibited elevated proportions of plasmacytoid dendritic cells, natural killer T cells, natural killer cells, regulatory T cells, and T follicular helper cells, alongside reduced proportions of type 17 T helper cells, in comparison to low-risk patients. Furthermore, high-risk patients demonstrated significantly increased expression of immunosuppressive molecules (PD-L1, TIM3, CD47, CD80, CD86, CD155, and Galectin-9), relative to the low-risk group. These findings suggest that the attenuated immune phenotype observed in the high-risk cohort may be attributed to a more pronounced immunosuppressive environment and heightened immune checkpoint expression compared to the low-risk group. Therefore, the risk model may serve as an indicator of the intensity of the immune response in CRC.

PNI is recognized as a poor prognostic factor in patients with colorectal cancer, correlating with reduced survival rates and an elevated risk of recurrence. NDRG4, an early detection marker for colorectal cancer, is specifically expressed in intestinal neurons. The deletion of NDRG4 in these neurons leads to an increased release of extracellular matrix proteins Nid1 and Fbln2, which in turn contributes to the progression of colorectal cancer^28,29. NGF is markedly overexpressed in colorectal cancer tissues. The interaction of NGF with phosphorylated TrkA receptor initiates the MAPK/ERK signaling pathway, thereby facilitating colorectal cancer metastasis³⁰. Han J et al. identified that NE augments the proliferation, invasion, and metastasis of colorectal cancer cells via activation of the CREB1-miR-373 axis³¹. Li et al. elucidated that miR-3679-5p, through its interaction with long non-coding RNA CYTOR, potentially enhances PNI by upregulating MACC1 expression in colorectal cancer³². Previous studies have indicated that this pathway is upregulated in pancreatic cancer patients with severe neuroinvasive disease, contributing to neural invasion³³. However, the precise role of nerves in CRC and the interactions between CRC cells and nerves have been relatively understudied.

In this study, we identified CLCF1 as a key gene associated with PNI in CRC patients through comprehensive bioinformatics analysis at the molecular level. The impact of CLCF1 on the biological behavior of CRC cells was subsequently validated through in vitro experiments. The findings demonstrated that CLCF1 promotes proliferation, invasion, migration, and other malignant biological behaviors of tumor cells. Currently, the role of CLCF1 in tumors is predominantly associated with the immune microenvironment, with a particular focus on lung cancer and liver cancer. In the context of non-small cell lung cancer, Zeng Z et al. integrated extracellular matrix-related genes to develop a prognostic model, revealing that CLCF1 secreted by CAFs was up-regulated in the high-risk group³⁴. In vitro experiments have demonstrated that mouse fibroblasts can secrete CLCF1, thereby promoting tumor progression. Furthermore, targeting the CLCF1-CNTFR signaling pathway with specific receptors has been shown to induce anti-tumor effects^35,36. In HCC, the secretion of CLCF1 by CAFs facilitates tumor cell desiccation, as well as the infiltration and polarization of tumor-associated neutrophils, thereby advancing tumor progression via the CLCF1-CXCL6/TGF-β axis. This mechanism contributes to a poor prognosis for liver cancer patients³⁷. Jiang et al. demonstrated that CLCF1 expression serves as an independent prognostic factor in glioma, with elevated levels correlating with unfavorable outcomes. Furthermore, glioma patients exhibiting low CLCF1 expression tend to show a more robust response to PD-1/PD-L1 immunotherapy³⁸. The findings suggest that CLCF1 may play a role in guiding immunotherapy.

However, this study has several limitations. Firstly, due to the paucity of information on PNI in many online databases, expression and survival analyses were confined to the TCGA cohort, precluding external dataset validation. Furthermore, the analysis may be subject to potential selection bias due to limited data availability. Secondly, the robustness and clinical applicability of the seven-gene signature require further validation through prospective trials. Additionally, further biological experimental research is necessary to elucidate the underlying mechanisms linking the selected gene in CRC to cancer hallmarks.

In this study, a prognostic model associated with PNI in CRC patients was developed utilizing bioinformatics approaches. The CLCF1 was identified as a potential biomarker for the diagnosis and prediction of PNI, thereby offering a prospective therapeutic target for CRC management.

Materials and methods

Data collection

Clinical and transcriptome data for TCGA COAD (Colon adenocarcinoma) and READ (Rectal adenocarcinoma) patients were sourced from the UCSC Xena (University of California Santa Cruz cancer database, https://xena.ucsc.edu/). PNI is characterized by tumor cells encircling at least one-third of the nerve bundle’s circumference or infiltrating any layer of the nerve sheath. The PNI status was extracted from the original TCGA pathology reports. Samples that lacked clinical or survival information were excluded from the analysis. IRGs data were obtained from the ImmPort database (https://immport.niaid.nih.gov).

Identification of DEIRGs

DEGs between tumor samples with PNI and those without PNI were identified using the limma package, with a threshold of |logFC| >0, P < 0.05 for statistical significance. DEIRGs were defined as the intersection of DEGs and IRGs.

Functional enrichment analysis

GO and KEGG pathway enrichment analysis were conducted using the clusterProfiler package in the R software. GO was used to annotate DEIRGs from BPs, MFs, CCs. KEGG is developed by Kanehisa Laboratories^39,40. The significance threshold was set at P < 0.05.

Construction and validation of the PNI-related risk model

The identified DEIRGs were used to construct LASSO Cox regression model. Each patient’s risk score can be obtained from the following formula: Risk score = 0.1343 × expression of SLPI + 0.0883 × expression of CRABP2 + 0.3100 × expression of IRF7 + 0.2282 × expression of UNC93B1 + 0.1059 × expression of CLCF1 + 0.3909 × expression of GDF9–0.7306 × expression of NUDT6. CRC patients were stratified into low- and high-risk groups based on the median risk score. PCA and t-SNE analysis were conducted to investigate the distribution of different clusters according to the expression profiles of the genes in the signature. Kaplan-Meier survival analysis and ROC analysis were employed to further evaluate the prognostic capability of the gene signature. For internal validation, the cohort was divided into two subsets, test cohort I and test cohort II, in a 1: 2ratio. Additionally, independent prognostic analysis was performed to determine whether this gene signature serves as an independent prognostic factor. Chi-square test was used to analyze the clinical characteristics proportion in the two groups. The violin plots were performed to show the relationships between the risk score and different clinical characteristics.

Correlation analysis between risk scores and TME

We conducted Single ssGSEA utilizing the R package GSVA to quantify the infiltration levels of 28 distinct immune cell types. The results were subsequently visualized using a heatmap. Additionally, violin plots were generated to illustrate the relative abundance of immune cells within high-risk and low-risk groups. The correlation between risk scores and immune cell infiltration was assessed using Pearson’s correlation method.

Establishment of nomogram

In this study, the R package rms was utilized to construct a nomogram incorporating a seven-gene signature alongside other clinical features, aimed at assessing the probability of OS in CRC patients at 1, 3, and 5 years. The prognostic and diagnostic accuracy of the nomogram was evaluated using the concordance index and calibration curves.

Expression and survival verification of key genes

Survival analysis was performed employing the Kaplan-Meier method and log-rank tests. Additionally, the R package maxstat was employed to determine the optimal cutoff value. Violin plots were generated to illustrate the differential gene expression levels between adjacent normal tissue samples and CRC samples. To compare the differential expression of the seven prognostic PNI-related genes between normal and colorectal tumor tissues, a Wilcoxon rank-sum test was conducted.

Cell lines and cell culture

The human colon carcinoma cell lines HT29, HCT116, SW480, and SW620, along with the normal colonic epithelial cell line NCM460, were obtained from Procell Life Science & Technology Company. All cell lines were maintained in DMEM (Gibco) supplemented with 10% FBS (Hyclone) and 1% penicillin/streptomycin, and incubated at 37 °C in a humidified atmosphere containing 5% CO2.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted using the FastPure Cell/Tissue Total RNA Isolation Kit V2 and subsequently reverse-transcribed with reverse transcription reagents (Vazyme). qRT-PCR was then performed on a CFX96 Touch System (Bio-Rad) with a total reaction volume of 20 µL. β-actin served as the internal reference for the analysis. The primers used in qRT-PCR are as follows:

β-actin,

F: 5’-ACAGAGCCTCGCCTTTGCC-3’.

R: 5’-GATATCATCATCCATGGTGAGCTGG-3’

CLCF1,

F: 5’-TTTCAACGAGCCAGACTTCAAC-3’.

R: 5’-GAGGCCACGCAAGTAACACA-3’

Western blot

Cells were lysed using RIPA buffer, and protein concentrations were determined using the standard BCA method. Equal amounts of protein were separated by SDS-PAGE and subsequently transferred onto PVDF membranes. The membranes were blocked and incubated with the specified primary antibodies (β-actin, Abclonol; CLCF1, Abmart), followed by incubation with secondary antibodies. Visualization of the bands was achieved using a Bio-Rad Infrared Imaging System.

SiRNA transfection

CLCF1 expression was knocked down using siRNAs via transfection with the Genemute transfection reagent (SignaGen). The siRNAs were synthesized by GenePharma. The sequences of the CLCF1 siRNAs are as follows:

CLCF1-Si1#,

5’-GGACCUAUCUGAACUACCUTT-3’.

5’-AGGUAGUUCAGAUAGGUCCTT-3’.

CLCF1-Si2#,

5’-UCAACCGGCUCAAGAAGAATT-3’.

5’-UUCUUCUUGAGCCGGUUGATT-3’.

CCK-8 viability assay

Cells were cultured in 96-well plates transfected with the appropriate plasmids or siRNAs, and were analyzed at designated time points using CCK8 Kits (Vazyme). Following the addition of the CCK8 reagent, cell viability was assessed (0 h, 24 h, 48 h, 72 h, 96 h and 120 h). The OD value was measured at 450 nm using a Multimode Plate Reader (SpectraMax CMAX Plus, Molecular Devices).

Colony formation assay

Cells (500 per well) were seeded in 6-well plates and cultured for 14 days to allow for colony formation. Subsequently, the cells were stained with 1% crystal violet (AMRESCO), washed with tap water, and photographed for counting.

Wound-healing assay

Cells were seeded in six-well plates (5 × 10⁵ cells/well) and grown to 100% confluence. A scratch wound was generated using a 10-µl pipette tip, followed by replacement of the medium with serum-free medium. Scratch wounds were photographed at specified time points.

Cell invasion and migration assays

Cells (5 × 10⁴ cells/well) were seeded in the upper chamber, either with (Transwell invasion assay) or without (Transwell migration assay) Matrigel (BD Biosciences), while growth media containing 20% FBS were added to the lower chamber. After a 24-hour incubation period, the cells were fixed with 4% PFA, stained with 0.1% crystal violet, and subsequently counted under a microscope.

Statistical analysis

Statistical analysis and data visualization were conducted using R software (version 4.1.2, http://www.r-project.org). Kaplan-Meier survival analysis was performed utilizing the survival and survminer packages. ROC analysis was conducted using the survivalROC package, with the AUC serving as an indicator of prognostic accuracy. Univariate and multivariate Cox proportional hazards regression analyses were performed utilizing the survival package. The Wilcoxon signed-rank test was employed for comparisons of means between two groups, while the Kruskal-Wallis test was used for comparisons involving more than two groups. Categorical data were analyzed using the chi-square test. Correlation coefficients between risk scores and immune cell infiltration were calculated using Spearman’s rank correlation analysis. The optimal cutoff value was determined using the maxstat package. All missing values were treated as NA without imputation. A p-value of less than 0.05 was considered statistically significant. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; ns, not significant.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AUC

Area under the ROC curve

BPs

Biological processes

CCs

Cellular components

CEA

Carcinoembryonic antigen

CLCF1

Cardiotrophin-like cytokine factor 1

COAD

Colon adenocarcinoma

CRC

Colorectal cancer

DCA

Decision curve analysis

DEGs

Differentially expressed genes

DEIRGs

Differentially expressed immune-related genes

GO

Gene Ontology

HCC

Hepatocellular carcinoma

ICPs

Immune checkpoints

IL-6

Interleukin-6

IRGs

Immune-related genes

KEGG

Kyoto Encyclopedia of Genes and Genomes

LASSO

Least absolute shrinkage and selection operator

LI

Lymphatic invasion

MFs

Molecular functions

MSI-H

Microsatellite Instability-High

MSI-L

Microsatellite Instability-Low

MSS

Microsatellite stability

OS

Overall survival

PCA

Principal component analysis

PNI

Perineural invasion

qRT-PCR

Quantitative real-time PCR

READ

Rectal adenocarcinoma

ROC

Receiver operating characteristic

ssGSEA

Single sample Gene Set Enrichment Analysis

TAMs

Tumor-associated macrophages

TCGA

The Cancer Genome Atlas

TICs

Tumor-infiltrating immune cells

t-SNE

T-distributed stochastic neighbor embedding

UCSC Xena

University of California Santa Cruz cancer database

VI

Vascular invasion

Author contributions

Conceptualization: WZ, SL, YW and FZ; Formal analysis: WZ, YG, NZ and JS; Investigation: WZ, NZ, JS, SL and YW; Methodology: WZ, YG, NZ and JS; Software: WZ, YG, SL and YW; Supervision: YW and FZ; Writing-original draft: WZ, YG, NZ and SL; Writing-review and editing: WZ, YW and FZ. All authors read and approved the final manuscript.

Data availability

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Consent for publication

The manuscript has received approval for submission from all authors.