Inflammation-Related Gene Profiling in Colorectal Cancer: A New Prognostic Signature

doi:10.21203/rs.3.rs-4619722/v1

Download PDF

Research Article

Inflammation-Related Gene Profiling in Colorectal Cancer: A New Prognostic Signature

https://doi.org/10.21203/rs.3.rs-4619722/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background Inflammation can influence the development of CRC as well as immunotherapy and plays a key role in CRC. Therefore, this study aimed to investigate the potential of inflammation-related genes in CRC risk prediction.

Methods The transcriptomic and clinical information of colorectal cancer patients was obtained from The Cancer Genome Atlas (TCGA) database and externally validated with the GSE39582 dataset. Consistency clustering was used to molecularly typify and genotype patients. Genes for model construction were screened using univariate Cox, LASSO Cox, and multivariate Cox regression, and model validation was performed by K‒M survival analysis and receiver operating characteristic (ROC) curve analysis. In addition, we combined nomograms for further prediction of patient prognosis. Finally, the possible mechanisms of inflammation-related genes in CRC were explored by functional enrichment analysis, immune microenvironment analysis and immune checkpoint analysis.

Results We identified two molecular subtypes and three genetic subtypes, two risk subgroups according to median risk values, constructeda prognostic model including thirteen genes (TIMP1, GDF15, UCN, KRT4, POU4F1, NXPH1, SIX2, NPC1L1, KLK12, IGFL1, FOXD1, ASPG, and CYP4F8), and validated the performance of each aspect of the model in an external database. Patients in the high-risk group had worse survival with reduced immune cell infiltration and a greater tumor mutational load. The risk score correlated strongly with the immune checkpoints PD1, PDL1, PDL2, and CTLA4, and it is possible that high-risk patients are more sensitive to treatment involving immune checkpoints. qPCR further verified that ASPG expression in the CRC tumors of our patients was significantly lower than that in the normal tissues and was a protective factor.

Conclusion In summary, we developed a prognostic marker associated with inflammatory genes to provide new directions for subsequent studies and to help clinicians assess the prognosis of CRC patients as well as to guide clinical treatment with different sensitive drugs.

Colorectal cancer (CRC)

inflammation

immune microenvironment

molecular subtypes

prognostic signature

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer death globally, with 3.2 million CRC cases expected to occur worldwide in 2040[1]. In recent decades, the view of cancer biology has shifted from a "cancer cell-centric" perspective to one in which tumor cells are located in a tumor microenvironment composed of fibroblasts, vascular cells, and inflammatory immune cells. All cells in the TME influence all stages of the tumor through close interactions, and the plasticity of tumor cells and surrounding cells in the TME is largely driven by inflammation[2]. Inflammation not only remodels the complex cellular components of the tumor microenvironment, creating tumor-supportive ecological niches through mutual crosstalk but also exerts direct protumorigenic signals and functions on epithelial and cancer cells to promote cancer[3]. Approximately 15–20% of cancer patients develop infection, chronic inflammation, or autoimmunity at the same tissue or organ site[2]. In particular, colorectal cancer, poorly controlled, long-standing inflammatory bowel disease and chronic inflammation of the bowel due to poor diet are associated with an increased incidence of colorectal cancer[4].

There are three forms of inflammation in colorectal cancer: chronic inflammation prior to tumorigenesis, tumor-induced inflammation, and treatment-induced inflammation[5]. Some dietary and physical activity habits, e.g., smoking, alcohol consumption, obesity, etc., are risk factors for colorectal cancer[6]. Colitis-associated colorectal cancer (CAC) is thought to arise from the expansion of protumorigenic clones that replace wild-type colorectal epithelium during chronic inflammation-induced field cancerization of the large intestine[7]. In tumors, interleukin-6 (IL-6) secreted by human colorectal cancer-derived mesenchymal stem cells (CC-MSCs) promotes colorectal cancer cell progression through IL-6/JAK2/STAT3 signaling[8]. IL-6 is associated with lymph node metastasis, TNM stage and degree of differentiation in colorectal cancer patients and is a prognostic biomarker in cancer immunotherapy [9]. In addition, inflammation and cancer can also be promoted in certain pathological states, where chronic oxidative stress promotes the activation of inflammatory signaling pathways, leading to the activation of multiple transcription factors or the dysregulation of gene and protein expression, which can lead to the development of tumors or the survival of cancer cells[10]. Several features of aging (i.e., genomic instability, epigenetic changes, chronic inflammation, and ecological dysregulation) are very similar to specific cancer features, making aging an important risk factor for colorectal cancer (CRC)[11, 12]. Therefore, inflammatory genes in different pathological states also affect colorectal cancer to different degrees, and further studies are needed to investigate the relationship between inflammation-related genes and the prognosis of colorectal cancer patients.

Colorectal cancer is one of the least effective tumors for immunotherapy. Although there is clear clinical evidence of a therapeutic role for immune checkpoint inhibitors in mismatch repair-deficient (MMR) or highly microsatellite unstable (MSI-H) metastatic colorectal cancers, the vast majority of patients with well-characterized MMR or microsatellite-stable (MSS) tumors do not benefit from immunotherapy[13, 14]. Inflammatory states are usually accompanied by the production of free radicals, stimulation of cytokines, chemokines, and growth and angiogenic factors that promote tumorigenesis by damaging DNA, stimulating angiogenesis, and inducing cell proliferation[15]. Recent studies have shown that IL-17A increases PD-L1 expression through the p65/nrf1/miR-15b-5P axis and enhances resistance to anti-PD-1 therapies. IL-17A has been identified as a promising therapeutic target for sensitization to immune checkpoint inhibitor therapy in MSS CRC patients[16]. In addition, blocking IL-6 enhances MHC-I and PD-L1 expression and improves the tumor response to anti-PD-1 therapy[17]. Inflammatory indexes (II), such as the systemic immune inflammation index (SII) and aspartate aminotransferase lymphocyte ratio index (ALRI), were identified as prognostic markers for progression-free survival (PFS) and overall survival (OS) in patients with metastatic colorectal cancer receiving chemotherapy and bevacizumab[18].

Thus, the inflammatory response plays an important role at any stage of the tumor, and multiple inflammatory indicators are associated with the prognosis of CRC patients. In this study, we constructed a robust signature of inflammatory genes (IRGs) in CRC by combining the expression of inflammatory response-related genes obtained from the TCGA database with bioinformatics data. We also validated the ability of the IRG risk score to predict the tumor microenvironment and immune cell infiltration. In conclusion, this study contributes to further understanding the relationship between the inflammatory response and CRC and provides more precise medication guidance for CRC patients.

2.1 Data collection and preprocessing

The gene expression data and clinical data used in this study were obtained from the TCGA and GEO databases and from the TCGA-COAD and GSE37582 cohorts, respectively. After the exclusion of samples with incomplete follow-up information and survival data, 456 samples were included in the TCGA cohort (including 415 tumor samples and 41 normal samples). We used the "TCGAbiolinks" R package to acquire RNA-seq data and selected gene transcript matrices with expression values in transcripts per million bases (TPM) for TCGA-COAD. The clinical information collected included sex, age, pathological stage, pathological TM stage (N stage was excluded because it was missing from GSE37582), survival time, and survival status.

2.2 IRG differential gene identification and molecular subtype characterization

We downloaded a total of 15,333 inflammation-related genes encoding proteins from GeneCards. The "limma" R package was used to analyze the gene differences with the filtering conditions of |log2FC|≥2 and FDR < 0.001, and we ultimately obtained 1204 DEGs. Next, the "ggVolcano" R package was used to map the volcanoes. We used the "ConsensusClusterPlus" R script to classify the combined samples into different molecular subtypes based on the expression of DEGs using the optimal classification range of K = 2–9. Next, the Kyoto Encyclopedia of Genes and Genomes (KEGG) database was selected for functional variant analysis (GSVA). K‒M curves were used to determine differences in survival between the two subtypes. Finally, differences in immune cell infiltration between subtypes were analyzed using single-sample gene set enrichment analysis (ssGSEA) to further understand the changes in the immune microenvironment in different subtypes.

2.3 Identification, functional enrichment analysis, and prognostic analysis of genes with differential molecular subtypes among the IRGs

We screened the DEGs of the two molecular subtypes of IRGs based on the following screening conditions: |log2FC|≥1 and FDR < 0.05. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses were performed to further explore the potential functions and pathways of the IRGs. Next, we analyzed DEGs and survival time using univariate Cox regression to screen for genes with prognostic features. The samples were categorized into different gene subtypes based on the expression levels of these prognostic genes. Finally, the prognostic ability of different gene subtypes was verified using K‒M.

2.4 Construction and validation of a prognostic signature for inflammation-related genes

First, we divided the samples from the two cohorts, TCGA-COAD and GSE39582, into a training set and a validation set at a ratio of 7:3. Supplementary Table 1 shows the baseline clinical characteristics after TCGA-COAD division. Next, we performed univariate Cox regression analysis of the 1204 DEGs related to overall survival (OS) to screen prognostic DEGs, used the "glmnet" function to perform least absolute shrinkage and selection operator (LASSO) analysis on the screened prognostic genes to further optimize the results, and finally used multivariate Cox regression to screen 13 characteristic genes for model construction. Modeling formulas: risk score = coef Σ (Genei) × exp (Genei).

The median risk value of the samples was calculated according to the model formula, and the samples were categorized into high-risk and low-risk groups. Next, we analyzed the differences in risk scores among molecular and genetic subtypes of IRGs to assess the predictive ability of risk scores in different subtypes. Sankey plots of sample allocation were drawn using the "ggalluvial" R package. K‒M survival analysis was performed to assess the difference in survival between patients in the different risk subgroups, progression-free survival (PFI) was estimated for each patient, and receiver operating characteristic (ROC) curves were plotted to assess the prognostic value of the genetic trait. In addition, we calculated the C-index of the risk model using the "pec" R package to assess the accuracy of the model.

2.5 Construction and verification of the nomogram

We used the "rms" R package to construct a line plot of risk scores combined with patient clinical characteristics to accurately predict patient survival and guide clinical practice. Calibration and decision curve analysis (DCA) were performed to validate the predictive power of the column chart. In addition, 1-, 3-, and 5-year ROC curves were plotted to further evaluate the performance of the column chart.

2.6 Functional enrichment analysis

We used the "cluster-Profiler" R package to analyze the GO and KEGG analyses as well as GSEA of the DEGs from the high- and low-risk groups to further understand the functions and pathways enriched by the IRGs.

2.7 Analysis of the tumor immune microenvironment and immune checkpoints

We used the CIBERSORT algorithm to quantify the infiltration of 22 immune cell types in the tumor microenvironment and assessed their relationships with characterized genes. Single-sample GSEA using a set of immune-related genes was used to assess multiple immune functions in the high- and low-risk groups and to compare the activity of these immune functions between the two risk groups. Then, to assess the relationships between different subgroups and immunotherapy efficacy, the expression levels of various immune checkpoint molecules in the two risk subgroups were analyzed, and Pearson correlation analyses between selected immune checkpoint molecules and risk scores were performed.

2.8 Tumor mutation and drug sensitivity assessment

The tumor mutational burden (TMB) is the sum of the number of mutations that occur in tumor cells. In general, a higher TMB indicates a greater number of mutations in the tumor cells. Therefore, tumors with a high TMB may be more likely to respond positively to immunotherapy. We analyzed the differences in TMB between patients in the high-risk and low-risk groups. The samples were grouped according to the median TMB, and survival analysis was used to assess the survival of patients in different subgroups. The "maftools" R package was used to construct a waterfall plot of the genes in the two risk groups. Finally, drug sensitivity data were downloaded from the GDSC database, the predictions of sensitivity to different therapeutic drugs in the two risk groups were analyzed using the "OncoPredict" R package, and the predicted drugs and risk scores were correlated.

2.9 Quantitative reverse transcription PCR (RT‒qPCR)

Total RNA from CRC patient tissues was extracted according to the instructions of the SteadyPure Quick RNA Extraction Kit. A PrimeScript™ RT kit (TaKaRa) was used for reverse transcription. Real-time quantitative polymerase chain reaction was performed according to the manufacturer's instructions to detect the mRNA expression of ASPG in tumor and paracancerous tissues; primer sequences are shown in Supplementary Table 2.

2.10 Statistical analyses

Statistical analyses were performed using version 4.3.1 of R, GraphPad Prism 9. Significant quantitative differences between and among groups were calculated by one-way ANOVA and two-tailed t-test, respectively, and Wilcoxon's rank sum test was used for nonanalytically normally distributed continuous variables. Pearson's analysis was applied for correlation analysis. The significance of the statistical results is indicated as ***P < 0.001, **P < 0.01, and *P < 0.05.

3.1 Construction and prognostic analysis of molecular isoforms of inflammation-related DEGs in COAD

Our workflow diagram is shown in Fig. 1. First, we obtained 10,333 genes encoding proteins from the GeneCards database. Compared with those in normal samples, 1024 DEGs were detected in COAD tumor samples (Fig. 2A). To evaluate the molecular subtypes of COAD based on the differential expression of IRGs, an unsupervised consensus clustering algorithm was applied to all 1024 patient cohorts. K = 2 was considered to indicate optimal stability, and the COAD patients were categorized into two subtypes: clusters A and B (Fig. 2B). Next, Kaplan‒Meier survival analysis revealed significantly different survival outcomes between subtypes. Patients in cluster A exhibited lower survival than those in cluster B (Fig. 2D). GSVA was performed on both isoforms to analyze their biological behavior. Cluster A was predominantly enriched in extracellular matrix-receptor interactions and cell-matrix adhesion, whereas the remaining pathways (e.g., drug metabolism by cytochrome P450, fatty acid metabolism, and butyric acid metabolism) were highly enriched in cluster B (Fig. 2C). In addition, we analyzed the differences in immune cell infiltration between the two subtypes using ssGSEA. Sixteen immune cells were highly infiltrated in Cluster A. The most prominent immune cells included myeloid-derived suppressor cells (MDSCs), regulatory T cells, and macrophages. Overall, 17 helper T cells and neutrophils exhibited high levels of immune cell infiltration in Cluster B (Fig. 2E).

3.2 Construction and validation of gene subtypes based on IRG molecular subtype DEGs

To further understand the potential biological activities of IRGs in COAD, we performed differential analysis of genes in IRG clustersA and clusterB using the "limma" software package and identified 193 DEGs under the following filtering conditions: log|FC|≥ 0.585 and corrected p value < 0.05 (Fig. 3A). Functional enrichment analysis showed that these DEGs were predominantly enriched for biological functions related to the extracellular matrix (Fig. 3B) and that matrix metalloproteinases (MMPs) are an important class of enzymes in vivo that may lead to the metastasis of tumor cells by disrupting the extracellular matrix [19]. Pathway enrichment analysis revealed that the DEGs were enriched mainly in the following pathways: protein digestion and uptake, ECM-receptor interaction, adhesion to the cellular matrix, and the PI3K-Akt signaling pathway (Fig. 3C).

Then, we performed univariate Cox regression analysis of the DEGs to screen 77 genes with prognostic value. We genotyped the 77 prognostic genes again and obtained k = 3 clusters by the algorithm, thus obtaining three genotypes named Cluster A, Cluster B, and Cluster C (Fig. 3D). K‒M analysis revealed that patients in Cluster A had the worst survival, and those in Cluster B had better survival than those in Cluster A and Cluster C (Fig. 3E). Subsequent clinicopathologic heatmaps showed a continuous downward trend in the degree of gene expression in the three subtypes, suggesting that the majority of genes in the Cluster A subtype are risk factors for colon cancer and that genes that are highly expressed in Cluster B are protective factors (Fig. 3F).

3.3 Construction and validation of the IRG-related prognostic signature

Previous IRG typing and genotyping methods have demonstrated great potential for determining the clinical prognosis of patients with colorectal cancer. To better utilize the predictive power of IRGs for COAD, we built a prognostic model. First, we performed a univariate Cox analysis on the 1024 DEGs obtained from the analysis of variance to obtain 97 prognostic genes. Then, the prognostic genes were subjected to LASSO regression analysis, and multivariate Cox regression was subsequently performed to avoid overfitting of the model (Fig. 4A, B). We eventually identified 13 key genes (Fig. 4C) and constructed an IRG-based risk score. IRG score = (0.327777843613415 * TIMP1 expression) + (-0.321505536641187 * GDF15 expression) + (0.519155635681503 * UCN expression) + (0.497663150446509 * KRT4 expression) + (0.704390071491466 * POU4F1 expression) + (0.979581063364134 * NXPH1 expression) + (0.302700609425077 * SIX2 expression) + (0.907534062215074 * CYP4F8 expression) + (0.342487252775221 * NPC1L1 expression) + (-0.360117110529824 * KLK12 expression) + (-1.05144522453045 * ASPG expression) + (0.361697071741414 * FOXD1 expression) + (0.319888833846153 * IGFL1 expression). The Sankey diagram shows the flow of data, with the low-risk group consisting only of cluster B in genotyping, which also corresponds to the fact that cluster B subtypes have the best survival. We then assessed the relationship between the two molecular typologies and the three genotypes and the risk scores and found that the risk scores for both the IRG typology and the Cluster A subtype of the genotypes were greater than those for the other subtypes, which also corresponded to the low survival of the Cluster A subtype (Fig. 4D-F).

The TCGA and GEO cohorts were then divided into training and test groups at a 7:3 ratio, and each cohort was divided into high- and low-risk groups based on the median value of the risk score. K‒M survival analyses revealed that patients in the low-risk group had good prognoses in all three cohorts (Fig. 4G-H). The area under the curve for subjects' work characteristics at 1, 3, and 5 years was 0.716, 0.693, and 0.763 for the TEST group and 0.822, 0.811, and 0.793 for the TRAIN group, respectively (Fig. 4J-K). Survival analysis for progression-free survival in patients in the TCGA cohort also revealed better survival in the low-risk group (Fig. 4L). These data demonstrate the good predictive power of IRG-based risk profiles.

3.4 Independent prognostic value of inflammatory risk profiles and the construction and validation of a nomogram

To further explore the prognostic potential of risk characteristics, we integrated three common clinical characteristics (age, sex, and stage) and subjected the risk scores to univariate Cox analysis and multivariate Cox regression analysis. Our analysis revealed that stage and risk score were independent prognostic factors for this disease (p < 0.001) (Fig. 5A). The C-index was primarily used to evaluate the ability of the model; therefore, we compared the C-indexes of the risk score and other clinical traits for disease prediction, and the C-indexes of the risk score were all greater than those of the other clinical traits and were close to 0.8 (Fig. 5B). These data provide further evidence of the ability of the prognostic model to predict disease.

Next, we combined risk scores and clinical traits to construct column line plots (Fig. 5C), and the calibration curves at 1, 3, and 5 years showed good agreement between the actual survival probabilities and the column line plot predictions (Fig. 5D). DCA analyses showed good validity for both column line plots and risk values (Fig. 5E). In addition, the one-, three-, and five-year risk scores and the ROC curve of the nomogram had the highest AUC values (AUC > 0.75) (Fig. 5F).

3.5 Correlation analysis of the IRG risk score and clinicopathologic characteristics of patients with COAD

To assess the clinical significance of the IRG risk score, we combined patient information from the TCGA and GEO cohorts, including sex, age, stage, and TN stage. We performed a chi-square test of each clinical characteristic against the risk score and found that all clinical characteristics except sex differed between the high- and low-risk groups (Fig. 6A). Next, we performed percentage statistics on the frequency of each clinical characteristic in the high- and low-risk groups, and the violin plots showed that the risk scores were able to distinguish the staging and TN staging well and that patients who were more likely to have lymph node metastasis (N staging) and advanced tumor staging (stage staging) had higher risk scores (Fig. 5B-F).

3.6 Functional enrichment analysis of different IRG risk subgroups

We identified DEGs between the high- and low-risk groups of IRGs by differential expression analysis, followed by GO and KEGG functional enrichment analysis. In the BP category, DEGs may play important roles in "epidermal development" and "epidermal cell differentiation". Among the cellular component (CC) terms, "corrupted envelope" was the most common term among the DEGs. In the field of molecular function (MF), "receptor ligand activity", "glycosaminoglycan binding", "cytokine activity", and "G protein-coupled receptor binding" were identified as possible DEGs (Fig. 7A). In addition, the results of KEGG analysis suggested that the DEGs might be involved in the "signaling pathways regulating pluripotency of stem cells" and "Hippo signaling pathway" (Fig. 7B). These data suggest that DEGs may be involved in cell signaling for epidermal growth and differentiation. Next, GSEA showed that pathways such as the citric acid cycle (TCA cycle), pentose and glucuronide interconversion, and peroxisomes were enriched in the low-risk group (Fig. 7C). The TCA cycle is a major intracellular energy production pathway, and colorectal cancer cells often exhibit abnormal TCA cycle activity[20]. Pathways enriched in the high-risk group include cell adhesion molecules and extracellular matrix-receptor interactions (Fig. 7D). ECM not only affects tumor malignancy and growth but also influences the tumor response to therapy [21].

3.7 Immune cell infiltration and immune checkpoint analysis between the two risk subgroups

The role of the immune microenvironment in tumor progression is closely related to patient prognosis. Therefore, in our study, the correlation of prognosis-related signature genes with 22 tumor immune infiltrating cells (TIICs) was first analyzed using the CIBERSORT algorithm (Fig. 8A). Each of the characterized genes correlated with a variety of immune cells; in particular, FOXD1 showed a strong positive correlation with CD8 + T cells and a significant negative correlation with M0 macrophages. Next, we quantified the enrichment scores of immune infiltrating cells and functions using the single sample gene set enrichment analysis (ssGSEA) approach. The results showed that four TIIC enrichment scores were greater in the high-risk affected group (iDCs, NKT cells, NK cells, and Tfh cells). Similar results were observed for two immune-related functions (macrophage and type I IFN response) (Fig. 8B-C). In addition, the risk score also correlated significantly with the initial CD4 + T-cell count (Fig. 8D). With the initial success of melanoma treatment, immunotherapy has rapidly become the mainstay of treatment for many types of solid cancers [22]. In our study, most of the immune checkpoint genes (CD276, CD70, HHLA2, etc.) were expressed at higher levels in the high-risk group than in the low-risk group, suggesting that patients in the high-risk group were more sensitive to treatment involving immune checkpoints (Fig. 8E). In addition, the risk score showed a significant positive correlation with the immune checkpoints PD1, PDL1, PDL2, and CTLA4 (Fig. 8F). These results suggest a potential link between the IRG model we constructed and immune infiltration, as well as the possibility that the risk score could predict the outcome of immunotherapy in patients.

3.8 Tumor mutation load and drug sensitivity analysis of IRG risk models

The tumor mutational burden (TMB) is the total number of mutations in a tumor cell and is usually expressed as the number of mutations per million base pairs (megabase, Mb). We analyzed the mutation frequency and TMB in the high- and low-risk groups. The 20 genes with the highest mutation frequencies are shown in Fig. 9A. TMB occurred in 92.97% of the samples in the low-risk group, with APC mutations being the most significant (71%). TMB occurred in 98.48% of patients in the high-risk group, and the APC gene was also mutated most significantly in the low-risk group (70%). The TMB was much greater in the high-risk group than in the low-risk group (Fig. 9B), and the survival analysis also revealed that the high-TMB group was able to improve survival (Fig. 9C). We then analyzed the ability of the risk score and TMB to predict patient survival, and we found that patients in the high-TMB and low-risk groups had a better prognosis (Fig. 9D).

Patients with colon cancer often require a combination of multiple therapies; therefore, we predicted the sensitivity of two subgroups of patients to commonly used chemotherapeutic agents and found that the high-risk group was more sensitive to oxaliplatin, sorafenib, and gefitinib (Fig. 9E), and correlation analyses revealed greater sensitivity to drugs with higher risk scores (Fig. 9F). Different IRG subgroups have significant differences in sensitivity to drugs, so the drug prediction results can provide guidance for the treatment of COAD patients.

Finally, we verified the expression level of ASPG, the model gene with the strongest coefficient, in colorectal cancer patients using qRT‒PCR (n = 7), and the expression of ASPG in tumors was significantly lower than that in normal tissues adjacent to the cancer (Fig. 10).

To date, inflammation has become a hotspot for all types of neoplastic diseases as well as preneoplastic diseases, the link between inflammation and tumorigenesis has been well established in the last two decades, and inflammatory bowel inflammation is an important risk factor for colon carcinogenesis[23]. The accumulation of inflammatory genetic alterations and loss of normal cellular regulatory processes are not only associated with cancer growth and progression but also lead to the expression of tumor-specific and tumor-associated antigens, which may activate antitumor immunity [24]. In the tumor microenvironment, antigen-specific adaptive immune responses are impeded, as soluble cytokines prevent T-cell infiltration, and cell surface signals are modulated to suppress lymphocyte-mediated killing activity[25]. Currently, MSI-H/dMMR is the only recognized biomarker that can be used to guide immunotherapy for colon cancer[26]. A small proportion of patients with microsatellite-stabilized (MSS) or proficient MMR (pMMR) mutations can benefit from immunotherapy, whereas a large proportion of MSI-H/dMMR patients exhibit intrinsic or acquired resistance to immunotherapy [27]. Therefore, the identification of new biomarkers or prognostic markers is crucial, and powerful biomarkers are needed for the stratification of colorectal cancer patients and assessment of the tumor immune microenvironment. Currently, studies have demonstrated the predictive ability of inflammation-related genes in colorectal cancer, and in this study, we screened genes in a novel way, identified new inflammation-related genes, and confirmed the value of the genes for prognostic prediction in patients with colorectal cancer through our validation.

In this study, we first classified patients into two different molecular subtypes based on IRG expression by an unsupervised consistent clustering method. Patients with IRG subtype A had poorer overall survival than those with subtype B, and there were also significant differences in pathways enriched in different subtypes and different immune cell infiltrates. We reclassified COAD patients according to the DEGs between the two subtypes. These results initially demonstrated the prognostic value of IRGs in COAD. In addition, we screened 13 genes with prognostic value using univariate Cox regression, LASSO regression, and multivariate Cox regression and then constructed risk-prognostic models. There were significant differences in survival between the high- and low-risk groups in the different cohorts, and the area under the curve (AUC) of the receiver operating characteristic (ROC) curve (AUC) curve (AUC) was also high. More importantly, the C-index, which evaluates the performance of the model, is superior to all other clinical characteristics. For better clinical application, we constructed a nomogram, and the ROC curves were able to predict the survival time of patients at 1, 3, and 5 years very well. In addition, there were significant differences in clinicopathologic correlation, TME, immune checkpoint analysis, TMB, and drug sensitivity between the high-risk and low-risk groups. Therefore, our constructed prognostic profile of inflammatory genes can assist in the clinical diagnosis of COAD patients, as well as the development of personalized and precise treatments for patients.

We screened 13 genes associated with inflammation and the prognosis of colon cancer patients. The TIMP1, GDF15, UCN, KRT4, POU4F1, NXPH1, SIX2, NPC1L1, KLK12, IGFL1, and FOXD1 genes have been studied in tumors in a correlative manner, and the ASPG and CYP4F8 genes were screened in other predictive models. For example, tissue inhibitor of metalloproteinases 1 (TIMP1) has been identified in a variety of genetic prognostic studies and was the first natural collagenase inhibitor to be discovered [28]. High TIMP1 expression is positively associated with poor prognosis in lung, brain, prostate, breast, colon and several other cancers [29]. TIMP1 causes tumor proliferation and metastasis through the FAK-PI3K/AKT and MAPK pathways in colon cancer [30]. ASPG, also known as 60 kDa lysophospholipase, is a protein with asparaginase, lysophospholipase, transacylase, and platelet-activating factor (PAF) acetylhydrolase activities and is an enzyme with antitumor activity [31]. In our model, ASPG is the gene with the largest coefficient with protective properties. To verify the impact of the database on different patients in our country, the results of our qPCR validation were consistent with the results in the database. FOXD1 plays a role in a wide range of tumors and can be of prognostic value in predicting the prognosis of patients with colorectal cancer through the ERK 1/2 pathway and promoting tumor progression[32, 33]. In conclusion, a portion of the genes for constructing predictive models are closely related to tumor development, and we also screened genes that are still lacking in tumor research; therefore, the molecular mechanisms and phenotypes of these biomarkers in cells need to be further investigated.

The role of the tumor microenvironment in oncology research has been clearly defined by researchers as a highly structured ecosystem containing cancer cells surrounded by multiple nonmalignant cell types. The TME includes a rich diversity of immune cells, cancer-associated fibroblasts (CAFs), endothelial cells (ECs), pericytes, and other cell types that vary by tissue[34]. In our analysis, the immune function of macrophages was upregulated in the high-risk group. Recent studies have shown that CXCL8 polarizes tumor-associated macrophages (TAMs) to the M2 phenotype, leading to increased M2 macrophage infiltration in CRC and reduced CD8 + T-cell infiltration, resulting in an immunosuppressive microenvironment [35]. In the low-risk group, the immune function of NK cells was upregulated, and NK cells also play a crucial role in bowel cancer. It is widely accepted that low levels of NK cell infiltration and/or impaired NK cell function are associated with poor overall patient survival and recurrence of CRC after treatment [36]. We also evaluated immune checkpoints that were differentially expressed between the high- and low-risk groups, and the risk scores showed significant positive correlations with important PD1, PDL1, PDL2, and CTLA4 checkpoints. Cytotoxic T lymphocyte-associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1), two inhibitory checkpoints commonly found on activated T cells, have been found to be the most reliable targets for the treatment of cancer. [37, 38]. The inflammatory cytokine interleukin 17 (IL-17) can upregulate PD-L1 protein expression in HCT116 CRC cells through activation of NF-κB and ERK1/2 signaling [39]. Smad7 is a negative regulator of TGF-β signaling, and its deficiency induces CD103 + PDL2/1 + DCs and Tregs to attenuate DSS-induced colitis [40]. Using single-cell sequencing, a recent study analyzed immune and stromal cell dynamics in neoadjuvant PD-1-blocked d-MMR/MSI-H CRC patients and showed that in patients with immunotherapy-pathologically complete remission (pCR) tumors, the proportions of CD8 + Trm-mitotic cells, CD4 + Tregs, proinflammatory IL1B + Mono cells, and CCL2 + fibroblasts after treatment were consistently decreased, whereas the proportions of CD8 + Tem cells, CD4 + Th cells, CD20 + B cells, and HLA-DRA + endothelial cells increased, providing yet another new insight into the mechanisms of ICI therapy[41]. When the level of inflammation in a patient's body is elevated, the response to ICIs is also decreased, and MSI-H CRC patients with inflammatory diseases are at increased risk of developing resistance to ICIs through neutrophil-associated T-cell depletion and hence ICI resistance [42]. In addition, in drug sensitivity analysis, patients at different risk levels have different sensitivities to drugs, and the treatment of cancer patients often requires a combination of multiple drugs; our risk scores can also guide the choice of drugs in the clinic. Our study suggested that higher expression of immune checkpoint molecules predicts that high-risk patients may benefit from these immunotherapies in the future.

Our study has several limitations: we performed external validation using only one database, GEO, and in the future, we will use several different databases for joint validation. Second, the TCGA database we analyzed contains Western case data and therefore lacks representative prospective data. We validated the expression level of only one model gene in vitro, and further validation of the mechanism of action of other genes in colorectal cancer is needed.

In this study, we comprehensively analyzed the IRGs of CRC patients and developed a prognostic model based on inflammatory gene correlation, which exhibited excellent ability to predict the OS of patients. Risk subgroups based on risk scores had their own unique characteristics in terms of the immune microenvironment, immune checkpoints, and drug sensitivity. In addition, we validated the expression of ASPG, a gene that plays an important role in the model, in our patients, but further studies are needed to determine its mechanism in CRC.

Ethics approval and consent to participate

The human tissues obtained in our study were approved by the Ethics Committee of the Affiliated Hospital of Southwest Medical University in accordance with the Declaration of Helsinki (1964).

Consent for publication

Informed consent was obtained from the participants and the study was approved by the Affiliated Hospital of Southwest Medical University (No. KY2024266).

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as potential conflicts of interest.

Author contributions

WY and XC conceived and designed the study. QJ, CZ, LY, and SL obtained the data and performed the statistical analysis. WX, GL, XS, and CX performed the experiments and analyzed the data. WY drafted the manuscript. MY and ZY revised the manuscript for intellectual content and language.

Funding

This study was supported by grants from the Science and Technology Planning Project of Sichuan Province (2022YFS0626), the "Luzhou Municipal People's Government-Southwest Medical University" Technology Strategy Project (2023LZXNYDJ001), and the Sichuan Science and Technology Program (2023RCM198).

Data availability

The transcriptomic data and clinical information for this study were obtained from the TCGA database (https://portal.gdc.cancer.gov/). External validation set data were downloaded from the GEO database (https://www.ncbi.nlm.nih.gov.). Because these are public databases, all patients were informed, and consent was obtained before the information was released.

Morgan E, Arnold M, Gini A, et al. Global burden of colorectal cancer in 2020 and 2040: incidence and mortality estimates from GLOBOCAN. Gut 2023; 72(2):338-344. https://doi.org/10.1136/gutjnl-2022-327736.
Greten FR, Grivennikov SI. Inflammation and Cancer: Triggers, Mechanisms, and Consequences. Immunity 2019; 51(1):27-41. https://doi.org/10.1016/j.immuni.2019.06.025.
Denk D, Greten FR. Inflammation: the incubator of the tumor microenvironment. Trends in Cancer 2022; 8(11):901-914. https://doi.org/10.1016/j.trecan.2022.07.002.
Birch RJ, Burr N, Subramanian V, et al. Inflammatory Bowel Disease-Associated Colorectal Cancer Epidemiology and Outcomes: An English Population-Based Study. American Journal of Gastroenterology 2022; 117(11):1858-1870. https://doi.org/10.14309/ajg.0000000000001941.
Schmitt M, Greten FR. The inflammatory pathogenesis of colorectal cancer. Nature Reviews Immunology 2021; 21(10):653-667. https://doi.org/10.1038/s41577-021-00534-x.
Marmol I, Sanchez-de-Diego C, Pradilla Dieste A, Cerrada E, Jesus Rodriguez Yoldi M. Colorectal Carcinoma: A General Overview and Future Perspectives in Colorectal Cancer. International Journal of Molecular Sciences 2017; 18(1). https://doi.org/10.3390/ijms18010197.
Zhou RW, Harpaz N, Itzkowitz SH, Parsons RE. Molecular mechanisms in colitis-associated colorectal cancer. Oncogenesis 2023; 12(1):48. https://doi.org/10.1038/s41389-023-00492-0.
Zhang X, Hu F, Li G, et al. Human colorectal cancer-derived mesenchymal stem cells promote colorectal cancer progression through IL-6/JAK2/STAT3 signaling. Cell Death & Disease 2018; 9(2). https://doi.org/10.1038/s41419-017-0176-3.
Zheng J, Wang X, Yu J, Zhan Z, Guo Z. IL-6, TNF-α and IL-12p70 levels in patients with colorectal cancer and their predictive value in anti-vascular therapy. Frontiers in Oncology 2022; 12. https://doi.org/10.3389/fonc.2022.997665.
Bardelcikova A, Soltys J, Mojzis J. Oxidative Stress, Inflammation and Colorectal Cancer: An Overview. Antioxidants 2023; 12(4). https://doi.org/10.3390/antiox12040901.
López-Otín C, Pietrocola F, Roiz-Valle D, Galluzzi L, Kroemer G. Meta-hallmarks of aging and cancer. Cell Metabolism 2023; 35(1):12-35. https://doi.org/10.1016/j.cmet.2022.11.001.
Crossland NA, Beck S, Tan WY, et al. Fecal microbiota transplanted from old mice promotes more colonic inflammation, proliferation, and tumor formation in azoxymethane-treated A/J mice than microbiota originating from young mice. Gut Microbes 2023; 15(2). https://doi.org/10.1080/19490976.2023.2288187.
Ciardiello D, Vitiello PP, Cardone C, et al. Immunotherapy of colorectal cancer: Challenges for therapeutic efficacy. Cancer Treatment Reviews 2019; 76:22-32. https://doi.org/10.1016/j.ctrv.2019.04.003.
Lichtenstern CR, Ngu RK, Shalapour S, Karin M. Immunotherapy, Inflammation and Colorectal Cancer. Cells 2020; 9(3). https://doi.org/10.3390/cells9030618.
Landi S, Gemignani F, Bottari F, et al. Polymorphisms within inflammatory genes and colorectal cancer. J Negat Results Biomed 2006; 5:15. https://doi.org/10.1186/1477-5751-5-15.
Liu C, Liu R, Wang B, et al. Blocking IL-17A enhances tumor response to anti-PD-1 immunotherapy in microsatellite stable colorectal cancer. J Immunother Cancer 2021; 9(1). https://doi.org/10.1136/jitc-2020-001895.
Li W, Wu Z, Meng W, et al. Blockade of IL-6 inhibits tumor immune evasion and improves anti-PD-1 immunotherapy. Cytokine 2022; 158:155976. https://doi.org/10.1016/j.cyto.2022.155976.
Passardi A, Azzali I, Bittoni A, et al. Inflammatory indices as prognostic markers in metastatic colorectal cancer patients treated with chemotherapy plus Bevacizumab. Ther Adv Med Oncol 2023; 15:17588359231212184. https://doi.org/10.1177/17588359231212184.
Omer U, Zeynep Y, Fatih T, Ibrahim Halil Y. Elevated expression levels of COX-2, IL-8 and VEGF in colon adenocarcinoma. Cellular and Molecular Biology 2023; 69(6):146-150. https://doi.org/10.14715/cmb/2023.69.6.22.
Pavlova NN, Thompson CB. The Emerging Hallmarks of Cancer Metabolism. Cell Metab 2016; 23(1):27-47. https://doi.org/10.1016/j.cmet.2015.12.006.
Henke E, Nandigama R, Ergün S. Extracellular Matrix in the Tumor Microenvironment and Its Impact on Cancer Therapy. Front Mol Biosci 2019; 6:160. https://doi.org/10.3389/fmolb.2019.00160.
Ganesh K, Stadler ZK, Cercek A, et al. Immunotherapy in colorectal cancer: rationale, challenges and potential. Nat Rev Gastroenterol Hepatol 2019; 16(6):361-375. https://doi.org/10.1038/s41575-019-0126-x.
Terzić J, Grivennikov S, Karin E, Karin M. Inflammation and colon cancer. Gastroenterology 2010; 138(6):2101-2114.e2105. https://doi.org/10.1053/j.gastro.2010.01.058.
Karin M, Shalapour S. Regulation of antitumor immunity by inflammation-induced epigenetic alterations. Cell Mol Immunol 2022; 19(1):59-66. https://doi.org/10.1038/s41423-021-00756-y.
Franzolin G, Tamagnone L. Semaphorin Signaling in Cancer-Associated Inflammation. Int J Mol Sci 2019; 20(2). https://doi.org/10.3390/ijms20020377.
Yang X, Lian B, Zhang N, et al. Genomic characterization and immunotherapy for microsatellite instability-high in cholangiocarcinoma. BMC Med 2024; 22(1):42. https://doi.org/10.1186/s12916-024-03257-7.
Hou W, Yi C, Zhu H. Predictive biomarkers of colon cancer immunotherapy: Present and future. Front Immunol 2022; 13:1032314. https://doi.org/10.3389/fimmu.2022.1032314.
Grünwald B, Schoeps B, Krüger A. Recognizing the Molecular Multifunctionality and Interactome of TIMP-1. Trends Cell Biol 2019; 29(1):6-19. https://doi.org/10.1016/j.tcb.2018.08.006.
Jackson HW, Defamie V, Waterhouse P, Khokha R. TIMPs: versatile extracellular regulators in cancer. Nat Rev Cancer 2017; 17(1):38-53. https://doi.org/10.1038/nrc.2016.115.
Song G, Xu S, Zhang H, et al. TIMP1 is a prognostic marker for the progression and metastasis of colon cancer through FAK-PI3K/AKT and MAPK pathway. J Exp Clin Cancer Res 2016; 35(1):148. https://doi.org/10.1186/s13046-016-0427-7.
Belviso S, Iuliano R, Amato R, Perrotti N, Menniti M. The human asparaginase enzyme (ASPG) inhibits growth in leukemic cells. PLoS One 2017; 12(5):e0178174. https://doi.org/10.1371/journal.pone.0178174.
Pan F, Li M, Chen W. FOXD1 predicts prognosis of colorectal cancer patients and promotes colorectal cancer progression via the ERK 1/2 pathway. Am J Transl Res 2018; 10(5):1522-1530.
Zong Y, Miao Y, Li W, et al. Combination of FOXD1 and Plk2: A novel biomarker for predicting unfavourable prognosis of colorectal cancer. J Cell Mol Med 2022; 26(12):3471-3482. https://doi.org/10.1111/jcmm.17361.
de Visser KE, Joyce JA. The evolving tumor microenvironment: From cancer initiation to metastatic outgrowth. Cancer Cell 2023; 41(3):374-403. https://doi.org/10.1016/j.ccell.2023.02.016.
Shao Y, Lan Y, Chai X, et al. CXCL8 induces M2 macrophage polarization and inhibits CD8(+) T cell infiltration to generate an immunosuppressive microenvironment in colorectal cancer. Faseb j 2023; 37(10):e23173. https://doi.org/10.1096/fj.202201982RRR.
Fionda C, Scarno G, Stabile H, et al. NK Cells and Other Cytotoxic Innate Lymphocytes in Colorectal Cancer Progression and Metastasis. Int J Mol Sci 2022; 23(14). https://doi.org/10.3390/ijms23147859.
Wei SC, Duffy CR, Allison JP. Fundamental Mechanisms of Immune Checkpoint Blockade Therapy. Cancer Discov 2018; 8(9):1069-1086. https://doi.org/10.1158/2159-8290.Cd-18-0367.
Ribas A, Wolchok JD. Cancer immunotherapy using checkpoint blockade. Science 2018; 359(6382):1350-1355. https://doi.org/10.1126/science.aar4060.
Wang X, Yang L, Huang F, et al. Inflammatory cytokines IL-17 and TNF-α up-regulate PD-L1 expression in human prostate and colon cancer cells. Immunol Lett 2017; 184:7-14. https://doi.org/10.1016/j.imlet.2017.02.006.
Garo LP, Ajay AK, Fujiwara M, et al. Smad7 Controls Immunoregulatory PDL2/1-PD1 Signaling in Intestinal Inflammation and Autoimmunity. Cell Rep 2019; 28(13):3353-3366.e3355. https://doi.org/10.1016/j.celrep.2019.07.065.
Li J, Wu C, Hu H, et al. Remodeling of the immune and stromal cell compartment by PD-1 blockade in mismatch repair-deficient colorectal cancer. Cancer Cell 2023; 41(6):1152-1169.e1157. https://doi.org/10.1016/j.ccell.2023.04.011.
Sui Q, Zhang X, Chen C, et al. Inflammation promotes resistance to immune checkpoint inhibitors in high microsatellite instability colorectal cancer. Nat Commun 2022; 13(1):7316. https://doi.org/10.1038/s41467-022-35096-6.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Inflammation-Related Gene Profiling in Colorectal Cancer: A New Prognostic Signature

Status:

Version 1

Abstract

Figures

1 Introduction

2 Methods

2.1 Data collection and preprocessing

2.2 IRG differential gene identification and molecular subtype characterization

2.4 Construction and validation of a prognostic signature for inflammation-related genes

2.5 Construction and verification of the nomogram

2.6 Functional enrichment analysis

2.7 Analysis of the tumor immune microenvironment and immune checkpoints

2.8 Tumor mutation and drug sensitivity assessment

2.9 Quantitative reverse transcription PCR (RT‒qPCR)

2.10 Statistical analyses

3 Results

3.1 Construction and prognostic analysis of molecular isoforms of inflammation-related DEGs in COAD

3.2 Construction and validation of gene subtypes based on IRG molecular subtype DEGs

3.3 Construction and validation of the IRG-related prognostic signature

3.4 Independent prognostic value of inflammatory risk profiles and the construction and validation of a nomogram

3.5 Correlation analysis of the IRG risk score and clinicopathologic characteristics of patients with COAD

3.6 Functional enrichment analysis of different IRG risk subgroups

3.7 Immune cell infiltration and immune checkpoint analysis between the two risk subgroups

3.8 Tumor mutation load and drug sensitivity analysis of IRG risk models

4 Discussion

5 Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1