StemCell Therapy

Construction of CSCC gene co-expression network and screening of immune-related module

Firstly, the unqualified samples and genes in the TCGA-CESC dataset were removed based on hierarchical clustering, and 253 samples and 4,553 genes were reserved for WGCNA to build the gene co-expression network. =3 (scale-free R2=0.92) was selected as an optimal soft threshold to construct a scale-free network, and finally 15 gene modules were obtained (Table 1). Then, the correlation between the feature genes of each module and four immune-related features (Stromal, Immune, Estimate Scores, and Tumor Purity) was calculated. It was found that brown module was significantly associated with immune status, presenting Immune Score (r=0.88, P=1e-84), Stromal Score (r=0.46, P=1e-14), ESTIMATE Score (r=0.79, P=2e-55) and Tumor Purity (r=-0.82, P=1e-63) (Fig.1). Therefore, the brown module was included in the subsequent study.

The immune-related modules based on WGCNA

Enrichment analysis was performed on 330 genes in the brown module to reveal relevant biological function. The results showed that the genes were largely enriched in the functions and pathways related to immune signal activation and immunomodulation, such as response to interferon-gamma, positive regulation of immune response, adaptive immune response, regulation of cytokine production, myeloid leukocyte activation, regulation of response to biotic stimulus, T cell activation, inflammatory response, negative regulation of immune system process, Type II interferon signaling (IFNG), Lysosome, response to tumor necrosis factor, regulation of viral process, etc. (Fig.2AC).

Enrichment analysis of genes in the brown module. A P value distribution of the top 20 enriched pathways and biological processes in the brown module; B The P value clustering network of genes in the brown module, with the redder the node color is, the more significant P value is; C Network analysis of enriched terms of genes in the brown module. Different node colors indicate different functional or pathway clusters that nodes belong to

The brown module was known to be highly correlated with cellular immunity. In the present study, consensus clustering was conducted on tumor samples based on the expression patterns of genes in the brown module to identify different immune subtypes. Since the grouping was suboptimal when using K=3 as the clustering value, we selected K=3 to divide the samples into three groups (Fig.3AC). The samples obtained were named as cluster A (38 cases), cluster B (132 cases) and cluster C (84 cases). To better understand the immune patterns of the three subgroups, we explored the expression of genes in the brown module in the three subgroups (Fig.3D). The results showed that most of the genes in the brown module were down-regulated in the cluster B subgroup, while most of the genes were up-regulated in the other two subgroups, and the overall level of gene up-regulation in the cluster A subgroup was more evident than that in the cluster C subgroup. Therefore, we assumed that the three subgroups might represent different immune patterns, which was further verified by subsequent analysis.

Consensus clustering analysis of gene expression pattern in the brown module. A Cumulative distribution function (CDF) of consensus clustering when K=2~9; B Relative change of AUC of CDF curve when K=2~9; C Tracking plot results of consensus clustering when K=3; D Heat map of gene expression in different subtypes in the brown module

GSVA was done to explore the biological behaviors of the three tumor immune subtypes. Cluster A enriched in the pathways associated with immune deficiency and disease development, such as PRIMARY IMMUNODEFICIENCY, TYPE I DIABETES MELLITUS, INTESTINAL IMMUNE NETWORK FOR IGA PRODUCTION, ALLOGRAFT REJECTION, etc. Cluster B was enriched in pathways related to immunosuppressive biological processes. Cluster C was mainly enriched in pathways associated with cell adhesion, cytokine and cytotoxic activation pathways, including CYTOSOLIC DNA SENSING PATHWAY, CELL ADHESION MOLECULES CAMS, HEMATOPOIETIC CELL LINEAGE, CYTOKINE NATURAL KILLER CELL MEDIATED CYTOTOXICITY, CYTOKINE RECEPTOR INTERACTION, etc. (Fig.4). These results indicated that the three subtypes have different enrichments in biological pathways, and it was speculated that these subtypes may have different biological behaviors.

Heat maps of GSVA among different subtypes. Red: up-regulated pathways; green: down-regulated pathways

The analysis of cell infiltration in TME showed that there were differences in the contents of B cells, T cells, NK cells, monocytes and macrophages among the three subtypes (Fig.5A). ssGSEA results showed significant differences in CD8 T cells, CD4 T cells, Treg cells, macrophage MD, M1 and dendritic cell contents among the three subtypes (Fig.5B). To further verify the classification, ESTIMATE was used to calculate Stromal Score, ESTIMATE Score, Immune Score, and Tumor Purity based on mRNA data. These indicators were used to distinguish the high, low and medium immune groups. Compared with low immune cell infiltration group, the high immune cell infiltration group had lower Tumor Purity and higher Stromal Score, Immune Score and ESTIMATE Score. Therefore, Cluster A was defined as high immune group, Cluster B as low immune group, and Cluster C as medium immune group (Fig.5C). High immune group was significantly positively correlated with ESTIMATE Score, Immune Score and Stromal Score, but negatively correlated with Tumor Purity (Fig.5D). human leukocyte antigen (HLA) is an expression product of human major compatibility complex and is also a highly polymorphic allogeneic antigen [23]. In the present study, the correlation between immune cell infiltration and HLA family proteins in different subgroups was analyzed to verify the rationality of typing. The results demonstrated that the expression of HLA family gene was significantly downregulated in high immune group compared with in low immune group (Fig.5E). The above results indicated that there were differences in the immune cell infiltration, immune-related scores and HLA family protein expression among subtypes, which also provided support for the rationality of the typing.

Analysis of immune cell infiltration and immune-related indices in different tumor subtypes. A CIBERSORT analysis of differences in immune cell composition among different subtypes; B Differences in the abundance of each immune infiltrating cell among different subtypes; C Heat map of immune cell typing; D Violin plot of the differential analysis of Tumor Purity, ESTIMATE Score, Immune Score and Stromal Score among the three subtypes; E Differences in the expression of HLA family gene among different subtypes

Subsequently, a prognostic model was constructed based on the genes in the brown module. In the TCGA-CESC dataset, the samples with survival time less than 30days were excluded. Then, for the 330 genes in the brown module, a univariate regression analysis was conducted, and 46 genes significantly associated with prognosis were obtained with P<0.01 as the screening condition (Additional file 1: Table S1). Next, lasso and multivariate regression analyses were done on these 46 genes, and 8 feature genes were obtained finally, including ISCU, MSMO1, GCH1, EEFSEC, SPP1, RHOG, LSP1 and TCN2 (Fig.6A, Additional file 2: Table S2). HRs of MSMO1 and SPP1 were higher than 1, which were risk factors for CSCC prognosis, while HRs of ISCU, GCH1, EEFSEC, RHOG, LSP1 and TCN2 were lower than 1, which could be protective factors for CSCC prognosis. The risk scores were calculated based on these 8 feature genes, and the samples were classified into high-risk and low-risk groups. According to the heat map, the expression levels of GCH1, EEFSEC, SPP1, RHOG, LSP1 and TCN2 were decreased overall with the increase of risk score (Fig.6B). Based on the risk score distribution and survival time of the high/low-risk group samples, we found that the number of patients dying increased and the survival time decreased with the increase of risk score (Fig.6CD). Survival curves of the high/low-risk groups also demonstrated that patients in the low-risk group had a higher survival rate (Fig.6E). ROC curve demonstrated the reliability of the risk assessment model in predicting 1-, 3- and 5- year survival rates of samples, with AUC values of 0.8, 0.77 and 0.75 respectively (Fig.6F). Also, the expression statuses of the 8 genes were examined using qRT-PCR, whose results showed that ISCU was downregulated, while MSMO1, GCH1, EEFSEC were upregulated in the tumor tissues (Fig.6G). In addition, this study assessed the correlation between the prognostic model and immune cell infiltration. As a result, the risk score was significantly negatively correlated with 6 immune cells, including B_ cell, CD8_ T cell, CD4_ T cell, neutrophil, dendritic cell and macrophage (Fig.7AF). To verify whether the risk score could be considered as an independent prognostic indicator, univariate and multivariate Cox regressions were introduced based on risk score and clinical features of the samples. As observed in Fig.7GH, risk score could independently serve as prognostic factor. In conclusion, we constructed an 8-feature gene risk assessment model to predict the prognosis of patients with CSCC and proved the favorable predictive ability of this model and revealed the association between the model and cellular immunity.

Construction and assessment of a prognostic model for CSCC. A Forest map of the 8-prognostic feature genes, *P<0.05; B Heatmap of expression of the 8 prognostic feature genes in the high- and low-risk groups; C Risk score distribution of CSCC patients, with green representing the low-risk group and red representing the high-risk group; D Scatter plot of survival status of CSCC patients, with green and red dots representing survival and death, respectively; E KaplanMeier survival curve of the high- and low-risk groups; F ROC curves of the prognostic model predicting 1-, 3-, and 5-year overall survival of patients.; G qRT-PCR was used to measure the mRNA expressions of the feature genes

Correlation between risk score and infiltration degree of 6 immune cells. A Correlation between risk score and B_cell infiltration; B Correlation between risk score and CD4 T cell infiltration; C Correlation between risk score and CD8 T cell infiltration; D Correlation between risk score and dendritic cell infiltration; E Correlation between risk score and macrophage infiltration; F Correlation between risk score and neutrophil infiltration. GH Univariate and multivariate Cox regression for risk score and the clinical features

Original post:
Identification of cervical squamous cell carcinoma feature genes and construction of a prognostic model based on immune-related features - BMC Women's...

This entry was posted on September 4, 2022 at 2:06 am and is filed under Immune System. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed. |