Method for constructing model for predicting prognosis of spermatocytoma based on immune-related genes
By constructing an immune-related prognosis model based on CTLA4, SNX17 and TMX1 genes, the problem of difficult to predict disease progression and immunotherapy response in TGCT patients in the prior art is solved, and accurate assessment of the risk of recurrence in patients with senotomy and immunotherapy is achieved. Optimization of effects.
Patent Information
- Application Number
- CN202410988951.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
AI Technical Summary
The prior art is difficult to effectively predict disease progression and response to immunotherapy in patients with testicular germ cell tumors (TGCT), especially in patients with semithromboma, and the lack of reliable immuno-related prognosis models, resulting in difficulty in chemotherapy resistance and recurrence risk. Evaluate.
An immune-related prognosis model based on CTLA4, SNX17 and TMX1 genes was constructed, and risk scoring formulas were established to predict the risk of recurrence and immunotherapy response in patients with senotomyoma.
This model can accurately predict 1-year, 3-year and 5-year survival rates in patients with serotonoma, with AUC values reaching 0.802, 0.899 and 0.943, exceeding traditional risk factors such as lymphovascular invasion and T-stage, providing a more powerful clinical Decision-making basis helps identify patients in high-risk groups and optimizes the effectiveness of immunotherapy.
Smart Images

Figure HDA0004958322750000011 
Figure HDA0004958322750000021 
Figure HDA0004958322750000031
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of gene engineering, and in particular relates to a method for constructing a seminoma prognosis prediction model based on immune-related genes. Background Art
[0002] Testicular germ cell tumors (TGCT) are considered one of the most common urinary malignancies, and despite their relatively low incidence, they are still showing an increasing trend in incidence year by year. In addition, they are the main type of solid tumors in young men aged 15-34 years, with about 50% of cases characterized by seminoma histology. In the United States alone, 9,190 new diagnoses and 470 deaths are expected in 2023, with deaths mainly attributed to tumor metastasis, recurrence or chemotherapy resistance. It is worth noting that at the time of TGCT diagnosis, 32.2% of patients have metastases, and 25% of these patients may relapse. The results of a survey of stage I testicular tumors showed that 18.9% of patients experienced tumor recurrence, and nearly half of the recurrent cases required additional chemotherapy or radiotherapy due to metastatic spread or recurrence, so seeking chemotherapy-resistant patients who benefit from established treatment patterns poses a huge challenge. In addition, the enhanced invasiveness of testicular tumor cells is an important prognostic indicator of tumor recurrence and poor clinical outcomes. A nationwide, population-based cohort study of 924 patients in Denmark showed that 148 (16%) patients relapsed during a median follow-up of 6.3 years. Factors including invasion of the hilum of the testis (including the rete testis and hilar soft tissue), lymphovascular invasion, and elevated levels of β-human chorionic gonadotropin (β-HCG) and lactate dehydrogenase (LDH) before orchiectomy emerged as independent predictors of recurrence. A study that included advanced germ cell tumors showed that all patients with progressive seminoma were resistant to chemotherapy. These findings highlight the need to explore additional treatment strategies for patients with advanced seminoma. The clinical efficacy of checkpoint inhibitors targeting the PD-1 / PD-L1 system is remarkable in various cancers, and previous studies have shown that tumor-infiltrating lymphocytes expressing PD-L1 in seminoma specimens have prognostic significance. In addition, there is considerable optimism about the potential of combination drug approaches targeting multiple immune checkpoint receptors to improve treatment efficacy.
[0003] Testicular germ cell tumors (TGCT) show high sensitivity to cisplatin-based chemotherapy, and comprehensive treatment with surgery as the mainstay has resulted in a high cure rate for the disease. However, patients who experience disease progression after surgery tend to have a low response to chemotherapy. Approximately 20% of patients diagnosed with clinical stage I testicular seminoma will relapse. In cases of advanced seminoma, up to 30% of patients may show cisplatin resistance and require aggressive salvage therapy, with a 50% risk of death in these patients. However, there are no more treatment options available clinically other than chemoradiotherapy.
[0004] Cytokine-related pathways were frequently enriched in this study, including “cytokine-mediated signaling pathways” in GO analysis, and “cytokine-cytokine receptor interaction” pathways based on hub genes and risk scores in KEGG and GSVA analyses, respectively. Treatment plans after orchiectomy include active surveillance, maintenance therapy or risk-adapted therapy for patients with relapse, plus surveillance for patients at low risk of relapse, and cisplatin-based adjuvant chemotherapy for patients at high risk of relapse. However, consensus remains elusive regarding the definition of high risk of relapse and the use of risk factors in patient stratification. Although Warde et al. introduced tumor size (>4 cm) and rete testis invasion as criteria for identifying patients at high risk of relapse in their clinical guidelines, the evidence supporting these risk factors in clinical decision making is significantly insufficient. The lack of consistency in the association between rete testis invasion and increased risk of relapse leaves uncertainty about its independence as a predictor of relapse. Although many studies have highlighted the prognostic relevance of tumor size, the evidence confirming the 4 cm risk stratification threshold remains unclear, and reliance on tumor size alone for risk stratification does not confirm its clinical utility. A systematic review showed that the 5-year recurrence risk for tumors >4 cm and ≤4 cm was 17.4%-27.0% and 4.5%-13.4%, respectively. In high-risk cases (tumors ≥6 cm), the 10-year cumulative risk of recurrence reached 32.0%, suggesting that adjuvant therapy based on tumor size risk assessment may lead to overtreatment of approximately two-thirds of patients.
[0005] Therefore, it is necessary to elucidate the underlying molecular mechanisms and prognostic markers associated with poor outcomes in this specific patient subpopulation. To date, validated immune-related models that can predict response to immunotherapy and PFS have not been reported. Summary of the invention
[0006] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for constructing a prognostic model based on immune-related genes. Subsequent verification and analysis showed that this model reliably predicted the outcome and sensitivity of immunotherapy (ICI). For the first time, an immune-related prognostic model was constructed to predict disease progression and response to immunotherapy in patients with seminoma. Further experimental studies are expected to advance our understanding of the pathogenesis of seminoma and promote the development of new immunotherapy methods.
[0007] Another technical problem to be solved by the present invention is to provide an application of the prognosis model constructed by the method described in the construction of a clinical decision-making model.
[0008] Technical solution: To solve the above technical problems, the present invention provides the application of CTLA4, SNX17 and TMX1 genes in constructing a model for predicting the prognosis of seminoma.
[0009] The present invention also includes a method for constructing a model for predicting the prognosis of seminoma based on immune-related genes, and the method for constructing a model for predicting the prognosis of seminoma comprises the following steps:
[0010] 1) Identification of immune-related genes; differential expression analysis was performed on seminoma samples and normal testicular tissue samples, and visualization |Log2(FC)|>2 and p value <0.01 were set as the criteria for identifying differentially expressed IRGs. Differentially expressed immune-related genes were identified, and candidate genes were analyzed by WGCNA to identify immune-related hub genes;
[0011] 2) Univariate Cox regression, lasso regression, and multivariate Cox regression were performed to determine the prognostic characteristics of immune-related hub genes;
[0012] 3) Construction of prognostic model: An immune-related prognostic model was constructed using CTLA4, SNX17 and TMX1 genes. The formula of the prognostic model was: risk score = CTLA4 expression * (-0.66) + SNX17 expression * 0.33 + TMX1 expression * 0.37; risk score threshold = 5.555; when the risk score is higher than the cutoff value of 5.555, it is high risk, and when the risk score is lower than the cutoff value of 5.555, it is low risk.
[0013] 4) The TCGA-TGCT and GSE99420 cohorts were used to evaluate the diagnostic efficacy of the model through PFS, C-index, DCA and ROC analysis.
[0014] Among them, the WGCNA analysis described in step 1) includes determining the average expression of each gene and removing genes with meanFPKM<0.5, filtering the gene expression matrix by the "goodSamplesGenes" function to check for missing values, using the "soft" threshold power to calculate the adjacency, and then converting the adjacency into a topological overlap matrix, followed by determining the gene proportion and dissimilarity, using hierarchical clustering and dynamic tree cutting functions to detect modules, using average linkage hierarchical clustering to classify genes with the same expression profile into gene modules under the minimum genome size and TOM-based dissimilarity metric, finally, calculating the dissimilarity of module feature genes, selecting the cutting line of the module dendrogram, and combining several module genes for further study, entering the module genes into the STRING database, constructing a protein-protein interaction network, and using the MCODE plug-in in Cytoscape to visualize and analyze the network to obtain important clusters of Hub genes.
[0015] Among them, step 1) also includes obtaining a turquoise module by determining the average linkage hierarchical clustering and the optimal soft threshold power, and the MM value of the clustered genes in the turquoise module is >0.5 or <-0.5, and further analysis is performed to obtain module genes that are highly correlated with seminoma. The module genes are input into the STRING database to construct a protein-protein interaction network, and the network is visualized and analyzed using the MCODE plug-in in Cytoscape to obtain important Hub gene clusters.
[0016] Among them, the p-value filtering value of Cox regression analysis in step 2) is 0.05.
[0017] The method specifically comprises the following steps:
[0018] 1) Data collection and preprocessing: Download transcriptome data of testicular germ cell tumors from TCGA, divide seminoma samples into recurrent and non-recurrent groups, download RNA-seq data and clinical information of independent external validation cohorts, and single-cell RNA-seq data from GEO datasets, obtain normal testicular sample data from GTEx database, and convert all original data sets into log2 form after calibration and normalization using algorithms in R;
[0019] 2) Identification of immune-related genes: Immune-related gene sets were obtained from the ImmPort and InnateDB databases. After normalizing the expression matrix using the “normalizeBetweenArrays” package in R, the differentially expressed immune-related genes between seminoma and normal testis were explored using the “limma” package and visualized using the “pheatmap” package. |Log2(FC)|>2 and p-value<0.01 were set as the criteria for identifying differentially expressed IRGs; weighted gene co-expression network analysis was then performed using the “WGCNA” package in R; gene proportions and dissimilarity were subsequently determined, and then, stratification was used. Sub-clustering and dynamic tree cutting functions were used to detect modules. Under the condition of minimum genome size and TOM-based dissimilarity metric, average linkage hierarchical clustering was used to classify genes with the same expression profile into gene modules. Finally, the dissimilarity of module characteristic genes was calculated, the cutting line of the module dendrogram was selected, and several modules were combined for further study. The String database was used to visualize the network between the most important module genes, and the Cytoscape (V3.9.0) plug-in mcode was used to construct key clusters in the network. The genes in the key clusters were considered to be hub genes, and GO and KEGG enrichment analysis were performed on the hub genes respectively.
[0020] 3) Construction of immune-related gene prognostic model: Progression-free survival analysis was performed on all hub genes. Survival-related genes met the model construction conditions. Univariate COX proportional hazard regression analysis, lasso regression, and multivariate COX regression analysis were performed to construct the prognostic model. Significant genes were further included for lasso regression. The multivariate COX proportional hazard model was used to calculate the gene (coef). The risk score formula was constructed as follows: risk score = ∑Gene(exp) × Gene(cofe), where Gene(exp) represents the expression of each single gene. According to the median risk score, the sample was defined as a high-risk sample or a low-risk sample;
[0021] 4) The differential expression of immune-related marker genes between seminoma and normal testis, recurrent (R) and non-recurrent (NR) was analyzed using the "limma" package. Kaplan-Meier curves and log-rank tests were used to evaluate the relationship between PFS and risk score based on the TCGA-TGCT dataset and the external validation cohort GSE99420 dataset. Univariate and multivariate COX regression analysis was performed to evaluate the risk factors of age, stage, T, lymphovascular invasion (LI) and risk score in patients with seminoma, followed by ROC analysis to evaluate the prognostic value of the signature according to PFS years and risk factors. The C-index was calculated using the "dplyr", "rms" and "pec" packages in R software, and the "Regplot", "rms" and "survival" packages were used to build nomograms. The "ggDCA" package was used to perform decision curve analysis (DCA). Principal component analysis (PCA) was performed using the R "scatterplot3d" package based on the TCGA-TGCT and GSE99429 datasets, respectively.
[0022] Among them, the method includes an immune landscape exploration method and a single-cell sequencing analysis method.
[0023] The single-cell sequencing method specifically includes: normalizing the scRNA-seq data, converting the normalized data into a Seurat object, and identifying highly variable genes. Subsequently, reducing the dimensionality of the scRNA-seq data of highly variable genes, determining important principal components, and selecting the top 15 principal components based on the explained variance ratio for cell clustering, using the "FindNeighbors" and "FindClusters" functions in the "Seurat" software package for cell clustering analysis, and constructing a k-nearest neighbor graph based on the Euclidean distance in the principal component analysis, where "Find The nearest neighbors of each cell were determined using the “ndNeighbors” function. Subsequently, t-distributed random neighbor embedding was performed using the “RunTSNE” function to visualize the cell clustering in t-SNE-1 and t-SNE-2 dimensions. Differential expression analysis was performed using the “FindAllMarkers” function in the “Seurat” package. The Wilcoxon–Mann–Whitney test was used to identify differentially expressed genes for each cluster. Marker genes for each cluster were determined based on a critical threshold with an adjusted p-value < 0.01 and |log2(Foldchange)| > 1.
[0024] The present invention also includes the application of the prognostic model constructed based on the method in constructing a clinical decision-making model.
[0025] The clinical decision model is implemented by R language, which includes the risk scoring factors and lymphovascular invasion factors.
[0026] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: The present invention constructs a prognostic model by selecting CTLA4, SNX17 and TMX1 genes, and the progression-free survival rate of patients in the high-risk group is reduced. The AUCs for predicting 1-year, 3-year and 5-year survival rates are 0.802, 0.899 and 0.943, respectively, which exceed other risk factors such as lymphovascular invasion and T stage. The C index of the risk score is 0.838. Decision curve analysis (DCA) shows that combining lymphovascular invasion and risk score produces the most favorable decision results for patients. In addition, individuals classified as high risk may benefit more from immunotherapy. The prognostic model constructed for the first time based on immune-related genes in the present invention is a promising model for predicting the prognosis of patients with seminoma and evaluating the efficacy of immunotherapy. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1To identify immune-related hub genes. Heatmap (a) and volcano plot (b) of differentially expressed genes in TCGA between seminoma and normal testicular tissue; (c) WGCNA showed significant correlation between module characteristic genes and tissues. (d) Scatter plot of gene significance and module membership in the most significant module (turquoise module), with a correlation coefficient of 0.96, P < 1e–200. (e) PPI network constructed by STRING database based on 690 module genes. (f) Significantly enriched key clusters constructed by Cytoscape software. GO (g) and KEGG (h) enrichment analysis of key cluster genes.
[0028] Figure 2 Construction of immune-related prognostic signature (IRPS) model. (a) Univariate Cox and Lasso regression analysis based on survival-related hub genes (b). (c) Progression-free survival analysis of three signature genes. (d) Expression patterns of three signature genes. (ef) Verification of mRNA and protein expression using clinical samples (g). (R: recurrence; NR: no recurrence; **, p<0.01; ***, p<0.001; ***, p<0.01.)
[0029] Figure 3 Validation of the diagnostic efficacy of the IRPS model. (a) Progression-free survival analysis of seminoma cohorts based on TCGA and GSE99420 (b). (cd) Risk score, survival status distribution between high-risk and low-risk groups, and expression heatmaps of three signature genes. (e) PCA and GSVA (f) analysis based on risk score.
[0030] Figure 4Construction and validation of the nomogram. (a) ROC analysis of the PFS risk score at 1, 3, and 5-year follow-up in the TCGA seminoma cohort. (b) ROC curves of the time dependence of risk, age, stage, T, lymphovascular invasion (LI), and risk score at 5-year follow-up. Forest plots of univariate (c) and multivariate (d) Cox regression analysis of age, stage, T, LI, and risk score. C-index plot (e) and restricted mean survival (RMS) analysis (f) of risk score, age, stage, T, and LI. (g) Nomogram and calibration plot for predicting 1-year, 3-year, and 5-year PFS. Forest plots of univariate (h) and multivariate (i) Cox regression analysis of age, stage, T, LI, and nomogram. (jk) DCA analysis of the four models. (Model 1: age, stage, T, LI, and serum markers; Model 2: expression of CTLA4, TMX1, and SNX17, risk score, and nomogram; Model 3: T, LI, risk score, nomogram, and serum markers; Model 4: LI and risk score).
[0031] Figure 5 Validation of immune landscape and immunotherapy efficacy based on risk score. (a) Cibersort infiltration analysis of 22 tumor immune cells in high-risk and low-risk groups. (b) Differences in immune function between high-risk and low-risk groups. (c) Correlation study between the expression of 3 signature genes and the infiltration of 22 immune cells. (df) Survival curves of PFS-related immune cells. Differences in TIDE (g), dysfunction (h), exclusion (i), and potential immunotherapy responders (j) between high-risk and low-risk groups. (k) Expression distribution and abundance of 3 signature genes based on single-cell RNA sequencing (GSE197778). (l) Expression heat map of 3 signature genes in seminoma tissues.
[0032] Figure 6 For WGCNA analysis, the logarithm of node connectivity (log(k)) and the logarithm of node probability (log(P(k)) are analyzed.
[0033] Figure 7 Figure 2 shows the genes that are significantly mutated in seminoma samples from high-risk and low-risk groups. DETAILED DESCRIPTION
[0034] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.
[0035] Antibodies and reagents sources:
[0036] The primary antibodies were from: mouse anti-CTLA4 (Santa Cruz, sc376016), rabbit anti-SNX17 (abcam, ab223046), rabbit anti-TMX1 (abcam, ab244263), anti-GAPDH, CD27, CD22, CD3, Vimentin, E-cad, N-cad were all from Servicebio technology. Secondary antibodies included anti-mouse IgG, anti-rabbit IgG, Alexa Fluor secondary antibodies (spectrum 440 aqueous solution, 488 green, 546 red, 594 orange, 647 yellow) were also obtained from Servicebio technology.
[0037] 1. Data collection, preprocessing and data analysis
[0038] Transcriptome data (counts) of TGCT were downloaded from TCGA, and after reviewing clinical data, seminoma samples were divided into recurrent (R) and non-recurrent (NR) groups. RNA-seq data and clinical information of the independent external validation cohort (GSE99420) were downloaded from the GEO dataset. Single-cell RNA-seq (scRNA-seq, GSE197778) data were also downloaded from the GEO dataset. Normal testicular samples were obtained from the GTEx database. After calibration and normalization, all raw datasets were converted to log2 form using algorithms in R.
[0039] Most analyses were performed using R software (version 4.0.0) or Perl scripts (version 5.32.0). Perl scripts were used primarily for data processing tasks, including gene ID conversion and data merging. Group comparisons were performed using Student's t test or chi-square test, depending on the comparison of two or three groups. A significance threshold of p < 0.05 was used unless otherwise specified.
[0040] 2. Sample collection:
[0041] Seminoma samples and adjacent normal tissues were collected from specimens of 38 patients (mean age: 34±4 years, range: 21-43 years) at the First Affiliated Hospital of Nanjing Medical University and Nanjing Jinling Hospital between 2011 and 2021. These patients underwent orchiectomy and were subsequently diagnosed with testicular seminoma based on postoperative pathology. Twelve cases had postoperative recurrence, including local recurrence in 7 cases, new primary tumors in 4 cases, and distant metastasis in 1 case. Samples for RNA extraction were immediately frozen in liquid nitrogen and stored at -80°C. Samples for immunohistochemical analysis were fixed in formalin. The study design and protocol were approved by the Ethics Committee of Jinling Hospital, and informed consent was obtained from all participants.
[0042] 3. RNA extraction, reverse transcription and qRT-PCR related reagents, etc.:
[0043] Total RNA was extracted from the testicular tissue and cell lines collected in step 2 using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). Total RNA was reverse transcribed into complementary DNA (cDNA) using HiScript II (Vazyme, Shanghai, China). qRT-PCR was performed on an ABI 7900 system (Applied Biosystems, Carlsbad, CA, USA) using SYBR Green I (Vazyme, Shanghai, China), and the specific primers were as follows:
[0044] CTLA4 forward: 5′-GCCCTGCCTCGTTT-3′;
[0045] CTLA4 reverse: 5′-IndexTermGGTTGCGCACAGACTTCA-3′.
[0046] SNX17 forward: 5′-CGCTACGTGGCCTATACAT-3′;
[0047] SNX17 reverse: 5′-CAATGGGTCTTGCCGAACAG-3′.
[0048] TMX1 forward: 5′-TTGCGAAAGTAGATGTCACAGAG-3′;
[0049] TMX1 reverse: 5'-CTGATAGCGCCTAATTCACCAT3'.
[0050] 4. Identification of immune-related genes
[0051] Immune-related gene sets were obtained from the ImmPort and InnateDB databases. After normalizing the expression matrix using the “normalizeBetweenArrays” package in R, the differentially expressed immune-related genes (IRGs) between seminoma and normal testis were explored using the “limma” package and visualized using the “pheatmap” package. |Log2(FC)|>2 and p-value<0.01 were set as the criteria for identifying differentially expressed IRGs. Weighted gene co-expression network analysis (WGCNA) was then performed using the “WGCNA” package in R. First, the mean expression (meanFPKM) of each gene was determined, and genes with meanFPKM<0.5 were removed. Second, the gene expression matrix was filtered by the “goodSamplesGenes” function to check for missing values. Third, the adjacency was calculated using a “soft” threshold power. The adjacency was then converted to a topological overlap matrix (TOM). Gene proportions and dissimilarity were subsequently determined. Modules were then detected using hierarchical clustering and dynamic tree cutting functions. In the case of a minimum genome size (n = 50) and a TOM-based dissimilarity metric, average linkage hierarchical clustering was used to classify genes with the same expression profile into gene modules. Finally, the dissimilarity of module eigengenes was calculated, the cutting lines of the module dendrogram were selected, and several modules were combined for further study. The String database (https: / / cn.string-db.org / ) was used to visualize the network between the most important module genes, and the plug-in mcode of Cytoscape (V3.9.0) was applied to construct the key clusters in the network. The selection criteria were as follows: rank cutoff = 2, maximum depth = 100, node score cutoff = 0.2, and K score = 2. Genes in the key clusters were considered hub genes. We performed GO and KEGG enrichment analysis on hub genes, respectively. Statistical significance was set at P < 0.05 and false discovery rate (FDR) q < 0.05. The procedure was performed using R packages, including “clusterProfiler”, “org.Hs.eg.db”, “circize”, and “dplyr”.
[0052] Identification of immune-related hub genes showed that in a differential expression analysis comparing 66 seminoma samples with 165 normal testes, a total of 766 differentially expressed immune-related genes (IRGs) were identified, of which 511 were upregulated and 255 were downregulated in seminoma compared with normal testes ( Figure 1 a and 1b). To identify immune-related hub genes, WGCNA analysis was performed on candidate genes (n = 766). The logarithm of node connectivity (log(k)) and the logarithm of node probability (log(P(k)) Figure 6). Two distinct modules were delineated by mean linkage hierarchical clustering and optimal soft thresholding power determination ( Figure 1 c). The turquoise module was highly correlated with seminoma (cor = 0.96, P < 1e-200), and the clustered genes within the turquoise module (MM value > 0.5 or < -0.5) were selected for further analysis. Subsequently, a total of 690 module genes were entered into the STRING database to construct a protein-protein interaction (PPI) network ( Figure 1 e). Visualization and analysis of the network using the MCODE plugin in Cytoscape revealed a significant cluster with an MCODE score of 31.947, consisting of 39 nodes and 1214 edges ( Figure 1 f). GO enrichment analysis based on these 39 Hub genes showed that “cytokine-mediated signaling pathway” and “outside the plasma membrane” were enriched ( Figure 1 g). In addition, according to KEGG analysis, the “cytokine-cytokine receptor interaction” pathway is the most important pathway and is located at the center of the pathway network ( Figure 1 h).
[0053] 5. Establishment and evaluation of immune-related prognostic models
[0054] All hub genes were analyzed for progression-free survival (PFS), and survival-related genes met the model construction conditions. We performed univariate COX proportional hazard regression analysis, lasso regression, and multivariate COX regression analysis to construct the prognostic model. The p filter value of univariate COX regression was 0.05, and significant genes were further included for lasso regression. The AIC value was used to optimize the multivariate COX model, and the gene with the lowest AIC was retained in the final constructed model. The risk score formula is as follows: risk score = ∑Gene(exp) × Gene(cofe) (Gene(exp) represents the expression of each single gene, and the gene (coef) is calculated using the multivariate COX proportional hazard model). According to the median risk score, the sample was defined as a high-risk sample or a low-risk sample. The differential expression of immune-related marker genes between seminoma and normal testis, R, and NR was analyzed using the "limma" software package. Kaplan-Meier curves and log-rank tests were performed to evaluate the relationship between PFS and risk score based on the TCGA-TGCT dataset and the external validation cohort GSE99420 dataset.
[0055] Univariate and multivariate COX regression analyses were performed to evaluate the risk factors of age, stage, T, lymphovascular invasion (LI), and risk score in patients with seminoma. ROC analysis was then performed to evaluate the prognostic value of the signature according to the number of years of PFS and risk factors (age, stage, T, LI, and risk score). The C-index was calculated using the "dplyr", "rms," and "pec" packages in R software. The "Regplot," "rms," and "survival" packages were used to build nomograms. The "ggDCA" package was used to perform decision curve analysis (DCA). Principal component analysis (PCA) was performed using the R "scatterplot3d" package based on the TCGA-TGCT and GSE99429 datasets, respectively.
[0056] IRPS construction and expression validation results. In order to identify independent prognostic genes, 39 immune-related hub genes were analyzed for PFS to evaluate their relationship with survival. Subsequently, these survival-related genes were subjected to univariate Cox regression analysis, and four genes (CTLA4, SNX17, TMX1, and TBX21) were selected for LASSO regression analysis ( Figure 2 a). After evaluation based on the AIC value ( Figure 2 b), multivariate Cox regression analysis was performed on these four genes. Finally, an immune-related prognostic model was constructed using CTLA4, SNX17, and TMX1, with the formula: CTLA4 expression*(-0.66)+SNX17 expression*0.33+TMX1 expression*0.37. Compared with normal testis, these three model genes were upregulated in seminoma tissues. However, although CTLA4 was downregulated in relapsed (R) samples, TMX1 and SNX17 were upregulated in relapsed samples compared with non-relapsed (NR) samples ( Figure 2 d). Notably, high expression levels of TMX1 and SNX17 were associated with an increased risk of recurrence ( Figure 2 c), which was classified as a risk gene. In contrast, CTLA4 was identified as a protective gene.
[0057] To verify the expression pattern of three model genes CTLA4, TMX1 and SNX17, qRT-PCR analysis was performed on 38 seminoma samples and their corresponding adjacent normal tissues. The reaction system was prepared according to the instructions of the Novozyme SYBR Green I reagent, 2×AceQ qPCR SYBR Green Master Mix (10μl), upstream primer (0.4μl), downstream primer (0.4μl), 50×ROX Reference Dye 1 (0.4μl), cDNA (1μl), and ddH2O was added to the total reaction system of 20μl. The qPCR reaction was performed according to the following conditions: Stage-1 95℃5min; Stage-2 first 95℃10s, followed by 60℃30s; Stage-3 melting curve.
[0058] According to the above analysis, all three signature genes CTLA4, TMX1 and SNX17 were found to be upregulated in seminoma samples ( Figure 2 e). In addition, TMX1 and SNX17 showed upregulation in relapse samples, while CTLA4 showed downregulation in relapse samples ( Figure 2 f).
[0059] Immunohistochemistry (IHC) staining experiments IHC staining and evaluation followed previous methods (Cui Y, Miao C, Liu S, Tang J, Zhang J, Bu H, et al. Clusterin suppresses invasion and metastasis of testicular seminoma by upregulating COL15a1. Mol Ther Nucleic Acids. 2021; 26: 1336-50.). Specifically, anti-CTLA4, anti-SNX17, and anti-TMX1 antibodies were used to evaluate the protein expression levels of CTLA4, SNX17, and TMX1 in seminoma and adjacent normal testicular tissue, respectively. These findings were confirmed by IHC staining, indicating that the results were consistent with mRNA expression ( Figure 2 g).
[0060] However, patients with CTLA4 immunoexpression above the 50% cutoff in immune cells showed significantly less LI and exhibited lower pT and N stages compared with patients with <50% immune cell positivity. We found that CTLA4 was downregulated in recurrent seminoma tissues, which resulted in a reduced proportion of γ-ΔT cells and poor outcomes.
[0061] IRPS prognostic evaluation: The samples were divided into high-risk group and low-risk group according to the IRPS risk score. Compared with the high-risk group, the low-risk group showed better PFS, which was shown in the TCGA-TGCT cohort ( Figure 3 a, P < 0.001) and validation GSE99420 cohort ( Figure 3 b, P = 0.012). In addition, as the risk score increased, a higher incidence of recurrence and a shorter progression-free survival were observed, such as Figure 3 c and 3d. The expression patterns of the three model genes CTLA4, TMX1 and SNX17 in the high-risk group and the low-risk group are shown in Figure 3 The principal component analysis (PCA) scatter plot demonstrates the effectiveness of IRPS in distinguishing samples ( Figure 3 e). In addition, gene set variation analysis (GSVA) showed that the "CYTOKINE_CYTOKINE_RECEPTOR_INTERACTION" pathway was significantly enriched in the low-risk group and the "MISMATCH_REPAIR" pathway was significantly enriched in the high-risk group ( Figure 3 f). Missense mutations of KIT and KRAS were found in both high-risk and low-risk groups ( Figure 7 ).
[0062] Specifically, take one example to illustrate. Take 0.1g of tumor tissue from a patient with testicular seminoma, cut and grind the tissue, and then use the operating steps of the tissue RNA extraction kit to extract total tissue RNA. Then perform RNA quality testing, and use the NanoDrop device to test the quality and concentration of the extracted RNA to ensure that the quality of the RNA meets the experimental requirements. Then the extracted RNA is reverse transcribed to synthesize cDNA for library construction. The constructed library is subjected to high-throughput sequencing, and the 1umin platform is selected for sequencing to obtain the RNA sequencing data of the patient. The expression levels of three genes were obtained by sequencing, CTLA4 expression level = 5.5618817, TMX1 expression level = 4.187253, SNX17 expression level = 10.042354;
[0063] TCGA-ZM-AA05
[0064] Risk score = CTLA4 expression 5.5618817*(-0.66)+TMX1 expression 4.187253*0.37+SNX17 expression 10.042354*0.33=1.1924. This risk score is lower than the cutoff value of 5.555, so it is a low-risk case. The result is consistent with the actual situation.
[0065] The AUC values of the IRPS model for 1-year, 3-year, and 5-year survival rates were 0.802, 0.899, and 0.943, respectively ( Figure 4 a). The risk score showed a higher AUC value of 0.943 ( compared with age (AUC = 0.636), stage (AUC = 0.507), T (AUC = 0.641) and LI (AUC = 0.603). Figure 4 b). Univariate Cox regression analysis showed that the risk score of patients with seminoma was an independent risk factor (p = 0.008, HR = 1.063, 95% CI = 1.016-1.112). However, based on multivariate Cox regression analysis, LI was also identified as a risk factor. The C index ranges from 0 to 1, where 1 indicates complete discrimination and 0.5 indicates random. The C index of the risk score was 0.838, which was lower than LI (0.896) but higher than other factors. Based on the restricted mean survival (RMS) analysis, only the risk score factor was identified as a risk factor for determining the prognosis of patients with seminoma, with a p value of 0.008, HR of 1.063, and 95% CI of 1.016-1.112 ( Figure 4 f).
[0066] The present invention establishes an accurate and stable mixed nomogram, including clinical pathological characteristics (age, stage, T, LI) and IRPS ( Figure 4 g). To use the nomogram, the patient's specific point (black dot) was positioned on each variable axis. The red line and red points rose to determine the point assigned to each variable; their cumulative sum (447) was positioned on the total score axis. Then, a line descended to the survival axis, and the probability of progression-free survival at 5 years (87.7%), 3 years (91.1%), and 1 year (96.6%) was calculated. In this comprehensive nomogram, the risk score model showed significant effects between these clinically relevant covariates, consistent with the results of RMS and multivariate Cox regression analysis. Therefore, IRPS can assist clinical management in predicting recurrence in patients with seminoma. The calibration curve of the PFS nomogram showed excellent agreement between prediction and observation. Subsequently, univariate and multivariate Cox regression analyses were performed to evaluate the nomogram. Notably, after integrating the nomogram with confounders, LI was no longer a risk factor, while the nomogram remained significant ( Figure 4 h, 4i).
[0067] The decision curve plots the sensitivity and specificity of the prediction model on the horizontal axis and the net benefit on the vertical axis. The reference strategy with a benefit of 0 ensures that its decision curve is always below the x-axis. In decision curve analysis (DCA), the closer the decision curve of the prediction model is to that of the reference strategy, the less clinically useful the prediction model is. In this study, four models were constructed and evaluated using DCA to assess their predictive ability. Model 1 included age, stage, T, LI, and serum markers (LDH, HCG, and AFP); Model 2 included CTLA4 expression, TMX1 expression, SNX17 expression, risk score, and nomogram; Model 3 combined T, LI, risk score, nomogram, and serum markers; Model 4 involved LI, risk score. Compared with the other three models, Model 4 showed superior net benefit ( Figure 4 j). In addition, the largest net reduction in interventions was observed in Model 4 ( Figure 4 k).
[0068] The DCA curve estimated by our nomogram indicated that including LI and risk score (model 4) in predicting PFS in patients with testicular tumors provided a net benefit over a complete treatment or no treatment strategy, thereby helping clinicians and patients make decisions on chemotherapy administration. Through a series of analyses and screening, IRPS was constructed using CTLA4, TMX1, and SNX17, in which CTLA4 was defined as a protective factor and SNX17 and TMX1 were considered risk factors. CTLA4 is generally considered to be an accomplice of tumor cells, which can be activated by tumor cells to inactivate activated T cells, thereby achieving immune escape of the tumor itself. CTLA-4 blockers, including Tremelimumab, Ipilimumab, etc., are designed to induce antitumor immune responses by promoting the activation and proliferation of tumor-specific T cells, and have been widely used in various malignancies, such as lung cancer and esophageal cancer. In contrast, for patients with advanced seminoma, CTLA4 may play a protective role. In a previous study, CTLA-4 positivity in immune cells was not associated with clinicopathological variables in testicular cancer.
[0069] 6. CIBERSORT and Immunotherapy Response Assessment
[0070] CIBERSORT was used to evaluate the infiltration pattern of 22 tumor immune cells in patients with seminoma. The Wilcoxon rank sum test was used to compare the differences in the levels of immune cell infiltration between the high-risk group and the low-risk group. The Spearman test was used to test the correlation between the model characteristic genes and immune cell infiltration. The differences in immune function between the high-risk group and the low-risk group were performed using the "GSVA" package. In addition, TIDE (http: / / tide.dfci.harvard.edu / ) was used to evaluate the clinical efficiency of immune checkpoint inhibition therapy. The higher the TIDE prediction score, the worse the treatment effect and the worse the prognosis.
[0071] CIBERSORT analysis was used to evaluate the differences in immune infiltration between the high-risk and low-risk groups. Figure 5 As shown in a, there are significant differences in the infiltration of naive B cells, memory B cells, and γ-ΔT cells. In addition, the different infiltration patterns of B cells may be related to the expression of TMX1 and SNX17 ( Figure 5 c). Increased expression of TMX1 and SNX17 was associated with a decrease in the proportion of memory B cells and an increase in the proportion of naive B cells, ultimately leading to PFS time ( Figure 5 d, 5e). In addition, increased CTLA4 expression was associated with a higher proportion of γ-ΔT cells and improved PFS ( Figure 5 f). The differences in immune function between the high-risk group and the low-risk group were also analyzed ( Figure 5 b). It is noteworthy that immune functions related to APC co-inhibition, APC co-stimulation, CCR, checkpoint regulation, cytolytic activity, HLA, inflammation promotion, parainflammation, T cell co-inhibition, and T cell co-stimulation were significantly suppressed in the high-risk group. In addition, TIDE ( Figure 5 g) and functional disability scores ( Figure 5 h) is lower, and the exclusion rate is higher ( Figure 5 i). The proportion of patients predicted to respond to immunotherapy was higher in the high-risk group (p<0.01, Figure 5 j).
[0072] 7. Single-cell sequencing analysis
[0073] We analyzed single-cell RNA sequencing (scRNA-seq) data using R packages including “Seurat” and “SingleR”. To ensure high-quality data, three filtering criteria were applied to each cell: (1) retaining genes expressed in at least 5 single cells, (2) excluding cells with more than 5% of mitochondrial genes, and (3) removing cells expressing less than 100 genes. Initially, the scRNA-seq data were normalized using the “LogNormalize” method. The normalized data were then converted into Seurat objects, and the top 1500 highly variable genes were identified using the “FindVariableFeatures” function. Subsequently, principal component analysis (PCA) was performed using the “RunPCA” function in the “Seurat” R package to reduce the dimensionality of the scRNA-seq data based on the top 1500 genes. Significant principal components were identified using JackStraw analysis, and the top 15 principal components were selected based on the proportion of variance explained for cell clustering. Cell clustering analysis was performed using the “FindNeighbors” and “FindClusters” functions in the “Seurat” package. A k-nearest neighbor graph was constructed based on the Euclidean distance in principal component analysis, where the “FindNeighbors” function determined the nearest neighbors of each cell. Subsequently, t-distributed stochastic neighbor embedding (t-SNE) was performed using the “RunTSNE” function to visualize the cell clusters represented in t-SNE-1 and t-SNE-2 dimensions. Differential expression analysis was performed using the “FindAllMarkers” function in the “Seurat” package, using the Wilcoxon–Mann–Whitney test to identify differentially expressed genes (DEGs) for each cluster. Marker genes for each cluster were determined based on a critical threshold with an adjusted p-value < 0.01 and |log2(fold change)| > 1. For cluster annotation, a reference-based annotation approach was adopted, utilizing the reference data of the Human Primary Cell Atlas.
[0074] In the present invention, testicular tumor homogenate was designated as control cells and renal hilar lymph node homogenate was designated as treated cells for single cell RNA sequencing (scRNA-seq) analysis. The results showed that CTLA4 is mainly expressed in T cells ( Figure 5k.5l). In addition, a decreased distribution of T cells was observed in the hilar lymph node tissue (P<0.001, avg_log2FC=7.99). Notably, TMX1 showed significant expression in smooth muscle cells, monocytes, and macrophages. TMX1 showed upregulation in smooth muscle cells and macrophages within the hilar lymph node homogenate compared with the testicular tumor homogenate. In addition, SNX17 showed significant expression in NK cells, B cells, and macrophages, and upregulation was observed in these cells compared with control cells from testicular tumors.
[0075] In summary, the present invention reports an IRPS model for the first time and verifies its potential in predicting recurrence in patients with seminoma. Model 4 constructed further by DCA can help clinicians and patients determine whether chemotherapy is needed. According to the model of the present invention, patients classified as high risk may benefit less from traditional chemotherapy regimens but may benefit more from ICI treatment.
Claims
1. Application of CTLA4, SNX17 and / or TMX1 genes in constructing a model to predict the prognosis of seminoma.
2. A method for constructing a prognostic model for seminoma based on immune-related genes, characterized in that: The following steps are involved: 1) Identification of immune-related genes; differential expression analysis was performed on seminoma samples and normal testicular tissue samples, and visualization |Log2(FC)|>2 and p value <0.01 were set as the criteria for identifying differentially expressed IRGs. After differentially expressed IRGs were identified, immune-related hub genes were identified by WGCNA analysis of candidate genes; 2) Univariate Cox regression, lasso regression, and multivariate Cox regression were performed to determine the prognostic characteristics of immune-related hub genes; 3) Construction of prognostic model: An immune-related prognostic model was constructed using CTLA4, SNX17 and TMX1 genes. The formula of the prognostic model was: risk score = CTLA4 expression * (-0.66) + SNX17 expression * 0.33 + TMX1 expression * 0.37; risk score cutoff value = 5.555; when the risk score is higher than the cutoff value of 5.555, it is a high risk, and when the risk score is lower than the cutoff value of 5.555, it is a low risk; 4) The TCGA-TGCT and GSE99420 cohorts were used to evaluate the diagnostic efficacy of the model through PFS, C-index, DCA and ROC analysis.
3. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 2, characterized in that: The WGCNA analysis described in step 1) includes determining the mean expression of each gene and removing genes with meanFPKM<0.5, filtering the gene expression matrix by the "goodSamplesGenes" function to check for missing values, using a "soft" threshold power to calculate the adjacency, and then converting the adjacency into a topological overlap matrix, followed by determining gene proportions and dissimilarity, using hierarchical clustering and dynamic tree cutting functions to detect modules, using average linkage hierarchical clustering to classify genes with the same expression profile into gene modules under the minimum genome size and TOM-based dissimilarity metric, finally, calculating the dissimilarity of module feature genes, selecting the cutting line of the module dendrogram, and combining several module genes for further study, entering the module genes into the STRING database, constructing a protein-protein interaction network, and using the MCODE plug-in in Cytoscape to visualize and analyze the network to obtain important clusters of Hub genes.
4. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 3, characterized in that: Step 1) also includes obtaining a turquoise module by average linkage hierarchical clustering and optimal soft threshold power, wherein the MM value of the clustered genes in the turquoise module is >0.5 or <-0.5, and further analysis is performed to obtain module genes highly correlated with seminoma, and a protein-protein interaction network is constructed by inputting the module genes into the STRING database, and the network is visualized and analyzed using the MCODE plug-in in Cytoscape to obtain important clusters of Hub genes.
5. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 2, characterized in that: The p-value filter value for Cox regression analysis in step 2) was 0.
05.
6. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 2, characterized in that: The method specifically comprises the following steps: 1) Data collection and preprocessing: Download transcriptome data of testicular germ cell tumors from TCGA, divide seminoma samples into recurrent and non-recurrent groups, download RNA-seq data and clinical information of independent external validation cohorts, and single-cell RNA-seq data from GEO datasets, obtain normal testicular sample data from GTEx database, and convert all original data sets into log2 form after calibration and normalization using algorithms in R; 2) Identification of IRGs: Immune-related gene sets were obtained from the ImmPort and InnateDB databases. After normalizing the expression matrix using the "normalizeBetweenArrays" package in R, the differentially expressed IRGs between seminoma and normal testis were explored using the "limma" package and visualized using the "pheatmap" package. |Log2(FC)|>2 and p-value <0.01 were set as the criteria for identifying differentially expressed IRGs; weighted gene co-expression network analysis was then performed using the "WGCNA" package in R; gene proportions and dissimilarity were subsequently determined, and then, hierarchical Clustering and dynamic tree cutting functions were used to detect modules. Under the condition of minimum genome size and TOM-based dissimilarity metric, average linkage hierarchical clustering was used to classify genes with the same expression profile into gene modules. Finally, the dissimilarity of module characteristic genes was calculated, the cutting line of the module dendrogram was selected, and several modules were combined for further study. The String database was used to visualize the network between the most important module genes, and the Cytoscape (V3.9.0) plug-in mcode was used to construct key clusters in the network. The genes in the key clusters were considered to be hub genes, and GO and KEGG enrichment analysis were performed on the hub genes respectively. 3) Construction of immune-related gene prognostic model: Progression-free survival analysis was performed on all hub genes. Survival-related genes met the model construction conditions. Univariate COX proportional hazard regression analysis, lasso regression and multivariate COX regression analysis were performed to construct the prognostic model. The p-value filtering value of univariate COX regression was 0.05, and significant genes were further included for lasso regression. The multivariate COX proportional hazard model was used to calculate the gene (coef); The risk score formula is constructed as follows: Risk score = ∑Gene(exp) × Gene(cofe), where Gene(exp) represents the expression of each single gene. According to the median risk score, the sample is defined as a high-risk sample or a low-risk sample; 4) The "limma" package was used to perform differential expression analysis of immune-related marker genes between seminoma and normal testis, recurrent (R) and non-recurrent (NR). Kaplan-Meier curves and log-rank tests were used to evaluate the relationship between PFS and risk score based on the TCGA-TGCT dataset and the external validation cohort GSE99420 dataset. Univariate and multivariate COX regression analyses were performed to evaluate the risk factors of age, stage, T, lymphovascular invasion (LI) and risk score in patients with seminoma, followed by ROC analysis to evaluate the prognostic value of the signature according to PFS years and risk factors. The C-index was calculated using the "dplyr", "rms" and "pec" packages in R software. The "Regplot", "rms" and "survival" packages were used to build nomograms. The "ggDCA" package was used to perform decision curve analysis (DCA). Principal component analysis (PCA) was performed using the R "scatterplot3d" package based on the TCGA-TGCT and GSE99429 datasets, respectively.
7. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 2, characterized in that: The method also includes an immune landscape exploration method and a single-cell sequencing analysis method.
8. The method for constructing a prognostic model for seminoma based on immune-related genes according to claim 7, characterized in that: The single-cell sequencing method specifically includes: using normalization of scRNA-seq data, then converting the normalized data into Seurat objects, and identifying highly variable genes, then reducing the dimensionality of the scRNA-seq data of highly variable genes, determining important principal components, and selecting the first 15 principal components based on the explained variance ratio for cell clustering, using the "FindNeighbors" and "FindClusters" functions in the "Seurat" software package for cell clustering analysis, and constructing a k-nearest neighbor graph based on the Euclidean distance in the principal component analysis, where The “FindNeighbors” function determined the nearest neighbors of each cell. Subsequently, t-distributed random neighbor embedding was performed using the “RunTSNE” function to visualize the cell clustering represented in t-SNE-1 and t-SNE-2 dimensions. Differential expression analysis was performed using the “FindAllMarkers” function in the “Seurat” package. The Wilcoxon–Mann–Whitney test was used to identify differentially expressed genes for each cluster. Marker genes for each cluster were determined based on a critical threshold with an adjusted p-value < 0.01 and |log2(Foldchange)|>1.
9. Use of the prognostic model constructed based on the method described in any one of claims 2 to 8 in constructing a clinical decision-making model.
10. The use according to claim 9, characterized in that: The clinical decision model is implemented by R language, and includes the risk scoring factors and lymphovascular invasion factors described in claim 2.
Citation Information
Cited By
Patient immune response monitoring method and system for cell drug clinical research
CN119361070A
Method and system for monitoring patient immune response in clinical research of cell-based drugs
CN119361070B
Protein joint detection-based viral ARDS prognosis evaluation system and method
CN120609937A
Portable single gene expression rapid detection system and tumor risk assessment method thereof
CN121826147A
