Constructing a Neutrophil Extracellular Trap Model Based on Machine Learning to Predict Clinical Outcomes and Immunotherapy Response in Renal Cell Carcinoma

doi:10.21203/rs.3.rs-4700747/v1

Download PDF

Article

Constructing a Neutrophil Extracellular Trap Model Based on Machine Learning to Predict Clinical Outcomes and Immunotherapy Response in Renal Cell Carcinoma

https://doi.org/10.21203/rs.3.rs-4700747/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Neutrophil extracellular traps (NETs) represent a novel form of inflammatory cell death within neutrophils. Emerging research indicates that NETs promote cancer progression and metastasis in various ways. This study aims to provide prognostic NETs characteristics and therapeutic targets for patients with renal cell carcinoma (RCC). NMF analysis was conducted on 89 NET-related genes in the training cohort. Subsequently, WGCNA networks were utilized to study the subtype feature genes. Six machine learning algorithms were assessed for model training, and the optimal model was selected based on 1-year, 3-year, and 5-year AUC values. A NETs signature was then constructed to predict overall survival in RCC patients. Furthermore, multi-omics validation was performed based on NETs signature. Finally, stable knockout key gene RCC cell lines were established to verify the biological function of KCNN4 both in vitro and in vivo. This study highlights the emerging hot topic of NETs in RCC. We provide a prognostic NETs signature and identify multiple roles of KCNN4 in RCC. This work contributes to risk stratification and the identification of new therapeutic targets for RCC patients.

Neutrophil Extracellular Traps

Renal Cell Carcinoma

KCNN4

Prognostic Model

Non-negative Matrix Factorization

Machine Learning

Renal cell carcinoma (RCC) is the most common kidney tumor in adults and ranks among the top ten causes of cancer-related deaths worldwide [1]. The pathological types of RCC include clear cell renal cell carcinoma, papillary renal cell carcinoma, chromophobe renal cell carcinoma, collecting duct carcinoma, and unclassified renal cell carcinoma [2]. Clear cell renal cell carcinoma is the most common subtype, originating from the proximal renal tubules and typically exhibiting an aggressive phenotype [3]. Surgical treatment has been an effective option for primary RCC, but it shows limited efficacy in metastatic cases. Approximately one-third of RCC patients have metastases at the time of diagnosis [4]. Given the promising results of immune checkpoint inhibitors (ICI) in treating RCC, researchers have focused on immune cells and immune responses in the tumor microenvironment (TME) of RCC [5]. As important members of the TME, neutrophils are functionally linked to the progression of RCC [6, 7].

Neutrophils in the TME are considered crucial pathogenic factors contributing to tumor progression [8, 9], especially through the formation of neutrophil extracellular traps (NETs), which are recognized as novel pro-tumor mechanisms in various cancers [10]. NETs are reticular structures composed of DNA, histones, and granules released by activated neutrophils [11]. Previous studies have indicated that the levels of NETs in healthy individuals' peripheral blood are lower than those in cancer patients [12, 13]. Additionally, in vitro studies have shown that co-culturing neutrophils with breast cancer and lung cancer cells can induce NETs formation [14–16].

Recent studies have revealed that NETs also play an anti-tumor role in the TME [17]. On one hand, NETs can repel cytotoxic cells and attract immunosuppressive cells. On the other hand, NETs can act as physical barriers to prevent antibody-dependent cell-mediated cytotoxicity (ADCC) [18]. Notably, NETs can express PD-L1, leading to the exhaustion of CD8 + T cells within the TME. Anti-NETs therapies may enhance immune cell-mediated tumor cell killing and synergize with immune checkpoint inhibitors (ICIs) [19, 20]. Therefore, combining anti-NETs therapy with ICIs might reduce the incidence of ICI resistance, providing a novel therapeutic strategy for cancer treatment.

Currently, a novel anti-cancer strategy involves inhibiting the formation of NETs in tumors or promoting their degradation [21]. Unfortunately, drugs specifically targeting NETs have yet to be developed, which limits improvements in cancer patient outcomes [22]. However, targeting drugs that modulate neutrophil recruitment and NET formation has shown promise in preclinical models, suggesting a new therapeutic strategy of targeting upstream mediators of NET formation in tumors [23]. Therefore, exploring the upstream mediators of NETs has become essential to targeted therapy.

In this study, we developed a prognostic NETs signature for RCC patients. Notably, we demonstrated that KCNN4 is a crucial mediator of NETs formation and plays multiple roles in tumor migration and invasion, making it a potential oncogenic factor in RCC patients.

Data downloading and processing

This study utilized publicly available data obtained from several online databases. Gene expression data and clinical data for RCC were downloaded from The Cancer Genome Atlas (TCGA) at https://portal.gdc.cancer.gov/, comprising 613 RNA-seq samples and 537 clinical data. Gene expression data for RCC were obtained from the Gene Expression Omnibus (GEO) database at https://www.ncbi.nlm.nih.gov/geo/, including GSE167573 (n = 77) [24] and GSE29609 (n = 39) [25]. Immunotherapy cohorts included CheckMate009/010 (n = 76) [26] and CheckMate025 (n = 207) [27].

The data obtained from GEO were processed using the R package "limma" (version 3.50.1) for normalization and log2 transformation. Batch effects between datasets were removed using the "ComBat" function, for the data obtained from TCGA, gene ID annotation was performed using the Ensembl database. In cases where multiple gene names existed for a single gene, only the gene with the highest expression level was retained. RNA-seq expression data downloaded from TCGA were provided as transcripts per million (TPM), which were normalized using log2(TPM + 1).

Non-negative matrix factorization clustering based on NETs related genes

In previous studies, a total of 89 NETs-related genes were identified [28, 29]. Non-negative matrix factorization (NMF) was employed to perform clustering analysis on the TCGA cohort based on the expression levels of these NETs-related genes. In NMF, the number of subtypes (k) must be manually specified. The optimal k value can be determined by evaluating the product of the cophenetic correlation coefficient and dispersion coefficient provided by the NMF algorithm. To determine the optimal k value, we iterate over multiple k values (k = 3 ~ 8), running NMF for each k value (50 iterations), and selecting the k value that maximizes the product of the cophenetic correlation coefficient and dispersion coefficient. After determining the optimal number of subtypes, we run NMF with the optimal k value as a parameter (500 iterations) to identify subtypes within the TCGA cohort. The R package "nmf" is used throughout this process.

Enrichment analysis and single sample gene set enrichment analysis

R package "clusterprofiles" (version 4.2.2) was used for conducting Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis to predict the biological functions of genes. Enrichment analysis results with a p-value < 0.05 were considered significant.

Hallmark and immune cell infiltration gene sets were collected from the MSigDB database (https://www.gsea-msigdb.org/gsea/index.jsp). Single-sample Gene Set Enrichment Analysis (ssGSEA) was performed using the "gsva" function from the R package "GSVA" (version 1.44.3) to evaluate the normalized enrichment scores (NESs) of gene sets in each patient. The R package "estimate" was utilized to calculate StromalScore, ImmuneScore, ESTIMATEScore, and TumorPurity for each subtype.

Construction of WGCNA network

Weighted Gene Co-expression Network Analysis (WGCNA) is a systems biology approach that identifies modules of highly correlated genes based on their associations with phenotypes, aiming to find candidate biomarker genes and potential therapeutic targets. In this study, we utilized gene expression profiles from RCC in TCGA to construct a WGCNA network. First, samples and genes were filtered, and then the R package "WGCNA" (version 1.71) was used to calculate the Pearson correlations between all pairs of genes in the selected samples, constructing an adjacency matrix. To ensure a scale-free network, we chose β = 5 (R² = 0.80) as the soft-thresholding parameter. To further identify functional modules in the weighted gene co-expression network, we calculated the Topological Overlap Measure (TOM) based on the adjacency matrix. Using the TOM values, a dynamic tree-cutting approach was employed to define gene modules and select module eigengenes (MEs). MEs are considered representative of the gene expression profiles within each module.

Building prognostic NETs signatures using machine learning frameworks

To establish a robust NETs signature, six machine learning algorithms were employed in the TCGA cohort, including CoxBoost, elastic network (Enet), lasso, random forest (RF), Ridge regression, and support vector machine (SVM). Patients were divided into training and validation sets in a 7:3 ratio, and 10-fold cross-validation was used to evaluate model performance. The 1 years, 3 years, and 5 years AUC values were computed and compared for each model to select the best NETs signature.

Construction of nomogram model

After selecting the best NETs signature, ssGSEA was used to calculate the NESs of the signature. Using the NESs of the NETs signature and other clinical features, a nomogram model was constructed to predict RCC patients' overall survival (OS) rates at 1 year, 3 years, and 5 years. Calibration curves were built to assess the consistency between actual survival and predicted survival. The R package "rms" was utilized to construct the nomogram model.

NETs signature annotation of tumor microenvironment, signaling pathways, and immune-related features

We collected 5 categories of immune modulators: antigen presentation, immune suppression, immune activation, chemokines, and receptors. Signal pathways related to targeted therapy and immunotherapy were gathered from the MSigDB database, and NESs were computed using ssGSEA. Five immune deconvolution methods assessed immune cell infiltration abundance in RCC patients, including QUANTISEQ, CIBERSORT, MCPCOUNTER, xCell, and EPIC. The Tumor Immune Phenotype (TIP) tracker (http://biocchrbmu.edu.cn) was used to evaluate anticancer immune activity [30]. The TIDE (http://tide.dfci.harvard.edu/) database was used for immune-related feature analysis [31].

Cell lines and culture conditions

Human kidney cell carcinoma cell lines (Caki-1, ACHN, OS-RC-2, 786-O, Caki-2, and 769-P) and normal renal cell lines (HK-2) were obtained from the ATCC (Manassas, VA, USA). The kidney cell carcinoma cell lines were grown in Ham’s F-12K medium (Jiangsu Kaiji Biotechnology Co, Ltd) or RPMI-1640 medium (Gibco Laboratories, Grand Island, NY). And HK-2 was cultivated in DMEM medium (Gibco Laboratories, Grand Island, NY).

Western blotting assay and cell counting kit-8 (CCK-8) assay

Western blot analysis was conducted to determine the expression levels of genes. The blots were incubated with an antibody overnight at 4°C. The following antibodies were used: KCNN4 (Abcam, ab215990), GAPDH (Abcam, ab8245), PCNA (Abcam, ab29), N-cadherin (Abcam, ab76011), E-cadherin (Abcam, ab314063), Vimentin (Abcam, ab92547). We used the CCK-8 assay (Beyotime, Shanghai, China) to measure cell proliferation according to a specific protocol.

Wound healing assay

As previously described [32], RCC cell lines were cultured in serum-free media. Monolayers were scratched using a pipette tip and washed with PBS. After 24 hours, wound healing was captured under a microscope (Beckman).

Animal experiments

All animal experiments were conducted in accordance with ethical standards and approved by the Ethics Committee of the Cancer Hospital, Chinese Academy of Medical Sciences. In vivo studies involved subcutaneous injection of 5×10⁶ Caki-1 and 769-P cell lines in 100 µL PBS into the flanks of nude mice. Tumor volumes were monitored weekly using the formula volume = (length×width²)/2. Mice were euthanized on day 28 post-transplantation. Lung and subcutaneous tumor tissues were collected for further analysis. Lung samples were embedded in paraffin, and sections were stained with hematoxylin and eosin (H&E) to visualize the size of lung metastases.

Immunohistochemistry

The mouse antibodies against KCNN4 (Abcam, ab215990), N-cadherin (Abcam, ab76011), E-cadherin (Abcam, ab314063), Vimentin (Abcam, ab92547, 1:200) and Ki-67 (Abcam, ab15580) were purchased from Abcam. IHC experiments were conducted according to standard procedures. ImageJ's IHC Profiler macro was used to quantify the protein intensity expression in the samples.

Statistical analysis

We used the “survminer” package (version 0.4.9) in R to compute the optimal cutpoint using the “surv_cutpoint” function, which grouped patients accordingly. Subsequently, we utilized both the “survminer” and “survival” packages (version 3.3-1) to plot Kaplan-Meier survival curves for different patient groups and to perform log-rank tests to assess the significance of differences between groups. ROC curve analysis was conducted using the “pROC” package (version 1.18.0). Graphical visualization was performed using the “ggplot2” package (version 3.3.5). All statistical analyses in this study were conducted using the R programming language (version 4.2.0). Unless otherwise stated, all statistical tests were two-sided, and P < 0.05 was considered statistically significant.

Development and validation of NETs related gene subtypes

According to the expression levels of 89 NETs-related genes, we performed NMF analysis on the TCGA cohort. The correlation plot indicated that k = 4 was the optimal number of subtypes, thus RCC patients were divided into C1-C4 subtypes (Fig. 1A-B). The clinical data summary of patients with different subtypes of RCC is shown in Table S1. There were no significant differences in age, gender, and N stage distribution among the four subtypes, whereas UICC stage, T stage, M stage, grade, and tumor location showed significant differences in distribution. The Kaplan-Meier survival curve analysis revealed significant differences in OS among the four subtypes (Fig. 1C). Specifically, patients with the C2 subtype had better prognosis, while patients with C1 and C3 subtypes had shorter survival times. Enrichment analysis using ssGSEA on hallmark pathways showed that the C1 subtype enriched in the Angiogenesis pathway, while oxidative phosphorylation was enriched in the C2 subtype (Fig. 1D). The PI3K_AKT_mTOR_signaling pathway was enriched in the C3 subtype, and the C4 subtype was significantly enriched in Interferon_gamma_response and Interferon_alpha_response (Fig. 1D). Deconvolution analysis revealed that the four subtypes enriched different immune cells (Fig. 1E). Additionally, the tumor immune infiltration status was also significantly enriched in different subtypes; for example, the C2 subtype had higher levels of StromalScore, ImmuneScore, and ESTIMATEScore, while showing lower TumorPurity. In contrast, the C4 subtype had higher TumorPurity (Fig. 1F-I).

WGCNA network identifies subtype-related genes

We set a β soft threshold of 5 and constructed the WGCNA network (R² = 0.8, Fig. 2A-B). We identified the ME green module as the most significant module (Fig. 2C), which contained 824 genes (correlation coefficient 0.63, p = 2.4e-92, Fig. 2D). The univariate Cox regression analysis was performed on genes in the green module, filtering for 330 genes with p < 0.01 for further analysis (Table S2). Subsequently, we applied six machine learning algorithms to filter the 330 genes and select the optimal NETs signature. We computed the AUC values for each model at 1 year, 3 years, and 5 years to evaluate their prognostic performance for RCC patients. The results showed that the Ridge model selected genes with the highest AUC values (Fig. 2E-G), identifying a total of 7 genes (Table S2).

Next, we conducted GO enrichment analysis and KEGG pathway analysis on the genes identified by the Ridge model. GO enrichment analysis revealed that these signature genes were enriched in tumor necrosis factor production (BP, Fig. 2H), tertiary granule (CC, Fig. 2I), and MHC class I receptor activity (MF, Fig. 2J). KEGG pathway analysis indicated enrichment of these genes in osteoclast differentiation (Fig. 2K).

Determination and verification of prognostic NETs signatures

Based on ssGSEA, we calculated the NESs of the signature genes in RCC patients from the Ridge model and defined this as the NETs signature. Using the optimal cutoff value for the NETs signature, we divided patients into high and low groups. We first evaluated the clinical differences between the high and low NETs signature groups (Table S3). It is evident that the high NETs signature group has a higher proportion of advanced-stage patients (Stages III-IV), while the low NETs signature group has a higher proportion of early-stage patients (Stages I-II). Further analysis of their impact on RCC prognosis revealed that the low NETs signature group has significantly longer survival times than the high NETs signature group (Fig. 3A). Additionally, ROC analysis demonstrated that the NETs signature has higher predictive efficacy in RCC patients for 1-year, 3-year, and 5-year prognosis (Fig. 3B). To ensure the robustness of the analysis, we further validated the results using four validation cohorts. The results showed that the NETs signature performed well across all four validation cohorts (Fig. 3C-F).

To provide a quantitative method for clinicians to assess patient prognosis, we further explored the potential associations between the NETs signature and clinical-pathological characteristics in the TCGA cohort. We constructed a nomogram model incorporating the NETs signature and clinical-pathological characteristics (Fig. 3G). Higher scores of the NETs signature in RCC patients were associated with poorer prognosis. Using the nomogram model's calibration curves, we predicted the survival probabilities of patients at 1 year, 3 years, and 5 years after diagnosis using the NETs signature. The calibration curves for 1-year, 3-year, and 5-year survival probabilities effectively predicted the actual survival probabilities at these time points (Fig. 3H-J). These results indicate that the nomogram model based on the NETs signature has strong discrimination and calibration capabilities.

Features of NETs signatures in tumor microenvironment

In RCC patients, the NETs signature showed significant correlations with five categories of immune regulators (Fig. 4A). Antigen presentation, immune suppression, immune activation, and receptor molecules were especially highly expressed in the high NETs signature group, whereas chemokines were highly expressed in the low NETs signature group. Various deconvolution algorithms were used to assess the abundance of immune cell infiltration between the two patient groups. In the high NETs signature group, RCC patients were enriched with immune-promoting cells, such as NK cells, CD8⁺ T cells, and CD4⁺ T cells. Conversely, in the low NETs signature group, RCC patients were enriched with immune-suppressive cells, such as myeloid-derived suppressor cells (MDSCs), neutrophils, mast cells, and fibroblasts (Fig. 4B-C). We used online tools to calculate the TIP scores of RCC patients to explore the biological mechanisms associated with the NETs signature. The cancer immune cycle was more activated in the high NETs signature group, including tumor antigen release (Step 1), tumor antigen presentation (Step 2), immune cell activation (Step 3), recruitment of tumor-infiltrating immune cells (Step 4), and infiltration of immune cells (Step 5). In contrast, the killing of tumor cells (Step 7) was enriched in the low NETs signature group (Fig. 4D). Additionally, the NETs signature was significantly positively correlated with PD-1 immunotherapy, tumor-related miRNAs, and mismatch repair (Fig. 4E). We also observed significant positive correlations between the NETs signature and the HER2 and KIT signaling pathways, while significant negative correlations were observed with the RET and FLT3 signaling pathways (Fig. 4F).

NETs related immune characteristics

We further explored the relationship between the NETs signature and various immunotherapy predictive factors. The TIDE score, Dysfunction score, IFNG levels, Merrck18, and CD8 molecules were higher in the high NETs signature group (Fig. 5A). In contrast, the Exclusion score, MDSCs, cancer-associated fibroblasts, and tumor-associated M2 macrophages were higher in the low NETs signature group (Fig. 5A). Additionally, we analyzed the relationship between the NETs signature and immunotherapy response rates. The results revealed that the low NETs signature group had a higher proportion of immunotherapy responders (Fig. 5B). The NETs signature was further analyzed in multiple validation cohorts. It was evident that the low NETs signature group had a higher response rate to immunotherapy, while the high NETs signature group was relatively resistant (Fig. 5C-F). Based on the TIDE algorithm, RCC patients with a low NETs signature had a better response to immunotherapy.

Knockdown of KCNN4 inhibits mRNA and protein expression levels of KCNN4

Since KCNN4 shows higher expression levels and survival correlation in RCC patients, we selected it for validation in both in vivo and in vitro experiments (Fig. 6A-B, Table S4). First, we analyzed the expression differences of KCNN4 in various RCC cell lines and normal renal cell lines. We found that KCNN4 has higher mRNA expression levels in RCC cell lines, especially in Caki-1 (Fig. 6C). To further analyze the role of KCNN4 in RCC patients, we first knocked down KCNN4 in the Caki-1 cell lines. After knocking down KCNN4, both the mRNA and protein expression levels of KCNN4 were significantly lower than those in the control group (Fig. 6D), and similar results were shown in the Western blotting experiment (Fig. 6E). Next, KCNN4 was overexpressed in the 769-P cell lines, significantly increasing both mRNA and protein expression levels of KCNN4 (Fig. 6F-G).

Knockdown of KCNN4 inhibits tumor cell growth and EMT ability

We further investigated KCNN4, first knocking it down (sh-KCNN4) in the Caki-1 cell lines. The CCK8 assay showed that sh-KCNN4 inhibited the proliferation of Caki-1 cells (Fig. 7A). The colony formation assay indicated that sh-KCNN4 disrupted the colony formation of Caki-1 cells (Fig. 7B-C). The wound healing assay results demonstrated that sh-KCNN4 suppressed the invasion of Caki-1 cells (Fig. 7D-E). Western blotting assay of the EMT ability in Caki-1 cells revealed a slight increase in E-cadherin and a downregulation of N-cadherin and Vimentin when KCNN4 was inhibited (Fig. 7F). Next, we overexpressed KCNN4 in the 769-P cell lines. The CCK8 assay showed that KCNN4 overexpression promoted the proliferation of 769-P cells (Fig. 7G). The colony formation assay indicated that KCNN4 overexpression enhanced the colony formation of 769-P cells (Fig. 7H-I). The wound healing assay results demonstrated that KCNN4 overexpression increased the invasion of 769-P cells (Fig. 7J-K). Western blotting assay of the EMT ability in 769-P cells revealed a downregulation of E-cadherin and a slight increase in N-cadherin and Vimentin when KCNN4 was overexpressed (Fig. 7L).

Inhibiting KCNN4 can inhibit tumor growth and metastasis in vivo

In the in vivo experiments, we established a subcutaneous xenograft tumor model in nude mice. Compared to the NC group, the sh-KCNN4 group exhibited significantly smaller and lighter tumors (Fig. 8A-B). The tumor growth curve indicated that sh-KCNN4 inhibited tumor growth (Fig. 8C). Furthermore, HE staining of lung tissues showed that sh-KCNN4 reduced the formation of lung metastatic nodules and tumor thrombi in vivo (Fig. 8D). IHC results demonstrated that the expression of KCNN4 was downregulated in the sh-KCNN4 group, along with a reduction in Ki-67 expression and suppression of EMT-related proteins such as N-cadherin and Vimentin, while E-cadherin was increased (Fig. 8E). In contrast, the tumors in the KCNN4 overexpression group were significantly larger and heavier than the NC group (Figure S1A-B). The tumor growth curve indicated that KCNN4 overexpression promoted tumor growth (Figure S1C). Additionally, HE staining of lung tissues showed that KCNN4 overexpression increased the formation of lung metastatic nodules and tumor thrombi in vivo (Figure S1D). IHC results demonstrated that the expression of KCNN4 was upregulated in the KCNN4 overexpression group, along with an increase in Ki-67 expression and enhancement of EMT-related proteins such as N-cadherin and Vimentin, while E-cadherin was decreased (Figure S1E).

As key mediators of extracellular matrix formation, angiogenesis, and immune response, NETs play a crucial role in tumor progression and metastasis. NETs-related genes have been shown to be promising therapeutic targets in various cancers. Therefore, establishing a robust prognostic signature and exploring genes that mediate NETs formation may provide new therapeutic strategies for treating RCC.

Based on previously identified NETs-related genes, this study classified RCC patients into four subtypes. The tumor staging differences among the four subtypes were statistically significant, with the C2 subtype having a better prognosis and the C1 subtype having a poorer prognosis. In our study, six machine learning methods were used to predict patient survival. The Ridge algorithm demonstrated the best performance and was used to establish the NETs signature. Prognostic analysis indicated that the NETs signature is a risk marker for OS in RCC patients. ROC analysis further revealed that the NETs signature has high accuracy in predicting 1 year, 3 years, and 5 years of OS in RCC patients.

Patients in the high NETs signature group exhibited a large presence of anti-tumor immune cells in the TME, such as NK cells, CD8 + T cells, and CD4 + T cells. Conversely, the low NETs signature group of RCC patients was enriched with immunosuppressive cells, including MDSCs, neutrophils, mast cells, and fibroblasts. Additionally, various immune regulators, such as antigen presentation, immunosuppression, immune stimulation, chemokines, and receptors, were upregulated in the high NETs signature group, inhibiting tumor cell recurrence and metastasis. The cancer immunity cycle was also more activated in the high NETs signature group. These factors suggest that patients in the high NETs signature group should have better prognoses. However, in our study, patients in the low NETs signature group achieved better outcomes. We need to explore further the mechanisms underlying this contradiction in future studies.

From the perspective of immunotherapy, the NETs signature can predict the response rate of RCC patients receiving anti-PD-1 or anti-PD-L1 treatment. Notably, patients in the high NETs signature group benefit less from immunotherapy. Several immunosuppressive markers are upregulated in the high NETs signature group, suggesting a potential association between the lower response rate and these immunosuppressive markers. Therefore, improving the expression levels of these immunosuppressive markers in the tumor microenvironment of the high NETs signature group should be a primary therapeutic focus.

Few studies have addressed the role of KCNN4 in RCC. Here, we identified the biological functions of KCNN4 through both in vitro and in vivo experiments. Briefly, KCNN4 is a risk factor for the survival of RCC patients and is associated with advanced pathological stages. Notably, the knockdown of KCNN4 not only inhibited tumor cell growth but also suppressed EMT capabilities. In a subcutaneous xenograft tumor model, we demonstrated that KCNN4 knockdown could inhibit tumor growth and reduce lung metastasis. Moreover, in vivo experiments also showed that NETs formation-related proteins, including NE and Vimentin, were downregulated in the sh-KCNN4 group.

KCNN4 can induce neutrophil infiltration in RCC and regulate the formation of NETs. There is some evidence supporting this hypothesis. KCNN4 plays a crucial role in type I IFN signaling activation. IFN-α/IFN-γ, as important stimuli, can induce NETs formation, suggesting that KCNN4 may be involved in regulating NETs. Additionally, research by Kantari et al. has shown that KCNN4 interacts with proteinase 3 (PR3) and inhibits macrophage clearance of apoptotic neutrophils. This process contributes to pro-inflammatory effects and NETs formation.

This study has several limitations. Firstly, it is based on publicly available bulk data, which do not accurately reflect the cell-cell interaction effects of neutrophils and other immune cells. Moreover, due to the short lifespan of neutrophils, single-cell sequencing faces challenges in sample acquisition and relatively low sequencing depth. Additionally, although this study reveals the association between KCNN4 and NETs formation, the underlying mechanisms still need further validation at the pathway level.

As mentioned above, NETs play an irreplaceable role in the tumor progression of RCC. We performed NMF analysis on RCC patients and identified four subtypes associated with NETs. Using various machine learning algorithms, this study established and validated a robust NETs signature for RCC patients. Subsequently, KCNN4 was further screened as a key gene and validated through in vitro and in vivo experiments. We ultimately demonstrated that KCNN4 is detrimental to RCC tumor growth and is associated with EMT capability and NETs formation.

Author contributions

Yihao Zhu: data curation, writing–original draft. Yajian Li and Xuwen Li: methodology, editing, validation. Yuan Yu: visualization, software. Can Chen: editing. Mingshuai Wang: IHC assistant. Dong Chen: cell cultural guidance. Nianzeng Xing: conceptualization. Xiongjun Ye and Feiya Yang: conceptualization, writing review, editing, check, and approval.

Funding

This research was funded by the National Natural Science Foundation of China (No. 62076007) and the Beijing Natural Science Foundation (No. 7232132).

Data availability statement

Public datasets were analyzed in this study. TCGA RCC cohort was downloaded from the TCGA database (http://cancergenome.nih.gov/). GSE167573 and GSE29609 were downloaded from the GEO database(http://www.ncbi.nlm.nih.gov/geo/). CheckMate009/010 and CheckMate025 were obtained from Tumor Immunotherapy Gene Expression Resource databases (http://tiger.canceromics.org/). The original data and material were available when required from the corresponding author Xiongjun Ye.

Conflict of interest

The authors declare that they have no known competing financial interests in this paper.

Ethics approval and consent to participate

This study was approved by the ethics committee of the Cancer Hospital, Chinese Academy of Medical Sciences (NCC2024A284) and was in accordance with ARRIVE guidelines.

Consent for publication

All the authors and patients signed the consent for publication.

Capitanio U, Bensalah K, Bex A, Boorjian SA, Bray F, Coleman J, Gore JL, Sun M, Wood C, Russo P: Epidemiology of Renal Cell Carcinoma. Eur Urol 2019, 75(1):74-84.
Ljungberg B, Bensalah K, Canfield S, Dabestani S, Hofmann F, Hora M, Kuczyk MA, Lam T, Marconi L, Merseburger AS et al: EAU guidelines on renal cell carcinoma: 2014 update. Eur Urol 2015, 67(5):913-924.
Liang T, Sang S, Shao Q, Chen C, Deng Z, Wang T, Kang Q: Abnormal expression and prognostic significance of EPB41L1 in kidney renal clear cell carcinoma based on data mining. Cancer Cell Int 2020, 20:356.
Rini BI, Campbell SC, Escudier B: Renal cell carcinoma. Lancet 2009, 373(9669):1119-1132.
Xiong W, Zhong J, Li Y, Li X, Wu L, Zhang L: Identification of Pathologic Grading-Related Genes Associated with Kidney Renal Clear Cell Carcinoma. J Immunol Res 2022, 2022:2818777.
Xu H, Zheng X, Zhang S, Yi X, Zhang T, Wei Q, Li H, Ai J: Tumor antigens and immune subtypes guided mRNA vaccine development for kidney renal clear cell carcinoma. Mol Cancer 2021, 20(1):159.
Bai J, Zheng A, Ha Y, Xu X, Yu Y, Lu Y, Zheng S, Shen Z, Luo B, Jie W: Comprehensive analysis of LAMC1 expression and prognostic value in kidney renal papillary cell carcinoma and clear cell carcinoma. Front Mol Biosci 2022, 9:988777.
Cui C, Chakraborty K, Tang XA, Zhou G, Schoenfelt KQ, Becker KM, Hoffman A, Chang YF, Blank A, Reardon CA et al: Neutrophil elastase selectively kills cancer cells and attenuates tumorigenesis. Cell 2021, 184(12):3163-3177 e3121.
Quail DF, Amulic B, Aziz M, Barnes BJ, Eruslanov E, Fridlender ZG, Goodridge HS, Granot Z, Hidalgo A, Huttenlocher A et al: Neutrophil phenotypes and functions in cancer: A consensus statement. J Exp Med 2022, 219(6).
Branzk N, Lubojemska A, Hardison SE, Wang Q, Gutierrez MG, Brown GD, Papayannopoulos V: Neutrophils sense microbe size and selectively release neutrophil extracellular traps in response to large pathogens. Nat Immunol 2014, 15(11):1017-1025.
Fuchs TA, Abed U, Goosmann C, Hurwitz R, Schulze I, Wahn V, Weinrauch Y, Brinkmann V, Zychlinsky A: Novel cell death program leads to neutrophil extracellular traps. J Cell Biol 2007, 176(2):231-241.
Rayes RF, Mouhanna JG, Nicolau I, Bourdeau F, Giannias B, Rousseau S, Quail D, Walsh L, Sangwan V, Bertos N et al: Primary tumors induce neutrophil extracellular traps with targetable metastasis promoting effects. JCI Insight 2019, 5(16).
Jin W, Xu HX, Zhang SR, Li H, Wang WQ, Gao HL, Wu CT, Xu JZ, Qi ZH, Li S et al: Tumor-Infiltrating NETs Predict Postsurgical Survival in Patients with Pancreatic Ductal Adenocarcinoma. Ann Surg Oncol 2019, 26(2):635-643.
Masucci MT, Minopoli M, Del Vecchio S, Carriero MV: The Emerging Role of Neutrophil Extracellular Traps (NETs) in Tumor Progression and Metastasis. Front Immunol 2020, 11:1749.
Wu L, Saxena S, Goel P, Prajapati DR, Wang C, Singh RK: Breast Cancer Cell-Neutrophil Interactions Enhance Neutrophil Survival and Pro-Tumorigenic Activities. Cancers (Basel) 2020, 12(10).
Park J, Wysocki RW, Amoozgar Z, Maiorino L, Fein MR, Jorns J, Schott AF, Kinugasa-Katayama Y, Lee Y, Won NH et al: Cancer cells induce metastasis-supporting neutrophil extracellular DNA traps. Sci Transl Med 2016, 8(361):361ra138.
Han AX, Long BY, Li CY, Huang DD, Xiong EQ, Li FJ, Wu GL, Liu Q, Yang GB, Hu HY: Machine learning framework develops neutrophil extracellular traps model for clinical outcome and immunotherapy response in lung adenocarcinoma. Apoptosis 2024.
Teijeira A, Garasa S, Ochoa MC, Villalba M, Olivera I, Cirella A, Eguren-Santamaria I, Berraondo P, Schalper KA, de Andrea CE et al: IL8, Neutrophils, and NETs in a Collusion against Cancer Immunity and Immunotherapy. Clin Cancer Res 2021, 27(9):2383-2393.
Kaltenmeier C, Yazdani HO, Morder K, Geller DA, Simmons RL, Tohme S: Neutrophil Extracellular Traps Promote T Cell Exhaustion in the Tumor Microenvironment. Front Immunol 2021, 12:785222.
Wang H, Zhang H, Wang Y, Brown ZJ, Xia Y, Huang Z, Shen C, Hu Z, Beane J, Ansa-Addo EA et al: Regulatory T-cell and neutrophil extracellular trap interaction contributes to carcinogenesis in non-alcoholic steatohepatitis. J Hepatol 2021, 75(6):1271-1283.
Papayannopoulos V, Metzler KD, Hakkim A, Zychlinsky A: Neutrophil elastase and myeloperoxidase regulate the formation of neutrophil extracellular traps. J Cell Biol 2010, 191(3):677-691.
Lewis HD, Liddle J, Coote JE, Atkinson SJ, Barker MD, Bax BD, Bicker KL, Bingham RP, Campbell M, Chen YH et al: Inhibition of PAD4 activity is sufficient to disrupt mouse and human NET formation. Nat Chem Biol 2015, 11(3):189-191.
Tang F, Tie Y, Tu C, Wei X: Surgical trauma-induced immunosuppression in cancer: Recent advances and the potential therapies. Clin Transl Med 2020, 10(1):199-223.
Sun G, Chen J, Liang J, Yin X, Zhang M, Yao J, He N, Armstrong CM, Zheng L, Zhang X et al: Integrated exome and RNA sequencing of TFE3-translocation renal cell carcinoma. Nat Commun 2021, 12(1):5262.
Edeline J, Mottier S, Vigneau C, Jouan F, Perrin C, Zerrouki S, Fergelot P, Patard JJ, Rioux-Leclercq N: Description of 2 angiogenic phenotypes in clear cell renal cell carcinoma. Hum Pathol 2012, 43(11):1982-1990.
Miao D, Margolis CA, Gao W, Voss MH, Li W, Martini DJ, Norton C, Bosse D, Wankowicz SM, Cullen D et al: Genomic correlates of response to immune checkpoint therapies in clear cell renal cell carcinoma. Science 2018, 359(6377):801-806.
Braun DA, Hou Y, Bakouny Z, Ficial M, Sant' Angelo M, Forman J, Ross-Macdonald P, Berger AC, Jegede OA, Elagina L et al: Interplay of somatic alterations and immune infiltration modulates response to PD-1 blockade in advanced clear cell renal cell carcinoma. Nat Med 2020, 26(6):909-918.
Wang J, Li Q, Yin Y, Zhang Y, Cao Y, Lin X, Huang L, Hoffmann D, Lu M, Qiu Y: Excessive Neutrophils and Neutrophil Extracellular Traps in COVID-19. Front Immunol 2020, 11:2063.
Zhang Y, Guo L, Dai Q, Shang B, Xiao T, Di X, Zhang K, Feng L, Shou J, Wang Y: A signature for pan-cancer prognosis based on neutrophil extracellular traps. J Immunother Cancer 2022, 10(6).
Xu L, Deng C, Pang B, Zhang X, Liu W, Liao G, Yuan H, Cheng P, Li F, Long Z et al: TIP: A Web Server for Resolving Tumor Immunophenotype Profiling. Cancer Res 2018, 78(23):6575-6580.
Avila Cobos F, Alquicira-Hernandez J, Powell JE, Mestdagh P, De Preter K: Benchmarking of cell type deconvolution pipelines for transcriptomics data. Nat Commun 2020, 11(1):5650.
Zheng S, Lin F, Zhang M, Fu J, Ge X, Mu N: AK001058 promotes the proliferation and migration of colorectal cancer cells by regulating methylation of ADAMTS12. Am J Transl Res 2019, 11(9):5869-5878.

No competing interests reported.

FigureS1.tif
Figure S1: In vivoexperiments of KCNN4 overexpression. A, Tumor images in null mice in NC group and KCNN4 overexpression group; B, Tumor weight in null mice in NC group and KCNN4 overexpression group; C, Tumor growth curves in null mice in NC group and KCNN4 overexpression group; D, HE staining of lung tissues in null mice in NC group and KCNN4 overexpression group; E, Immunohistochemical staining of EMT-related proteins in null mice in NC group and KCNN4 overexpression group.
TableS1.docx
TableS2.docx
TableS3.docx
TableS4.docx
gelsandblots.pdf

Download PDF

Version 1

posted

You are reading this latest preprint version

Constructing a Neutrophil Extracellular Trap Model Based on Machine Learning to Predict Clinical Outcomes and Immunotherapy Response in Renal Cell Carcinoma

Status:

Version 1

Abstract

Figures

Introduction

Methods

Data downloading and processing

Non-negative matrix factorization clustering based on NETs related genes

Enrichment analysis and single sample gene set enrichment analysis

Construction of WGCNA network

Building prognostic NETs signatures using machine learning frameworks

Construction of nomogram model

NETs signature annotation of tumor microenvironment, signaling pathways, and immune-related features

Cell lines and culture conditions

Western blotting assay and cell counting kit-8 (CCK-8) assay

Wound healing assay

Animal experiments

Immunohistochemistry

Statistical analysis

Results

Development and validation of NETs related gene subtypes

WGCNA network identifies subtype-related genes

Determination and verification of prognostic NETs signatures

Features of NETs signatures in tumor microenvironment

NETs related immune characteristics

Knockdown of KCNN4 inhibits mRNA and protein expression levels of KCNN4

Knockdown of KCNN4 inhibits tumor cell growth and EMT ability

Discussion

Conclusions

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1