Identification and Validation of Biomarkers in Chronic Rhinosinusitis With Nasal Polyps and Study of Their Associated Mechanisms

ABSTRACT

Background

Cell senescence (CS) and lipid metabolism (LM) disorders have been reported in chronic rhinosinusitis with nasal polyps (CRSwNP). However, the mechanism is still unclear.

Methods

Data were obtained from public databases; differential expression analysis and machine learning were performed to identify biomarkers and to understand the pathways in which it involved. Subsequently, immune infiltration analysis was performed to observe alterations in the CRSwNP immune microenvironment. Additionally, founded on single cell data, clustering and annotation were performed to acquire key cells for pseudo‐time analysis, and to understand the interactions between different cells. Finally, reverse transcription quantitative PCR (RT‐qPCR) and immunofluorescence (IF) staining were performed to further validate the results.

Results

The biomarkers VDR, CPT1C, and PLA2G2A were distinguished. Notably, the VDR was primarily associated with the neutrophil extracellular trap formation pathway, suggesting a correlation with CRSwNP pathology. Next, the immune microenvironment was altered in CRSwNP patients; in total, seven cell types were annotated for the band, with monocytes and epithelial cells being the key cells. Interestingly, with the differentiation of monocytes and epithelial cells, the expression level of VDR gradually increased. Moreover, epithelial cells had the most frequent and strongest interactions with other cells, followed by monocytes. Finally, the observed expression trends of the VDR and PLA2G2A were consistent with the results of the bioinformatics analyses; however, the opposite was true for the CPT1C mRNA.

Conclusion

The three CRSwNP‐related biomarkers (VDR, CPT1C, PLA2G2A) associated with CS and LM were acquired, supplying novel targets for clinical diagnosis and treatment of CRSwNP.

1. Introduction

Chronic rhinosinusitis (CRS) is a common chronic inflammatory disease, affecting 2%–8% of the Chinese population [1, 2]. Clinically, CRS can be divided into two types based on the presence or absence of nasal polyps (NPs): CRS without nasal polyps (CRSsNP) and CRS with nasal polyps (CRSwNP) [3, 4]. CRSwNP is a severe form of CRS, accounting for about 20% of the overall CRS population [5], and typically has a more severe disease severity and poorer treatment response compared to CRSsNP [6]. Moreover, CRSwNP is becoming increasingly prevalent and is particularly difficult to treat with conventional therapies, making its study more important. The long‐lasting symptoms of CRSwNP include nasal congestion, rhinorrhea, facial pain or tenderness, and reduced sense of smell [3], which severely impair the quality of life and productivity of patients [7] and cause a significant economic burden [8]. These symptoms are closely related to the destruction of the nasal epithelial barrier, increased exposure to pathogenic microbiota, immune system dysfunction, and tissue remodeling. Additionally, environmental factors, genetic factors, lipid metabolism (LM) [9], and cellular senescence (CS) [10] are also believed to be involved in the development of CRSwNP, but the specific pathogenesis is still largely unclear.

Cellular senescence (CS) is a process that is mainly designed to eliminate unwanted cells by inducing tissue remodeling [11]. CS leads to a decline in the ability to control persistent inflammation, significantly reducing the body's capacity to resolve it [12]. CRSwNP is a multifactorial disease infiltration of inflammatory cells in the lamina propria, predominantly eosinophils [13]. In the early and chronic phases of inflammation, excessive amounts of lipid mediators can be released and affect LM [14]. It has been found that leukotrienes (LTs) [15, 16] and prostaglandins (PGs) [9, 17] are related to the nasal polyps (NPs). One study revealed that prostaglandin E2 (PGE2) concentrations were lower in NP tissues, and higher production of LT metabolites, such as LTC4, LTD4, and LTE4, was correlated with lower PGE2 concentrations [13]. Supporting this finding, the mRNA level of PGES was lower in NPs; moreover, the low PGES levels [18] and enhanced LTD4 production [19] were correlated with increased eosinophil infiltration, as well as disease severity, suggesting that PGs or LTD4 might have a role in eosinophilic inflammation in NPs.

Recently, it was found that LM disorders also regulate the CS process, and found that the changes in LM in aging cells are mainly concentrated in the metabolism of phospholipids, fatty acids, and cholesterol [10, 20]. Lipids are key participants in cellular aging; there is a complex interplay between LM and CS [21]. On the one hand, alterations in LM are an important characteristic of cellular aging. On the other hand, abnormalities in LM, such as abnormally elevated levels of lipids in the plasma or functional impairments, may lead to the occurrence and progression of cellular aging [22]. Nevertheless, little is known about their role in modulating the pathogenesis of NPs.

Our study focused on the identification and validation of biomarkers related to CS and LM in CRSwNP by bioinformatics methods based on transcriptome and single‐cell data from a public database, with the aim of providing new insights into the study of pathological mechanisms related to CRSwNP and the development of clinical drugs.

2. Materials and Methods

The datasets were all obtained from the Gene Expression Omnibus (GEO) database. Following differential expression analysis, enrichment analysis, and the identification of candidate genes, the study proceeded through functional enrichment of these candidates, construction of a protein‐protein interaction (PPI) network, and the application of machine learning to obtain candidate biomarkers, culminating in the identification of biomarkers and construction of a nomogram in the CRSwNP and control groups. Furthermore, biomarkers were respectively subjected to Gene Set Enrichment Analysis (GSEA). Also, friend analysis and immune infiltration analysis were performed. Besides, for the construction of molecular regulatory networks, the MicroRNA Data Integration Portal and miRNet were applied to anticipate the microRNAs (miRNAs) of biomarkers, respectively. In addition, we explored correlation analysis of biomarkers with inflammatory response, and drug forecasting was applied. The GSVA package [23] (v 1.46.2) was adopted to calculate single‐sample GSEA (ssGSEA) scores for labeled IRGs in whole‐tissue clinical samples based on the Genotype‐Tissue Expression (GTEx) database. In addition, single‐cell analysis and the pseudo‐time and cellular communication analyses were performed. Last but not least, to confirm the consistency of biomarker expression with bioinformatics results, the biomarker expression was verified by reverse transcription quantitative PCR (RT‐qPCR) and immunofluorescence (IF) staining. NPs biopsies were collected during functional endoscopic sinus surgery (FESS) procedures, while inferior turbinate (IT) mucosa specimens from control subjects were obtained during septoplasty. CRSwNP was classified as eosinophilic if the average histologic eosinophil count exceeded 10 per high‐power field (HPF) upon hematoxylin‐eosin (HE) staining. The mean number of eosinophils per HPF for the samples were 4, 7, 15, 27, and 50, respectively. Detailed information can be found in the Supporting Information: Appendices.

2.1. Statistical Analysis

The R software (v 4.2.3) was applied to perform statistical analysis, with statistical significance defined as a p value < 0.05.

3. Results

3.1. Identification and Enrichment Analysis of 14 Candidate Genes

After threshold screening, 5236 DEGs were obtained between the CRSwNP and control groups, of which 2946 were upregulated DEGs and 2290 were downregulated DEGs in the CRSwNP group (Figure 1A). Expression of the top 20 upregulated and downregulated DEGs displayed different expression patterns (Figure 1B). Subsequently, the 1551 functional pathways were markedly enriched in GO by DEGs (Table S6). Among them, BP, CC, and MF were markedly enriched to 1312, 112, and 127 entries, respectively. Specifically, there was a marked enrichment mainly to the entries of leukocyte‐mediated immunity (BP), immunoglobulin complex (CC), and immune receptor activity (MF), suggesting a correlation between DEGs and immune‐related functions (Figure 1C). Moreover, the 74 functional pathways were markedly enriched in KEGG, the DEGs were markedly pertinent to the pathways of chemokine signaling pathway and cytokine‐cytokine receptor interaction (Figure 1D and Table S7). Finally, a total of 14 candidate genes were retained by taking intersections (Figure 1E ). Notably, the candidate genes were mainly involved in the pathways of metabolism of lipids, regulation of lipid transport, and fatty acid metabolic process, revealing the relevance of the candidate genes to LM (Figure 1F). Furthermore, all 14 candidate genes encoded proteins that interacted with each other; there were no discrete proteins, with AGT, AKR1B1, and CAV1 interacting with a large number of proteins (Figure 1G).

Figure 1.

Identification and enrichment analysis of DEGs. (A) Volcanic map of differential expression genes, of which 2946 were upregulated DEGs (purple) and 2290 were downregulated DEGs (blue) in the CRSwNP group. (B) The top 20 upregulated and downregulated DEGs are shown as a heat map. (C) GO‐enriched tree diagram of DEGs. (D) KEGG pathway enrichment map of DEGs. (E) Venn diagram of candidate genes. (F) Network map of candidate gene enrichment biological function. (G) PPI network of candidate genes.

3.2. VDR, CPT1C, and PLA2G2A Are Recognized as Biomarkers

Subsequently, nine LASSO‐feature genes were acquired after analysis (lambda.min = −3.964) (Figure 2A,B), which were PLA2G2A, PON1, CAV1, PRKD1, ACER2, INPP4B, VDR, CPT1C, and PIK3R5. Meanwhile, in the SVM‐RFE algorithm, the model prediction accuracy was highest when the forecasted genes were ACER2, VDR, PLA2G2A, CPT1C, CAV1, AKR1B1, INPP4B, and MIF (Figure 2C). Then, six candidate biomarkers were acquired by intersection, namely ACER2, VDR, PLA2G2A, CPT1C, CAV1, and INPP4B (Figure 2D). Next, three biomarkers were acquired in the CRSwNP and control groups; their expression trends were the same in GSE136825 and GSE179265, specifically, the VDR and CPT1C were markedly upregulated in the CRSwNP group, while the PLA2G2A was markedly downregulated (p < 0.05) (Figure 2E,F).

Figure 2.

Types and biological characteristics of key characteristic genes screened by machine learning. (A, B) LASSO model screening for signature genes. (C) Support vector machine model screening for feature genes. (D) Venn diagram for screening key characteristic genes by machine learning. (E, F) Box diagram of key characteristic gene expression in two datasets.

3.3. Construction and Evaluation of Nomogram

Founded on CPT1C, PLA2G2A, and VDR, the forecasted incidence of CRSwNP was displayed in the nomogram (Figure 3A). Subsequently, the calibration curves displayed that the nomogram almost overlapped with the reference line, and the area under the curve (AUC) was 0.884 (Figure 3B,C). This demonstrated the favorable accuracy of the nomogram in anticipating the incidence of CRSwNP. Besides, the DCA curve displayed that the net benefit of the nomogram was superior to 0 and higher than the net benefit of all and none (Figure 3D). Overall, biomarker‐based construction of the nomogram had a favorable prediction of CRSwNP incidence, which supplied an effective decision‐making tool for the clinical aspects of CRSwNP.

Figure 3.

Construction and evaluation of a nomogram. (A) Nomogram for predicting morbidity. (B) Calibration curve for nomogram. (C) ROC curve for nomogram. (D) Decision curve for nomogram.

3.4. Friend Analysis and GSEA of CPT1C, PLA2G2A, and VDR

The results of friend analysis displayed a strong functional correlation between CPT1C and PLA2G2A (cor > 0.3), with PLA2G2A being the most important of the three biomarkers and most similar to the other two biomarkers, followed by CPT1C (Figure 4A).

Figure 4.

Friend analysis and GSEA of CPT1C, PLA2G2A, and VDR. (A) The results of friend analysis. (B, C) The CPT1C was markedly enriched to 2230 GO and 188 KEGG entries. (D, E) The PLA2G2A was markedly enriched for 535 GO and 107 KEGG entries. (F, G) The VDR was markedly enriched to 1265 GO and 175 KEGG entries.

Additionally, GSEA results showed that the CPT1C was markedly enriched to 2230 GO (Figure 4B and Table S8) and 188 KEGG entries (Figure 4C and Table S9), such as mitochondrial protein‐containing complex and mitochondrial translation (GO), as well as oxidative phosphorylation and ribosome pathway (KEGG). The PLA2G2A was markedly enriched for 535 GO (Figure 4D and Table S10) and 107 KEGG entries (Figure 4E and Table S11), such as microtubule bundle formation and cilium movement (GO), as well as protein processing in the endoplasmic reticulum and valine, leucine, and isoleucine degradation pathway (KEGG). The VDR was markedly enriched to 1265 GO (Figure 4F and Table S12) and 175 KEGG entries (Figure 4G and Table S13), such as keratinocyte differentiation and microtubule cytoskeleton organization involved in mitosis (GO), as well as cell cycle and neutrophil extracellular trap formation, among other pathways (KEGG). In summary, by understanding the patterns in which biomarkers interact and the pathways involved it helped to understand how biomarkers function, which in turn supplied ideas for targeting these pathways to treat CRSwNP.

3.5. Altered Immune Microenvironment in Patients With CRSwNP

Infiltration of immune cells in CRSwNP and control groups is displayed in Figure 5A; the eosinophils, mesenchymal stem cells (MSC), and epithelial cells were more predominant in the CRSwNP group, and the top 3 in the control group were MSC, multipotent progenitors (MPP), and hematopoietic stem cells (HSC). Subsequently, 36 immune cells were markedly different between the two groups. Among them, in the CRSwNP group, the infiltration level of the 24 differential immune cells were markedly higher than the control group, namely aDC, B‐cells, memory CD4⁺ T‐cells, naive CD4⁺ T‐cells, CD4⁺ effector memory T cells (Tem), CD8⁺ T‐cells, central memory CD8⁺ T cells (Tcm), CD8⁺ Tem, class‐switched memory B‐cells, dendritic cells (DC), eosinophils, epithelial cells, fibroblasts, immature dendritic cells (iDC), keratinocytes, macrophages, M1 macrophages, mast cells, mesangial cells, monocytes, pericytes, T gamma delta (Tgd) cells, T‐helper 1 (Th1) cells, and T‐helper 2 (Th2) cells (Figure 5B). Additionally, CPT1C showed the strongest positive correlation with fibroblasts (cor = 0.52, p < 0.05) and the strongest negative correlation with naive CD8⁺ T‐cells (cor = −0.51, p < 0.05). The PLA2G2A showed strong positive correlations with melanocytes, adipocytes, and neurons (cor > 0.6, p < 0.05), while it had the strongest negative correlation with eosinophils (cor = −0.52, p < 0.05). Moreover, the VDR had the strongest positive correlation with Th1 cells (cor = 0.5, p < 0.05), whereas it presented a strong negative correlation with neurons and adipocytes (cor < −0.7, p < 0.05) (Figure 5C). This suggested that the biomarkers might influence the onset and development of CRSwNP by altering the immune microenvironment.

Figure 5.

Altered immune microenvironment in patients between healthy subjects and CRSwNP. (A) Expression thermogram of different immune cells in normal and disease samples. (B) Box diagram of immune cell proportion difference, indicating that 36 immune cells were markedly different between the two groups. (C) The relationship between the differential immune cells with CPT1C, PLA2G2A, and VDR, respectively.

3.6. Molecular Regulatory Networks of CPT1C, PLA2G2A, and VDR, Correlation of Inflammatory Responses, and Drug Prediction

The CPT1C, PLA2G2A, and VDR were forecasted 3, 2, and 31 miRNAs (Figure 6A), respectively. The miRNA–mRNA regulatory network composed 34 nodes and 36 interaction pairs; among them, the CPT1C and VDR together forecasted hsa‐miR‐3195, hsa‐miR‐598‐5p, and hsa‐miR‐1343‐3p. The PLA2G2A and VDR together forecasted hsa‐miR‐6736‐5p and hsa‐miR‐6745 (Figure 6B). Meanwhile, the CPT1C, PLA2G2A, and VDR were forecasted 3, 15, and 6 TFs, respectively, of which GATA2 was jointly forecasted by PLA2G2A and VDR. At last, the TF‐mRNA regulatory network displayed a total of 26 nodes and 24 interaction pairs (Figure 6C). In a word, finding out the molecular regulation mechanisms upstream of biomarkers suggested an idea for targeting and regulating biomarkers.

Figure 6.

Molecular regulatory networks of CPT1C, PLA2G2A, and VDR. (A) Venn diagram of the key miRNA. (B) The regulatory network of key miRNA and biomarkers (blue nodes for key genes; purple for miRNA). (C) TF‐mRNAs interaction network (blue nodes for key genes; pink for TF). (D–F) Correlation analysis between biomarkers and inflammatory response. (G) Targeted drug network predicted by CTD database.

Moreover, it was further observed that there was mostly a marked positive pertinence between the biomarkers and the inflammatory response (cor > 0.3, p < 0.05) (Table S15). Specifically, the CPT1C was positively correlated with the blood vessel, pancreas, and salivary gland, while having the strongest negative correlation with the brain and skin (Figure 6D), and the VDR was positively correlated with the salivary gland, liver, and ovary (Figure 6E). Notably, the PLA2G2A was positively correlated with 31 inflammatory responses in all tissue samples (Figure 6F). Inflammatory response is a protective response of the body to stimuli involving immune cells, blood vessels, and a variety of molecular mediators. By understanding its correlation with biomarkers, it is clinically important for the diagnosis and monitoring of inflammatory diseases (CRSwNP). Finally, a total of 84 targeted drugs were obtained from the CPT1C, PLA2G2A, and VDR. The network of biomarkers and drugs contains 87 nodes totaling 100 interaction pairs. Specifically, including 9 drugs targeting CPT1C, 24 drugs targeting PLA2G2A, and 67 drugs targeting VDR. There were 4 drugs targeting 3 biomarkers simultaneously, namely benzo(a)pyrene, bisphenol A, trichostatin, and tetrachlorodibenzodioxin (Figure 6G). This targeting of biomarkers for CRSwNP treatment provided novel ideas.

3.7. Annotated to Seven Cell Types

In the GSE276503, after quality control of the scRNA‐seq data, 104,614 cells and 37,099 genes were obtained (Figure S1A,B). A total of 2000 highly variable genes were identified (Figure 7A). The results of the principal component analysis (PCA) showed that 17 cell samples were concentratedly distributed (Figure S2). Meanwhile, among 50 PCs, the top 30 PCs were selected for analysis (Figure 7B,C). Through UMAP clustering analysis, 22 different cell clusters were identified in total (with a resolution of 0.3) (Figure 7D). According to the expression of marker genes, cell annotation was carried out for different cell clusters, and seven types of cells were obtained, such as epithelial cells and basal cells (Figure 7E,F).

Figure 7.

Single‐cell data analysis annotated to seven cell types. (A) Visualized scatter plot of hypervariable genes. Red dots are hypervariable genes, and blue dots are non‐hypervariable genes. (B, C) Principal component analysis results of a single‐cell data set. (D) UMAP diagram of 22 different cell clusters. (E) The expression of marker genes in 22 different cell clusters. (F) Annotation of 7 cell types.

3.8. Monocytes and Epithelial Cells Were Regarded as Key Cells

VDR, PLA2G2A, and CPT1C were highly expressed in epithelial cells and monocytes (Figure 8A and Figure S3). The expression levels of VDR and PLA2G2A were significantly different in both epithelial cells and monocytes (Figure 8B), while the expression level of CPT1C was significantly different in monocytes (Figure 8D). Therefore, monocytes and epithelial cells were regarded as key cells.

Figure 8.

Monocytes and epithelial cells were regarded as key cells. (A) VDR, PLA2G2A, and CPT1C were highly expressed in epithelial cells and monocytes. (B, C) The high expression level of VDR and PLA2G2A in epithelial cells and monocytes. (D) The high expression level of CPT1C in monocytes.

3.9. Results of Pseudo‐Time and Cellular Communication Analyses

Monocytes were clustered into 15 cell clusters (resolution = 0.4). VDR, PLA2G2A, and CPT1C were expressed in different cell clusters, among which VDR had the highest expression in cluster 8 (Figure S4). Pseudo‐time analysis of monocytes showed that the darker the color, the more the initial developmental state was represented, and the lighter the color, the more mature the development was. Cell clusters 2, 3, 4, and 5 were in the early stage of cell differentiation, while cell clusters 6, 7, 8, and 9 had the latest differentiation (Figure 9A,B). With the differentiation of monocytes, the expression level of VDR gradually increased, and the expression levels of PLA2G2A and CPT1C gradually decreased (Figure 9C). Endothelial cells were clustered into 15 subgroups (resolution = 0.4). VDR, PLA2G2A, and CPT1C were expressed in these subgroups, among which VDR had the highest expression in cell clusters 0 and 8 (Figure S5). Pseudo‐time analysis of endothelial cells showed that cell clusters 5, 6, and 7 were in the early stage of cell differentiation, while cell clusters 3, 11, 12, 13, and 14 had the latest differentiation (Figure 9D,E). With the differentiation of endothelial cells, the expression level of VDR gradually increased, and the expression levels of PLA2G2A and CPT1C genes first gradually increased and then gradually decreased (Figure 9F). Cell communication analysis indicated that in CRSwNP samples, epithelial cells had the most frequent and strongest interactions with other cells, followed by monocytes (Figure 9G,H). Endothelial cells had strong interactions with NK cells, and the ligand‐receptor pairs involved in these interactions were MDK−NCL, MIF−(CD74 + CXCR4), and so forth (Figure 9I and Figure S6). Monocytes also had strong interactions with basal cells, with ligand−receptor pairs, such as MIF−(CD74 + CD44), MDK−NCL, and so forth. Additionally, monocytes had strong interactions with epithelial cells, and the ligand−receptor pairs were MIF−(CD74 + CD44), MDK−NCL, and so forth (Figure 9J and Figure S6).

Figure 9.

Results of pseudo‐time and cellular communication analyses. (A, B) The pseudo‐time analyses of monocytes. (C) The expression levels of the top 20 highly variable genes, VDR, PLA2G2A, and CPT1C in endothelial cells. (D, E) The pseudo‐time analyses of endothelial cells in single‐cell datasets. (F) The expression levels of the top 20 highly variable genes and VDR in endothelial cells. (G, H) Cellular communication analyses showed that epithelial cells had the most frequent and strongest interactions with other cells, followed by monocytes in the CRSwNP group. (I) The reciprocal ligand and receptor pairs in endothelial cells. (J) The reciprocal ligand and receptor pairs in monocytes.

3.10. Validation of CPT1C, VDR, and PLA2G2A Expression

The results showed that both mRNA and protein levels of VDR were significantly upregulated expression in the CRSwNP group, which were consistent with the results of bioinformatic analysis (Figure 10A,D,G). Meanwhile, in the CRSwNP group, PLA2G2A mRNA and protein expression were downregulated, though this change did not reach statistical significance (Figure 10B,E,H). Notably, CPT1C mRNA level was markedly upregulated in control groups (Figure 10C,E), which was different from the results of bioinformatic analysis. However, IF analysis revealed significantly higher CPT1C protein levels in CRSwNP compared to controls (Figure 10F, I). The inconsistency in mRNA and protein expression of CPT1C needed to be verified further.

Figure 10.

Validation of CPT1C, VDR, and PLA2G2A expression. (A) Compared with the control group, the VDR mRNA was significantly (*) upregulated expression in the CRSwNP group. (B) The PLA2G2A mRNA expression was downregulated in the CRSwNP group without a significant difference (ns), yet. (C) The mRNA level of CPT1C was significantly (**) downregulated in the CRSwNP group. (D–F) Representative lF staining images of CRSwNP and the control group, immunostained with VDR, PLA2G2A, and CPT1C (green) and counterstained with DAPI (blue, ×400 magnification). (G–I) Comparisons of protein levels of VDR, PLA2G2A, and CPT1C between CRSwNP patients (n = 4) and the control group (n = 4, 10 fields of each slide were evaluated at ×400 magnification).

4. Discussion

It is well known that even after standardized drug and surgical treatments, a portion of NPs tend to recur. As reported, the recurrence rates, defined as the number of patients undergoing revision endoscopic sinus surgery, are 20% within a 5‐year period after surgery [24]. Thus, there is an urgent need for effective drugs to overcome this challenge. Recent studies have found that CS and LM are involved in the pathogenesis of Chronic Rhinosinusitis with Nasal Polyps (CRSwNP), but the specific regulatory mechanisms are not yet clear. In this study, we obtained transcriptomic data of CRSwNP and genes related to lipid metabolism (LMRGs) and cellular senescence (CSRGs) from databases. Through multiple methods of analysis and validation, we identified three biomarkers: Vitamin D Receptor (VDR), Carnitine Palmitoyltransferase 1C (CPT1C), and Phospholipase A2 Group IIA (PLA2G2A). Based on single‐cell data of CRSwNP, we found that monocytes and epithelial cells are the key cell types primarily associated with these three biomarkers. Building on this foundation, we successfully constructed a reliable nomogram model for predicting the occurrence of CRSwNP, which provides a potentially effective tool for the early diagnosis of CRSwNP.

The VDR is an intracellular macromolecule [25]. VDR forms a hormone‐receptor complex by binding to 1alpha,25‐dihydroxyvitamin D (1,25(OH)2D), acting on specific DNA sequences in target genes, thereby regulating the expression of structural genes [25]. 1,25VD3 has antibacterial and anti‐inflammatory actions that have contributed to CRSwNP [26]. VD3 is a precursor form of vitamin D that needs to be metabolically activated to form 1,25VD3 [27]. VD3 modulates the immune system by activating innate antibacterial responses, while generally has a tolerogenic effect with respect to adaptive immunity [26]. As reported, patients with CRSwNP have been proven to be VD3 deficient [28, 29] because the absence of 1‐α‐hydroxylase expressed by sinonasal epithelial cells likely leads to reduced antibacterial responses, increases elaboration of inflammatory cytokines, and increases fibroblast proliferation in patients with CRSwNP [26]. It was indicated that low VD3 levels can result in CRS in NPs [30]. Consistently, the reduced expression of VDR leads to impaired activation of the VD3 signal in nasal epithelial cells, thereby increasing the production of IL‐36y and neutrophil inflammation in CRSwNP [31]. VD impedes eosinophil chemotaxis by inhibiting glycolysis‐induced CCL26 expression via attenuating the activation of the ERK pathway and reducing lactate production [32]. Interestingly, the results of our study were contrary to those reported in the literature; the GSEA results found that VDRs are significantly increased in CRSwNP, which is strongly related to neutrophil extracellular trap formation (NETs). Indeed, NETs are significantly increased in NPs, and IL‐37 induces NET formation in CRSwNP patients [33]. Besides, NETs contribute to disease pathogenesis, exacerbating inflammation and host tissue damage [34] and may contribute to the refractoriness in refractory non‐Type 2 NPs (non‐T2‐NPs), mainly in Asian countries [35]. The PCR results and IF staining also showed that VDR was significantly upregulated expression in the CRSwNP group, which was consistent with the results of bioinformatic analysis. Fortunately, there was a study that can support our results that epithelial cells from patients with CRSwNP demonstrated an intrinsic reduction in conversion of 25VD3 to 1,25VD3 in vitro, but total sinonasal VDR expression by PCR is not statistically different between control and CRSwNP [28]. The reason why the result is different may be due to the complex VDR regulation, which is determined by genetic and environmental factors [36], including diet, exposure to sunlight, infections, and pollution [25]. It has been postulated that these environmental factors could modify vitamin D levels, which regulate the receptor. The mechanism is not clearly understood. Therefore, different results are worth considering and need further study.

CPT1C is mainly expressed in neurons [37] and can bind to carnitine to sense the nutritional status of cells [38], primarily involved in the regulation of CS [39], cell migration, and endoplasmic reticulum–mitochondrial contact sites [40], a function similar to our GSEA results. CPT1C was significantly enriched in mitochondria‐associated processes, including mitochondrial protein complexes and mitochondrial translation, as well as critical metabolic pathways such as oxidative phosphorylation and ribosome pathways. These findings suggest that CPT1C may contribute to the pathogenesis of CRSwNP by modulating mitochondrial activity and energy metabolism [41]. CPT1C is mainly associated with fatty acid oxidation (FAO), and CPT1C has been identified as a new biomarker and a key regulator of cancer CS [42]. According to current research, the high expression of CPT1C may promote the activation and proliferation of inflammatory cells [43]. In addition, inhibition of FAO with etomoxir caused lipid accumulation and suppressed human bladder cancer (BCa) cell growth by affecting epithelial–mesenchymal transition (EMT)‐related proteins [44]. As we all know, EMT plays a vital role in the pathological mechanism of CRSwNP [45]. The expression of CPT1C in CRS has not been reported yet, and our study is the first to find that CPT1C expression is upregulated in CRSwNP. Although the specific mechanism is unclear, we speculate that CPT1C may also affect the pathological process of EMT through FAO, or as explored in breast and pancreatic cancer cells [46], CPT1C may interact with microRNA (miRNA), regulating the expression of specific genes, and then affect the inflammatory response and immune response, and ultimately participate in the pathogenesis and development of CRSwNP diseases. The above studies revealed that CPT1C is closely related to cell proliferation and inflammatory response.

However, the CPT1C mRNA was observed to be significantly downregulated in the CRSwNP group, which was different from the results of bioinformatic analysis and the above research, probably due to the small sample size and heterogeneity of NPs types, and needed to be verified further. Actually, CPT1 inhibitor, Etomoxir, has been developed and has been demonstrated to suppress cell proliferation in vitro. Nevertheless, there have been no studies that have analyzed the roles that CPT1C might play in NPs, and the functions of CPT1C. Based on this, it was suggested that inhibition of CPT1C may represent a new therapeutic strategy for NPs through induction of cell senescence [47], which is worth further study.

Secretory PLA2G2A is a member of a family of secretory phospholipases [48]. PLA2G2A catalyzes the hydrolysis of phospholipids to produce free fatty acids and lysophospholipids, which can be further converted into PGs, LTs, and other inflammatory mediators, promoting the occurrence and development of inflammatory response [49]. So far, secreted group IIA‑phospholipase A2 (sPLA2‑IIA) from the PLA2G2A gene has been proposed to play a role in antibacterial defense, inflammation, and eicosanoid generation, in clearance of apoptotic cells, and in the Wnt signaling pathway [50]. Recent studies have uncovered that, as a modulator of the gut microbiota in the intestine, sPLA2‐IIA could bind to gram‑positive bacteria and degrade bacterial membranes, controlling microbiota in the intestine, thereby secondarily affecting inflammation, allergy, and cancer in proximal and distal tissues [51]. sPLA2‑IIA interaction of human neutrophils with the mitochondria/sPLA2‐IIA complex promotes the release of pro‐inflammatory mediators.

The long‐term persistent symptoms of CRSwNP include nasal congestion, rhinorrhea, facial pain or tenderness, and reduced sense of smell, which severely impair the quality of life and productivity of patients and create a significant economic burden. These symptoms are closely related to the dysfunction of the mucociliary clearance (MCC) system. The MCC is mainly composed of two functional parts: active cilia and secreted mucus. Motile cilia beat rhythmically on the surface of cells to power the flow of fluid. In CRSwNP, frequent infections or inflammations can lead to the loss and dysfunction of cilia, impairing MCC function. Even cilia that are not directly damaged may not function properly due to ongoing microbial and/or inflammatory damage, thereby causing clinical symptoms. Thus, we think that as long as the cilia function is improved, the symptoms of sinusitis can be improved. As reported, ciliary movement is generated by the axoneme, a molecular machine consisting of microtubules, ATP‐powered dynein motors, and regulatory complexes [52]. Based on our analysis, we found that PLA2G2A is primarily enriched in axoneme, ciliary plasm, axoneme assembly, microtubule bundle formation, cilium movement, and 536 other GO pathways. Thus, PLA2G2A might play a vital role in the ciliary movement, and targeting the PLA2G2A for mechanism research and corresponding drug development holds some promise.

The immune infiltration analysis revealed that the top three immune cells are eosinophils, MSCs, and epithelial cells. Additionally, fibroblasts and eosinophils show higher infiltration levels in the CRSwNP group. Some research has found that in Western countries, CRSwNP exhibits high eosinophil infiltration. The presence of mucosal eosinophilia is correlated with disease severity [53, 54]. Fibroblasts have been identified as key cells that trigger tissue‐specific inflammatory responses, and dysregulated inflammatory signals in tissue‐resident fibroblasts may support type‐2 inflammatory responses [55]. Similarly, they found that inhibiting fibroblast proliferation could prevent CRSwNP recurrence [24]. In this study, we performed endotyping analysis of CRSwNP based on eosinophil infiltration levels. Using a diagnostic threshold of ≥ 10 eosinophils per high‐power field (HPF, eos/HPF) on average [3, 56], three ECRSwNP were identified and included in the analysis. This finding aligns with the immune infiltration profiling results demonstrating eosinophils as the predominant immune cell population, consistent with previous reports characterizing eosinophil‐dominated inflammation in NP pathogenesis. Although substantial heterogeneity in cut‐off values was observed across studies, a systematic review and meta‐analysis confirmed that a threshold of > 55 eos/HPF reliably predicts disease recurrence in ECRSwNP [57], which need for at least two courses of systemic corticosteroids or continuous use of systemic corticosteroids ≥ 2 courses per year OR > 3 months low‐dose steroids [3], significantly impaired quality of life, anosmic on smell test and/or a diagnosis of comorbid asthma needing regular inhaled corticosteroids. High tissue eosinophilia is typically associated with high levels of cytokines IL‐4, IL‐5, and IL‐13, which are also known as Type 2 cytokines [58]. Above all, in this study, transcriptome data analysis was conducted. Building on established connections, the study made breakthrough findings. Through high‐resolution xCell algorithm analysis [59], it systematically quantified infiltration changes of 64 immune cells in CRSwNP for the first time. The high expression of eosinophils, a key finding, offers new evidence for their role in CRSwNP's pathologic process.

Benzo[a]pyrene is a potent carcinogen that exerts toxic effects primarily through activation of the aryl hydrocarbon receptor (AhR) pathway [60]. Studies demonstrate its ability to inhibit weight gain and reduce lipid content in mice [61]. In respiratory inflammation, benzo[a]pyrene activates phospholipase A2 (PLA2) via a Ca²⁺‐dependent mechanism, increasing PLA2G2A mRNA expression and mediating PM2.5‐induced pulmonary inflammation and injury [62]. However, no direct evidence currently establishes its regulatory effects on CPT1C and VDR genes in respiratory inflammatory contexts.

Bisphenol A (BPA), an endocrine‐disrupting chemical [63], has been shown in in vitro studies to reduce lipolysis rates in adipocytes, leading to significant intracellular triglyceride accumulation [63]. Additional research indicates that BPA may influence the expression of CS‐associated proteins, potentially accelerating cellular aging [64]. Notably, no direct studies have confirmed BPA's impact on CPT1C, PLA2G2A, or VDR expression in respiratory inflammatory diseases.

Trichostatin A (TSA), a histone deacetylase (HDAC) inhibitor, modulates cholesterol reverse transport through regulation of related genes [65]. For instance, HDAC inhibitors such as TSA may induce SR‐BI expression by acting on the PPRE element in its promoter region, thereby influencing LM and promoting clearance of excess cholesterol—particularly in arterial fatty streak lesions [66, 67]. Mechanistically, TSA primarily affects cellular functions by altering chromatin structure and gene expression. It is generally accepted that through HDAC inhibition, TSA induces chromatin relaxation, potentially upregulating tumor suppressor genes and cell cycle regulators, which may subsequently inhibit CS [68].

In the CRSwNP group, the single‐cell analysis confirmed that the key cells are monocytes and epithelial cells. Monocytes, as essential components of the immune system, play a pivotal role in the pathogenesis of CRSwNP. Studies demonstrate that under inflammatory conditions, monocytes are recruited to sinonasal mucosal tissues via chemokine signaling [69]. Furthermore, monocytes can differentiate into dendritic cells, activate T cells, and initiate adaptive immune responses, thereby driving CRSwNP progression [70]. These findings highlight monocytes as central mediators bridging innate immune recruitment and adaptive immune activation in CRSwNP, offering potential therapeutic targets to modulate disease‐specific immune cascades. The increase in glucose in nasal secretions promotes glucose uptake and primary glycolysis in epithelial cells, enhancing the pro‐inflammatory function of epithelial cells in CRSwNP [71]. These findings reinforce our findings that epithelial cells and monocytes are key actors in persistent Type 2 inflammation in NPs.

Despite this, there are still some limitations in this study. First, the results of RT‐qPCR and IF staining are not entirely consistent with the bioinformatics analysis results, which may be related to the small sample size and the heterogeneity of the samples. Second, the mechanisms of action of these biomarkers in the disease still need to be confirmed through more experiments, particularly eosinophil‐associated signaling pathways warrant in‐depth investigation. Future studies will employ multidimensional validation and multi‐omics approaches to advance mechanistic understanding. CRISPR‐based gene editing will be utilized to construct CRSwNP cellular models, enabling systematic interrogation of VDR, CPT1C, and PLA2G2A in regulating CS, LM, and inflammatory responses. In this study, we identified biomarkers (VDR, CPT1C, and PLA2G2A) co‐associated with cell aging and LM in CRSwNP through a series of bioinformatic analyses, and initially but relatively comprehensively explored their potential molecular mechanisms in CRSwNP. This will provide a new perspective and reference for the diagnosis and pathogenesis of CRSwNP.

Author Contributions

All authors made a significant contribution to conception, study design, execution, acquisition of data, analysis, and interpretation; drafted, revised, and reviewed the article; and gave the final approval.

Ethics Statement

The consent procedure was approved by the Ethics Committee for Human Study at the Third Affiliated Hospital of Sun Yat‐sen University (China). The patients of the specimen for PCR and IF staining experiments were informed of the purposes and procedures of the study and provided written informed consent. Registry and the Registration No. of the study: ChiCTR2300075072. Animal Studies: N/A.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Quality control of the scRNA‐seq data.

Figure S2: The principal component analysis (PCA).

Figure S3: VDR, PLA2G2A, and CPT1C were highly expressed in epithelial cells and monocytes.

Figure S4: VDR, PLA2G2A, and CPT1C were expressed in different monocyte clusters.

Figure S5: VDR, PLA2G2A, and CPT1C were expressed in different endothelial cell clusters.

Figure S6: Cellular communication of endothelial cells and monocytes with other cells.

Table S1: 866 Cellular senescence‐related genes (CSRGs). Table S2: 871 lipid metabolism‐related genes (LMRGs). Table S3: 200 Inflammation‐related genes (IRGs). Table S4: Primers for biomarkers. Table S5: Enrolled participant characteristics. Table S6: Significant enrichment of DEGs to GO entries. Table S7: Significant enrichment of DEGs to KEGG entries. Table S8: GO results for CPT1C. Table S9: GSEA results for CPT1C. Table S10: GO results for PLA2G2A. Table S111: GSEA results for PLA2G2A. Table S12: GO results for VDR. Table S13: GSEA results for VDR. Table S14: Correlation of biomarkers with the inflammatory response. Table S15: The marker gene of the different cell types.

Supporting Information

Data Availability Statement

The data underlying this study can be accessed by contacting the corresponding author with a formal request.