WGCNA identifies key modules in psoriasis and cervical cancer
The investigators merged two psoriasis-related GEO datasets, GSE14905 and GSE13355. The data sets were merged and normalized to ensure uniformity for principal component analysis and to rectify batch effects.The final training dataset consisted of 91 patients and 85 matched controls, and the evaluation showed that the data preprocessing was valid and reliable. From the density plot, we can observe that the sample distributions of the individual datasets before removing the batch effect varied greatly, suggesting a batch effect, and after removing the batch effect the data distributions between the individual datasets converged, with similar means and variances (Fig. 2A, B). Weighted gene co-expression network analysis was conducted utilizing the R package WGCNA, the genes with expression variance in the top 50% were used as the screening conditions, and the genes with less volatility were excluded, and the co-expression network was constructed for 20547 genes of psoriasis and 10,275 genes of cervical cancer.
Combining the analysis of scale independence and average connectivity, in the psoriasis samples, b=12 was chosen (Fig. 2C, D) as the soft threshold. The minimum module size was set to 30 and 15 gene modules were obtained. The results showed that the brown module had the highest correlation with psoriasis (correlation coefficient = 0.92, p= 2.2e-72, Fig. 2I). brown module contained 4917 genes; 969 genes were identified as core genes with high MM (> 0.8) and GS (> 0.1) values, and ultimately, 969 psoriasis-significantly correlated genes were identified in brown color module. Key genes. In the cervical cancer samples, 8 was chosen as the optimal soft threshold to build a scale-free network (Fig. 2E, F). Subsequently, cluster analysis was used to identify highly similar modules with the minimum module size set to 30, sensitivity set to 3, and modules with distances less than 0.25 were merged to obtain 18 gene modules (Fig. 2K). The correlation between cervical cancer and gene modules (Fig. 2J) showed that the green module had the highest correlation with cervical cancer (2270 genes, r=0.71, p=5.4e-9), and the green module was taken as the key module. The genes in the green module: MM>0.8 and GS>0.1 were selected as pivotal genes, and a total of 421 key genes significantly associated with cervical cancer were identified.
Identification of Differentially Expressed Genes and Machine Learning Screening of Key Genes
By LIMMA analysis, 2066 differentially expressed genes (DEGs) between psoriasis patients and healthy controls were identified in the integrated dataset, of which 1134 genes were up-regulated and 932 genes were down-regulated. These DEGs were presented by volcano plot visualization (Fig. 3A). In addition, the cervical squamous cell carcinoma dataset generated 6573 DEGs, including 2689 up-regulated genes and 1586 down-regulated genes (Fig. 3B). The DEGs from the cervical squamous cell carcinoma and psoriasis samples were intersected with key genes taken from the WGCNA to obtain a total of 27 genes for subsequent analysis (Fig. 3C). 27 genes were uploaded to the STRING database to construct a protein-protein interaction network (Fig. 3D), then we analyzed the top 10 genes by using the "degree" algorithm with the CytoHubba application in Cytoscape to identify the key genes, and the color of the nodes indicated the strength of the correlation (Fig. 3E). The color of the nodes indicates the strength of the correlation. Random forest pairs were used for screening and finally 16 characterized genes were identified in psoriasis samples, including MELK, AURKA, CENPN, CDCA5, KIF2C, NDC80, PRSS3P2, PRC1, DEPDC1B, FOXM1, UHRF1, WDR53, MCM10, BUB1B, NCAPH, CDCA2(Fig. 3F). Meanwhile, 10 cervical squamous cell carcinoma signature genes were also identified using RF algorithms, including NCAPH, SMC4, CENPN, UHRF1, STIL, CDCA2, MELK, ZNF665 (Fig. 3G). Next, the study found that these algorithms identified five overlapping genes (Fig. 3H), namely NCAPH, UHRF1, CENPN, CDCA2, MELK which were used for subsequence analysis (Table 3).
Table 3 Overview of the five hub genes.
Symbol
|
Description
|
Aspect
|
References
|
NCAPH
|
Non-SMC Condensin I Complex Subunit H
|
Interferes with plasmids and affects cell proliferation and migration
|
(39)
|
UHRF1
|
Ubiquitin Like With PHD And Ring Finger Domains 1
|
Required for G1/S phase transition;
Regulation of DNA methylation, chromatin modification, cell proliferation and DNA repair
|
(40, 41)
|
CENPN
|
Centromere Protein N
|
Binds to filaments in S and G2 phases and recruits proteins
|
(42)
|
CDCA2
|
Cell Division Cycle Associated 2
|
Affects tumor cell proliferation and regulates the G0 / G1 phase of the cell cycle
|
(43)
|
MELK
|
Maternal Embryonic Leucine Zipper Kinase
|
Induces inflammatory responses through secretion of pro-inflammatory factors
Involved in mitosis, proliferation, apoptosis, differentiation and tumorigenesis
|
(44, 45)
|
GO and KEGG enrichment analyses were performed to identify biological pathways and diseases associated with key genes.
For biological processes in GO enrichment analysis, biological processes were highly enriched in mitotic cell cycle processes (Fig. 4A, biological processes (BP)). And for the cellular components in GO, it involves intracellular non-membrane-bound organelles, chromosomes, and mitotic regions (Fig. 4B, cellular components (CC)). For the molecular functions enriched in GO, including nucleotide binding, phosphoribosylation, chromatin binding (Fig. 4C, Molecular Functions (MF)). Based on the KEGG database further to decipher the biological pathways behind. The enriched molecular pathways included cell cycle, microRNAs in cancer, oocyte meiosis, breast cancer, gastric cancer, and mTOR signaling pathway (Fig. 4D). These findings are in line with the results of GO enrichment analysis, providing further evidence of the association between cervical squamous cell carcinoma and psoriasis.
The CIBERSORT analysis tool calculated the proportions of 22 types of leukocyte subpopulations in psoriasis and CESC samples, respectively, including naïve B cells, memory B cells, plasma B cells, CD8 T cells, CD4 naive T cells, CD4 memory quiescent T cells, CD4 memory-activated T cells, follicular helper T cells, regulatory T cells (Tregs), γ δ T cells, resting natural killer (NK) cells, activated NK cells, monocytes, M0, M1 and M2 macrophages, resting and activated myeloid dendritic cells, and resting and activated mast cells. We also did the relationship of key genes to immune infiltrating cells in both diseases and found that genes associated with psoriasis can also play a role in cervical squamous cell carcinoma. Stacked bar graphs of the two datasets show the percentage of 22 immune cells in each sample (Fig. 5A). Analysis of the immune microenvironment in psoriasis patients revealed significant differences in the abundance of 20 immune cells. Analysis of the immune microenvironment in patients with cervical squamous cell carcinoma revealed notable variations in the abundance of five immune cells. These differences were statistically significant (Fig. 5B). In summary, patients with psoriasis and cervical squamous cell carcinoma have varying degrees of multiple immune cell infiltrations, and these immune cell infiltrations may be potential regulatory points for therapy. Then, the spearman correlation coefficient between hub genes and the infiltration level of the immune cell was calculated. As a result, resting mast cells and CD8T cells were negatively correlated with the expression of NCAPH, UHRF1, CDCA2, CENPN and MELK in patients with psoriasis and cervical squamous carcinoma, respectively (Fig. 5C).
Identification of candidate small molecule compounds for the treatment of psoriasis and cervical squamous cell carcinoma
The intersection of DEGs genes upregulated in psoriasis and cervical squamous cell carcinoma was taken with hub genes in the WCGNA module, and 24 relevant pathogenic genes were obtained (Fig. 6A). The screened 24 relevant pathogenic genes were imported into connectivity map (cMAP) database to predict small molecule compounds that could reverse the gene expression alterations in psoriasis-related pathogenesis and cervical squamous cell carcinoma. Phloretin, antimycin-a, palbociclib, purvalanol-a, aminopurvalanol-a, PD-102807, 7b-cis, pyrvinium-pamoate, angiogenesis-inhibitor, roscovitine were the top 10 compounds with the highest negative scores as potential drugs for therapy (Fig. 6B). The targeting pathways and chemical structures of these 10 compounds are described in Fig. 6C, D.
Validation of hub genes with GEO and TCGA databases and cellular experimental validation
To further confirm the accuracy of the comprehensive bioinformatics analysis described above, we first examined the expression patterns of the five hub genes in the patients of the two validation cohorts, and chose the psoriasis dataset, GSE63514 and the cervical squamous cell carcinoma dataset, TCGA-CESC, as the validation datasets. Multi-group box plots showed that the expression levels of NCAPH, UHRF1, CDCA2, CENPN and MELK were significantly higher in psoriasis patients and cervical squamous cell carcinoma patients than in normal controls (Fig. 7A, B). RT-qPCR results confirmed that the expression patterns of the five pivotal genes were consistently up-regulated in cervical cancer samples as compared to the control samples (Fig. 7D), and that the expression levels of CENPN and MELK mRNA levels were increased (Fig. 7C).
Cohort validation of hub genes and enrichment analysis
We plotted ROC curves based on the five candidate genes to assess the diagnostic value of each gene. The calculated AUCs and 95% confidence intervals were as follows: NCAPH (AUC 0.92, CI 0.97–0.88), UHRF1 (AUC 0.89, CI 0.94–0.84), CDCA2 (AUC 0.96, CI 0.99–0.92), CENPN (AUC 0.94, CI 0.98–0.90) and MELK (AUC 0.96, CI 1.00–0.93). The findings indicated that the acquired genes had a significant diagnostic value in Psoriasis (Fig. 8A). To investigate the potential functions of common central genes, we divided the samples from the psoriasis dataset into groups with high and low expressions based on median levels. We then identified DEGs between these groups and conducted GO/KEGG enrichment analysis. The significant enriched genes include "lysosomes, phagocytosis, SLE, pyrimidine metabolism, arachidonic acid metabolism, complement and coagulation cascades, and natural killer cell-mediated cytotoxicity (Fig. 8B, C).
The regulatory signatures analysis
We applied the miRNet database to screen the targeted miRNAs of NCA NCAPH, UHRF1, CDCA2, CENPN and MELK. As depicted in Fig. 9A, the prediction identifies three miRNAs: hsa-miR-124-3p, hsa-mir-129-2-3p, and hsa-mir-147a. The Network analysis tool explored 9 transcription factors namely FOXC1, NFKB1, RELA, SREBF1, NRF1, GATA2, TFAP2A, USF1, USF2 (Fig. 9B). The TFs and miRNAs related to three hub genes via network analysis were shown in Table 4.
Table 4 Top transcription factors and miRNA predicted from miRNA-mRNA, TFs-mRNA regulatory networks
TFs/miRNAs
|
Description
|
Biological function
|
Reference
|
FOXC1
|
Forkhead
|
Regulation of cell proliferation, migration and invasion through PI3K / AKT signaling
|
(46)
|
NFKB1
|
nuclear factor kappa B subunit 1
|
Inhibition of cell proliferation, colony formation and migration in cervical cancer
|
(47)
|
RELA
|
v-rel avian reticuloendotheliosis viral oncogene homolog A
|
Control of NF-κB activity by autophosphorylation in inflammatory diseases and cancer。
|
(48)
|
SREBF1
|
sterol regulatory element binding transcription factor 1
|
Stimulates ubiquitination of SREBP1 and inhibits endoplasmic reticulum stress in CESC cells.
|
(49)
|
NRF1
|
nuclear respiratory factor 1
|
Leads to severe oxidative stress, genomic instability
|
(50)
|
GATA2
|
GATA binding protein 2b
|
A common regulatory elements in cervical cancer
|
(51)
|
TFAP2A
|
transcription factor AP-2 alpha
|
Promotes the growth of cervical tumors
|
(52, 53)
|
USF1/2
|
upstream transcription factor 1/2
|
Enhancement of cervical cancer cell malignancy by transcriptional activation of p65
|
(54) (55)
|
hsa-miR-124-3p
|
MicroRNA 124
|
Direct targeting of IGF2BP to inhibit cervical cancer growth and metastasis is considered to be an important marker and target for CC prognosis
|
(56)
|
hsa-mir-129-2-3p
|
MicroRNA 129
|
The methylation process of mir-129-2-3p increases cervical (pre)cancerous lesions.
|
(57)
|
hsa-mir-147a
|
MicroRNA 147a
|
Interacts with circ_0018289 binding and Linc00319 to promote cervical cancer progression.
|
(58, 59)
|