Identification of potential immune-related circRNA-miRNA-mRNA regulatory network in cutaneous squamous cell carcinoma

Abstract

Circulating RNAs (circRNAs) are involved in tumor development and progression by participating in immune regulation. Nevertheless, the circRNAs expression profiles and their roles on the immunomodulatory effects in cutaneous squamous cell carcinoma (cSCC) have rarely been studied. In our study, we identified the differentially expressed circRNAs (DEcircRNAs), miRNAs (DEmiRNAs), mRNAs (DEmRNAs) in cSCC and established the circRNA competing endogenous RNAs (ceRNAs) network. Subsequently, the hub differentially expressed immune-related genes were identified and validated by immunochemistry as well as the GO and KEGG pathway analysis were performed. 54 differentially expressed circRNAs were identified and hub differentially expressed immune-related genes were identified and they were mostly associated with immune response in the progression of cSCC. Our results indicated that the potential immune-related circRNA-miRNA-mRNA network may assist in understanding the molecular mechanisms underlying the carcinogenesis and progression in cSCC. Moreover, the immune-related genes may provide an insight into the pathogenesis, molecular biomarkers, and potential therapeutic targets for cSCC patients.

Introduction

Cutaneous squamous cell carcinoma (cSCC) is one of the most common non-melanoma skin malignant tumors derived from epidermal keratinocytes, accounting for about 20% of skin tumors, with high metastasis and mortality rates [1,2]. But due to lack of typical early presentation, most patients in clinical diagnosis often have an aggressive cSCC that usually spreads to other tissues and organs when they were confirmed [3]. Currently, the main treatment options are surgical resection, radiotherapy and chemotherapy. Most patients have better outcomes with surgical resection at an early stage [4]. However, extensive locally destructive or metastatic lesions remain difficult to manage and the prognosis is poor [5,6]. Therefore, intensive investigation of the regulatory factors involved in cSCC progression is imperative. Meanwhile, more reliable diagnostic and therapeutic biomarkers should be further studied.

Circular RNA, one of endogenous long-stranded non-coding RNAs, formed by reverse splicing in a class of covalently closed single-stranded circular, is widespread in eukaryotic cells [7]. Moreover, it has the characteristics of stability, richness, conservation, localization, and disease-specificity [8,9]. Notably, it has been revealed that circRNAs play a critically important role in the progression of numerous diseases, such as tumors [10,11], cardiovascular diseases [12,13], metabolic diseases [14] and digestive system diseases [15,16]. Moreover, the investigators have discovered that circRNAs were rich in miRNA binding sites, which could competitively adsorb miRNAs and influence their functions [17]. Furthermore, a multitude of researches have shown that aberrant circRNA expression played an essential role in cancer through the circRNA-miRNA-mRNA regulatory axis and thereby influence the development and progression of cancer [18-20]. For example, Ma et al. revealed that hsa_circ_0004872 could suppresses gastric cancer progression by competitively binding miR-224/Smad4/ADAR1 pathway [21]. Zhang et al. found that circFGFR1 could sponge miR-381-3p to promote the progression of non-small cell lung cancer (NSCLC) and anti-PD-1 resistance [22]. Nevertheless, the tumor microenvironment is extremely complex, and it is not optimal to focus on one or more relationships of competing endogenous RNAs (ceRNAs) in a single isolated way.

Recently, increasing evidences have pointed that the immune response was involved in the development of cSCC [23,24]. More importantly, the latest developments suggested that circRNAs were not only engaged in the innate immune responses but also exert their role in tumor immunity [25,26]. Wei et al. found that circ_0020710 could facilitate the progression and immune evasion of melanoma by the regulation of miR-370-3p/CXCL12 axis [27]. Another study also suggested that circ-CPA4/let-7 miRNA/PD-L1 axis could modulate immune escape in NSCLC [28]. Li et al. discovered that circ_0000284 could contribute to the development of NSCLC by competitively binding miR-377 to upregulate PD-L1 expression [29]. However, previous studies have just concentrated on individual circRNAs, there has been few report regarding the roles of circRNAs acting as a miRNA sponge to impact the immune-related genes of cSCC thereby affecting its progression.

Recognizing the essential of the immune system in malignant cancer, there is an urgency for credible prognosticators and effectively therapeutic targets for cSCC, we identified DEcircRNAs, DEmiRNAs, DEmRNAs and established the circRNA ceRNA network associated with cSCC. From the ceRNA network, we identified immune-related genes and established a PPI network to identify hub immune-related genes to provide new insights for improving clinical diagnosis and potential immunotherapy targets of cSCC patients.

Materials and methods

Data collection and inclusion criteria

We retrieved the microarray data through the GEO database (http://www.ncbi.nlm.nih.gov/geo/). The following search terms were used: ‘cSCC’ and ‘circRNA’ or ‘miRNA’ or ‘mRNA’. Moreover, the inclusion criteria of the characteristics of GEO database: (1) the samples including cSCC and normal controls; (2) organism: homo sapiens; (3) experiment type: non-coding RNA profiling or expression profiling were detected by microarray; (4) platforms: circRNA and miRNA were detected by Agilent; mRNA by Affymetrix Human Genome U133 Plus 2.0 Array; (5) sample source: tissue.

Identification of DEcircRNAs and annotation of circRNAs

The DEcircRNAs were identified via limma R package according to the criteria of |log₂ (fold change)| (|log₂ FC|)>2.0 and P-value <0.05 after standardization and log₂ transformation of the raw microarray data. Then, we annotated DEcircRNAs using circBase [30]. Moreover, the details information of circRNA were downloaded from circBase. Next, we used the website tools of Cancer-Specific CircRNA (CSCD) to predict their basic structural patterns. Finally, circRNAs’ parental genes were used to GO and KEGG pathway analysis by FunRich software (version 3.1.3) and KOBAS, respectively.

Identification of DEmiRNAs and prediction the target miRNAs of circRNAs

The DEmiRNAs were obtained according to |log₂ FC|>2.0 and P-value <0.05. Furthermore, “miRNA sponge” is widely considered to be a potential explanation for the work of circRNAs. In order to investigate the miRNAs sponged to DEcircRNAs, the circular RNA Interactome (https://circinteractome.nia.nih.gov/) was used to predict target miRNAs for DEcircRNAs. The above target miRNAs were further filtered by DEmiRNAs based on GEO34536, and these target miRNAs and DEmiRNAs were overlapped by Venny 2.1.0 to find the intersection.

Identification of DEmRNAs and prediction of miRNA-mRNA pairs

The differentially expressed mRNAs (DEmRNAs) was screened out with the |log₂ FC|>1.0 and P-value <0.05. Moreover, the targeting miRNAs of DEcircRNAs and DEmiRNAs were overlapped as the ultimate miRNAs to further study. Subsequently, the miRNA-mRNA interactions were predicted using miRDB, miRTarBase and TargetScan. Only target mRNAs identified by the three databases were selected for intersection with DEmRNAs to obtain the final functional genes (FmRNAs).

Constructing the circRNA-miRNA-mRNA network

The ceRNA network was constructed with the final circRNAs, miRNAs, and mRNAs, which obtained from the above steps. Visualization of ceRNA network used R software.

GO and KEGG pathway of mRNAs in the network

To assess functional enrichment, FunRich3.1.3 was used to perform GO analysis and KOBAS was adapted to conduct KEGG pathway analysis of final mRNAs (FmRNAs) in the circRNA-miRNA-mRNA network.

Constructing PPI network and identification of hub genes

The PPI network of the DEmRNAs in ceRNAs network was established using the STRING database (http://string-db.org/) with the minimum required interaction score was set to 0.4, and then visualized using Cytoscape 3.7.1 software. The apps in Cytoscape (CytoHubba and MCODE) were used to clarify hub genes and analyze modules, respectively. Then, the expression of some hub genes was confirmed by immunohistochemistry in HPA database and clinical samples.

Immune-related gene set

The lists of immune-related genes were obtained from the ImmPort (https://www.immport.org/shared/home) databases. Then, the differentially expressed immune-related genes were obtained by overlapped with DEmRNAs and the lists of immune-related genes. Then they were analyzed by GO and KEGG analyses to explore their functions. Similarity, the differentially expressed immune-related genes were also used to conduct a PPI network and selected the hub genes and validated by immunohistochemistry in HPA database and clinical samples.

Tissue samples

28 cases of Formalin-fxed, parafn-embedded cSCC tissue and 27 cases of non-cancerous tissue were collected from the Department of Pathology, West China Hospital, Sichuan University from 2018 to 2021. The Ethical Committee of West China Hospital approved this study and waived informed consent.

Immunohistochemistry

The paraffin-embedded tissue sections were used for immunohistochemical staining. Briefly, the tissue sections were deparaffinized and then rehydrated. The second step was antigen retrieval and followed by blocking endogenous peroxidase activity. Next, the primary antibodies were incubated and the main information as follows: anti-ISG15 (1:100 dilution, Cat no#15981-1-AP, Proteintech), anti-RSAD2 (1:200 dilution, Cat no# 28089-1-AP, Proteintech), anti-PI3 (1:200 dilution), anti-OAS3 (1:100 dilution, Cat no#21915-1-AP, Proteintech), anti-CXCL10 (1:200 dilution, Cat no#10937-1-AP, Proteintech), anti-USP18 (1:1000 dilution, Cat no#GB111923, Servicebio), anti-IL7R (1:100 dilution, Cat no#DF6362, Affinity). The semi-quantitative criterion of evaluating protein levels was based on the intensity of staining and the proportion of positive cells. According to the staining intensity, the results were as follows: 0 (no stain), 1 (light yellow), 2 (brown) and 3 (tan). Additionally, based on the proportion of positive cells, there were five grades: 0 (0%), 1 (1-25%), 2 (26-50%), 3 (51-75%) and 4 (76-100%). The final result was multiplied the two scores. Finally, if the score was more than 6, it was considered high, otherwise the opposite.

Results

Characteristics of eligible GEO database

A total of 13 GEO databases related to keywords were searched. The flow chart diagram of the screening process is shown in Figure 1. Three of them are circRNA databases (GSE74758, GSE139505, GSE136113), four are miRNAs (GSE57768, GSE34536, GSE130767, GSE101192), and six are mRNAs (GSE45164, GSE66359, GSE42677, GSE45216, GSE150727, GSE66412). According to the inclusion criteria, in the end only GSE74758, GSE34536, GSE45164, GSE66359, and GSE42677 met the criteria for inclusion in this study. The following key information was collected from each GEO database: organism, platform, contributor(s), experiment type, last update year, region, sample size, detected RNA, which were listed in Supplementary Table 1.

Figure 1.

A flow chart displaying the screening process in this study.

Identification of differentially expressed circRNAs and annotation of circRNAs

A total of 54 DEcircRNAs were identified by comparing the 3 cSCC and 3 non-lesional skin tissues in the GSE74758 dataset (Figure 2). Then, we clarified the detailed information regarding these DEcircRNAs through the circBase database, such as the position, best transcript and gene names (Supplementary Table 2).

Figure 2.

The expression profiles of circRNAs. (A) The heatmap of DEcircRNAs, (B) The volcano plots of DEcircRNAs.

Enrichment analysis of circRNAs’ parental genes

To explore the functions of DEcircRNAs based on their parental genes in cSCC patients, GO and KEGG analysis were performed. The results showed that the most significant GO terms for biological process (BP) were energy pathways, and metabolism (Figure 3A). For cellular component (CC), endoplasmic reticulum membrane, COPII vesicle coat and gamma-tubulin ring complex were the most significant terms (Figure 3B). Furthermore, CoA-ligase activity, mannosyltransferase activity, and transporter activity were the most enriched term for molecular function (MF) (Figure 3C). Moreover, the enriched pathways analysis by KEGG were most related to mitophagy, renal cell carcinoma, PPAR signaling pathway, prostate cancer, HIF-1, and AMPK signaling pathway (Figure 3D).

Figure 3.

Top 10 Gene Ontology terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of 54 differentially expressed circRNAs’ parental genes. A-C. The biological process, cellular component and molecular function of GO terms, respectively. D. KEGG pathway enrichment analysis of the genes.

Predicted the targets miRNAs of circRNAs

To further elucidate the functions of circRNAs, the top 6 up and 6 down-regulated circRNAs were identified as candidate circRNAs to further study according to the |logFC| value, including hsa_circ_0068631, hsa_circ_0070933, hsa_circ_0067772, hsa_circ_0003528, hsa_circ_0070934, hsa_circ_0001955, hsa_circ_0022392, hsa_circ_0022383, hsa_circ_0005085, hsa_circ_0046449, hsa_circ_0072279, and hsa_circ_0000375. Their basic structural patterns were shown in Supplementary Figure 1 based on the data from CSCD. Next, the above most 12 significantly changed circRNAs were used to predict downstream miRNAs and we identified 215 circRNAs-miRNAs pairs in Table 1.

Table 1.

The target miRNAs of DEcircRNAs

CircRNA	Position	Target microRNA number	Target microRNA
hsa_circ_0068631	chr3:195802029-195803993	8	hsa-miR-1303, hsa-miR-139-5p, hsa-miR-515-5p, hsa-miR-599, hsa-miR-615-3p, hsa-miR-640, hsa-miR-659, hsa-miR-758
hsa_circ_0070933	chr4:128995614-128996148	7	hsa-miR-1290, hsa-miR-1304, hsa-miR-136, hsa-miR-1825, hsa-miR-431, hsa-miR-515-5p, hsa-miR-526b
hsa_circ_0067772	chr3:155547476-155560408	26	hsa-miR-1183, hsa-miR-1225-5p, hsa-miR-1238, hsa-miR-1256, hsa-miR-1276, hsa-miR-1283, hsa-miR-1286, hsa-miR-1289, hsa-miR-1294, hsa-miR-409-3p, hsa-miR-421, hsa-miR-488, hsa-miR-492, hsa-miR-494, hsa-miR-512-5p, hsa-miR-518a-5p, hsa-miR-527, hsa-miR-548c-3p, hsa-miR-567, hsa-miR-581, hsa-miR-623, hsa-miR-646, hsa-miR-649, hsa-miR-651, hsa-miR-1827, hsa-miR-892a
hsa_circ_0003528	chr5:134032815-134044578	26	hsa-miR-1236, hsa-miR-1248, hsa-miR-1252, hsa-miR-1253, hsa-miR-1279, hsa-miR-1287, hsa-miR-1288, hsa-miR-192, hsa-miR-215, hsa-miR-217, hsa-miR-223, hsa-miR-224, hsa-miR-324-5p, hsa-miR-330-3p, hsa-miR-370, hsa-miR-421, hsa-miR-488, hsa-miR-495, hsa-miR-548, hsa-miR-574-5p, hsa-miR-587, hsa-miR-606, hsa-miR-668, hsa-miR-890, hsa-miR-891b, hsa-miR-892b
hsa_circ_0070934	chr4:128995614-129012667	33	hsa-miR-1208, hsa-miR-1229, hsa-miR-1234,hsa-miR-1236,hsa-miR-1238, hsa-miR-1247, hsa-miR-1263, hsa-miR-1290, hsa-miR-1304, hsa-miR-1305, hsa-miR-136, hsa-miR-1825, hsa-miR-197, hsa-miR-203, hsa-miR-383, hsa-miR-431, hsa-miR-487a, hsa-miR-502-5p, hsa-miR-515-5p, hsa-miR-520g, hsa-miR-520h, hsa-miR-526b, hsa-miR-556-3p, hsa-miR-558, hsa-miR-578, hsa-miR-598, hsa-miR-600, hsa-miR-626, hsa-miR-636, hsa-miR-647, hsa-miR-668, hsa-miR-889, hsa-miR-942
hsa_circ_0001955	chr15:64495280-64508912	32	hsa-miR-1229, hsa-miR-1243, hsa-miR-1252, hsa-miR-1256, hsa-miR-1287, hsa-miR-1296, hsa-miR-1299, hsa-miR-1305, hsa-miR-145, hsa-miR-149, hsa-miR-188-3p, hsa-miR-326, hsa-miR-330-5p, hsa-miR-338-3p, hsa-miR-496, hsa-miR-502-5p, hsa-miR-516a-5p, hsa-miR-520g, hsa-miR-520h, hsa-miR-532-3p, hsa-miR-568, hsa-miR-583, hsa-miR-599, hsa-miR-607, hsa-miR-630, hsa-miR-644, hsa-miR-646, hsa-miR-659, hsa-miR-766, hsa-miR-767-5p, hsa-miR-769-5p, hsa-miR-873
hsa_circ_0022392	chr11:61630443-61631258	8	hsa-miR-1288, hsa-miR-1299, hsa-miR-136, hsa-miR-548b-3p, hsa-miR-657, hsa-miR-665, hsa-miR-766, hsa-miR-873
hsa_circ_0022383	chr11:61605249-61615756	17	hsa-miR-1206, hsa-miR-1245, hsa-miR-1258, hsa-miR-1287, hsa-miR-1324, hsa-miR-197, hsa-miR-495, hsa-miR-498, hsa-miR-524-3p, hsa-miR-525-3p, hsa-miR-624, hsa-miR-629, hsa-miR-635, hsa-miR-665, hsa-miR-766, hsa-miR-7, hsa-miR-876-3p
hsa_circ_0005085	chr2:9419445-9437574	11	hsa-miR-186, hsa-miR-515-5p, hsa-miR-521, hsa-miR-548b-3p, hsa-miR-579, hsa-miR-589, hsa-miR-609, hsa-miR-626, hsa-miR-661, hsa-miR-766, hsa-miR-873
hsa_circ_0046449	chr17:80714040-80758872	32	hsa-miR-1182, hsa-miR-1184, hsa-miR-1205, hsa-miR-1277, hsa-miR-1247, hsa-miR-1299, hsa-miR-186, hsa-miR-361-3p, hsa-miR-433, hsa-miR-450b-3p, hsa-miR-769-3p, hsa-miR-512b-5p, hsa-miR-517a, hsa-miR-517c, hsa-miR-548c-3p, hsa-miR-548p, hsa-miR-587, hsa-miR-1270, hsa-miR-620, hsa-miR-624, hsa-miR-636, hsa-miR-639, hsa-miR-646, hsa-miR-654-3p, hsa-miR-665, hsa-miR-668, hsa-miR-758, hsa-miR-873, hsa-miR-885-3p, hsa-miR-885-5p, hsa-miR-891b, hsa-miR-940
hsa_circ_0072279	chr5:37352838-37358253	7	hsa-miR-139-5p, hsa-miR-488, hsa-miR-518a-5p, hsa-miR-527, hsa-miR-654-3p, hsa-miR-665, hsa-miR-934
hsa_circ_0000375	chr12:6657590-6657991	8	hsa-miR-1182, hsa-miR-1184, hsa-miR-1205, hsa-miR-296-5p, hsa-miR-330-3p, hsa-miR-487a, hsa-miR-490-5p, hsa-miR-622

Differential expression of miRNAs in cSCC patients

Based on the GSE34536 database, we obtained 21 DEmiRNAs, including 11 upregulated and 10 down-regulated (Figure 4A, 4B). Then, intersection of miRNAs predicted by 12 top circRNAs and DEmiRNAs. Finally, two miRNAs of hsa-miR-136 and hsa-miR-766 were obtained as candidate miRNAs (Figure 4C).

Figure 4.

The expression profiles of miRNAs and mRNAs. (A) the heatmap of DEmiRNAs, (B) The volcano plots of DEmiRNAs, (C) intersection of miRNAs predicted by 12 top circRNAs and DEmiRNAs, (D) DEmRNAs in GSE45164, (E) DEmRNAs in GSE66359, (F) DEmRNAs in GSE42677, (G) Common DEmRNAs in both three GEO database, (H) target genes of predicted miRNAs by miRDB, miRTarBase, and TargetScan databases, (I) common genes were obtained by interact 129 target genes of predicted miRNAs and 66 genes among three GEO databases.

Differential expression of mRNAs in patients with cSCC and the targets of miRNAs

After comparing the 10 tumor tissues and 3 normal tissues in the GSE45164 database, we filtered and obtained 974 DEmRNAs, which were made up of 502 up-regulated and 472 down-regulated genes (Figure 4D). Meanwhile, 1252 DEmRNAs were selected in the GSE66359 dataset, including 656 up and 596 down-regulated genes (Figure 4E). Moreover, 2888 DEmRNAs were identified in the same way from the GSE42677 dataset, and it included 1432 up-regulated and 1456 down-regulated genes (Figure 4F). The number of common genes in both three GEO database was 66 genes (Figure 4G). There were 129 target genes of predicted miRNAs through miRDB, miRTarBase, and TargetScan databases (Figure 4H). Finally, 36 common genes (FmRNAs) were obtained by interact 129 target genes of predicted miRNAs and 66 genes among three GEO databases (Figure 4I).

The circRNA-microRNA-mRNA interaction network

Increasing evidence has suggested that circRNAs may play a role in regulating the functions of target mRNAs as miRNA sponges. In order to further make sense the potential roles of the DEcircRNAs in cSCC, the circRNA-microRNA-mRNA interaction network was constructed. Among the top 12 DEcircRNAs, the predicted miRNAs of six circRNAs were not found in the DEmiRNAs, so only 6 circRNAs, 2 miRNAs and 36 target mRNAs were used to construct the network (Figure 5A). We found that hsa-miR-136 was predicted to be common downstream of hsa_circ_0070933, hsa_circ_0022392 and hsa_circ_0070934, which indicated that there might be some functional connection between them. These RNA interactions may provide new insight into the mechanism underlying cSCC.

Figure 5.

ceRNA network and GO and KEGG pathway analysis of the genes in ceRNA network. (A) ceRNA network, (B) GO terms of BP category, (C) CC category, (D) MF category, and (E) The significant enriched KEGG pathways.

Enrichment analysis of the genes in the network

The 36 genes in the network were used to evaluate the biological roles of the circRNAs. The top 10 GO terms of each group are shown in Figure 5B-D. In the BP category, the main enriched categories were immune response, spindle assembly and cell growth and/or maintenance (Figure 5B). In CC, extracellular, membrane raft, and plasma membrane were the main enriched terms (Figure 5C). In MF, the mostly enriched terms were cell adhesion molecule activity, protease inhibitor activity, ATP binding, and deaminase activity (Figure 5D). Furthermore, the most enriched KEGG pathways were epstein-barr virus infection, human papillomavirus infection, hepatitis C, influenza A, chemokine signaling pathway, RIG-I-like receptor signaling pathway, and microRNAs in cancer. The Figure 5E showed the most enriched cancer-related pathways.

Identification of hub genes in the PPI network

According to STRING database, we established a PPI network to show the interactions of the 36 target genes (Figure 6A). MCODE in Cytoscape was applied to identify hub module. Two modules were identified, Module 1 consisted of 7 genes and 19 edges (Figure 6B), while module 2 consisted of 6 genes and 15 edges (Figure 6C). According to cytoHubba plugin’s MCC ranking, the top 10 hub genes were USP18, OAS3, ISG15, RSAD2, PI3, CXCL10, BST2, SPRR3, DSC2 and DSG3, which was shown in Figure 6D. Immunohistochemical assays from the HPA database was implemented to conforme the expression of proteins encoded by these hub genes, the result was presented in Figure 7. However, PI3 and CXCL10 protein expression was not found in HPA database. In order to better reveal the reality of our results, we have collected 28 cases of cSCC tissue and 27 cases of non-cancerous tissue from our hospital to validate the top hub gens and the genes that were not found in HPA database (PI3 and CXCL10), the results showed the expression level of USP18 was expressed higher (17/28, 60.7%) than that in controls (9/27, 33.3%). OAS3 was elevated in tumor tissue (19/28, 67.9%) compared with non-cancerous tissue (10/27, 37.0%). Similarly, ISG15 was expressed higher in cSCC (18/28, 64.3%) than controls (10/27, 37.0%). CXCL10 was also up-regulated in tumor tissue (13/28, 46.4%) compared with controls (5/27, 18.5%). Conversely, RASD2 was expressed higher in non-cancerous tissue (12/27, 44.4%) than tumor tissues (5/28, 17.9%). PI3 was expressed lower in tumor (9/28, 32.1%) than controls (15/27, 55.6%). The above results were presented in Figure 8.

Figure 6.

PPI analysis. A. PPI network of 36 genes was constructed in STRING database. B. Module 1 of the PPI network. C. Module 2 of the PPI network. D. Top 10 hub genes selected by the CytoHubba.

Figure 7.

Validation of hub genes by immunohistochemistry in HPA databases. The scale bars represent 200 μm.

Figure 8.

The clinical samples were performed to confirm the protein expression of USP18, OAS3, ISG15, CXCL10, IL7R, PI3, and RSAD2 by immunohistochemistry. The scale bars represent 50 μm.

Identification of immune-related genes and enrichment analysis

10 immune-related differentially expressed genes were obtained by overlapped with selected DEmRNAs in the ceRNA network (FmRNAs) and the lists of immune-related genes (Figure 9A and Supplementary Table 3). Next, the PPI network was constructed based on immune-related differentially expressed genes and shown in Figure 9B. According to cytoHubba plugin’s MCC ranking, the top hub genes were CXCL10, BST2, ISG15, RSAD2, LYN, SHC1, PI3, IL7R, which was shown in Figure 9C. CXCL10, BST2, ISG15, RSAD2 and PI3 have been verified above. The expression level of SHC1 and LYN was validated by immunohistochemical assays from the HPA database, the result showed that they were expressed higher in tumor than in controls (Figure 7). However, IL7R protein expression was not found in HPA database, our own clinical samples were collected to validate it, the result found that IL7R was expressed higher in tumor (15/28, 53.6%) than in controls (7/27, 25.9%) and presented in Figure 8. The GO enrichment analysis of different expressed immune-related genes was carried out to further clarify the biological functions. The genes closely related to the BP terms were immune response and cell communication (Figure 9D). Moreover, extracellular, plasma membrane part, and sin3 complex were enriched in CC (Figure 9E). Additionally, protein-tyrosine kinase activity, cytokine activity and chemokine activity were mostly enriched in MF terms (Figure 9F). Furthermore, KEGG pathway enrichment analysis was significant enriched in RIG-I-like receptor, chemokine signaling pathway, microRNAs in cancer and chronic myeloid leukemia (Figure 9G).

Figure 9.

Identification of differentially expressed immune-related genes, constructed PPI network and GO enrichment analysis and KEGG pathway analysis. (A) Identification of differentially expressed immune-related genes, (B) PPI network of 10 genes was constructed in STRING database, (C) top 8 hub genes selected by the CytoHubba, (D) GO terms of BP category, (E) CC category, and (F) MF category, and (G) the significant enriched KEGG pathways.

Discussion

circRNAs are a cluster of long-stranded non-coding RNAs with a highly stable and conserved structure [31]. Acting as microRNA (miRNA) sponge, it can inhibit miRNA function and regulate its downstream target genes expression in a variety of cancers [10,32]. In addition, a bunch of studies have suggested that circRNAs could be found commonly expressed in tumor tissues and engaged in the regulation of cancer-related pathways [33,34]. While there has been just a small number of studies so far exploring the specific role of circRNAs in cSCC [35,36]. In this study, we identified 54 differentially expressed circRNAs in cSCC and found that their parental genes were most enriched in mitophagy, PPAR, HIF-1, and AMPK signaling pathway. Zhou et al. found that HOXA9 could inhibit the HIF-1α-mediated glycolytic pathway by interacting with CRIP2 to suppress the development of cSCC [37]. An et al. reveled that HIF-1α overexpression might play an crucial role in the occurrence of cSCC [38]. The above results suggested that the parental genes of DEcircRNAs might exert a significant role cSCC progression. However, there are little studies about cSCC and PPAR, AMPK signaling pathway. It is deserved further studied.

The hsa_circ_0070934 was up-regulated in cSCC of 54 differentially expressed circRNAs. Zhang et al. found that the circ_0070934 expression was higher in various cSCC cell lines than in normal human keratinocyte cell lines [39]. This result was similar to a previous study, which has revealed that circ_0070934 was expressed higher in the cSCC samples compared to the corresponding control samples [35]. An et al. also found that circ_0070934 was overexpressed in cSCC tissue and cell lines [36]. The above results suggested that circ_0070934 might serve as an early diagnostic biomarker of cSCC. Moreover, hsa_circ_0001955, hsa_circ_0070933, hsa_circ_0067772, and hsa_circ_0003528 were found up-regulated in cSCC. while hsa_circ_0022392, hsa_circ_0022383, hsa_circ_0005085, hsa_circ_0046449, hsa_circ_0072279, and hsa_circ_0000375 were most down-regulated, which was consistent with our source data research [35]. Moreover, zhang et al. found that hsa_circ_0001955 was expressed higher in hepatocellular carcinoma [40]. Ding et al. confirmed that hsa_circ_0001955 expression was largely increased in HCC cells and tissues than in corresponding normal controls [41]. Ding et al. identified that hsa_circ_0001955 was significantly up-regulated in colorectal cancer [42]. However, there was no relevant research in cSCC about it. Moreover, the other top differently expressed circRNAs in cSCC were little known about their expression and function. Further researches shall be implemented to elucidate the underlying mechanisms concerned.

MiRNAs belong to small RNAs and are 20 nucleotides in length. Its main function is to regulate the expression of mRNAs, and thereby regulating cell proliferation, apoptosis, and migration, and thus promoting tumor development [43,44]. Moreover, some researchers have found that circRNAs could competitively bind to miRNAs and regulate tumor progression [15,45]. Hao et al. reported that circ_0007534 could promote pancreatic ductal carcinoma progression through binding miR-892b and miR-625 [46]. Niu et al. showed that hsa_circ_0001829 could facilitate gastric cancer development via the miR-155-5p/SMAD2 axis [47]. To elucidate the complex mechanism of circRNAs, we predicted their target miRNAs and identified DEmiRNAs in cSCC patients. We obtained 21 differentially expressed miRNAs, including 11 up-regulated (hsa-miR-135b, hsa-miR-31, hsa-miR-130b, hsa-miR-21*, hsa-miR-766, hsa-miR-31*, hsa-miR-424, hsa-miR-18a, miR-H16, hsa-miR-455-5p, hsa-miR-7) and 10 down-regulated miRNAs (hsa-miR-30a*, hsa-miR-30a, hsa-miR-101, hsa-miR-497, hsa-miR-378, hsa-miR-214, hsa-miR-4324, hsa-miR-127-3p, hsa-miR-136, hsa-miR-1). Rock et al. observed that miR-21-5p and miR-31-5p expression were increased in invasive cSCC compared to normal tissue [48]. Olasz et al. discovered that miR-135b expression was upregulated in three cSCC cell lines [49]. Wang et al. demonstrated that microRNA-31 was highly expressed in cSCC and could regulate tumor biological processes [50]. Ma et al. indicated that miR-214 expression was low in both cSCC tissues and cells, and patients with higher miR-214 expression had a better prognosis [51]. Yamane et al. found that miR-214 was down-regulated in cSCC [52]. Shao et al. revealed that miR-30a-5p expression was reduced in SCL-1 and A431 cells [53]. Wei et al. also suggested that miR-497 was expressed lower in cSCC tissues and cells, and could promote the progression of cSCC [54]. The above results were consistent with our study. Then in order to further investigate circRNA-miRNA interactions. We intersected the miRNAs predicted by circRNAs and DEmiRNAs, and hsa-miR-136 and hsa-miR-766 were identified as the candidate miRNAs. Hsa-miR-136 was significantly down-regulated, and hsa-miR-766 was up-regulated in cSCC, which is consistent with our source data research [55]. Moreover, another study implied that miR-766 was upregulated in cSCC tissues and cells [56]. Sand et al. discovered that hsa-miR-766 was up-regulated in cSCC [55], which is consistent with our research.

A large amount of studies demonstrated that circRNAs could act as miRNA sponges and regulate miRNA-mediated gene expression, hence leading to the progression and development of tumors [57]. Chen et al. found that circ-ERBIN could contribute to the growth and metastasis of colorectal cancer by regulating miR-125a-5p and miR-138-5p [58]. Ma et al. discovered that hsa_circ_0004872 could inhibit gastric cancer progression by regulating miR-224/Smad4/ADAR [21]. To further evaluate the roles of DEcircRNAs in cSCC, the circRNA-microRNA-mRNA network was constructed. We found that the circRNA ceRNAs may provide new insight into the mechanism underlying cSCC. Gao et al. found that silencing circRNA_001937 might inhibit proliferation and induce apoptosis in cSCC by blocking sponging of the miRNA-597-3p/FOSL2 pathway [59]. Chen et al. revealed that circPVT1 might an important target for the treatment of cSCC [60]. Zhang et al. showed that hsa_circ_0070934 could regulate HOXB7 expression in cSCC by competitively sponging miR-1236-3p [39]. An et al. have found that circ_0070934 could promote the growth and invasion of cSCC cells by sponging miR-1238 and miR-1247-5p [36]. These studies indicated that the circRNA-related ceRNA network played an important role in the progression of cSCC. However, there is currently little research on it and our research has laid a good foundation for follow-up research.

GO and KEGG enrichment analyses were executed for the 36 genes in the network to study the biological functions of the circRNAs. Then we found the most enriched KEGG pathways were chemokine, RIG-I-like receptor signaling pathway, and microRNAs in cancer. Mittal et al. discovered that the CCR2/MCP-1 chemokine pathway played an important role in lung adenocarcinoma [61]. Zhu et al. found that chemokine signaling pathways were involved in immune infiltration and the tumor microenvironment of sarcoma [62]. Next, the upregulated-hub genes of USP18, OAS3, ISG15, CXCL10 were identified in cSCC compared with normal tissues and PI3, RSAD2 were downregulated. In addition, the GO analysis showed that gens were most enriched in immune response in biological process. The previous study also showed the immune system played a significant role in cSCC progression [23,63]. Then, 10 immune-related differentially expressed genes were obtained by overlapped with 36 differently expression genes in the ceRNA network and the lists of immune-related genes. The PPI network was constructed and the top hub genes were selected. CXCL10, ISG15 and IL7R were found to be upregulated in cSCC. PI3 and RSAD2 were downregulated. Moreover, Liu et al. also found that CXCL10 and ISG15 was a hub genes and might be potential targets for the diagnosis and therapy of patients with cSCC [64]. Toriseva et al. found that CXCL10 was associated with tumor progression in cSCC [65], which was consistent with our study.

In order to validate the above results and explore the biological functional implication of hub differentially expressed immune-related genes, the GO analysis found that they were mostly associated with immune response in the biological process and the most significant KEGG pathways were chemokine signaling pathway, microRNAs in cancer, and RIG-I-like receptor signaling pathway, which were consistent with the above results. These results suggest that circRNAs might compete binding to miRNAs to regulate the expression of immune-related genes to promote tumor progression. Liu et al. concluded that circRNAs could competitively bind PKR so that interfere the cellular immune signaling pathways [26]. Ng WL et al. revealed that macrophages could be activated under certain conditions and could produce circRNAs which could influence the immune functionality of macrophages [66]. Furthermore, circ-ANRIL might be involved in the apoptosis of macrophages and inhibit the proliferation of them [67]. It is clear that circRNAs could play a role in the activation and function of immune cells and influence tumor progression. Our study provides a good basis for further research on the immunomodulatory effects of circRNA on cSCC and related regulatory mechanisms.

Conclusion

Our study identified the DEcircRNAs, DEmiRNAs, DEmRNAs in cSCC and established the ceRNA network. Then, the immune-related genes were identified based on mRNAs in the ceRNA network and hub immune-related genes were validated by immunochemistry, which might act as potential diagnostic biomarkers, therapeutic targets, as well as new immune checkpoints for patients with cSCC.

Disclosure of conflict of interest

None.