Identification of Significant Genes and Pathways for the Chronic and Subacute Cutaneous Lupus Erythematosus via Bioinformatics Analysis

Background Chronic cutaneous lupus erythematosus (CCLE) and subacute cutaneous lupus erythematosus (SCLE) are both common variants of cutaneous lupus erythematosus (CLE) that mainly involve the skin and mucous membrane. Oral mucosal involvement is frequently observed in patients of CLE. Despite that they have different clinicopathological features, whether there is a significant difference in pathogenesis between them remains unclear. Herein, we investigated specific genes and pathways of SCLE and CCLE via bioinformatics analysis. Methods Microarray expression datasets of GSE109248 and GSE112943 were both retrieved from the GEO database. Differentially expressed genes (DEGs) between CCLE or SCLE skin tissues and health controls were selected by GEO2R. Common DEGs were picked out via the Venn diagram software. Then, functional enrichment and PPI network analysis were conducted, and the top 10 key genes were identified via Cytohubba. Results Totally, 176 DEGs of SCLE and 287 DEGs of CCLE were identified. The GO enrichment and KEGG analysis of DEGs of SCLE is significantly enriched in the response to virus, defense response to virus, response to IFN-gamma, cellular response to IFN-γ, type I IFN signaling pathway, chemokine activity, chemokine receptor binding, NOD-like receptor signaling pathway, etc. The GO enrichment and KEGG analysis of DEGs of CCLE is significantly enriched in the response to virus, regulation of multiorganism process, negative regulation of viral process, regulation of lymphocyte activation, chemokine receptor binding, CCR chemokine receptor binding, NOD-like receptor signaling pathway, etc. The top 10 hub genes of SCLE and CCLE, respectively, include STAT1, CXCL10, IRF7, ISG15, and RSAD2 and CXCL10, IRF7, IFIT3, CTLA4, and ISG15. Conclusion Our finding suggests that SCLE and CCLE have the similar potential key genes and pathways and majority of them belong to IFN signatures and IFN signaling pathway. Besides, the NOD-like receptor signaling pathway might also have an essential role in the pathogenesis of SCLE and CCLE. Together, the identified genes and signaling pathways have enhanced our understanding of the mechanism underlying the occurrence and development of both SCLE and CCLE.


Introduction
Lupus erythematosus (LE), as a chronic autoimmune disorder, is induced by an interplay of genetic, hormonal, and environmental factors and characterized with a wide range of clinical forms from skin and mucosal lesions to lifethreatening systemic manifestations. Oral involvement ranges from 9% to 45% in patients of SLE and 3% to 25% in patients of CLE [1,2]. Among CLE patients, higher cutaneous disease activity might have a significant correlation with the following oral manifestations: discoid plaques, cobblestone, and red/brown-pigmented macules. Additionally, the presence of gingivitis was associated with systemic inflammation [3]. According to the clinical and histological features, CLE was generally categorized into acute CLE (ACLE), SCLE, and CCLE. The most common type of CCLE is discoid LE (DLE), which is characterized with erythema, telangiectasias, atrophy, and regression with scarring. Approximately 30% of DLE patients have the oral mucosal involvement. SCLE is characterized with annular and lesions in sun-exposed regions [4,5]. Oral ulcers are the common clinical presentation. Above all, oral involvement acts as an indicative role to the diagnosis and evaluation of CLE disease course.
For the individuals with a genetic predisposition, multiple environmental factors can activate innate and adaptive immune responses that contribute to the formation of CLE skin lesions. There is strong evidence that clarifies the function of cytotoxic T cell-mediated immune response against the epidermis with the result of keratinocyte death and nuclear antigens release [6]. Thereby, B cells might act as antigen-presenting cells initiating the activation of autoreactive T cell [7]. Additionally, it has been suggested that keratinocytes themselves may be involved with the selfperpetuating cycle of lesions by producing type 1 and 2 IFN and IFN-regulated proinflammatory cytokines [8,9]. SCLE and CCLE are the major two categories of CLE. Although they have different clinicopathological features, the specific pathogenesis of the two and whether there is a huge difference between them have not been completely clarified. Harrris et al. [10] have indicated that the gene expression of CLE subtypes differs from health control skin with the predominance of upregulated expression of type 1 IFN and T cell chemotactic genes. However, there is more similarity among the subtypes of CLE. In our study, we selected the gene microarray (GSE109248 and GSE112943) from the Gene Expression Omnibus databases (GEO) to screen and identify the DEGs of SCLE and CCLE compared with the healthy skin control. Then, enrichment analysis and protein-protein interaction (PPI) network analysis were performed to select possible pathways and key genes that participated into the pathogenesis of SCLE and CCLE. This study might provide a new understanding of pathogenesis of SCLE and CCLE.

Materials and Methods
2.1. Microarray Information. The microarray expression profiles of GSE109248 and GSE112943 were retrieved from NCBI-GEO (https://www.ncbi.nlm.nih.gov), a free public database of microarray profile. Titles related to SCLE and CCLE were screened, and the details were further evaluated. The datasets of GSE109248 and GSE112943 were both based on the GPL10558, Illumina HumanHT-12 V4.0 expression beadchip platform. GSE109248 collected 6 skin samples from the CCLE, 12 skin samples from the SCLE, and 14 control skin samples. GSE112943 collected 6 skin samples from the CCLE, 10 skin samples from the SCLE, and 10 control skin samples. The volcano plot analysis was performed by GraphPad Prism 9.0.

Data
Processing of DEGs. The DEGs between CCLE and SCLE skin samples and skin control samples were identified via GEO2R online tool via the standard jLogFCj > 2 and adjusted P value < 0.05 [11]. Then, the raw data in TXT were processed in Venn software online to select the common DEGs of CCLE and SCLE among the two datasets.
2.3. GO Enrichment and KEGG Pathway Analysis. Gene ontology (GO) analysis is a commonly applied method to define genes and its RNA or protein product to identify unique biological properties of high-throughput transcriptome or genome data. The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis is a collection of datasets managing the genomes, diseases, biological pathways, drugs, and chemical materials. We utilize the Database for Annotation, Visualization, and Integrated Discovery (DAVID 6.8, https://david.ncifcrf.gov) to perform the GO analysis and KEGG analysis in helping classify the DEGs (P value < 0.05). The GO enrichment analysis comprises biological processes (BP), cellular components (CC), and molecular functions (MF).

PPI Network Formation and Identification of Key Genes.
The PPI network of DEGs was created by Search Tool for the Retrieval of Interacting Genes (STRING; https://www .string-db.org). The top 10 hub genes were picked out via the Cytohubba, a plugin of Cytoscape software (Cytoscape, 3.9).

Identification of DEGs in SCLE and CCLE.
Through the GEO2R online tools, we picked out 288 DEGs of CCLE and 283 DEGs of SCLE from GSE109248 and 357 DEGs of SCLE from GSE112943 ( Figure 1). Then, we utilized the Venn diagram software to select the common DEGs in the two categories of DEGs from GSE109248 and GSE112943. Results indicated that a summary of 176 commonly expressed DEGs of SCLE was identified ( Figure 2). Then, we selected the 358 DEGs of SCLE from GSE112934 to perform the following analysis.

DEGs of GO and KEGG Pathway Analysis in SCLE and
CCLE. 288 DEGs of CCLE and 176 DEGs of SCLE were analyzed for GO and KEGG analysis by DAVID software. The BP, CC, and MF analysis indicated DEGs of defense response to virus, regulation of multiorganism process, negative regulation of viral process, regulation of lymphocyte activation, cornified envelope, secretory granule membrane, chemokine receptor binding, cytokine activity, CCR chemokine receptor binding, etc. The KEGG analysis indicated that DEGs of CCLE were significantly enriched in the viral protein interaction with cytokine and cytokine receptor, primary immunodeficiency, NOD-like receptor signaling pathway, etc. The BP and MF analysis demonstrated that DEGs of SCLE were significantly enriched in the response to virus, response to IFN-gamma, cellular response to interferon-gamma, type I interferon signaling pathway, MHC 2 Disease Markers class I protein binding, chemokine activity, chemokine receptor binding, and MHC protein binding. The KEGG analysis showed that DEGs of SCLE were significantly enriched in the cytokine-cytokine receptor interaction, cytosolic DNA-sensing pathway, NOD-like receptor signaling pathway, etc. (Tables 1 and 2 and Figure 3).

PPI Network Analysis and Identification of Key Genes.
The PPI network of DEGs was created by STRING to screen the most significant clusters of the DEGs. As revealed in

Discussion
CCLE and SCLE are both common variants of CLE that may occur independently or as clinical manifestations of SLE [12,13]. Concern about the appearance of cutaneous rashes leads  The pathogenesis of SLE is multifactorial always with the abnormal expression of molecules that are associated with the type I IFN signaling pathway [14]. In our study, the results of GO analysis revealed that the DEGs of SCLE were primarily enriched in the response to IFN-gamma, cellular response to IFN-gamma, and type I IFN signaling pathway.
Inappropriate activation of IFN family IFN and its immune regulatory pathway might act a significant role in the formation of LE as well as CLE. In skin disease of LE, the type I IFNs are believed to participate in amplifying the lesional inflammation by eliciting the IFN-inducible chemokines and recruiting potentially autoreactive T cells into skin lesions. Additionally, interplays between the chemokine receptor CXCR3 and its IFN-inducible ligands CXCL9 and CXCL10 induce lesional inflammation in this condition. This is consistent with our results that the top 10 hub genes of SCLE and CCLE both include CXCL10 [15,16]. Increasing evidence has recognized the anti-type I IFN receptor antibodies as a therapeutic strategy for the clinical treatment of SLE. Nico et al. once clarified the identity and similarities between skin and oral lesions and also demonstrated that oral LE was characterized with stronger expression of IFNgamma [17,18]. Above all, they might become a promising Additionally, the result of our study demonstrated that the top 10 genes of CCLE and SCLE mostly overlap that include CXCL10, IRF7, ISG15, OAS1, OAS2, RASD2, GBP, and IFIT3. 2 ′ -5 ′ -Oligoadenylate synthetase (OAS), a type of important antiviral proteins, is induced by IFN. The OAS gene family includes OAS1, OAS2, OAS3, and OASL genes, which have similar structure and function. OAS1, OAS2, and OASL were all found to be significantly overexpressed in PBMC of SLE using gene microarray. Moreover, their expression was supposed to become the indicators of  Disease Markers disease activity of SLE [19]. Interferon-stimulated gene 15 (ISG15), an ubiquitin-like protein, is one of the IFNinducible gene signatures. ISG15 also participated in the pathogenesis of SLE and is correlated with the disease activity at baseline [20]. Braunstein et al. [21] reported that OAS1, OASL, and ISG15 were found to have an increase in SCLE and DLE regardless of concomitant SLE. Interferon regulatory factor 7 (IRF7) is a type I IFNdependent key transcriptional regulator of immune responses and is present in an inactive form in the cytoplasm of normal cells when viruses infect cells. Although increasing studies have confirmed the function and role of genetic polymorphism of IRF7 gene in the formation of SLE, more researches are acquired in the human body to further clarify the precise relationship between IRF7 and SLE [22,23]. Radical Sadenosyl methionine domain-containing 2 (RSAD2) is a host protein with extensive antiviral activity. Previous studies have shown that RSAD2 is closely correlated with the immune response. Further qRT-PCR verification in mouse B lymphocytes showed that RSAD2 was still significantly upregulated in SLE B lymphocytes compared with normal mouse B lymphocytes [24,25]. The large GTPase human guanylatebinding protein 1 (GBP1) is a pivotal mediator of angiostatin effects of inflammation and is elicited by IFN-α and IFN-γ in endothelial cells (ECs). The relationship between GBP1 and LE is rarely reported. Similar to our study, a bioinformat-ics analysis conducted by Liu et al. [26] also found that GBP1 is one of the top 10 hub genes of SLE. Interferon (IFN) inducible gene 3 (IFIT3) is also elevated in SLE. Wang et al. [27] proposed that IFIT3 is one of the genes that elicits the cGAS-STING signaling pathway in SLE monocytes to be overactive. It may be acted as a treatment target to stop the type I IFN and other proinflammatory cytokine production via the cGAS-STING signaling pathway. However, the genes mentioned above have not been reported to have relationship with CLE that needs more studies.
The GO analysis of DEGs with SCLE and CCLE both indicated that the DEGs were mostly enriched into the chemokine receptor binding and NOD-like receptor signaling pathway. Chemokines are a category of chemotactic cytokines that participate in immune and inflammatory responses through the chemoattraction and leukocyte activation. Except CXCL9 and CXCL10 mentioned above, other members of the chemokine family include CXCR2, CXCR3, CXCR5, CCR2, CCR6, CCL2, and CCL17 that were also reported to have abnormal expression in CLE [28]. NODlike receptors (NLRs) are an evolutionarily conserved family of innate immune receptors that were initially supposed to be responsive to intracellular pathogens (bacterial wall components, damaged membrane, toxins, uric acids, etc.) and endogenous byproducts of tissue damage. Recently, some studies have revealed that NLRs also act crucial roles in   [29]. Yu et al. [30] indicated that compared with healthy controls the expression of NOD2 in plasmacytoid dendritic cells and monocytes of SLE patients was greatly increased. Bacterial exposure increased the expression of NOD2 in monocytes contributing to proinflammatory cytokine production by PBMCs that aggravated the condition of SLE. However, the limitations of our study could not be ignored. Firstly, some other important factors including age, race, and regions should be considered as well. Additionally, the potential key genes need further experimental verification by RT-qPCR in clinical samples, contributing to the final conclusions just speculative. Finally, the specific mechanisms in which these crucial genes play are not completely clarified that need more evidence.

Conclusions
The top hub genes of both CCLE and SCLE are almost induced by IFN, which support the significant role of IFN gene signatures and IFN signaling pathway in CLE. However, the hub genes like IRF7, IFIT3, and GBP1 are only reported to be related to SLE but not CLE. Although the hub genes like OAS1, OASL, and ISG15 and chemokine like CXCL10 are reported to have an abnormal expression in CLE, the specific roles of them in the pathogenesis of CLE are still not clarified and required more studies. The findings of our study also gave a new insight into the onset of CLE. The NOD-like receptor signaling pathway might provide a new therapeutic target pathway of CLE. Owing that our study is only a bioinformatics study, our results require further validation.