Transcriptome Analysis Reveals Hub Genes Regulating Autophagy in Patients With Severe COVID-19

Background: The COVID-19 pandemic has currently developed into a worldwide threat to humankind. Importantly, patients with severe COVID-19 are believed to have a higher mortality risk than those with mild conditions. However, despite the urgent need to develop novel therapeutic strategies, the biological features and pathogenic mechanisms of severe COVID-19 are poorly understood. Methods: Here, peripheral blood mononuclear cells (PBMCs) from four patients with severe COVID-19, four patients with mild COVID-19, and four healthy controls were examined by RNA sequencing (RNA-Seq). We conducted gene expression analysis and Venn diagrams to detect specific differentially expressed genes (DEGs) in patients with severe disease compared with those with mild conditions. Gene Ontology (GO) enrichment analysis was performed to identify the significant biological processes, and protein–protein interaction networks were constructed to extract hub genes. These hub genes were then subjected to regulatory signatures and protein–chemical interaction analysis for certain regulatory checkpoints and identification of potent chemical agents. Finally, to demonstrate the cell type-specific expression of these genes, we performed single-cell RNA-Seq analyses using an online platform. Results: A total of 144 DEGs were specifically expressed in severe COVID-19, and GO enrichment analysis revealed a significant association of these specific DEGs with autophagy. Hub genes such as MVB12A, CHMP6, STAM, and VPS37B were then found to be most significantly involved in the biological processes of autophagy at the transcriptome level. In addition, six transcription factors, including SRF, YY1, CREB1, PPARG, NFIC, and GATA2, as well as miRNAs, namely, hsa-mir-1-3p, and potent chemical agents such as copper sulfate and cobalt chloride, may cooperate in regulating the autophagy hub genes. Furthermore, classical monocytes may play a central role in severe COVID-19. Conclusion: We suggest that autophagy plays a crucial role in severe COVID-19. This study might facilitate a more profound knowledge of the biological characteristics and progression of COVID-19 and the development of novel therapeutic approaches to achieve a breakthrough in the current COVID-19 pandemic.


INTRODUCTION
The current COVID-19 pandemic, caused by novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has led to urgent healthcare issues worldwide. According to the World Health Organization, 223 countries or regions had reported 456,797,217 confirmed cases of COVID-19 by 14 March 2022, including 6,043,094 deaths. The manifestations of COVID-19 vary, and most infected individuals have only mild symptoms similar to typical pneumonia or even no symptoms (Wu and McGoogan, 2020). Furthermore, mortality is mainly observed in patients with severe COVID-19 with severe respiratory failure associated with interstitial lung pneumonia and acute respiratory distress syndrome (Berlin et al., 2020). In countries that did not implement active control measures, the case fatality rate of COVID-19 was as high as~10% (Iype and Gulati, 2020). However, treatment options are limited to symptomatic treatment to reduce the severity of symptoms, and no curative treatment is available. Moreover, in COVID-19, especially in the severe forms, the characteristics and effects of biological reactions are still poorly understood, which prompts researchers to search for better predictors of clinical outcomes and tools to provide information for developing new therapeutic targets and appropriate therapeutic measures. Transcriptome profiling by RNA sequencing offers sufficient gene expression analysis for characterizing COVID-19 and explains biological pathways and key genes that are not yet targeted by current therapies. In this way, Mahmud et al. (2021) identified transcription factor-gene interactions, protein-drug interactions, and DEG-miRNA coregulatory networks with differentially expressed genes (DEGs) for effective treatment of COVID-19. Auwul et al. (2021) identified that common gene signatures and pathways between COVID-19 and chronic kidney disease (CKD) could be therapeutic targets in COVID-19 patients with CKD as a comorbidity using the RNA sequencing (RNA-Seq) transcriptomic dataset of peripheral blood mononuclear cells (PBMCs) infected with SARS-CoV-2.
Autophagy refers to the process of sealing a part of the cytoplasm in the double-membrane autophagosome and delivering it to the lysosome for degradation; it is an essential cellular mechanism to cope with various stress conditions (such as starvation, energy deprivation, and pathogen invasion) and maintain a steady-state balance (Feng et al., 2014). As a monitoring mechanism, autophagy is also involved in resisting the foreign invasion of viruses. In response to viral infection, the autophagic activity is activated by host cells through virusencoded activators, cellular stresses provoked by infection, and sensing of viral constituents mediated by Toll-like receptors (TLRs) (Viret et al., 2018). As a defense mechanism during viral infection, the autophagic activity could deliver the virus or viral protein to the lysosome for degradation, transport viral nucleic acids and antigens to endolysosomal compartments for innate and adaptive immune responses, and regulate virusinduced cell death (Levine et al., 2011). SARS-CoV-2 is an enveloped, approximately 30 kb single-stranded RNA βcoronavirus . Several studies have demonstrated that infection with SARS-CoV-2 may be associated with autophagy. Miao et al. (2020) demonstrated that SARS-CoV-2 virus infection would block autophagy, resulting in the accumulation of autophagosomes and causing late endosomal sequestration of the homotypic fusion and protein sorting (HOPS) component VPS39. In contrast, Hui et al. (2021) provided evidence that SARS-CoV-2 promotes autophagy to suppress type I interferon response.
Here, to explore the biological characteristics and progression in patients with severe COVID-19 as opposed to those with mild COVID-19, we first pre-processed raw data on GSE167930 and screened specific DEGs for severe COVID-19. Gene Ontology analysis of these DEGs was performed to gain knowledge regarding their biological processes. Subsequently, we examined the most significant term and performed protein-protein interaction (PPI) network analysis to extract the hub genes regulating autophagy. Furthermore, we identified transcription factors (TFs) and microRNAs (miRNAs) at regulatory checkpoints using these hub genes. We then analyzed the protein-chemical interaction network, to determine potent chemical agents. Finally, to determine the cell type-specific expression of these genes, we performed singlecell RNA-Seq analysis using an online dataset. The sequential workflow of the processes in this study is shown in Figure 1.

Sample Collection, Data Processing, and Differential Expression Analysis
The study cohort comprised peripheral blood mononuclear cells from four mild COVID-19 patients, four severe COVID-19 patients, and four healthy controls. All samples were subjected to RNA-Seq analysis, and the results could be obtained from the GEO database of the National Center for Biotechnology Information (NCBI). The GEO accession ID of the dataset was GSE167930, which was already deposited for early published article by our team (Zhou et al., 2021). The limma R package was used for RNA-Seq to identify significant DEGs (the cut-off value of fold change >2 and fold change <0.5; p-value < 0.05). To screen specific DEGs for severe COVID-19, we set two gene clusters. In cluster 1, DEGs were significantly expressed in severe COVID-19 patients compared with healthy controls. In cluster 2, DEGs were significantly expressed in mild COVID-19 patients compared Frontiers in Genetics | www.frontiersin.org July 2022 | Volume 13 | Article 908826 with severe COVID-19 patients. The Venn diagram of cluster 1 and cluster 2 was used to specifically distinguish DEGs associated with severe COVID-19 patients. This differential expression analysis was performed and figures were obtained using the SangerBox tools, a free online platform for data analysis (http://vip.sangerbox.com/).

Gene Ontology Enrichment Analysis
Gene Ontology (http://geneontology.org/) stores a database of gene annotations that participate in biological processes; it calculates the probability of obtaining at least as many genes with the observed annotations. DAVID (https://david.ncifcrf. gov) was used as a data source for Gene Ontology enrichment analysis of the 144 DEGs specifically expressed in severe COVID-19, and the significant enrichments were filtered based on p-value < 0.05 and FDR (q-value) < 0.05. The enrichment analysis was performed, and figures mentioned earlier were obtained using the SangerBox tools (http://vip.sangerbox.com/).

Protein-Protein Interaction Network Analysis and Hub Gene Cluster Identification
The genes involved in the crucial biological process of severe COVID-19 were included in the STRING database (https://string-db.org/) (version 11.0) to construct a PPI network. Cytoscape v.3.7.1 was then used for the visual presentation of the results from STRING. Hub gene cluster analysis was conducted using the Molecular Complex Detection (MCODE) plugin.

Transcriptional and Post-transcriptional Network Analysis
The hub genes involved in the crucial biological processes of severe COVID-19 were used to recognize its TF gene and gene-miRNA network using the JASPAR database and TarBase database v8.0 on the NetworkAnalyst platform, respectively. Subsequently, Cytoscape v.3.7.1 was used to obtain a visual presentation of the TF gene and gene-miRNA interaction network.

Gene-Chemical Interaction Network Analysis
The Comparative Toxicogenomics Database in the NetworkAnalyst tool was further used to recognize the relationship of potential chemical agents and hub genes in the crucial biological process of severe COVID-19. For a visual presentation of the gene-chemical interaction network, Cytoscape v.3.7.1 was used.

Single Cell RNA-Seq Analysis
To indicate the cell type-specific expression of hub genes involved in autophagy in this study, we performed single-cell RNA-Seq analysis on a free online database platform: COVID-19 Cell Atlas Data Mining Site (http://www.covidcellatlas.com/) (Unterman et al., 2022). We set the comparison as COVID-19 Stable versus Progressive on the website, where "Stable" refers to patients hospitalized in internal medicine wards who eventually recovered and were discharged, that is, mild patients in our study, and "Progressive" refers to severe patients who required admission to the ICU and eventually succumbed to the disease. To show the identified cell types that express the gene, the UMAP Explorer was used for plotting the gene expression. We then imported the five most specifically expressed cell types into interactive connectome to explore the intercellular ligand-receptor pair interactions. All the figures were obtained from this online platform.

Identification of Differentially Expressed Genes Specific for Severe COVID-19 Patients
To identify the DEGs specific for severe COVID-19, we first compared genes expressed in severe COVID-19 to those expressed in healthy controls and set these 262 DEGs in cluster 1 (the cut-off value of fold change >2 and fold Frontiers in Genetics | www.frontiersin.org July 2022 | Volume 13 | Article 908826 change <0.5; p-value < 0.05). Following this, in cluster 2, a total of 1751 DEGs were significantly changed in mild COVID-19 patients compared with severe COVID-19 patients. As shown in the Venn diagram of cluster 1 and cluster 2, there were 144 DEGs specifically expressed in severe COVID-19, and these 144 DEGs were employed to accomplish further determination of biological process enrichment analysis ( Figure 2).

Gene Ontology Enrichment Analysis
To gain insight into the regulation of genes and the transmission of signals that occur during the progression of severe COVID-19, we performed Gene Ontology enrichment analysis. In biological process enrichment with DAVID at p-value < 0.05 and FDR (q-value) < 0.05, we found a total of 11 GO terms enriched significantly for the specific 144 DEGs in severe COVID-19 patients ( Figure 3). Table 1 presents the 11 GO terms based on the number of genes included. The biological process terms "Autophagy" (GO:0006914, p-value 0.0001, FDR 0.0327) and "Process utilizing autophagic mechanism" (GO:0006919, p-value 0.0001, FDR 0.0327) were considered the crucial biological process as they contained maximum 12 genes including ATM, CHMP6, EP300, RIPK2, ATP6V0E1, VPS37B, ATP6V1E2, PLEKHM1, ZC3H12A, STAM, MVB12A, and RALB.

Protein-Protein Interaction and Identification of Hub Genes Involved in Autophagy
To further acquire the core genes associated with autophagy in severe COVID-19, we conducted a PPI network of the total 12 genes involved in GO terms "Autophagy" and "Process utilizing autophagic mechanism" using STRING. In this manner, 12 nodes and eight edges were obtained with a local clustering coefficient of 0.667 and a PPI enrichment p-value of 0.00032 ( Figure 4B). The data file was calculated by MCODE, a novel Cytoscape plugin to identify significant gene clusters. Subsequently, we obtained only one gene cluster, which consisted of four nodes (MVB12A, CHMP6, STAM, and VPS37B) and six edges. Therefore, MVB12A, CHMP6, STAM, and VPS37B were regarded as hub genes regulating autophagy in severe COVID-19.

Determination of Regulatory Signatures
The selected hub genes as mentioned earlier (MVB12A, CHMP6, STAM, and VPS37B) were evaluated with TF gene and gene-miRNA interaction network analysis to detect transcriptional signatures and post-transcriptional regulatory signatures. Figure 5 shows the TF gene interaction network drawn by Cytoscape. The transcription factors, namely, SRF, YY1, CREB1, PPARG, and NFIC, linked with VPS37B and MVB12A and GATA2 connected with MVB12A and STAM. The gene-miRNA interaction network has 66 nodes and 72 edges ( Figure 6). Among these miRNAs, hsa-mir-26b-5p connected with CHMP6 and VPS37B, and hsa-mir-103a-3p and hsa-mir-107 were associated with CHMP6 and STAM. Subsequently, hsa-mir-124-3p and hsamir-191-5p were commonly linked with VPS37B, CHMP6, and STAM. Furthermore, we considered hsa-mir-1-3p to be the most pivotal miRNA as it was the common posttranscriptional factor for all the four hub genes.

Construction of the Gene-Chemical Interaction Network and Identification of Potent Chemical Agents
We used the four hub genes to analyze their interactions with different chemical agents in addition to using them for identifying potential antiviral agents associated with autophagy. Using the FIGURE 5 | Regulatory interaction network of TF gene was identified using the NetworkAnalyst tool. Herein, the circle nodes are genes (blue); the square nodes are TFs (green); TFs targeting more than two genes simultaneously are shown in orange.
FIGURE 6 | Regulatory interaction network of gene-miRNA was identified using the NetworkAnalyst tool. Genes are shown in blue; miRNAs are shown in pink; miRNAs targeting two genes are shown in yellow; miRNAs targeting three genes are shown in orange; miRNAs targeting four genes are shown in green.
Frontiers in Genetics | www.frontiersin.org July 2022 | Volume 13 | Article 908826 6 Comparative Toxicogenomics Database in the NetworkAnalyst tool, 40 chemical agents were predicted (Figure 7). Among these, thimerosal and valproic acid were linked with VPS37B and MVB12A, cyclosporine and phenobarbital were connected with VPS37B and CHMP6, and sodium selenite was associated with MVB12A and CHMP6. In addition, (+)-JQ1 compound was connected with VPS37B, CHMP6, and MVB12A, and arsenic was linked with STAM, CHMP6, and MVB12A. Most importantly, copper sulfate and cobaltous chloride were associated with all the four hub genes and could be considered to be the most relevant potent chemical agents.

Single-Cell RNA-Seq Analysis
An online single-cell RNA-Seq platform was used to assess the cell type-specific expressions of hub genes (MVB12A, CHMP6, STAM, and VPS37B). MVB12A was highly expressed in classical monocytes and effector T cells ( Figure 8B). CHMP6 was highly enriched in plates and effector T cells ( Figure 8C). STAM was highly expressed in Tregs and dying T & NK cells ( Figure 8D). VPS37B was highly expressed in dying T & NK cells and effector T cells ( Figure 8E). Subsequently, the interactive connectome tool in this platform was used to explore the intercellular ligand-receptor pair interactions between classical monocytes, effector T cells, plates, Tregs, and dying T & NK cells. Figure 8F shows that the ligands of effector T cells, plates, Tregs, and dying T & NK cells among PBMCs of severe COVID-19 patients are targeted to match the receptors of classical monocytes compared with the ligands of those cells of mild COVID-19 patients, and classical monocytes also secrete ligands to induce a cellular response through cognate receptors.

DISCUSSION
The COVID-19 pandemic is a major threat to a safe and healthy living environment of human beings and has resulted in more than 4 million deaths worldwide. It is a type of pneumonia, an infection with a virus named SARS-CoV-2 in the lungs. However, the condition is more complex. Severe COVID-19 strains display more aggressive symptoms, consequently resulting in a high mortality rate, especially the delta variant initially discovered in India in December 2020. Furthermore, the Omicron variant was discovered in South Africa a few months ago. Mild COVID-19 patients show moderate or even no symptoms. Therefore, identifying the characteristics of severe COVID-19 by comparison with mild COVID-19 would be more helpful for FIGURE 7 | Gene-chemical interaction network. Genes are colored in red; chemical agents are colored in purple; chemical agents targeting two genes simultaneously are shown in yellow; chemical agents targeting three genes simultaneously are shown in orange; chemical agents targeting all four genes simultaneously are shown in green.
Frontiers in Genetics | www.frontiersin.org July 2022 | Volume 13 | Article 908826 developing potential biomarkers and even new therapeutic targets. In this study, we first identified 144 specific DEGs in severe COVID-19 cases and performed the GO biological process enrichment analysis to acquire insight into the biological characteristics. Following this, we suggested that autophagy plays a key role in severe COVID-19, which corresponded to provide evidence in several studies. SARS-CoV-2 virus could block autophagy by infection or expression of ORF3a to sequestrate the HOPS component VPS39 and impaired the assembly of the STX17-SNAP29-VAMP8 SNARE complex (Miao et al., 2020). Hui et al. (2021) reported that SARS-CoV-2 M protein induced mitophagy to block the downstream innate immunity signaling for inhibiting the type I IFN response. Thus, autophagy may crucially contribute to the SARS-CoV-2 viral lifecycle. Assessment of the PPI network is considered a key pattern of protein affiliation and interaction. A total of 12 genes in the biological process of autophagy were involved in PPIs and the determination of hub genes. Here, four hub genes including MVB12A, CHMP6, STAM, and VPS37B were considered to be involved in the core regulation of autophagy in severe COVID-19 cases. Among these, Multivesicular Body Subunit 12A (MVB12A) is a component of the endosomal sorting required for transport I (ESCRT-I) complex. Its depletion and overexpression inhibit HIV-1 infectivity by inducing aberrant virion morphologies and altering viral Gag protein processing (Morita et al., 2007). The charged multivesicular body protein 6 (CHMP6) gene is the core component of endosomal sorting required for the transport III (ESCRT-III) complex, which is considered essential for virallike particles (VLPs) and virion release (Kumar et al., 2016). The signal transducing adapter molecule (STAM) gene forms the endosomal sorting complex required for transport-0 (ESCRT-0), and vacuolar protein sorting-associated protein 37B (VPS37B) is a component of the ESCRT-I complex. All the four genes are components of the ESCRT complex, and several viruses take advantage of the ESCRT system for proliferation, budding, and transmission in infected cells (Ju et al., 2021;Meng et al., 2021). Consequently, we supposed the infection of SARS-CoV-2 may be allied to the ESCRT system. Subsequently, the hub genes specialized for autophagy in severe COVID-19 were selected to predict their potential function at transcriptional and post-transcriptional levels. A number of transcription factors were detected. The serum response factor (SRF) has been demonstrated to modulate asymmetrical cardiac myocyte hypertrophy by constituting an epigenomic switch balancing the growth of adult ventricular myocytes in width versus length . The transcription factor Yin-Yang 1 (YY1) played an essential role in apoptosis and angiogenesis, and its cardioprotective effects were associated with T helper 2 cytokine production and M2 macrophage polarization . In addition, cAMP responsive element binding protein 1 (CREB1) and its target genes identified by the recombinant canarypox vector ALVAC + Alum could augment immunogenicity and reduce the HIV-1 infection rate (Tomalka et al., 2021). Furthermore, NFIC is related to digestive system carcinoma ) (Liang et al., 2021) and regulates renal inflammation and renal fibrosis in patients with diabetic nephropathy . Eventually, hepatitis B virus x protein can interact with GATA binding protein 2 (GATA2) to influence the activity of the ST2 promoter. We detected several significant miRNAs as latent post-transcriptional factors. We believed hsa-mir-1-3p to be the most pivotal miRNA in the process of autophagy in severe COVID-19 as it was targeted by four hub genes, and hsamir-1-3p has been identified to have a relationship with COVID-19 in several studies (Sardar et al., 2020;Sarma et al., 2020). It can inhibit influenza A virus replication by targeting the supportive host factor ATP6V1A (Peng et al., 2018). In addition, hsa-mir-1-3p was related to tumors such as endometrial cancer, (Czerwiński et al., 2021), metastatic prostate cancer (Mukherjee and Sudandiradoss, 2021), and breast cancer (Yan et al., 2021). In addition, hsa-mir-124-3p and hsa-mir-191-5p were commonly linked with VPS37B, CHMP6, and STAM.
Hsa-miR-124-3p was considered a potential candidate for treating COVID-19 (Prasad et al., 2021) and regulating ACE2 networks (Wicik et al., 2020). Then, hsa-miR-191-5p showed an inhibitory effect on HIV-1 replication (Zheng et al., 2021); it is associated with cervical lesions and can serve as a non-invasive biomarker (Ning et al., 2021). Next, the chemical agents that may target the common hub genes have been detected using the Comparative Toxicogenomics Database. Among significant chemical agents, copper sulfate has been proposed as a locally applied fungicide, bactericide, and astringent in medical practice (https://go.drugbank.com/ drugs/DB06778). It may induce pulmonary fibrosis through EMT activation induced by the TGF-β1/Smad pathway and MAPK pathways (Guo et al., 2021). Moreover, copper sulfate has also been identified as a potential chemical agent in pathogenetic profiling of COVID-19 (Nain et al., 2021). Furthermore, cobaltous chloride is a chemical agent that has been found to have an application in certain insecticides and fungicides (https://www.britannica.com/ science/cobaltous-chloride). As the evidence of potent chemical agents in severe COVID-19 is indirective, their roles need to be further studied to be confirmed. However, although these critical factors lack experimental verification, the correlations to autophagy suggest that they play a role in the prognosis of severe COVID-19.
Eventually, for determining whether the localization of these genes regulates autophagy, we assessed the cell typespecific expressions of MVB12A, CHMP6, STAM, and VPS37B using an online single cell RNA-Seq database platform. The results showed that classical monocytes, effector T cells, plates, Tregs, and dying T & NK cells play roles in autophagy. Furthermore, classical monocytes exhibit a central role among the five cell types that constitute cellular communication because all ligands match their receptors. Monocytes are phagocytic innate immune cells in blood circulation and depending on their respective expressions of CD14 and CD16 are traditionally divided into classical monocytes (CD14 ++ CD16 − ), non-classical monocytes (CD14 + CD16 ++ ), and intermediate monocytes (CD14 ++ CD16 + ). In acute patients with severe COVID-19, the number of non-classical and intermediate monocytes is found to be significantly reduced, whereas circulating classical monocytes display clear signs of activation (Knoll et al., 2021). Vanderbeke et al. (2021)have demonstrated that classical proinflammatory monocytes (based on the expressions of S100A8, S100A9, and S100A12 markers) dominate COVID-19 immunopathology in most critical cases. The results also indicated that classical monocytes were the primary source of major COVID-19-mediating cytokines, including the monocyte chemoattractant CCL2 and its receptor CCR2, the neutrophil chemoattractant CXCL8, and TNF-α (Vanderbeke et al., 2021). In addition, the expression level of the monocyte chemoattractant CCR2, which is a classical monocyte, was higher than that of non-classical monocytes, and anti-CCR2 treatment improved the course of the disease in preclinical trials (Channappanavar et al., 2016). Thus, these results demonstrate a correlation between classical monocytes and COVID-19, which could contribute to the design of novel therapeutics for this pandemic. However, because the samples used in this experiment were already used before, our conclusions may be limited by direct experimental validation. There are also few studies reported on the relationship between COVID-19 and autophagy. To the best of our knowledge, this is the first study to propose that MVB12A, CHMP6, STAM, and VPS37B are crucial genes associated with autophagy of PBMCs in patients with severe COVID-19 as opposed to those with a mild condition. Classical monocytes may play a central role in this disease; accordingly, subsequent studies should deeply explore the insight into the relationship between autophagy and classical monocytes in severe COVID-19.

CONCLUSION
The present study highlights the potential specific pathogenic processes in severe COVID-19 relative to mild COVID-19 and identifies hub genes, regulatory components, and chemical agents that may help develop novel and efficacious clinical therapeutic targets. We first identified 144 specific DEGs in severe COVID-19 patients. Subsequently, using these DEGs, we identified autophagy as a critical biological process. Next, based on the PPI network, we identified the most significant gene cluster involving the hub genes of MVB12A, CHMP6, STAM, and VPS37B. Consequently, we determined that the most pivotal miRNA hsa-miR-1-3p may play a role at the regulatory level. Copper sulfate and cobaltous chloride were considered relevant potent chemical agents. Eventually, we reported that classical monocytes may play a central role in genes regulating autophagy in severe COVID-19 cases compared with mild ones. Overall, our findings will shed light on the knowledge regarding biological characteristics of severe COVID-19 cases, as well as help find novel therapeutic strategies enabling us to achieve breakthroughs in the current pandemic.