Abstract
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the
main cause of COVID-19, causing hundreds of millions of confirmed cases
and more than 18.2 million deaths worldwide. Acute kidney injury (AKI)
is a common complication of COVID-19 that leads to an increase in
mortality, especially in intensive care unit (ICU) settings, and
chronic kidney disease (CKD) is a high risk factor for COVID-19 and its
related mortality. However, the underlying molecular mechanisms among
AKI, CKD, and COVID-19 are unclear. Therefore, transcriptome analysis
was performed to examine common pathways and molecular biomarkers for
AKI, CKD, and COVID-19 in an attempt to understand the association of
SARS-CoV-2 infection with AKI and CKD. Three RNA-seq datasets
([49]GSE147507, [50]GSE1563, and [51]GSE66494) from the GEO database
were used to detect differentially expressed genes (DEGs) for COVID-19
with AKI and CKD to search for shared pathways and candidate targets. A
total of 17 common DEGs were confirmed, and their biological functions
and signaling pathways were characterized by enrichment analysis. MAPK
signaling, the structural pathway of interleukin 1 (IL-1), and the
Toll-like receptor pathway appear to be involved in the occurrence of
these diseases. Hub genes identified from the protein–protein
interaction (PPI) network, including DUSP6, BHLHE40, RASGRP1, and TAB2,
are potential therapeutic targets in COVID-19 with AKI and CKD. Common
genes and pathways may play pathogenic roles in these three diseases
mainly through the activation of immune inflammation. Networks of
transcription factor (TF)–gene, miRNA–gene, and gene–disease
interactions from the datasets were also constructed, and key gene
regulators influencing the progression of these three diseases were
further identified among the DEGs. Moreover, new drug targets were
predicted based on these common DEGs, and molecular docking and
molecular dynamics (MD) simulations were performed. Finally, a
diagnostic model of COVID-19 was established based on these common
DEGs. Taken together, the molecular and signaling pathways identified
in this study may be related to the mechanisms by which SARS-CoV-2
infection affects renal function. These findings are significant for
the effective treatment of COVID-19 in patients with kidney diseases.
Keywords: SARS-CoV-2, acute kidney injury, chronic kidney disease,
differentially expressed genes, gene ontology, protein-protein
interaction, hub gene, drug molecule
1. Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a novel
coronavirus that belongs to the Coronaviridae family and
Pisoniviricetes class. SARS-CoV-2 has been found to cause severe
respiratory problems when infecting the respiratory tract and is the
main cause of COVID-19 ([52]1, [53]2), which was estimated to have
resulted in 18.2 million deaths worldwide during the pandemic in 2020
and 2021 ([54]3). COVID-19 was initially deemed a febrile respiratory
disease, but increasing evidence suggests that it is a complex
multisystem disease ([55]4, [56]5). Indeed, COVID-19 patients often
exhibit manifestations of renal involvement in addition to respiratory
symptoms ([57]6). Acute kidney injury (AKI) is a common complication of
COVID-19 that increases mortality, especially in intensive care unit
(ICU) settings. Patients with chronic kidney disease (CKD) have a high
risk of SARS-CoV-2 infection and COVID-19-related mortality
([58]7–[59]9).
AKI is the second most common complication in critically ill COVID-19
patients and is characterized by elevated serum creatinine, renal
inflammation, and tubular necrosis. Epidemiologically, the incidence of
AKI in COVID-19 patients is variable and depends on the severity of
COVID-19, ranging from 10.5% to 37% ([60]10). The pathophysiology of
COVID-19-associated AKI is complex, and an increasing number of studies
suggest that factors such as systemic inflammation and immune
responses, activation of coagulation pathways, the renin–angiotensin
system, and endothelial injury are involved in the process of renal
damage that occurs in COVID-19 ([61]9, [62]11, [63]12). Early reports
indicated underlying CKD as a risk factor for COVID-19 severity and
mortality ([64]8, [65]9). The largest study included data from 17
million electronic health records and identified CKD as a risk factor
for mortality in COVID-19 patients, with glomerular filtration rate
(GFR)<30 ml/min/1.73 m^2 and organ transplantation conferring a high
risk in multivariate analyses ([66]13). Additionally, a nationwide
study in a US dialysis center reported higher seroprevalence of
SARS-CoV-2 antibodies than in the general US population ([67]14). AKI
and CKD are often considered two separate stages of the same disease
class ([68]15, [69]16). Although most COVID-19 patients have improved
renal function at discharge, the complex renal damage mechanisms of
COVID-19 and the use of nephrotoxic drugs and mechanical ventilation
during hospitalization suggest that further investigation is required
to determine the long-term prognosis of renal function in COVID-19
patients ([70]17–[71]19).
The exact mechanism of SARS-CoV-2-related renal damage is not known.
The main binding site for SARS-CoV-2, i.e., angiotensin-converting
enzyme 2 (ACE2), is expressed at much higher levels in the kidney than
in the lung ([72]20–[73]22). ACE2 is expressed apically in primary
human airway epithelia ([74]23), and previous studies have demonstrated
that in COVID-19, pneumonia occurs as ACE2 levels increase in the cell
membrane. In connection with a viral infection, the density level of
ACE2 is extremely progressive in the lungs ([75]24). Single-cell RNA
sequencing analysis indicated that ACE2 is mainly expressed by
glomerular parietal epithelial cells and proximal tubular cells. Other
studies have suggested that SARS-CoV-2 can directly invade human kidney
organoids through the ACE2 receptor ([76]25). The infectivity of cells
depends on not only ACE2 expression but also the types of proteases
expressed. The cellular components required for virus entry into the
kidney, such as cellular cathepsin L (CTSL) and transmembrane serine
protease 2 (TMPRSS2), are also highly expressed, suggesting favorable
conditions for the presence of SARS-CoV-2 in the kidneys ([77]26). In
addition, SARS-CoV-2 contributes to an imbalance in the
renin–angiotensin–aldosterone system (RAAS) via ACE2, which may also
exert deleterious hemodynamic effects involved in lung and kidney
injury ([78]27). Moreover, SARS-CoV-2 may target lymphocytes because
they express ACE2, leading to lymphocyte activation, which consequently
results in lymphocyte death and decreased immune protection ([79]28).
In patients with CKD, especially those with diabetic kidney disease
(DKD), baseline downregulation of ACE2 and upregulation of ACE, a
combination of proinflammatory and profibrotic states in the kidneys,
might lead to CKD progression ([80]11, [81]29). Therefore, the human
kidney is a main target for SARS-CoV-2 infection, and it is necessary
for researchers to further explore the complicated interactions between
SARS-CoV-2 infection, AKI, and CKD.
In this study, three datasets were used to explore the biological
relationship between COVID-19, AKI, and CKD. These datasets were
collected from the Gene Expression Omnibus (GEO) database, with
[82]GSE147507, [83]GSE1563, and [84]GSE66494 being used for COVID-19,
AKI, and CKD, respectively. First, differentially expressed genes
(DEGs) were confirmed using these datasets, and then common DEGs for
the three diseases were identified and served as the main experimental
genes for the entire study. These common DEGs were utilized for further
experiments and analyses, including pathway and enrichment analyses, to
understand the biological processes of genome expression studies.
Extracting hub genes from common DEGs is essential for potential drug
prediction, and a network of protein–protein interactions (PPIs) was
also constructed via common DEGs to collect hub genes. Transcriptional
regulators were also explored based on the common DEGs of
[85]GSE147507, [86]GSE1563, and [87]GSE66494, and potential drugs are
suggested ([88] Figure 1 ).
Figure 1.
[89]Figure 1
[90]Open in a new tab
This diagram illustrates the overall workflow of the study. The author
first found the common differentially expressed genes of COVID-19, AKI,
and CKD and then analyzed the enriched functions, pathways, PPI
networks, transcription factors and miRNAs, related diseases, and
potential drugs of these differential genes. The three datasets,
[91]GSE1563, [92]GSE66494, and [93]GSE147507, in the figure represent
the datasets of AKI, CKD, and COVID-19, respectively. AKI, acute kidney
injury; CKD, chronic kidney disease; PPI, protein–protein interaction.
2. Materials and methods
2.1. Datasets employed in this study
To identify common genetic interactions among SARS-CoV-2, AKI, and CKD,
microarray, and RNA-seq data were obtained from the National Center for
Biotechnology Information (NCBI)
([94]https://www.ncbi.nlm.nih.gov/geo/) GEO database ([95]30). The
SARS-CoV-2 dataset (GEO accession ID: [96]GSE147507) involves
transcriptional analysis of COVID-19 lung biopsies for respiratory
infections using the Illumina NextSeq 500 platform for high-throughput
sequencing. The AKI dataset (GEO accession ID: [97]GSE1563) comprises
human kidney tissue containing nine normal renal tissue samples and
five AKI renal samples (samples from transplant patients with renal
dysfunction without rejection), which were sequenced by Affymetrix
Human Genome U95 Version 2 Array ([98]31). The CKD dataset (GEO
accession: ID [99]GSE66494) was obtained from eight subjects with
normal renal function and 54 CKD subjects ([100]32); Agilent-014850
Whole Human Genome Microarray 4x44K G4112F was used to measure gene
expression.
2.2. Identification of DEGs and mutual DEGs among AKI, CKD, and COVID-19
Genes are defined as distinctively expressed when statistically
significant differences exist between the different levels of
transcripts tested ([101]33). DEGs for the acquired datasets
[102]GSE147507, [103]GSE1563, and [104]GSE66494 were first identified
from long expression values with the LIMMA package and
Benjamini–Hochberg calibration to control for the false discovery rate
and DESEq2 in the R programming language (v 4.0.2) for multiple test
options. Significant DEGs in the dataset were detected by cutoff
criteria (p-value<0.05 and |logFC| ≥ 1.0), and mutual DEGs were
obtained for [105]GSE147507, [106]GSE1563, and [107]GSE66494 by the
online VENN analysis tool Jvenn.
2.3. Gene Ontology and pathway enrichment analyses
The purpose of gene set enrichment analysis is to identify common
biological insights, such as biological processes or chromosomal
locations related to different diseases ([108]34). Gene Ontology,
functional enrichment, and pathway enrichment studies were performed
with EnrichR ([109]https://maayanlab.cloud/Enrichr/) to characterize
the biological mechanisms and signaling pathways of the shared DEGs.
Kyoto Encyclopedia of Genes and Genomes (KEGG), WikiPathways, and
BioCarta were also used to identify shared pathways between AKI-, CKD-,
and COVID-19-related metabolic processes. The top pathways were
selected based on p-value<0.05.
2.4. Protein–protein interaction network analysis
AKI, CKD, and COVID-19 functional and physiological interactions were
mapped with STRING ([110]https://string-db.org/) (version 11.0). PPIs
were examined using channels such as text mining, experimental
databases, coexpression, culture, gene fusion, and co-occurrence under
different settings of classification confidence scores (low, medium,
and high) ([111]35). Then, a medium confidence score of 0.5 was set to
generate PPI networks for common DEGs. Cytoscape (v.3.7.1) was used to
visualize PPIs, genetic interactions, protein–DNA interactions, and
other types of interactions ([112]36).
2.5. Hub gene extraction and submodule analysis
Nodes, edges, and connections exist between PPI networks, and nodes
with high levels of cross-linking can be considered hub genes.
CytoHubba ([113]http://apps.cytoscape.org/apps/cytohubba) is a
Cytoscape plug-in for ranking and extracting the key or potential
targeted elements of biological networks based on various network
characteristics. There are 11 methods for studying networks from
different perspectives in cytoHubba, among which maximal clique
centrality (MCC) is the best ([114]37). By using the MCC method, the
top 15 central genes were identified from the PPI network. The shortest
possible paths between central genes were classified according to the
closest neighboring feature of cytoHubba.
2.6. Recognition of TFs and miRNAs interacting with common DEGs
A transcription factor (TF) is a protein that binds to specific genes
and controls the rate at which genetic information is transcribed.
Therefore, TFs are crucial for molecular profiling. Topologically
plausible TFs that tend to bind to our common DEGs were identified in
the JASPAR database via the NetworkAnalyst platform. JASPAR is an
openly available resource that collects profiles of TFs for numerous
species in six taxonomic groups ([115]38). NetworkAnalyst is an online
platform for meta-analyzing gene expression data and obtaining insight
into biological mechanisms, roles, and explanations. Furthermore,
miRNAs targeting gene interactions are included to track the
detrimental effects of miRNAs that target gene transcripts to affect
protein expression ([116]39). Both TarBase and mirTarBase are
experimental validity databases for miRNA–target gene interactions
([117]39, [118]40). MiRNAs interacting with common DEGs were obtained
from TarBase and miRTarBase through miRNA–gene interactions from
NetworkAnalyst. Topological analysis was performed by Cytoscape, and
TF–gene and miRNA–gene interaction networks were identified. Using this
tool, researchers can screen miRNAs with high rankings and detect
biological functions and features to develop valid biological
hypotheses.
2.7. Evaluation of applicant drugs and molecular docking
Through Enrichr, drug molecules were identified using the drug
signature database (DSigDB) in relation to COVID-19, AKI, and CKD.
Enrichr is a popular portal with a large number of different gene set
libraries for exploring genome-wide enrichment of gene sets ([119]41).
DSigDB is a global archive for the identification of targeted drugs
associated with DEGs ([120]42). DSigDB, which contains 22,527 gene
sets, can be accessed via Enrichr.
After drugs for the common DEGs were predicted by Enrichr, we
downloaded the MOL2 form of these drugs from the ZINC
([121]https://zinc.docking.org/) database (due to the lack of the MOL2
form of dimethyloxalylglycine in ZINC, we downloaded its SDF format
from PubChem ([122]https://pubchem.ncbi.nlm.nih.gov/)). We used
openBabel software to convert the MOL and SDF formats of these small
molecules to PDB formats. We downloaded the PDB format of DUSP6,
BHLHE40, RASGRP1, TAB2, ACE2 (the functional host receptor of
SARS-CoV-2), and 3CLpro (an enzyme necessary for SARS-CoV-2
replication) from Protein Data Bank ([123]https://www.rcsb.org/). We
used Autodock tools (version 1.5.4) to dock eight drugs and three
proteins and then visualized the results with PyMOL Molecular
Visualization System 2020 (PyMOL).
2.8. Molecular dynamics simulation
Based on the docking results for each protein and drug molecule, the
drug–protein complex with the lowest binding energy was used as the
initial structure for all-atom molecular dynamics simulations, and the
simulation was performed using AMBER 18 software. Before the
simulation, charges of the small molecules are calculated by the
Hartree-Fock (HF) SCF/6-31G* of the antechamber module and Gaussian 09
software. Afterward, drug molecules and proteins are described using
the GAFF2 small molecule force field and ff14SB protein force field,
respectively. Each system utilizes the LEaP module to add hydrogen
atoms to the system, add a truncated octahedral TIP3P solvent box at a
distance of 10 Å in the system, add Na^+/Cl^− to the system to balance
the charge of the system, and output the topology and parameter file.
Molecular dynamics simulations were performed using AMBER 18 software.
Before simulations, energy optimization of the system was carried out,
including the steepest descent method with 2,500 steps and the
conjugate gradient method with 2,500 steps. After the system energy
optimization was completed, the temperature of the system was raised
slowly from 0 to 298.15 K by 200 ps at a fixed volume and a constant
heating rate. Under the condition that the system maintained a
temperature of 298.15 K, a 500-ps NVT (isothermal isotropic) system
simulation was performed such that the solvent molecules were further
uniformly distributed in the solvent box. In the case of NPT
(isothermal and isobaric), a 500-ps equilibrium simulation of the
entire system was performed. Finally, under periodic boundary
conditions, the two composite systems were simulated by 4-ns NPT
(isothermal and isobaric) systems. During the simulation, the non-bond
cutoff distance was set to 10 Å, the particle mesh Ewald (PME) method
was used to calculate the long-range electrostatic interaction, the
SHAKE method was applied to limit the length of hydrogen atomic bonds,
and the Langevin algorithm ([124]43) was used for temperature control,
where the collision frequency γ was set to 2 ps-1. The system pressure
is 1 atm, the integration step is 2 fs, and the trajectory is saved
every 10 ps for subsequent binding energy calculations.
2.9. MM/GBSA binding free energy calculation
The free energies of binding between proteins and ligands in all
systems were calculated by the molecular mechanics generalized Born
surface area (MM/GBSA) method. In this study, the above molecular
dynamics (MD) trajectory was used for calculation, and the specific
formula is as follows:
[MATH:
ΔGbind
=ΔGcomplex
−(ΔGreceptor<
/mtext>+ΔGl
igand),
:MATH]
[MATH:
=ΔEinternal+ΔEVDW+ΔE
elec+ΔGGB+ΔGSA :MATH]
where ΔE[internal] represents the internal energy, ΔE[VDW] represents
the van der Waals interaction, and ΔE[elec] represents the
electrostatic interaction. The internal energy includes the bond energy
(E[bond]), angular energy (E[angle]), and torsion energy (E[torsion]);
ΔG[GB] and ΔG[SA] are collectively referred to as solvation-free
energy. Among them, G[GB] is the free energy of polar solvation, and
G[SA] is the free energy of non-polar solvation. For ΔG[GB], we used
the GB model developed by researchers such as Nguyen ([125]44) for
calculation (igb = 2). The non-polar solvation free energy (ΔG[SA]) was
calculated based on the surface tension (γ) multiplied by the solvent
accessible surface area (surface area, SA), ΔG[SA]= 0.0072 × ΔSASA.
Entropy change was neglected in this study due to high computational
resource consumption and low precision.
2.10. Gene–disease association analysis
The DisGeNET project is a centralized database of gene–disease
interactions obtained from a variety of sources and features various
biomedical aspects of diseases. It highlights novel views of human
genetic disorders ([126]45). The network-analyst program was used to
study gene–disease associations to discover the relationship between
related diseases and chronic complications for the shared DEGs.
2.11. Construction of the COVID-19 diagnostic model
We used [127]GSE147507 expression matrix information to establish a
COVID-19 diagnosis model with fivefold cross-validation. We set 17
common DEGs as model key variables. Six different machine learning
algorithms (“extreme gradient boosting (XGBoost)”, “light gradient
boosting (LGBM)”, “RandomForest”, “Adaboost”, “support vector machine
(SVC)”, and “k-nearest neighbor (KNN)”) were employed for modeling. The
performance of each model was compared by a multimodel calibration
curve and the area under the curve (AUC), and the best model was
selected. After filtering out the best-performing models, we used the
“SHapley Additive exPlanations (SHAP)” package in Python to explain the
importance of key variables to the model and the contribution of each
variable.
2.12. Statistical analysis
DEGs for three GEO datasets were first identified from long expression
values with the LIMMA package and Benjamini–Hochberg calibration to
control for the false discovery rate and DESEq2 in the R programming
language (v 4.0.2) for multiple test options. Significant DEGs in the
dataset were detected by cutoff criteria (p-value<0.05 and |logFC| ≥
1.0).
Python software (version 3.7) was used to build the COVID-19 diagnostic
model. During the modeling of various machine learning algorithms, the
xgboost 1.2.1 package was applied to run the XGBoost algorithm, the
lightgbm 3.2.1 package to run the LightGBM algorithm, and the sklearn
0.22.1 package to run other machine learning algorithms. The shap
0.39.0 package was used to demonstrate model interpretability. All
statistical analyses in constructing the COVID-19 diagnostic model were
carried out with Python version 3.7 and the Extreme Smart Analysis
platform ([128]https://www.xsmartanalysis.com/).
3. Results
3.1. Identification of DEGs and common DEGs among COVID-19, AKI, and CKD
To discover the interrelationships and implications of AKI and CKD with
COVID-19, we analyzed human RNA-seq and microarray datasets from NCBI
to classify DEGs related to COVID-19, AKI, and CKD. We assessed the
RNA-seq and microarray dataset experiments in the R language
environment using the DESeq2 and limma packages with the
Benjamin–Hochberg false discovery rate. In total, we identified 2199
genes differentially expressed in COVID-19, and we also detected the
most significant DEGs for AKI and CKD: 200 in the AKI dataset and 5,211
in the CKD dataset. All significant DEGs were extracted on the basis of
p-value<0.05 and |logFC| ≥ 1. After performing cross-comparative
analysis with Jvenn, a reliable web portal for Venn analysis, 17 common
DEGs from the AKI, CKD, and SARS-CoV-2 datasets were identified,
including HBD, HBB, TANK, RNF6, TAB2, WTAP, PNRC1, ING3, TNFAIP8,
S1PR1, SEC24A, NRIP1, MARCKS, BHLHE40, DUSP6, EIF2AK2, and RASGRP1. The
expression levels of these 17 common DEGs based on the three datasets
are shown in heatmaps ([129] Supplementary Figure 1 ). However, the
upregulation and downregulation of these 17 DEGs in the cluster
heatmaps of the three diseases were not completely consistent. Overall,
these genes may be affected by certain pathways, resulting in
inconsistent upregulation and downregulation, and we will further
investigate how the upregulation and downregulation of these genes are
affected in future studies. The three diseases correlate with each
other because they share one or more common genes ([130]46) ([131]
Figure 2 ).
Figure 2.
[132]Figure 2
[133]Open in a new tab
This study incorporates two microarray datasets and one RNA-seq
dataset, which together encompass AKI ([134]GSE1563), CKD
([135]GSE66494), and SARS-CoV-2 ([136]GSE147507). This integrated
analysis identified 17 DEGs that are common to SARS-CoV-2, AKI, and
CKD. AKI, acute kidney injury; CKD, chronic kidney disease; SARS-CoV-2,
severe acute respiratory syndrome coronavirus 2; DEGs, differentially
expressed genes.
3.2. Gene Ontology and pathway enrichment analyses
Gene Ontology and pathway enrichment analyses were used to identify the
biological significance and enriched pathways for the shared DEGs. Gene
Ontology analysis is performed within three categories (biological
process, cellular component, and molecular function) ([137]
Figures 3A–C ); pathway analysis reveals the functional pathways in
which genes are enriched. The most affected pathways of the DEGs common
to AKI, CKD, and COVID-19 were gathered from three global databases,
including KEGG, WikiPathways, and BioCarta. The top 10 pathways in
WikiPathways include the structural pathway of interleukin 1 (IL-1),
MAPK signaling pathway, TNF-α signaling pathway, vitamin D receptor
pathway, mammary gland development pathway–puberty (Stage 2 of 4),
circadian rhythm-related genes, small ligand GPCRs, serotonin receptor
2 and ELK-SRF/GATA4 signaling, transcription factor regulation in
adipogenesis, and signal transduction of the S1P receptor. The top 10
pathways in KEGG were the MAPK signaling pathway, measles, protein
processing in the endoplasmic reticulum, NOD-like receptor signaling
pathway, pathogenic Escherichia coli infection, Epstein–Barr virus
infection, lipid and atherosclerosis, coronavirus disease, circadian
rhythm, and African trypanosomiasis. The top 10 pathways in BioCarta
include the Toll-like receptor pathway, regulation of MAP kinase
pathways through dual specificity phosphatases, regulation of elF2,
TNFR2 signaling pathway, hemoglobin’s chaperone, effects of calcineurin
in keratinocyte differentiation, double-stranded RNA-induced gene
expression, TNF/stress-related signaling, phospholipids as signaling
intermediaries, and signal transduction through lL1R ([138]
Figures 4A–C ).
Figure 3.
Figure 3
[139]Open in a new tab
The ontological bar graphs of the DEGs that are shared among
SARS-CoV-2, AKI, and CKD using the Enricher online tool. The GO
function is divided into three parts: (A) biological processes, (B)
molecular function, and (C) cellular component. Each bar graph
represents a function in GO. DEGs, differentially expressed genes;
SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; AKI, acute
kidney injury; CKD, chronic kidney disease; GO, Gene Ontology.
Figure 4.
Figure 4
[140]Open in a new tab
Bar graphs showing pathway enrichment analysis of DEGs shared by
SARS-CoV-2, AKI, and CKD as performed by Enricher: (A) KEGG 2019 human
pathway, (B) WikiPathways, and (C) BioCarta. Each bar represents a
pathway in KEGG/WikiPathways/BioCarta. DEGs, differentially expressed
genes; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2;
AKI, acute kidney injury; CKD, chronic kidney disease; KEGG, Kyoto
Encyclopedia of Genes and Genomes.
3.3. Classification of hub proteins and submodules
We carefully checked the PPI network from STRING and visualized it in
Cytoscape to predict common DEG interactions and related pathways. The
majority of interconnected nodes are considered hub genes of a PPI
network. Based on PPI network analysis incorporating the cytoHubba
plugin in Cytoscape, we classified the top 4 DEGs as the most
influential genes; DUSP6, BHLHE40, RASGRP1, and TAB2 were detected as
hub genes ([141] Figure 5A ).
Figure 5.
[142]Figure 5
[143]Open in a new tab
(A) PPI network with nodes representing DEGs and edges representing
interactions between nodes among SARS-CoV-2, AKI, and CKD. (B)
Determination of hub genes from the PPI network by using the cytoHubba
plugin in Cytoscape. The latest MCC procedure of the cytoHubba plugin
was pursued to obtain hub genes. Here, the red nodes indicate the
highlighted top 4 hub genes and their interactions with other
molecules. PPI, protein–protein interaction; DEGs, differentially
expressed genes; SARS-CoV-2, severe acute respiratory syndrome
coronavirus 2; AKI, acute kidney injury; CKD, chronic kidney disease;
MCC, maximal clique centrality.
These hub genes are potential biomarkers, and the results may lead to
new therapeutic strategies to study diseases. As hub genes are
potential markers, we also constructed a submodule network with the
cytoHubba plugin to better understand their near connectivity and
proximity ([144] Figure 5B ).
3.4. Determination of regulatory signatures
To determine substantial changes at the transcriptional level and
understand the hub proteins’ regulatory molecules and common DEGs, we
adopted a network-based approach to decode regulatory TFs and miRNAs.
From TF–gene and miRNA–gene interaction network analyses, it was
ascertained that 53 TF ([145] Figure 6 ) and 34 posttranscriptional
miRNA ([146] Figure 7 ) regulatory signatures are regulated by more
than one common DEG, indicating that they strongly interact with each
other.
Figure 6.
[147]Figure 6
[148]Open in a new tab
A regulatory interaction network of DEG–TFs derived from
NetworkAnalyst. Here, the square nodes represent TFs, and gene symbols
are circled as they interact with TFs. The larger the square or circle,
the more important the TFs or DEGs are in this network. DEG,
differentially expressed gene; TFs, transcription factors.
Figure 7.
[149]Figure 7
[150]Open in a new tab
The regulatory interaction network of DEGs and miRNAs. The square nodes
represent miRNAs, and gene symbols interact with miRNAs as circles. The
larger the square or circle, the more important the miRNAs or DEGs are
in this network. DEGs, differentially expressed genes.
3.5. Identification of candidate drugs
Evaluating protein–drug interactions is crucial to understand the
structural features recommended for receptor sensitivity ([151]46,
[152]47). Regarding common DEGs as potential drug targets in AKI, CKD,
and COVID-19, we identified eight possible pharmaceutical molecules
based on transcriptome signatures from the DSigDB database using
Enrichr. A list of the top 8 chemical compounds for three diseases
according to p-value and potential drugs for DEGs, including common
chemical compounds, is presented in [153]Table 1 .
Table 1.
List of the suggested drugs for COVID-19 with AKI or CKD.
Name Adjusted p-value Odds ratio Combined score
1 Tanespimycin ssMCF7 DOWN 0.000926 203.69 2809.91
2 Camptothecin MCF7 DOWN 0.001249 10.91 122.15
3 Niclosamide HL60 UP 0.001249 35.03 393.6
4 Camptothecin PC3 DOWN 0.001249 11.06 124.87
5 Pyrvinium MCF7 UP 0.001249 36.73 419.4
6 Staurosporine MCF7 DOWN 0.001249 16.41 188.24
7 Dimethyloxalylglycine PC3 UP 0.001249 94.94 1107.22
8 Sulpiride PC3 DOWN 0.001249 17.29 203.43
9 Daunorubicin MCF7 DOWN 0.001249 12.97 160.49
10 Niclosamide MCF7 UP 0.001249 51.36 652.36
[154]Open in a new tab
MCF and PC3 represent different cell lines. Adjusted p-value<0.05 has a
statistical difference.
AKI, acute kidney injury; CKD, chronic kidney disease.
The results of docking analyses are shown in [155]Supplementary Table 1
, in which lower binding energy indicates a more stable docking result.
The docking results for pyrvinium with BHLHE40, RASGRP1, and ACE2 are
the most stable, with binding energies of −7.46, 7.77, and −6.61
kcal/mol, respectively. The most stable drug molecules docked with
DUSP6 and TAB2 are tanespimycin and niclosamide, with binding energies
of −5.67 and −9.26 kcal/mol, respectively; the most stable drug binding
to 3CLpro is camptothecin, with a binding energy of −5.59 kcal/mol
([156] Figure 8 ).
Figure 8.
[157]Figure 8
[158]Open in a new tab
Results of molecular docking of drug and protein. Each figure shows the
overall picture of the docking of protein and drug molecules and the
enlarged picture of the docking part. (A) The docking diagram of DUSP6
with tanespimycin, with binding energy of −5.67 kcal/mol. (B) The
binding energy between ACE2 and pyrvinium is −6.61 kcal/mol. (C) 3CLpro
and camptothecin docking diagram, with binding energy of −5.59
kcal/mol.
3.6. MM/GBSA results
Based on the trajectory of the molecular dynamics simulation, we used
the MM/GBSA method to calculate binding energy, which can accurately
reflect the binding effect of a drug molecule and target protein.
As shown in [159]Supplementary Table 2 , the binding energies of drug
molecule ligands and proteins in the tanespimycin–DUSP6,
pyrvinium–RASGRF1, niclosamide–TAB, pyrvinium–BHLHE40, pyrvinium–ACE2,
and camptothecin–3CLpro systems were found to be −15.7857 ± 1.3991,
−27.6909 ± 0.9977, −14.8572 ± 0.5838, −21.5866 ± 0.9644, −28.5042 ±
1.4538, and −13.2160 ± 1.4146, respectively. Negative values indicate
that the two molecules have binding affinity for the target protein,
and lower values indicate stronger binding. Obviously, our calculations
show that all systems have the potential to bind, with the binding
affinities of pyrvinium–BHLHE40 and pyrvinium–TAB2 being significantly
lower than 20 kcal/mol, suggesting that the two complexes have better
binding effects. Through energy decomposition, we can determine that in
the pyrvinium–ACE2 complex, the electrostatic energy (EEL) has a strong
contribution; in contrast, the electrostatic energy has a weak
contribution in the tanespimycin–DUSP6 and camptothecin–3CLpro
complexes. The van der Waals energy (VDW) plays a role in all
combinations. In addition, the polar solvation energy of 24 is 331.4498
± 1.3050 kcal/mol, indicating that it is not conducive to binding, with
the non-polar solvation energy playing a weak role. The remaining polar
or non-polar solvation energy contribution of the other systems is not
significant and has little effect on binding.
3.7. Identification of disease association
Different diseases can correlate with each other and usually share one
or more similar genes ([160]46). Therapeutic design strategies for
combating disease have begun to uncover relationships between genes and
disorders ([161]48). According to NetworkAnalyst, studies have reported
an impaired sense of smell, heart failure, testicular hypogonadism, and
mood disorders associated with COVID-19. Persistent loss of smell or
taste without an obvious cause (e.g., typhoid) is called olfactory
failure. The most common causes of olfactory loss are allergic
sinusitis, nasal polyps, colds, and viral infection. Heart failure is a
syndrome of impaired cardiac circulation due to impaired systolic or
diastolic function, which is not an independent disease but rather the
end stage of various heart diseases, resulting in blood stagnation in
the venous system and inadequate perfusion in the arterial system. In
the majority of heart failure cases, the common initial manifestation
is pulmonary congestion. In addition, many COVID-19 patients experience
symptoms of renal tissue ischemia, which probably progresses to AKI or
even CKD ([162] Figure 9 ).
Figure 9.
[163]Figure 9
[164]Open in a new tab
The gene–disease association networks show diseases associated with
mutual DEGs. Diseases are represented by square nodes, and their
associated gene symbols are represented by circular nodes. The larger
the square or circle, the more important the diseases or DEGs are in
this network. DEGs, differentially expressed genes.
3.8. COVID-19 diagnostic model
The receiver operating characteristic (ROC) values of each machine
learning model for training and validation sets are shown in
[165]Supplementary Figures 2A, B , respectively. Calibration curves for
each model are shown in [166]Supplementary Figure 2C . XGBoost had the
highest AUC in both the training set and validation set at 1 and 0.792,
respectively ([167] Supplementary Figures 2A, B ). The calibration plot
in [168]Supplementary Figure 2C shows that the XGBOOST model was also
the most accurate. [169]Supplementary Tables 3 , [170]4 show that the
AUC, cutoff, accuracy, sensitivity, specificity, positive predictive
value, negative predictive value, and F1 score of XGBoost for the
training set were 1.000, 0.683, 0.984, 1.000, 1.000, 1.000, 0.978,
1.000, and 0.961, respectively. In brief, XGBoost was the
best-performing model, and we used it to build a diagnostic model for
COVID-19.
After filtering out the best-performing XGBoost model, we used the
“SHAP” package to explain the importance of key variables to the model.
[171]Supplementary Figure 2D shows the contribution of each variable,
with red dots indicating positive contributions and blue dots
indicating negative contributions. A shorter distance from the point to
the left indicates a smaller value and a larger value at a longer
distance. For example, a higher expression value of TANK predicts a
higher risk of COVID-19, whereas a lower value predicts a lower risk.
4. Discussion
AKI and CKD are currently considered to be two stages of renal disease
progression; the former is a common complication and mortality risk
factor in COVID-19 patients, and the latter is an independent risk
factor for COVID-19 and poor prognosis of COVID-19 ([172]49, [173]50).
Decreased GFR is strongly associated with the prevalence and mortality
of COVID-19, and chronic metabolic diseases resulting in CKD, such as
diabetes, hypertension, and obesity, are also related to COVID-19
mortality ([174]9). In this study, we collected three datasets and used
a computational network data analysis method to discover gene
expression patterns and molecular pathways of AKI, CKD, and COVID-19
and identify molecular targets of potential biomarkers, providing more
treatment options for different disease conditions ([175]51–[176]53).
By analyzing transcriptional profiles of SARS-CoV-2, AKI, and CKD to
identify genes with altered expression in SARS-CoV-2 infection
implicated in the pathogenesis of AKI and CKD, we report novel
interaction mechanisms. Seventeen common DEGs were revealed that showed
similar expression patterns in the three diseases and were evaluated by
Gene Ontology (GO) pathway analysis functions based on p-values to
acquire insight into the pathophysiology of AKI, CKD, and COVID-19.
GO involves a genetic adjustment context based on a general theoretical
model that promotes genes and their internal relationships.
Evolutionary studies have gradually provided biological knowledge of
genetic functions and their regulation in different ontological
categories ([177]54). From Enrichr, three categories of GO analysis,
namely, biological process (molecular activities), molecular function
(activities at the molecular level), and cellular component (genes that
regulate function), were evaluated through the GO database as a source
of annotation for ontological processes ([178]55). In the biological
process category, hydrogen peroxide catabolic and hydrogen peroxide
metabolic processes were among the top GO terms. Hydrogen peroxide has
emerged as a major redox metabolite that functions in redox sensing,
signaling, and redox regulation ([179]56). Hydrogen peroxide catabolism
contributes to limiting or repairing oxidative damage ([180]57).
SARS-CoV-2-infected individuals are susceptible to oxidative stress,
and their ability to resist oxidative stress may be associated with the
inflammatory status and may have little association with the severity
of the disease ([181]58). One study revealed that SARS-CoV-2 captures
iron and generates reactive oxygen species to injure the human immune
system while promoting the catabolism of hydrogen peroxide to oxygen
and water in phagocytes to reduce killing capacity ([182]59). Excessive
peroxide causes a renal oxidative stress response, inducing
mitochondrial metabolism and kinetic dysfunction and causing
inflammation and apoptotic cell death, which induce AKI and aggravate
CKD ([183]60). Chen et al. and Huang et al. found massive infiltration
of CD4+ T cells, CD56+ natural killer cells, and CD68+ macrophages in
the tubular stroma in the renal tissue of COVID-19 patients and that
activated T cells migrate to the location of infection to exert their
function ([184]61, [185]62). Under these conditions, SARS-CoV-2 may
promote necrosis or apoptosis of T cells by activating reactive oxygen
species metabolism, consequently hindering viral clearance, and excess
peroxide production can trigger the oxidative stress response in kidney
tissue, causing inflammation, cell death, and the deterioration of
renal function. Regarding molecular function, hemoglobin-α binding and
heme-binding activity were the two top GO pathways. Endothelial cell
expression of hemoglobin-α regulates nitric oxide signaling, impacting
blood perfusion and oxygen supply ([186]63). Kronstein-Wiedemann et al.
found that SARS-CoV-2 infects red blood cell progenitors and
dysregulates hemoglobin and iron metabolism, impairing hemoglobin
homeostasis and exacerbating COVID-19 ([187]64). It has also been
demonstrated that an abnormal hemoglobin phenotype is directly
associated with a decreased renal function ([188]65). Moreover, free
heme is a pro-oxidant that can disrupt homeostasis in vivo through
proinflammatory and cytotoxic effects ([189]66). AKI causes renal
hemopexin accumulation, potentially impacting heme Fe-mediated tubular
injury and leading to disease progression ([190]67). Therefore, it
cannot be ruled out that SARS-CoV-2 infection may initiate AKI by
disrupting hemoglobin metabolic homeostasis, which in turn can
aggravate this vicious cycle, leading to sustained progression of renal
function impairment.
Pathway analysis is a key step to reflect the internal reaction process
of an organism with a viral infection. KEGG, WikiPathways, and BioCarta
pathways of 17 common DEGs were identified to find similar pathways for
AKI, CKD, and COVID-19. Our analysis found that the MAPK signaling
pathway, the structural pathway of IL-1, and the Toll-like receptor
pathway may have pivotal roles in the occurrence mechanisms of these
three diseases. The MAPK signaling pathway activated in viral
infections links cell-surface receptors to the transcription machinery,
transducing extracellular signals into several outputs, which may also
affect the mechanisms of host defense and apoptosis ([191]68). A
variety of studies have demonstrated that the MAPK signaling pathway is
associated with cell injury, inflammation, and fibrosis, all of which
result in acute and chronic kidney diseases ([192]69–[193]73). Weckbach
et al. and Saheb et al. found that MAPK pathway activation is one of
the important mechanisms of organ inflammation in SARS-CoV-2 infection
and may affect sensitivity to steroid treatment ([194]74, [195]75). In
general, the MAPK pathway is vital for regulating organ inflammation
and function and is probably involved in the occurrence of
multiple-organ dysfunction in COVID-19 patients. COVID-19 is suggested
to involve a proinflammatory factor pattern similar to that of some
autoimmune diseases; therefore, a potential way to treat COVID-19 may
be by inhibiting increases in cytokine and chemokine levels ([196]76).
IL-1 binds to specific receptors, which leads to increases in
coreceptor and intracellular signal conduction, thereby inducing an
effective inflammatory response ([197]77). In the chronic inflammatory
mechanism underlying the progression of AKI to renal fibrosis, IL-1
signaling plays an important role ([198]78, [199]79). Following
secretion of chemokines and cytokines such as IL-1β, IL-6, TNF-α,
IL-21, and IL-8, the SARS-CoV-2-induced cytokine storm and
hyperinflammatory response have pivotal roles in infection severity,
AKI development, and death ([200]80). Bowe et al. even pointed out that
survival in COVID-19 somehow predisposes patients to worsening
subsequent long-term kidney function ([201]81). Toll-like receptors
(TLRs) are activated by foreign and host molecules to initiate the
immune response. TLR agonists are able to serve as a possible
therapeutic agent or a vaccine adjuvant for cancers or infectious
diseases; TLR inhibitors may be a promising approach to the treatment
of autoimmune diseases and bacterial and viral infections ([202]82). In
AKI caused by ischemia and reperfusion, researchers have discovered
that proximal tubule TLR4 expression is linked to inflammation and
apoptosis following hypoxia–reoxygenation injury ([203]83). Activation
of TLR4 signaling regulates the transcription of numerous
proinflammatory cytokines and chemokines, resulting in renal
inflammation ([204]84, [205]85). Therefore, the Toll-like receptor
pathway is involved in the pathogenesis of SARS-CoV-2 infection and
kidney diseases. Nevertheless, the mechanism by which SARS-CoV-2
triggers inflammation is not clear. Recently, a study discovered that
antibody-mediated SARS-CoV-2 uptake by monocytes and macrophages causes
inflammatory cell death that eliminates the production of infectious
viruses and results in systemic inflammation that contributes to
COVID-19 pathogenesis. This strong inflammatory effect may be the main
cause of severe illness and death ([206]86). The underlying
inflammatory pathways identified in these three diseases once again
demonstrate that inflammation is a significant mechanism by which
SARS-CoV-2 infection leads to damage in multiple organs.
Based on the analysis of DEGs, we established a PPI network showing
protein biology and predicting relevant drug targets at the proteomic
level and identified hub proteins expressed by topology metrics that
may serve as biomarkers or key treatment targets of COVID-19 and are
associated with various pathobiological mechanisms. The top hub
proteins represent different diseases, most of which are risk factors
for AKI, CKD, and COVID-19. The top 4 topological metric hub proteins
(DUSP6, BHLHE40, RASGRP1, and TAB2) are clearly involved in these
diseases. In this step, the cutoff (parameter) of the topological
metric for hub proteins is 15 (degree). DUSP6, a negative regulator of
the extracellular signaling-regulated kinase (ERK) signaling pathway,
is a broadly expressed dual-specificity phosphatase protein and has
roles in apoptosis inhibition and cellular protection ([207]87,
[208]88). Han et al. found that H[2]O[2] potentially promotes heart
regeneration in zebrafish by stimulating MAPK signaling through a
depression mechanism involving DUSP6 ([209]89). Missinato et al. also
suggested that DUSP6 attenuates Ras/MAPK signaling during regeneration
and that suppressing DUSP6 can enhance cardiac repair ([210]90). In
contrast, dual inactivation of DUSP4 and DUSP6 selectively impairs
growth in NRAS and BRAF mutant cells in cancer through hyperactivation
of MAPK signaling ([211]91). These studies demonstrate that DUSP6 plays
a vital role in tissue damage and repair by regulating hydrogen
peroxide metabolism and the MAPK signaling pathway. Moreover, in
diabetic nephropathy patients who have the highest prevalence of CKD,
DUSP6 has been found to mediate protection against high glucose-induced
inflammation ([212]92). Interestingly, Hsu et al. demonstrated that
DUSP6 also plays a positive role in the pathological process of
endothelial inflammation through TNF-α-induced endothelial
intercellular adhesion molecule-1 (ICAM-1) expression, a process that
is independent of ERK signaling ([213]93). Expression of ACE2 in
vascular endothelial cells provides the pathophysiological basis for
viral invasion. Histopathological examination of COVID-19 patients has
revealed that SARS-CoV-2 directly invades endothelial cells, causing
diffuse endothelial cell inflammation and microvascular damage, which
most likely leads to the failure of multiple organs, including the
kidneys ([214]94, [215]95). Hence, manipulation of DUSP6 holds great
potential for the treatment of acute inflammatory diseases, such as AKI
and COVID-19. There are more studies on DUSP6 in oncology,
demonstrating that its expression improves tumor proliferation and drug
resistance ([216]96–[217]99). The factor BHLHE40 has emerged as an
important regulator of immunity during infection, autoimmunity, and
inflammatory conditions, especially in cytokine production and
proliferation ([218]100). BHLHE40 also plays an important role in the
transcriptional regulation of immune cell infiltration ([219]101). As
mentioned above, the cytokine storm and infiltration of immune cells in
tissues and organs are pivotal causes of the aggravation and organ
dysfunction occurring in COVID-19. As an important immune regulator,
BHLHE40 is significant in regional and systemic inflammatory responses
to AKI, CKD, and SARS-CoV-2 infection. Feng et al. identified that
17β-estradiol (E2) regulates BHLHE40 expression to exert a protective
effect on carotid artery ligation and that upregulation of BHLHE40 in
vascular smooth muscle cells (VSMCs) results in suppression of MAPK
signaling ([220]102). One study found that BHLHE40 plays an important
role as a transcription factor in autoreactive T helper (Th) cell
pathogenicity. Lin et al. showed that BHLHE40 expression induced by the
IL-1 signaling pathway can identify encephalitogenic Th cells and
defines a pertussis toxin (PTX)-IL-1-BHLHE40 pathway active in
autoimmune neuroinflammation ([221]103). In addition, Camponeschi
et al. indicated that B-cell receptor (BCR) or TLR9 activation induces
expression of BHLHE40, a key negative regulator of activation-induced
proliferation of human B cells and highly expressed in anergic cells
([222]104). Therefore, as a regulator of many significant
immune-inflammatory signaling pathways, BHLHE40 participates in the
pathogenic process of immune-inflammatory diseases. RASGRP1 is an
important guanine nucleotide exchange factor and activator of the
RAS-MAPK pathway following T-cell antigen receptor (TCR) signaling, and
its deficiency causes immunodeficiency with impaired cytoskeletal
dynamics ([223]105). Moreover, Zhang et al. found that RASGRP1 mediates
TLR2-induced ERK1/2 activation and inhibition of IL-12p40 production,
which regulates TLR9 activation to induce an appropriate protective
IL-12 response ([224]106). By promoting lymphocyte proliferation,
RASGRP1 activity is also indispensable to autoimmunity ([225]107).
Thus, RASGRP1 activity is essential to the innate protective immune
response. In addition, a study discovered that its expression in
vascular endothelial cells maintains vascular health ([226]108). Based
on this evidence, RASGRP1 has great potential to become a pivotal
regulatory target of COVID-19 with AKI and CKD. Nuclear TAB2 is a
repressor of NF-κB-mediated gene regulation. The TAB2 protein is
expressed in the vascular endothelium of most tissues ([227]109), and
its downregulation has a significant effect in inhibiting the
inflammatory response and protecting tissue from acute injury, and it
can serve as a target of manipulation for multiple cytokines ([228]110,
[229]111). TAB2 gene may be one of the target genes for COVID-19
infection and organ injury. Taken together, we reveal that the top 4
hub genes are all involved in the regulation of microvascular
endothelial cell function. Many published studies support that
endothelial inflammation is the key mechanism promoting COVID-19
progression and multiorgan dysfunction. Therefore, the hub genes
identified in this study are potential biomarkers and therapeutic
targets for COVID-19.
Transcriptional and posttranscriptional modifications are important
aspects of epigenetics, influencing gene expression. Therefore, we
analyzed TF–gene and miRNA–gene interactions to identify the
transcriptional and posttranscriptional regulators of common DEGs. TFs
control transcriptional processes and proportions, and miRNAs play key
roles in gene regulation at the posttranscriptional level and in RNA
silencing. The discovery of relationships between DEGs, TFs, and miRNAs
is conducive to an understanding of the molecular-level progression of
diseases. The identified TFs, such as FOXC1, FOXL1, POU2F2, NFIC,
NFkB1, MEF2A, GATA2, and E2F1, are mainly associated with different
types of cancers and congenital disorders. DUSP6, BHLHE40, RASGRP1, and
TAB2, the top 4 topological metric hub genes, appear to be pivotal
molecular targets of COVID-19, AKI, and CKD. The TF–gene interaction
network indicates that FOXC1 is involved in BHLHE40, RASGRP1, and TAB2
expression but that FOXL1 only regulates TAB2. Additionally, POU2F2,
NFIC, NFkB1, and MEF2A regulate the expression of DUSP6 gene, and NFkB1
manipulates the transcription of BHLHE40 and TAB2 genes. GATA2 is
involved in TAB2 expression, and E2F1 regulates the expression of the
other three hub genes. In detail, FOXC1 and FOXL1 belong to the human
Forkhead-box (FOX) gene family, which is widely involved in cellular
activities ([230]112). For example, Koo et al. reported that FOXC1
appears to contribute to pathological angiogenesis by regulating
vascular endothelial growth factor signaling ([231]113). Additionally,
a study by Zhang et al. showed that FOXC1, as an ischemia-inducible TF,
upregulates the expression of TLR members in myocardial ischemia,
promoting cardiac inflammation and playing a detrimental role in
myocardial ischemia ([232]114). Although studies of FOX genes in
infection and kidney disease are scarce, current evidence suggests that
FOXC1 activates inflammation under hypoxia, which may have a regulatory
role in SARS-CoV-2 infection and renal dysfunction. POU2F2 is a member
of the POU transcription factor family and is involved in the immune
response by regulating B-cell proliferation and differentiation genes
([233]115). NFIC belongs to the family of transcription factors
involved in various morphogenetic processes during development
([234]116). A number of studies have found that NFIC controls cell
proliferation by regulating TGF-β1 signaling in adult regenerative
processes, such as tooth root development, hair follicle cycling, and
hepatocyte proliferation ([235]117–[236]119). Overall, the roles of
these transcription factors in kidney disease and infection have been
poorly investigated to date. MiRNAs are small non-coding RNAs that
serve as central players that regulate the posttranscriptional
processes of gene expression. They bind to target mRNAs and repress
their translation by inducing their degradation or inhibiting their
translation to control mRNA expression. RNA sequencing is becoming
popular in the postgenomic era, but high-throughput experimental
technologies for miRNA target identification are still expensive and
time-consuming. Therefore, an increasing number of bioinformatics
approaches are being developed for miRNA studies, especially for miRNA
target prediction. In this study, we successfully used bioinformatics
tools to accurately identify miRNAs targeting the DEGs of the three
diseases. Some of these miRNAs are closely related to regulating the
expression of the hub genes. For instance, hsa-mir-181a-5p,
hsa-mir-125b-5p, and hsa-mir-603 participate in the expression of the
DUSP6 gene; both hsa-mir-329-3p and hsa-mir-335-5p are associated with
BHLHE40 expression, and hsa-mir-125b-5p, hsa-mir-335-5p, and
hsa-mir-21-5p are involved in RASGRP1 expression; hsa-mir-181a-5p,
hsa-mir-155-5p, and hsa-mir-21-5p manipulate TAB2 expression.
Specifically, miRNA mutations often lead to the development of various
diseases. Some miRNAs are involved in lung cancer (e.g., hsa-mir-665,
hsa-mir-30a-5p, hsa-mir-150-5p, and hsa-mir-181a-5p)
([237]120–[238]123), immune disorders (e.g., hsa-mir-92a-3p,
hsa-mir-665, and hsa-mir-155-5p) ([239]124–[240]126), and different
types of chronic inflammation or infection (e.g., hsa-mir-483-3p,
hsa-mir-92a-3p, and hsa-mir-335-5p) ([241]127–[242]129). Most miRNAs
are related to cancer and congenital diseases, though some specific
miRNAs are related to the pathogenesis of AKI. For example, Zhang
et al. discovered that miR-181a-5p inhibits pyroptosis through the
downregulation of NEK7 in lipopolysaccharide (LPS)-induced HK-2 cells
and cecum ligation and puncture (CLP)-induced mice and indicated that
miR-181a-5p is a new potential therapeutic target for sepsis-induced
AKI therapy ([243]130). The research results of He et al. confirmed
that miR-122 directly targets vitamin D receptor (VDR) in renal tubular
cells, which strongly suggests that miR-122 upregulation contributes to
LPS-induced kidney injury by downregulating VDR expression ([244]131).
Moreover, hsa-mir-122-5p has been proven to regulate the ASF1A, BRWDM,
and PFKFB2 signaling pathways, a potential mechanism for the
development of AKI in transplanted kidneys ([245]132). As a novel small
molecule, miR-665-3p regulates autophagy by targeting ATG4B, indicating
that miR-665-3p inhibition is a potential therapeutic approach against
inflammation and apoptosis for the treatment of ischemia–reperfusion
([246]133). hsa-miR-483-3p is associated with diabetic renal vascular
injury and lupus nephritis ([247]134), and hsa-mir-186-5p is involved
in a variety of acute organ injury processes ([248]135). Accordingly,
TFs and miRNAs target major proteins to alter particular diseases
([249]136). SARS-CoV-2 infection possibly induces transcriptional
regulator mutations regulating primary signaling pathways, thus
activating inflammatory responses and leading to impairment of renal
function.
Mutations in genes are often closely related to multiple diseases, and
we performed gene–disease (GD) analysis to predict associations between
significant DEGs and various diseases. The results revealed various
diseases from the common DEGs of AKI, CKD, and COVID-19, which include
DUSP6, NRIP1, TNFAIP8, S1PR1, and TANK. It is notable that most of
these diseases are involved in reproductive phylogenetic problems,
psychophysiological disorders and cancers, and occasionally heart
dysfunctions. DUSP6, as an important DEG, has been discovered to be
associated with gonad development, altered sexual signs, and
psychophysiological disorders, such as hypogonadism, absence of
secondary sex characteristics, and mood and depressive disorders.
COVID-19 has a strong relationship with hypogonadism. Recent studies
have demonstrated the mechanisms by which secondary immune responses
govern endocrine function in SARS-CoV-2 infection and can hinder
testosterone synthesis in male patients, affecting male reproductive
health; there is also a possibility of inflammation due to the
infection, direct viral invasion of the testis, and drug-related damage
([250]137, [251]138). These findings indicate that men should be
considered at higher risk of poor prognosis or death. Anxiety and
depression are common manifestations in COVID-19 patients, and the
immune system perturbation caused by infection and the roles of
inflammatory and clinical predictors may induce psychopathology. The
COVID-19 pandemic might be associated with psychiatric disorders
([252]139, [253]140). TNFAIP8 and NRIP1 are mostly associated with
tumors, with breast cancer appearing more frequently in our GD network.
Some studies have suggested that estrogen levels in COVID-19 patients
can affect the inflammatory state and microbiome, which may be a
mechanism of breast tumor production; psychological factors also have
certain effects on female patients ([254]141, [255]142). Additionally,
heart diseases, such as heart failure and myocardial infarction, may be
regulated by NRIP1. Cardiac injury in COVID-19 patients seems to be
associated with higher mortality. Myocardial infarction,
cardiomyopathies, arrhythmias, fulminant myocarditis, and venous
thromboembolism are the most common cardiovascular complications of
COVID-19 ([256]143). Excessive secretion of inflammatory cytokines
(IL-6 and TNF-α) leads to systemic inflammation and multiple organ
dysfunction syndrome, severely affecting the cardiovascular system
([257]144). Furthermore, SARS-CoV-2 tropism and interaction with the
RAAS system may enhance inflammatory responses and cardiac aggression
([258]145). It is clear that COVID-19 is a systemic disease complicated
by multiorgan dysfunction, and organ crosstalk plays a key role in this
process. The involvement of the kidney, as the main organ leading to
organ crosstalk, was first defined in patients with acute respiratory
distress syndrome (ARDS) and COVID-19. The lungs and the kidneys
cooperate to maintain the electrolyte balance and the acid–base balance
in the body, and impairment of renal function disrupts the balance and
affects lung function ([259]95). In addition to the lungs, the kidneys
engage in crosstalk with multiple other organs. Progression of CKD is
also often accompanied by reproductive health challenges, including
menstrual abnormalities, impaired sexual health, and reduced fertility
([260]146). Crosstalk between the gonad and the kidney may also be
related to reproductive system problems in COVID-19 patients.
Depression has been reported to be the most common psychological
problem in patients with CKD and is influenced by biological,
psychological, and socioeconomic factors ([261]147, [262]148).
Furthermore, psychological symptoms in CKD are independent predictors
of adverse clinical outcomes, including faster GFR decrease, dialysis
therapy initiation, death, or hospitalization ([263]149). It is
possible that the bidirectional relationship between the progression of
COVID-19 and depression is also affected by the complex interplay
between biopsychosocial factors. Kidney diseases and cancers are
intertwined in many ways. On the one hand, underlying kidney disease
appears to increase cancer risk and its associated morbidity and
mortality ([264]150). Jørgensen et al. showed that an elevated urinary
albumin/creatinine ratio at baseline correlates with subsequent cancer
incidence ([265]151). Albuminuria is also associated with an increased
risk of cancer death from all causes and lung and prostate cancers in
men aged 50 and older in the USA ([266]152). As with albuminuria,
end-stage renal disease (ESRD) is associated with an increased risk of
renal and urinary tract cancer, and increased rates of endocrine
cancer, viral infection-related cancer, skin cancer, and liver cancer
have also been reported in ESRD patients ([267]153, [268]154). On the
other hand, carcinoma, paraneoplastic renal manifestations, and
nephrotoxicity of chemotherapeutic- and molecular-targeted drugs can
lead to the development of AKI and sustained impairment of renal
function ([269]155). GD analysis demonstrates the common underlying
molecular mechanisms of various comorbidities in COVID-19 and kidney
disease and highlights a possible reason why the kidney is able to act
as the main organ for organ crosstalk in COVID-19.
Currently, some drugs have been approved for the treatment of COVID-19
with few adverse effects. For example, remdesivir and chloroquine have
been demonstrated to prevent SARS-CoV-2 infection and COVID-19
([270]156). Furthermore, baricitinib, which shows antiviral effects by
interfering with viral entry into cells, shows improved therapeutic
effects in combination with remdesivir ([271]157). Casirivimab and
imdevimab (REGN-COV2), neutralizing antibodies, also have shown
promising results for SARS-CoV-2 infection by inhibiting viral
receptor-binding domain binding to host cells ([272]158). In addition,
drugs such as dexamethasone, tocilizumab, and interferon have been
shown to have significant effects against COVID-19 ([273]159–[274]161).
The protein–drug interaction and molecular dynamics analyses of this
study indicate eight possible chemical compounds targeting common DEGs,
with different binding affinities for the four hub proteins: DUSP6,
BHLHE40, RASGRP1, and TAB2. Pharmaceutical molecules strongly binding
to TAB2 were the most abundant, including tanespimycin, camptothecin,
niclosamide, pyrvinium, and daunorubicin; molecules strongly binding to
RASGRP1 and BHLHE40 were the second most abundant, with the former
including camptothecin, niclosamide, and pyrvinium and the latter
including tanespimycin, niclosamide, and pyrvinium. The least abundant
was the DUSP6 protein, but all pharmaceutical molecules can bind to
this protein. We identified the heat shock protein 90 (HSP90) inhibitor
tanespimycin as a host-dependent factor of SARS-CoV-2 and an effective,
broad-spectrum antiviral drug against human coronavirus ([275]162).
Another drug is camptothecin, a quinoline alkaloid originally isolated
from the Chinese happy tree, and has been found to have anticancer and
antiviral properties. Regarding SARS-CoV-2, camptothecin potentially
blocks the interaction of the spike glycoprotein with the ACE2 receptor
on host cells ([276]163). Another identified drug is niclosamide, an
anthelminthic drug, which is widely used to treat a variety of diseases
due to its pleiotropic anti-inflammatory and antiviral activities. An
effect via interruption of the viral life cycle or induction of the
cytopathic effect renders it a possible candidate for COVID-19
([277]164). Pyrvinium has anthelmintic properties and therapeutic
functions against fungi and is a potential novel agent for tumor
therapy ([278]165). Daunorubicin belongs to the anthracycline group and
is widely used in human cancer chemotherapy ([279]166, [280]167).
Moreover, staurosporine, dimethyloxalylglycine, and sulpiride were
found to be potential drugs in this study. Staurosporine is a very
potent inducer of apoptosis because it inhibits many different kinases.
Staurosporine-induced apoptosis has been discussed for various tumor
therapies ([281]168). The prolyl-hydroxylase inhibitor
dimethyloxalylglycine activates the hypoxia-inducible factor (HIF)-1
pathway by stabilizing HIF-1α and has a protective effect against
ischemia/reperfusion injury ([282]169). Dimethyloxalylglycine may be
protective against AKI. Sulpiride, an antipsychotic with selective
dopaminergic antagonist properties, has a therapeutic effect in
COVID-19 patients with psychiatric disorders ([283]170, [284]171). With
regard to diseases as risk factors of COVID-19 infection, such as
cancer, other infections, and organ damage, the above drugs all have
potential therapeutic effects. Further investigation is needed for
confirmation.
5. Conclusion
Our study summarizes relationships among COVID-19, AKI, and CKD through
bioinformatics analysis and identifies the potential molecular
mechanism by which SARS-CoV-2 infection affects renal function. We
examined 17 DEGs from three datasets by GO analysis and identified
oxidative metabolism as the major biological function of these genes.
Moreover, pathway enrichment analysis revealed that the MAPK signaling
pathway, the IL-1 structural pathway, and the Toll-like receptor
pathway, which are important pathways of systemic and organ
inflammation pathology, are pivotal in the occurrence of AKI, CKD, and
COVID-19. This study suggests that these pathways are involved in the
mechanisms of AKI in COVID-19 patients and the deterioration of renal
function. Then, the four most significant hub genes were screened from
the PPI network and found to be closely related to the inflammatory
response and tissue injury. In addition, the TFs and miRNAs identified
play crucial roles in different functional disorders. Different types
of diseases related to DEG mutations, mainly reproductive phylogenetic
problems, psychophysiological disorders, and cancers, are shared
complications of the three diseases. Analysis of COVID-19, AKI, and CKD
provides a way to identify the pathogenesis of various diseases and
helps in further understanding the underlying mechanisms of the
development of AKI and the progression of CKD in COVID-19 patients.
Therefore, it is possible to reduce the risk of SARS-CoV-2 infection
resulting in AKI and CKD. However, COVID-19 is a newly discovered
disease that has not been thoroughly studied, and more data are needed
for further research. Multiomics analysis of COVID-19 is becoming
important with the availability of bioinformatics approaches. Further
cohort follow-up may help to elucidate the molecular mechanisms of AKI
and CKD development in COVID-19 patients. This study provides promising
pathways and molecular biomarkers for the association of COVID-19 with
kidney diseases, and the findings are significant for the effective
treatment of COVID-19.
Data availability statement
The datasets presented in this study can be found in online
repositories. The names of the repository/repositories and accession
number(s) can be found in the article/[285] Supplementary Material .
Author contributions
QY and FL conceived and designed the research. WZ and LL wrote the
paper. XX, HZ, ZP, WW, LH, YX, and LT revised the paper. HX, WN, and XY
generated and provided analytical tools. WZ, LL, and XX analyzed data.
All authors contributed to the article and approved the submitted
version.
Funding Statement
This work was supported by the National Key R&D Program of China
(2020YFC2005004), the National Natural Science Foundation of China
(82070717), the Natural Science Foundation of Hunan Province China
(2020JJ5942, 2019JJ40515 and 2019JJ20035), the Major Program of the
National Natural Science Foundation of China (82090024), the General
Programs of the National Natural Science Foundation of China
(82173877), and the Key Research and Development Program of Hunan
Province (2021SK2015).
Conflict of interest
The authors declare that the research was conducted in the absence of
any commercial or financial relationships that could be construed as a
potential conflict of interest.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or claim
that may be made by its manufacturer, is not guaranteed or endorsed by
the publisher.
Supplementary material
The Supplementary Material for this article can be found online at:
[286]https://www.frontiersin.org/articles/10.3389/fimmu.2023.961642/ful
l#supplementary-material
Supplementary Figure 1
Expression of 17 common DEGs in three datasets. Panels (A–C) represent
the expression of common DEGs in [287]GSE1563, [288]GSE66494, and
[289]GSE147507, respectively. In the heatmap, each row represents the
expression of a gene in different samples; red indicates that the gene
is upregulated, and blue indicates that the gene is downregulated.
[290]Click here for additional data file.^ (346.7KB, jpeg)
Supplementary Figure 2
Building COVID-19 diagnostic models and model interpretability (A) AUC
of the 6 machine learning models in the training set. (B) AUC of the 6
machine learning models in the validation set. (C) Calibration plots
were used to assess the agreement between predicted and observed values
in different percentiles of predicted values. (D) Figure a is a SHAP
diagram showing the relationship between each variable and the outcome.
Each point represents a patient; the redder the color of the point
indicates a larger value, and the bluer the color of the point
indicates a smaller value. The farther to the right of the abscissa of
the point, the greater the contribution to the predicted positive
outcome, and the farther to the left of the abscissa of the point, the
greater the contribution to the predicted negative outcome.
[291]Click here for additional data file.^ (272.4KB, jpeg)
Supplementary Table 1
Results of molecular docking of drugs and proteins (kcal/mol).
[292]Click here for additional data file.^ (12KB, docx)
Supplementary Table 2
Binding free energies and energy components predicted by MM/GBSA
(kcal/mol).
[293]Click here for additional data file.^ (11.7KB, docx)
Supplementary Table 3
Comparison of multiple models in training set.
[294]Click here for additional data file.^ (11.7KB, docx)
Supplementary Table 4
Comparison of multiple models in validation set.
[295]Click here for additional data file.^ (11.9KB, docx)
References