Abstract

   Marburg virus (MARV) disease (MVD) is an uncommon yet serious viral
   hemorrhagic fever that impacts humans and non-human primates. In
   humans, infection by the MARV is marked by rapid onset, high
   transmissibility, and elevated mortality rates, presenting considerable
   obstacles to the development of vaccines and treatments. Bats,
   particularly Rousettus aegyptiacus, are suspected to be natural hosts
   of MARV. Previous research reported asymptomatic MARV infection in
   bats, in stark contrast to the severe responses observed in humans and
   other primates. Recent MARV outbreaks highlight significant public
   health concerns, underscoring the need for gene expression studies
   during MARV progression. To investigate this, we employed two models
   from the Gene Expression Omnibus, including kidney cells from Rousettus
   aegyptiacus and primary proximal tubular cells from Homo sapiens. These
   models were chosen to identify changes in gene expression profiles and
   to examine co-regulated genes and pathways involved in MARV disease
   progression. Our analysis of differentially expressed genes (DEGs)
   revealed that these genes are mainly associated with pathways related
   to the complement system, innate immune response via interferons
   (IFNs), Wnt/β-catenin signaling, and Hedgehog signaling, which played
   crucial roles in MARV infection across both models. Furthermore, we
   also identified several potential compounds that may be useful against
   MARV infection. These findings offer valuable insights into the
   mechanisms underlying MARV's pathophysiology and suggest potential
   strategies for preventing transmission, managing post-infection
   effects, and developing future vaccines.

   Keywords: Marburg virus (MARV), Rousettus aegyptiacus, Homo sapiens,
   zoonotic disease, bioinformatics

Introduction

   Marburg virus (MARV) was first identified in 1967 and is classified as
   a negative-sense RNA virus from the Filoviridae family. It shares a
   close genetic relationship with Ebola virus, both known to cause severe
   hemorrhagic fever in humans and non-human primates, generally known as
   MARV disease (MVD) and Ebola virus disease (EVD), respectively. The
   MARV genus consists of two variants, MARV and Ravn virus (RAVV), whose
   overall genetic sequences exhibit a genetic divergence of approximately
   20%. However, both variants lead to a clinically indistinguishable form
   of MVD characterized by multiple hemorrhagic manifestations with a case
   fatality rate that can range up to 80%-90% as recorded in previous
   outbreaks[45]^1. In the last two decades, subsequent outbreaks of MVD
   were found to be sporadic and endemic to sub-Saharan Africa,
   characterized by the continuous introduction of both MARV and RAVV from
   fruit bats, especially the cave-dwelling Egyptian fruit bat (Rousettus
   aegyptiacus), which was identified as a wildlife reservoir[46]^2.
   African green monkeys are the primate spillover hosts whose infected
   tissues were the cause of seven fatalities in the first outbreaks that
   were simultaneously documented in Marburg, Germany, and Belgrade,
   Yugoslavia (now Serbia). As of 30 September 2024, in Disease Outbreak
   News reported by the WHO
   ([47]https://www.who.int/emergencies/disease-outbreak-news/item/2024-DO
   N537), during the most recent and largest outbreak in new geographical
   areas such as Rwanda, Equatorial Guinea, and Tanzania, MAVR was
   documented to spread through human-to-human transmission via direct
   exposure to blood, bodily fluids, and secretions of sick patients.
   Indirect transmission occurs through close contact with materials or
   surfaces contaminated with the virus from infected humans and
   animals[48]^3^-[49]^5. This has led to increased public health concerns
   and heightened media attention on MVD, despite the fact that MARVs have
   historically received less attention compared to the Ebola
   virus[50]^6^, [51]^7.

   As of now, documented information regarding the fundamental processes
   of viral transmission and disease pathways remains limited, thus
   presenting challengs to research into specific antiviral treatments and
   vaccines for MVD. Therefore, comprehending the dynamics of MARV
   transmission between its natural host reservoirs and humans has become
   crucial for outbreak prevention and control. Although accurate animal
   models that mirror the human pathogenesis are indispensable for drug
   development, the US FDA's animal efficacy rule provides a framework for
   utilizing data from such models when conducting human trials is not
   feasible[52]^8. Notably, differences in a host's responses to viral
   infection among infected animals have been recorded, suggesting that
   the relevant pathogenesis and immunity following MARV replication may
   vary among reservoirs. In the case of MARV, previous research reported
   the asymptomatic infection of this virus in natural bat reservoirs,
   which was completely opposed to the severe immune responses and
   inflammatory gene dysregulation recorded in their spillover hosts
   including humans and non-human primates following viral
   infection[53]^9^-[54]^11. This difference is believed to primarily be
   caused by the reverse regulation pattern of interferon (IFN) responses
   and proinflammatory disease mediators in these two hosts, resulting in
   either the induction or suppression of antiviral responses[55]^12.
   However, to date, no studies have investigated similar patterns of
   viral infection in bats and humans during the incubation period, which
   could help characterize the entry and dissemination of MARV in diferent
   hosts as well as subsequent disease progression.

   In addition to lymphoid tissues and the liver, which are well-known
   targets of filovirus infection, previous laboratory findings showed a
   link between renal dysfunction and patients in the late stages of MVD,
   as MARV virions were detected in renal tissues and MARV antigens were
   present in proximal tubular cells[56]^13^, [57]^14. Supporting these
   findings, Martini et al. in 1971 and Smith et al. in 1982 respectively
   reported the presence of live virus isolated from kidney tissues and
   urine[58]^15^, [59]^16. In 2018, Arnold et al. conducted a
   transcriptomic analysis of two models: an Egyptian rousette bat (ERB)
   kidney-derived cell line (RoNi 7.1) infected with either a wild-type
   MARV or a recombinant mutant MARV (VP35mut) compared to their
   respective mock-infected groups[60]^17. The major finding of that work
   was that immune suppression remained a critical feature of MARV
   infection in the in vitro ERB model via upregulation of antiviral
   genes. Another study conducted by Koch et al. in 2023 applied
   transcriptome analyses at multiple time points in primary human
   proximal tubular cells and revealed that the in vitro productive
   replication of MARV was correlated with elevated levels of IFN-related
   factors and cytokines in the early phase of viral infection. This
   strong inflammatory and antiviral response was linked to kidney
   injury[61]^18. Therefore, investigating how reverse regulation patterns
   of related genes and pathways in the two MARV-infected in vitro models
   of humans and bats are related to antiviral responses would be a highly
   valuable research study.

   High-throughput technologies have become essential in the systematic
   investigation of expression differences of thousands of genes across
   various biological and genomic systems[62]^19^-[63]^25. In recent
   years, the integration of multiple high-throughput databases and
   analytical tools has paved the way for big data approaches in
   biomedical research[64]^26^-[65]^28, offering new insights and
   enhancing the precision of data-driven discoveries[66]^29^-[67]^34. In
   this work, we aimed to investigate key similarities and distinctions in
   gene expression patterns of two in vitro MARV-infected kidney
   models—one from humans and one from bats—using sequencing data from
   previous studies. This includes data from the MARV-infected
   immortalized RoNi/7.1 kidney cell line, derived from the fruit bat R.
   aegyptiacus, which was identified as a reservoir host of MARV
   ([68]GSE117367), and from MARV-infected H. sapiens primary proximal
   tubular cells, compared to a mock-infected group ([69]GSE226148).
   Pathways regulated by top exclusive and shared differentially expressed
   genes (DEGs) between these two MARV-infected cell line models may
   provide a more-insightful understanding of key factors that contribute
   to the disparate outcomes between reservoir hosts and spillover hosts
   during MARV progression (Figure [70]1), guiding the development of
   effective antiviral therapies in the near future.

Figure 1.

   [71]Figure 1
   [72]Open in a new tab

   Graphical abstract of the study design. Data were obtained from the
   Marburg virus (MARV)-infected RoNi/7.1 cell line, derived from
   Rousettus aegyptiacus, and from MARV-infected primary proximal tubular
   cells from Homo sapiens in the Gene Expression Omnibus (GEO) database.
   By crossing fold change > 2.0 upregulated genes (MARV-infected groups
   vs. a mock-infected control groups) in each MARV-infected model using a
   Venn diagram analysis, common genes were projected to pathway analyses
   and functional interpretations using bioinformatics approaches.

Materials and Methods

Data acquisition and processing methods

   In the current study, our objective was to investigate the effects of
   MARV infection on gene expressions from two distinct MARV-infected
   subjects: the RoNi/7.1 immortalized kidney cell line derived from a
   MARV-infected Egyptian fruit bat (R. aegyptiacus) and primary proximal
   tubular cells derived from H. sapiens. Relevant transcriptomic datasets
   were sourced from the NCBI's Gene Expression Omnibus (GEO;
   www.ncbi.nlm.nih.gov/geo). Transcriptomic profiles of the 24-h
   MARV-infected R. aegyptiacus RoNi/7.1 cell line and their mock-infected
   counterparts were extracted from the [73]GSE117367 dataset[74]^18,
   while similar information for 20-h MARV-infected H. sapiens proximal
   tubular cells was retrieved from the [75]GSE226148 dataset[76]^17. Two
   NCBI-generated RNA-sequencing (RNA-Seq) count datasets underwent
   processing and normalization using DESeq2[77]^35. Comparisons with
   mock-infected groups were conducted to identify DEGs in each
   model[78]^36^-[79]^38. Gene groups with significant annotations were
   selected using a cutoff of 0.05 (p < 0.05). Clustering based on
   expression profiles was performed using Kallisto following guidance of
   Jayaprakash et al.[80]^39^, [81]^40. The most substantial fold changes,
   with absolute values exceeding 1.5, were then mapped onto the Uniform
   Manifold Approximation and Projection (UMAP)[82]^41, and subjected to
   gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   analyses[83]^42^, [84]^43., all integrated using Omics Playground
   v.3.4.1[85]^44.

MetaCore pathway analysis

   An additional functional enrichment analysis was conducted using
   MetaCore™ (GeneGo, St. Joseph, MN, USA) to comprehensively analyze
   biological pathways associated with the distinctive and common DEGs
   (with fold changes of > 1.5) between the two models. The significant
   pathways in each case were ranked in descending order of -log10(p
   value) as we previous described[86]^45^, [87]^46..

Gene set enrichment analysis (GSEA)

   A GSEA was performed to compare enriched gene sets in the two
   MARV-infected datasets using the Python GSEApy package[88]^47. The
   cutoff for significant enrichment was set to a false discovery rate
   (FDR) of < 0.25[89]^48^, [90]^49.

Protein-protein interaction (PPI) network construction using STRING and
Cytoscape

   The STRING database, vers. 12.0 (accessible at
   [91]https://string-db.org/), functions as both a search engine and a
   resource for protein-related information. It offers an extensive
   collection of proteins and established interaction data used to analyze
   DEGs in this study. Additionally, to visualize and interpret the
   complex interaction networks among the identified proteins, Cytoscape
   software was utilized to construct a detailed PPI network, and the
   k-means clustering algorithm was employed to further categorize
   proteins into distinct clusters based on their interaction patterns,
   facilitating a more-granular analysis of their roles and relationships
   within the biological processes under study.

Constructing a connectivity map (CMap) of potential therapeutic chemical
compounds

   CMap employs a systematic methodology to analyze changes in gene
   expressions to uncover disease interactions and match those with
   compounds listed in the LINCS L1000 database[92]^50. Common DEGs shared
   between the two MARV-infected models, compared to their mock-infected
   groups, were projected to the CMap platform to construct a connectivity
   map of potential compounds that may induce or reverse the biological
   effects or counteract gene expression changes observed in the query.
   Normalized connectivity scores indicate both the direction and
   intensity of the relationship between a query (such as a gene
   expression signature) and the perturbagen profiles (such as drugs or
   small molecules) found in the LINCS database. A positive score suggests
   that the gene expression signature induced by the perturbagen closely
   resembles the query signature, whereas a negative score indicates that
   the gene expression signature induced by the perturbagen is the
   opposite of the query signature.

Statistical analysis

   Statistical analyses in this study were conducted using R vers. 4.1.3.
   The Wilcoxon test, a non-parametric method ideal for comparing two
   independent samples, was used to evaluate differences between various
   phenotypic groups. To correct for multiple comparisons, p values
   obtained from the Wilcoxon test were adjusted using the FDR method. A p
   value of < 0.05 was regarded as statistically significant, indicating
   meaningful differences between the groups analyzed.

Results

Clustering and GO analytical results of DEGs from the two MARV-infected
models

   In the model involving the MARV-infected bat-derived RoNi/7.1 cell
   line, we conducted a comparative analysis of DEGs using transcriptomic
   data from the [93]GSE117367 dataset. Figure [94]2A illustrates the
   results of the GO analysis associated with MARV infection. Cluster 2
   contained the most significantly upregulated genes and pathways in the
   MARV-infected group, while cluster 1 exhibited the opposite trend.
   These pathways within cluster 1 were linked to intracellular components
   and cell development, such as intracellular parts (GO:0044424),
   DNA-dependent (GO:0045893), GTPase regulator activity (GO:0030695),
   phospholipid binding (GO:0005543), cilium movement (GO:0003341), cell
   projection assembly (GO:0030031), heart development (GO:0007507),
   pronephros development (GO:0048793), primary metabolic processes
   (GO:0044238), cellular response to lithium ions (GO:0071285),
   cytoskeletal protein binding (GO:0008092), fatty acid binding
   (GO:0005504), cell cycle processes (GO:0022402), regulation of eye
   pigmentation (GO:0048073), phosphatidylcholine catabolic processes
   (GO:0010899), negative regulation of interleukin (IL)-6 processes
   (GO:0045409), regulation of the vascular endothelium (GO:0030947),
   peptide hormone secretion (GO:0030072), embryonic skeletal
   morphogenesis (GO:0048704), embryo development (GO:0009790), cell
   adhesion involved in gastrulation (GO:0070587), and regulation of cell
   migration (GO:0030334). In contrast, the GO analysis revealed that
   cluster 2 was associated with inflammatory responses, chemokine
   activities, and immune system processes (Figure [95]2A). These included
   regulation of cell migration (GO:0030334), regulation of activated
   T-cell proliferation (GO:0046006), regulation of inflammatory responses
   (GO:0050727), toll-like receptor 9 signaling pathway (GO:0034162),
   mitogen-activate protein kinase (MAPK) kinase kinase (MAPKKK) cascade
   (GO:0000165), and drug transport (GO:0015893).

Figure 2.

   [96]Figure 2
   [97]Open in a new tab

   Hierarchical clustered heatmaps of gene ontology (GO) enrichment. A GO
   enrichment analysis was performed on sets of differentially expressed
   genes (DEGs) within the two Marburg virus (MARV)-infected models
   compared to corresponding mock-infected groups, including (A) a
   MARV-infected bat-derived cell line at 24 h post-infection, and (B) a
   MARV-infected human-derived cell line at 20 h post-infection.

   For the MARV-infected human-derived cell line model, transcriptomic raw
   data were obtained from the [98]GSE226148 dataset. In this case,
   pathways within cluster 2 were strongly associated with MARV infection,
   while cluster 1 displayed an opposing trend (Figure [99]2B). The GO
   analysis indicated that pathways within cluster 1 were related to
   endosome function and metabolism; these included processes such as
   endosome to endosome to lysosome transport (GO:0008333), late endosome
   membranes (GO:0031902), pigment biosynthetic processes (GO:0046148),
   late endosomes (GO:0005770), positive regulation of osteoblast
   differentiation (GO:0045669), cytoplasm (GO:0005737), RNA processing
   (GO:0006396), phosphoric ester hydrolase activity (GO:0042578),
   nucleoplasm (GO:0005654), regulation of macromolecule metabolic
   processes (GO:0060255), mitochondrial parts (GO:0044429), regulation of
   primary metabolic processes (GO:0080090), cellular protein metabolic
   processes (GO:0044267), and cellular protein metabolic processes
   (GO:0044267). On the other hand, pathways within cluster 2 were linked
   to regulation of viral infections and stress-related signaling, such as
   stress-activated protein kinase signaling cascade (GO:0070304),
   negative regulation of viral genome replication (GO:0045071), defense
   response to viruses (GO:0051607), cholesterol biosynthetic processes
   (GO:00066695), RNA processing (GO:0006396), cytoplasm (GO:0005737),
   late endosomes (GO:0005770), mitochondrion organization (GO:0007005),
   response to unfolded proteins (GO:0006986), methyltransferase activity
   (GO:0008168), protein transport (GO:0015031), RNA binding (GO:0003723),
   and regulation of macromolecule metabolic processes (GO:0060255).

UMAP clustering of gene sets and relative GO annotations of unique and shared
highly expressed genes in the two MARV-infected models

   Gene expression profiles of the MARV-infected bat-derived RoNi/7.1
   immortalized kidney cell line ([100]GSE117367) and MARV-infected
   human-derived primary proximal tubular cell line ([101]GSE226148) were
   analyzed using the GO platform integrated with Omics Playground for
   biological process, molecular function, and cellular component
   annotations specific to MARV infection conditions. In the former model,
   these genes primarily functioned in endothelial cell development
   (GO_0001885), tumor necrosis factor (TNF) receptor superfamily binding
   (GO_0032813), regulation of c-Jun N-terminal kinase (JUNK) activity
   (GO_0043506), negative regulation of protein localization to membranes
   (GO_1905476), determination of bilateral symmetry (GO_0009855), RNA cap
   binding (GO_0000339), mitotic spindle assembly checkpoint signaling
   (GO_0007094), mitotic spindle checkpoint signaling (GO_0071174),
   spindle assembly checkpoint signaling (GO_0071173), and axoneme
   assembly (GO_0035082) (Figure [102]3A). In the latter model, the most
   significant biological terms involved were negative regulation of viral
   processes (GO_0048525), innate immune responses (GO_0045087), positive
   regulation of cytosolic calcium ion concentrations (GO_0007204),
   defense responses to viruses (GO_0051607), defense responses to
   symbionts (GO_0140546), cytokine activity (GO_0005125), regulation of
   cell adhesion (GO_0030155), defense responses to gram-positive
   bacterium (GO_0050830), positive regulation of the MAPK cascade
   (GO_0043410), and negative regulation of cytokine production
   (GO_0001818) (Figure [103]3B). Major characteristics of BP annotations
   for each MARV-infected model are summarized in Figure [104]5A with
   details in Supplementary Table A2. The annotations are organized in
   descending order of -log(p values), highlighting the most significantly
   enriched processes associated with the DEGs in each model.

Figure 3.

   [105]Figure 3
   [106]Open in a new tab

   UMAP clustering of top enriched gene ontology (GO) terms represented
   for the top distinct diferentially expressed genes in the two Marburg
   virus (MARV)-infected models. The top 10 most enriched GO terms across
   biological process, molecular function, and cellular component
   categories in (A) a MARV-infected bat-derived cell line at 24 h
   post-infection, and (B) a MARV-infected human-derived cell line at 20 h
   post-infection.

Figure 5.

   [107]Figure 5
   [108]Open in a new tab

   Summary visualization of common and distinct biological process (BP)
   gene ontology (GO) terms and pathway analysis associated with the top
   differentially expressed genes (DEGs) in two Marburg virus
   (MARV)-infected in vitro models. The top enriched GO biological process
   terms in (A) the MARV-infected Rousettus aegyptiacus in vitro model,
   and (B) the MARV-infected Homo sapiens in vitro model. (C) Distinctive
   and common pathways regulated by the top DEGs in the two MARV-infected
   models at a cutoff log2 fold change of > 1.5 and a p value of < 0.05.

Pathway analysis results and pathway enrichment analysis results of distinct
and common DEGs shared between the two MARV-infected models

   Since the MARV is a zoonotic disease observed in the context of
   emerging zoonotic infectious diseases, the dynamics underlying the
   shifting epidemiology can be attributed to multiple factors. These
   include variations related to pathogen evolution, characteristics of
   the zoonotic reservoir, and nuances of the human-animal interface. In
   order to explore the co-regulation of genes for both in vitro
   MARV-infected bat and human models, we employed Venn diagrams to
   identify common DEGs shared between the MARV-infected human model from
   the [109]GSE226148 dataset and the MARV-infected R. aegyptiacus model
   from the [110]GSE117367 dataset, specifically retrieving overlapping
   DEGs with a fold change of > 1.5 (Figure [111]4A, Supplementary Table
   A3). Altogether, we identified 463 common DEGs shared between these two
   MARV-infected models, and then subsequently subjected these genes to a
   MetaCore analysis, which revealed several standard maps related to MARV
   infection, including pathways such as: “Alternative complement cascade
   disruption in age-related macular degeneration (AMD)”, “Immune
   response_Alternative complement pathway”, “Immune
   response_IFN-alpha/beta signaling via JAK/STAT”, “SHH signaling in
   colorectal cancer”, “Development_Embryonal epaxial myogenesis”, “SHH
   signaling in melanoma”, “Development_Early embryonal hypaxial
   myogenesis”, “SHH signaling in oligodendrocyte precursor cells
   differentiation in multiple sclerosis”, “Stem cells_Aberrant Hedgehog
   signaling in medulloblastoma stem cells”, “Development_Positive
   regulation of WNT/Beta-catenin signaling at threceptor level”,
   “Development_Positive regulation of WNT/Beta-catenin signaling in the
   nucleus”, “Stellate cells activation and liver fibrosis”, “Immune
   response_IFN-alpha/beta signaling via MAPKs”, and “Stem cells Aberrant
   Wnt signaling in medulloblastoma stem cells. Hedgehog signaling in
   breast cancer”. Notably, signaling pathways related to complement
   signaling, IFN-alpha/beta response, WNT/Beta-catenin signaling, and
   Hedgehog (HH) signaling were highly enriched among the common DEGs
   between these two models (Figure [112]4C).

Figure 4.

   [113]Figure 4
   [114]Open in a new tab

   Comparison of differentially expressed gene (DEG)-enriched signaling
   pathways in Marburg virus (MARV)-infected models. (A) The Venn diagram
   illustrates the number of distinct and common DEGs (with fold change of
   > 1.2) when comparing MARV-infected groups to their respective
   mock-infected control groups within the [115]GSE117367 and
   [116]GSE226148 datasets. The overlapping region represents genes shared
   between the two datasets, while the non-overlapping areas indicate
   genes unique to each dataset. (B) List of the top 15 DEG-enriched
   signaling pathways regulated by the common DEGs, generated by MetaCore,
   and sorted in descending order of log(p values). (C) Visualization of
   the global signal transduction pathway network generated by MetaCore
   confirmed that the “Immune response_IFN-alpha/beta signaling via
   JAK/STAT” pathway was highly enriched in both MARV-infected models.

   To gain a deeper understanding of the biological processes in each
   model, we conducted separate analyses of pathways related to MARV
   infection. In the case of the MARV-infected R. aegyptiacus model, 3212
   DEGs with an absolute fold change of > 1.5 (MARV-infected groups vs.
   mock-infected control groups) were subjected to a MetaCore analysis
   which revealed significant maps such as: “Role of ER stress in obesity
   and type 2 diabetes”, “DNA damage_p53 activation by DNA damage”,
   “Apoptosis and survival_p53 and p73-dependent apoptosis”, “SCAP/SREBP
   Transcriptional Control of Cholesterol and FA Biosynthesis”,
   “Development_Positive regulation of WNT/Beta-catenin signaling in the
   cytoplasm”, “Development_Negative regulation of WNT/Beta-catenin”,
   “Transcription_Epigenetic regulation of gene expression”, “Regulation
   of metabolism_Negative regulation of insulin signaling”, “Signal
   transduction_mTORC1 downstream signaling”, “Signal transduction_ESR1
   (membrane) and ESR2 (membrane) signaling”, “Signal transduction_mTORC2
   downstream signaling”, “Development_Negative regulation of
   WNT/Beta-catenin signaling in the nucleus”,
   “Chemotaxis_Lysophosphatidic acid signaling via GPCRs”, “G-protein
   signaling_CDC42 activation,” and “Autophagy_Autophagy”
   ([117]Supplementary Figure 1A-B, Supplementary Table A4). For the
   MARV-infected H. sapiens model, 682 DEGs with an absolute fold change
   of > 1.5 (MARV-infected groups vs. mock-infected control groups) were
   subjected to a MetaCore analysis which revealed significant maps such
   as “Immune response_IFN-alpha/beta signaling via JAK/STAT”, “COVID-19:
   Regulation of antiviral response by SARS-CoV-2”, “Immune
   response_IFN-alpha/beta signaling via MAPKs”, “Cigarette smoke-mediated
   regulation of NRF2-antioxidant pathway in airway epithelial cells”,
   “Attenuation of IFN type I signaling in melanoma cells”, “Inhibition of
   Ephrin receptors in colorectal cancer”, “Immune response_IFN-gamma
   signaling via MAPKs”, “Immune response_Antiviral actions of
   Interferons”, “Immune response_Antimicrobial actions of IFN-gamma”,
   “Zika virus infection mechanism” “Immune response_Induction of
   apoptosis and inhibition of proliferation mediated by IFN-gamma”, “WNT
   signaling in proliferative-type melanoma cells”, “Glomerular injury in
   Lupus Nephritis”, “Transcription_HIF-1 targets”, and “Role of
   Apo-2L(TNFSF10) in Prostate Cancer cell apoptosis gamma”
   ([118]Supplementary Figure 2A, B, Supplementary Table A5).

   In summary, pathways related to complement signaling, IFN-alpha/beta
   response, WNT/Beta-catenin signaling, and HH signaling were identified
   as enriched pathways regulated by common leading edge genes under MARV
   infection conditions in both species. Conversely, endoplasmic reticular
   (ER) stress, DNA damage, WNT/Beta-catenin, and sterol regulatory
   element-binding protein (SREBP) cleavage-activating protein
   (SCAP)/SREBP were significantly enriched signaling-related pathways in
   the MARV-infected R. aegyptiacus-derived RoNi/7.1 immortalized kidney
   cell line, whereas pathways related to viral infection response,
   IFN-alpha/beta signaling, and oxidative stress response were notably
   upregulated in the MARV-infected H. sapiens primary proximal tubular
   cell line (Figure [119]5B). In addition, pathways identified through
   the GSEA for each MARV-infected model were consistent with the above
   findings ([120]Supplementary Figure 3).

Construction of a PPI Network and potential inhibitory compounds from the
connectivity map

   To further investigate the significance of commonly shared DEGs between
   the two MARV-infected models, we used a Venn diagram to identify 90
   genes that were differentially expressed in both MARV-infected groups
   (compared to mock-infected control groups) with a fold change of > 1.5.
   These 90 genes were then imported into the STRING platform to construct
   PPI networks, and we applied a k-means clustering algorithm with n=10,
   aiming to understand their potential roles in manifestations following
   a MARV infection (Figure [121]6A). Subsequently, we employed a CMap to
   identify compounds that could potentially mimic or reverse the 90 gene
   expression profiles in these two MARV-infected models. We screened over
   1309 compounds and more than 7000 treatments with various dosages
   across different cell lines, selecting potential therapeutic agents
   based on an FDR q-value threshold of 2.0 (negative log10-transformed
   FDR q-values) and sorting by normalized connectivity scores (norm_cs).
   Compounds were considered promising if they demonstrated a negative
   norm_cs, as they may reverse the biological effects or counteract gene
   expression changes observed in the query. The top 25 compounds with the
   highest positive and negative connectivity scores were each
   respectively visualized in red and green in the CMap in the form of a
   heatmap (Figure [122]6B). In total, 117 perturbagens showed a positive
   correlation, while 98 perturbagens showed a negative correlation to the
   90 queried genes (Supplementary Table A6, Supplementary Table A7).
   Among the 98 compounds that were likely to reverse gene expression
   changes induced by MARV infection, the top 25 perturbagens with most
   negative connectivity scores were enalapril, GR-127935, KU-0063794,
   BRD-K14329163, westcort, KU-0063794, clinofibrate, AZ-628,
   carbenoxolone, trimetazidine, zolpidem, orlistat, K-858, PI-103, PP-2,
   letermovir, BRD-K95992530, ebselen, lucitanib, ranitidine, IOX2,
   selamectin, MEK-162, equilin, and efaproxiral.

Figure 6.

   [123]Figure 6
   [124]Open in a new tab

   A key protein-protein-interacting (PPI) network and list of potential
   compounds from the connectivity map (CMAP)-based analysis. (A) The 90
   differentially expressed genes (DEGs) shared between two MARV-infected
   groups, compared to the mock control group of both models, namely the
   Marburg virus (MARV)-infected Rousettus aegyptiacus ([125]GSE117367)
   and MARV-infected Homo sapiens ([126]GSE226148), were subjected to a
   functional protein association network analysis using the STRING
   database and the k-means clustering algorithm (n=10). (B) The CMap
   analysis presents a significant connectivity score when querying gene
   expression changes of 90 genes of interest against compounds from the
   LINCS L1000 database. Small-molecule compounds that caused similar gene
   expression signatures resulted in a positive correlation (red), while
   those causing opposing gene expression signatures resulted in a
   negative correlation (green). The top 50 potential compounds were
   sorted in descending order based on normalization of connectivity
   scores (norm_cs) at a cutoff false discovery rate (FDR)_q_nlog10 = 2.

Discussion

   The persistence of viruses in bat cells and populations is a complex
   and multifaceted phenomenon. The complexity of this issue is further
   compounded by the diversity of bat species and their ecological roles,
   which influence virus transmission dynamics and persistence. Ongoing
   field research focused on bat-borne emerging pathogens coupled with
   advancements in next-generation sequencing technologies is vital for
   uncovering broader implications of these viral reservoirs. The rapid
   onset and high transmission rates of MVD present unique challenges,
   especially given the sporadic and unpredictable nature of MARV
   outbreaks, posing challenges for conducting human efficacy trials of
   MARV vaccines and treatments[127]^51^-[128]^54. Therefore,
   understanding factors driving dynamic patterns in the transmission of
   MARV between bats and humans is essential for preventing and
   controlling MVD. Compared to previous studies, our current findings
   indicated that ER stress signaling, DNA damage, apoptosis, and
   SCAP/SREBP signaling are significantly enriched pathways in the in
   vitro MARV-infected bat model. In contrast, viral infection signaling,
   pathways related to IFN-alpha/beta signaling and oxidative stress
   responses were notably upregulated in the in vitro MARV-infected human
   model. This clarifies key differences between humans, who suffer from
   severe flu-like symptoms in response to the viral infection, and R.
   aegyptiacus, which, as a well-adapted reservoir, shows no notable
   illness during MARV infections in either natural or experimental
   settings[129]^5. When comparing the two models, pathways related to
   complement signaling, IFN-alpha/beta signaling, WNT/Beta-catenin
   signaling, and HH signaling were identified as enriched pathways
   regulated by common leading edge genes under MARV infection conditions
   in both species. Targeting these pathways offers a promising
   opportunity to develop innovative treatments, supplementing current
   antiviral therapies and bringing new hope against the challenges of
   MARV.

   Complement activation is a key component of the innate immune system,
   whose activation leads to the rapid clearance of viral pathogenesis in
   the early phase of infection[130]^55. In general, the complement system
   can be activated through three primary pathways: the antibody-induced
   classical pathway, a lectin pathway via mannose-binding lectin (MBL),
   and an antibody-independent alternative pathway[131]^56. Earlier
   studies established a connection between filoviral infection and the
   first two pathways, particularly the binding of MBL to envelope
   glycoproteins of EBOV and MARV, as reported by Ji et al. and Michelow
   et al., which contributed to the development of antibody-based
   therapies[132]^57^-[133]^59. To date, no research findings have yet
   linked the third pathway to MARV infection, but there is evidence of
   such a link with the influenza virus and Chandipura virus[134]^60^,
   [135]^61. Therefore, targeting alternative pathways, in addition to the
   lectin pathway, may offer new therapeutic strategies for combating
   viral infections. The relationship between IFN and viral replication is
   complex and involves a dynamic interplay between a host's immune
   defenses and the virus' strategies to evade them. Type I IFNs (IFN-α/β)
   are key components of the innate immune response. When a virus infects
   a host cell, the cell detects the presence of viral components (such as
   double-stranded RNA) and triggers the production of type I IFNs.
   IFN-α/β then bind to their receptors on the surface of both infected
   and neighboring cells, initiating a signaling cascade that leads to
   expressions of IFN-stimulated genes (ISGs). These genes produce
   proteins that inhibit various stages of viral replication, including
   viral entry, replication, and assembly. Many viruses, including
   filoviruses like EBOV and MARV, have evolved mechanisms to evade or
   suppress the IFN response. In 2003, a study of Bray et al. explored the
   role of the type I IFN (IFN-α/β) responses in protecting against
   filovirus infections, such as EBOV and MARV, and highlighted that mice
   lacking the IFN-α/β receptor or treated with antibodies against IFN-α/β
   could be lethally infected with EBOV or MARV[136]^62. More-recent
   studies, including those by Cao et al. and Basler et al., demonstrated
   how filoviruses evade a host's IFN responses. Both studies focused on
   how viral proteins (VPs), particularly VP35 and VP24, interfere with
   IFN signaling pathways, allowing the virus to suppress immune
   responses, thereby facilitating robust viral replication and spread
   within the host[137]^63^, [138]^64. Thus, a strong and early IFN
   response can control viral replication, limiting the spread of the
   virus and reducing disease severity. In 2011, Guito et al. discovered
   differences in the potency of IFN antagonism among various filoviruses.
   They confirmed that VP35 is the primary antagonist of IFN production,
   while other VPs, such as VP24 and MARV VP40, are responsible for
   inhibiting downstream IFN signaling pathways across different
   filoviruses, including EBOV and MARV[139]^65. An in silico study by
   Alsaady et al. later identified six compounds—estradiol benzoate,
   paliperidone, isosilybin, protopanaxadiol, permethrin, and bufalin—as
   potential inhibitors of the MARV VP35 protein; however, further
   experimental validation is required to confirm their efficacy[140]^66.

   The Wnt/β-catenin signaling pathway was reported to facilitate viral
   replication, notably in the case of severe acute respiratory syndrome
   (SARS)-coronavirus (CoV)-2[141]^67^, [142]^68. However, this pathway
   has shown opposing roles, either promoting or inhibiting viral
   replication, as observed with bovine parainfluenza virus type, herpes
   simplex virus, and pseudorabies virus[143]^69^-[144]^72. In the context
   of filoviruses, RNA-Seq results of MARV-Angola variant-infected
   macaques that were pre-exposed to recombinant vesicular stomatitis
   virus (rVSV) vectors expressing the MARV-Angola-glycoprotein (GP)
   (Woolsey et al.) confirmed upregulation of the Wnt signaling pathway in
   survivors compared to those that died[145]^73. Recently, RNA-Seq
   results of human monocyte-derived macrophages inoculated with
   bat-isolated MARV repeatedly confirmed upregulation of the
   Wnt/β-catenin pathways in MARV-infected groups compared to both the
   lipopolysaccharide (LPS)-treated (positive control) and mock-infected
   (negative control) groups[146]^74. These consistent research findings
   once again highlight the significant role of Wnt signaling pathways,
   particularly in the case of MARV infection. These pathways could be
   targeted for drug development or used in treatment evaluation, in
   addition to the use of IFNs as discussed above. HH signaling refers to
   key pathways regulated by the HH gene family, which includes the
   secreted proteins Sonic HH (SHH), Indian HH (IHH), and Desert HH (DHH)
   in mammals. Dysregulated HH signaling was previously reported to result
   in developmental disorders and cancer. However, specific functions of
   this pathway in immunomodulation and viral infection have not been
   thoroughly explored[147]^75. Kim et al., Yang et al., and Arzumanyan et
   al. reported enrichment of the HH signaling pathway in heaptitis B
   virus (HBV) infection; Singh et al. highlighted the relationship
   between the disruption of HH signaling in the central nervous system
   (CNS) and human immunodeficiency virus (HIV)-associated neurocognitive
   disorder (HAND); Thawani et al. explored implications of HH signaling
   in Zika virus (ZIKV)-induced microcephaly, while other previous
   findings reported the relationships between this signaling pathway and
   oncoviruses[148]^76^-[149]^81. Existing evidence suggests that
   impairing various cellular pathways related to the cell cycle, DNA
   damage repair mechanisms, and cellular metabolism can lead to
   uncontrolled cell growth, the accumulation of genetic mutations, and
   metabolic reprogramming, all of which are hallmarks of cancer. Notably,
   impaired HH, Notch, and WNT signaling pathways are significantly
   associated with oncoviral infection. To date, eight major human viruses
   have been linked to oncogenesis and are therefore referred to as
   oncoviruses. These include human T-cell lymphotropic virus type I
   (HTLV-1), hepatitis C virus (HCV), HBV, Merkel cell polyomavirus
   (MCPV), human cytomegalovirus (CMV), human papillomavirus (HPV),
   Epstein-Barr virus (EBV), and Kaposi's sarcoma-associated herpesvirus
   (KSHV)[150]^82^, [151]^83. So far, no research has established a link
   between the MARV and hallmarks of cancer, nor has there been any
   examination of the connection between HH signaling and MARV infection.
   These areas appear to be unexplored and could benefit from further
   investigation.

   In the context of drug discovery, among the top five compounds with the
   most negative connectivity scores, enalapril ranked first. Enalapril is
   a prodrug belonging to the angiotensin-converting enzyme (ACE)
   inhibitor class that targets the renin-angiotensin-aldosterone system-a
   key regulator of blood pressure and fluid-electrolyte balance. Its
   active metabolite, enalaprilat, inhibits the ACE, reducing angiotensin
   II production, which decreases vasoconstriction and sodium
   reabsorption, ultimately lowering blood pressure and fluid
   volume[152]^84. Significant previous findings hightlighted enalapril as
   a potential compound related to drug repurposing against dengue virus
   and chikungunya virus[153]^85^, [154]^86. Notably, a significant number
   of previous findings also indicated that the ACE, being predominantly
   expressed by proximal tubules and glomerular podocytes, is a key
   receptor for SARS-CoV-2 entry into human cells via interaction with its
   spike (S) protein. This contributes to early kidney involvement during
   the course of infection, which is associated with worse outcomes in
   SARS-CoV-2 patients[155]^87^-[156]^89. Although past systematic reviews
   and meta-analyses showed conflicting results regarding the use of ACE
   inhibitors and outcomes in SARS-CoV-2 patients, subsequent in silico,
   in vitro, and in vivo approaches demonstrated promising
   results[157]^90^-[158]^93. Remarkably, a RNA-Seq analysis of livers
   from bats infected with EBOV and MARV (raw data deposited at
   [159]GSE152728) revealed decreased expression of angiotensinogen (AGT)
   but increased expression of prostaglandin I2 synthase (PTGIS),
   resulting in elevated ACE levels, all of which are blood
   pressure-regulating genes[160]^39. Other field studies confirmed the
   improved outcomes of EBRV-infected inpatients treated with angiotensin
   receptor blockers (ARBs), namely Irbesartan, as reported from an
   observational study[161]^94. The WHO also lists ACE inhibitors in their
   clinical management standard operating procedures (SOPs) for EBOV
   disease, suggeting that enalapril can serve as a highly promising
   candidate for treating filovirus infections in general[162]^95. The
   compound [163]GR127935, which ranked second, is a 5-hydroxytryptamine
   (5-HT) receptor antagonist that primarily binds to the 5-HT1A, 5-HT2A,
   and 5-HT7 receptors and has been widely studied in neurological disease
   research. Various studies, including those by Hu et al. and Riva et
   al., underscored its potential against both the HCV and
   SARS-CoV-2[164]^96^, [165]^97. Additionally, Ahmad et al.'s in silico
   drug repurposing study identified [166]GR127935 as one of the top
   scoring compounds targeting the ACE2 host-virus interface, highlighting
   its promise in blocking SARS-CoV-2 entry[167]^98. Another target,
   mammalian targe or rapamycin (mTOR), has become increasingly recognized
   for its potential as a target in antiviral therapies. KU0063794 is a
   selective inhibitor of the mTOR, a key pathway that plays a crucial
   role in regulating cell growth, proliferation, and survival. mTOR
   inhibitors can interfere with a virus's ability to manipulate host cell
   processes, thereby hindering viral replication, making mTOR a promising
   candidate in antiviral treatment strategies. In summary, although the
   aforementioned compounds and targets show potential, no studies to date
   have evaluated their antiviral efficacy against MARV infection. Further
   experimental studies are necessary to confirm these findings.

   Additionally, our screening results identified several potential
   targets, although the supporting evidence is currently limited and
   requires further validation to confirm their effectiveness in antiviral
   therapies. The identified targets included acetylcholine receptor
   antagonists (biperiden and naphazoline), adrenergic receptor agonists
   (reboxetine and urapidil), cyclooxygenase inhibitors (nefopam,
   valdecoxib, and ebselen), epidermal growth factor receptor (EGFR)
   inhibitors (WZ-4002 and lapatinib), fibroblast growth factor receptor
   (FGFR) inhibitors (AZD-4547 and lucitanib), histamine receptor agonists
   (ranitidine, doxylamine, and SNX-2112), heat shock protein (HSP)
   inhibitors (SNX-2112 and BIIB-021), insulin-like growth factor (IGF)-1
   inhibitors (BMS-536924 and linsitinib), lipase inhibitors (clinofibrate
   and orlistat), MAPK kinase (MEK) inhibitors (MEK-162, U-0126, TAK-733,
   selumetinib, AS-703026, PD-98059, and PD-0325901), phosphatidylinositol
   3 kinase (PI3K) inhibitors (AS-605240, GDC-0941, and PI-103), RAF
   inhibitors (AZ-628, vemurafenib, and AZ-628), Rho-associated kinase
   inhibitors (BRD-K23875128, fasudil, and GR-127935), an interleukin
   receptor agonist (pidotimod), a DNA inhibitor (pibenzimol), an estrogen
   receptor agonist (equilin), and a glucocorticoid receptor agonist
   (westcort). Further research is needed to fully explore and validate
   the potential of these compounds in antiviral therapies. The original
   study by Arnold et al. ([168]GSE117367) found that the Egyptian
   rousette bat serves as a natural reservoir for MARV and effectively
   controls viral infections with low viremia and no disease symptoms,
   unlike humans who suffer severe disease due to MARV-induced IFN
   antagonism. Analysis of the ERB genome and transcriptome revealed that
   while MARV inhibited antiviral responses in ERB cells, it did not
   induce specific IFN genes. However, there were increased expressions of
   unannotated antiviral genes and higher basal expressions of key
   antiviral genes in uninfected ERB cells. In contrast, findings of Koch
   et al. ([169]GSE226148) highlighted the impact of MARV infection on
   human primary proximal tubular cells (PTCs). They observed a
   significant inflammatory response characterized by the activation of
   signaling pathways associated with IFNs (IFN-α and IFN-y) and tumor
   necrosis factor (TNF)-α, along with downregulation of gene sets related
   to energy metabolism and mitochondrial function, which may contribute
   to the acute kidney injury observed in MVD. In our current study, by
   analyzing common DEGs between the two studies, we discovered that these
   genes are primarily associated with pathways related to the complement
   system, the innate immune response involving IFNs, Wnt/β-catenin
   signaling, and HH signaling, all of which appeared to play significant
   roles in the context of MARV infection in both models. Additionally,
   our analysis identified several potential compounds, notably the ACE
   inhibitor, enalapril, that could be used to combat MARV infection.

   As the first-ever report of its kind, this study provides valuable
   insights into the underlying mechanisms of MARV's pathophysiology and
   offers suggestions regarding modes of transmission, post-infection
   consequences, and potential avenues for future vaccine development. The
   findings represent a significant step in bridging the gap between
   laboratory research and clinical applications, a crucial aspect in the
   rapid development of effective treatment strategies for MARV outbreaks.
   Additionally, the outcomes of our study offer critical guidance for
   experimental efforts aimed at developing vaccines against MARV,
   contributing to the ongoing global efforts to combat the potential
   threat of MARV outbreaks. The primary limitation of this study was its
   reliance on in vitro models, specifically MARV-infected bat and human
   cell lines. While these models offer valuable insights into cellular
   responses to viral infection, they do not fully replicate the complex
   immune and systemic dynamics of a whole organism. In vitro systems are
   inherently limited in mimicking intricate interactions among different
   cell types, tissues, and the immune system. Consequently, this can lead
   to an overrepresentation of systemic pathways, which might not be as
   prominently activated during localized or tissue-specific responses
   observed in actual viral infections in vivo. As a result, the study's
   findings, particularly regarding activation of certain immune or
   signaling pathways, might not fully reflect localized responses seen in
   living organisms. Furthermore, because the study did not account for
   the immune system's role in viral clearance or drug efficacy, the
   identified pathways and compounds should be validated in more-complex
   models that better mimic human physiology, such as animal models or ex
   vivo tissue systems, before progressing to clinical trials.

Supplementary Material

   Supplementary figures and tables.
   [170]ijmsv22p0001s1.pdf^ (3.6MB, pdf)

Acknowledgments