Abstract

Background

   The inclusion of exosomes enters the recipient cells by means of
   endocytosis or direct fusion for information exchange between cells and
   cells. The inclusion of BMSCs-exo helps to guide the diagnosis and
   prognosis of cancer, especially cancer.

Purpose

   This research was to systematically elucidate the prognostic value of
   mRNAs of the exosomes derived from bone marrow stromal cells (BMSCs) in
   common malignant neoplasms, such as breast cancer, ovarian cancer, lung
   cancer, and gastric cancer.

Methods

   Gene expression data ([27]GSE78235) for the exosomes derived from BMSCs
   were extracted from the Gene Expression Omnibus database. Firstly, the
   differentially expressed genes were detected by comparing the RNA
   expression from exosomes derived from BMSCs between four tumor patients
   and two healthy controls using the limma package. Subsequently,
   functional enrichment analysis, including Gene Ontology and Kyoto
   Encyclopedia of Genes and Genomes of differentially expressed genes,
   was performed using the Functional Enrichment analysis tool (FunRich
   v3.1.3) software followed by the construction of a protein–protein
   interaction (PPI) network via STRING v3.6.0. Molecular Complex
   Detection was used to screen the hub proteins by setting up the
   following threshold score ≥4 and nodes ≥10. Cytoscape v3.6.1 was used
   to for visualizing PPI network. Finally, Kaplan–Meier analysis for hub
   proteins was performed by Kaplan–Meier plotter online platform.

Results

   A total of 386 genes originating from the exosomes derived from BMSCs
   were identifed as statistically signifcant (P < 0.05, FDR <0.05), which
   consisted of 150 upregulated genes and 236 downregulated genes. Also,
   32 pathways were identifed as signifcant (P < 0.05, FDR < 0.05). The
   PPI network of exosomes derived from BMSC proteins included 100 protein
   nodes with 579 interaction edges. The hub proteins, including PODN,
   ZNF521, and CFI, which interacted with ten or more other proteins, were
   indicated as the hub proteins of PPN of exosomes derived from BMSCs.

Conclusion

   Taken together, our findings revealed the prognostic roles of mRNAs of
   exosomes derived from BMSCs and provided implications for targeted
   therapy for common malignant neoplasms. However, further studies
   require large samples and experimental verification.

   Keywords: bioinformatics, BMSCs, clinical outcome, exosome, malignant
   neoplasms

Introduction

   Mesenchymal stromal cells (MSCs), also known as mesenchymal stem cells,
   have the potential for self-renewal and multidirectional
   differentiation.[28]^1 Considering the source, MSCs can be isolated
   from multiple tissues, such as bone marrow, umbilical cord, blood,
   placenta, and adipose tissue.[29]^2^–[30]^5 It was reported that bone
   marrow mesenchymal cells (BMSCs) that originated from the bone marrow
   can secrete a diverse variety of bioactive factors in an autocrine
   and/or paracrine way, which was expected to become a non-cellular
   therapy for the treatment of various diseases, especially
   tumors.[31]^6^–[32]^9 Of them, an important paracrine factor of BMSCs
   is the exosome, which is a lipid membrane wrapped vesicle with a
   diameter of 40–200 nm. Exosome is abundant in miRNA, mRNA, lipids, and
   proteins. It has been reported that exosomes enter the recipient cells
   by endocytosis, or directly fuse with the recipient cells to exchange
   information between cells.[33]^10^–[34]^13 Exosomes derived from BMSCs
   have been able to treat proangiogenic, cardiovascular disease, liver
   disease, lung injury, and kidney injury.[35]^14^–[36]^18 Indeed, there
   are two opposite viewpoints on the role of exosome in the treatment of
   tumors. Some studies have suggested that exosomes have a strong
   anti-tumor effect, while other studies have shown that exosomes derived
   from MSCs can promote the growth and metastasis of cancer cells.
   Disappointingly, the mechanism of antitumor effect of MSC-derived
   exosomes remains unclear.[37]^19

   Malignant tumors cause an extremely large burden around the world,
   second only to cardiovascular disease.[38]^20^,[39]^21 Despite using
   combinations of various strategies of diagnosis and treatment, the
   prognosis of cancer patients remains poor.[40]^22 Therefore, it is
   necessary to explore multiple effective early diagnostic and prognostic
   markers. Recently, it has been reported that exosomes are involved in
   cancer development, progression, and resistance; therefore, they may be
   used as multiple targets for the development of a diagnostic tool for
   early detection of cancer.[41]^23^–[42]^26 Along with the advances in
   microarray and life molecules sequencing technology, high-throughput
   research methods are widely used in biomedicine. High-throughput and
   corresponding bioinformatics analysis of exosomes is helpful to guide
   the diagnosis and prognosis of diseases, especially
   cancer.[43]^27^,[44]^28 In this study, our goal was to identify the
   prognostic roles of differentially expressed genes originating from
   exosomes derived from BMSCs in common cancers, which provide new
   diagnostic and therapeutic targets for cancers.

Materials and methods

Microarray data extraction and identifying differentially expressed genes

   We downloaded the gene expression profile ([45]GSE78235) from the
   public availability repository Gene Expression Omnibus (GEO;
   [46]https://www.ncbi.nlm.nih.gov/geo/) database. GEO2R online tool
   allows researchers to carry out R-based analysis of GEO data to
   identify differentially expressed genes (DEGs). Subsequently, the
   Database for Annotation, Visualization, and Integrated Discovery
   ([47]https://david.ncifcrf.gov/) v6.8 was used to perform re-annotation
   for DEGs. A gene is defined as a DEG between the tumor sample and the
   normal control sample when the false discovery rate (FDR) is <0.05 (FDR
   ≤0.05) and the fold change (FC) is at least two times higher or lower
   (|log2FC|≥1).

Functional enrichment analysis for DEGs

   Functional enrichment analysis was performed for DEGs to give a
   functional overview of the DEGs by calculating the whole significance
   of the gene expression. Gene ontology (GO) comprises molecular
   function, cellular component, and biological process. Kyoto
   Encyclopedia of Genes and Genomes (KEGG) was used to clarify how DEGs
   perform function through a certain path. We selected Functional
   Enrichment analysis tool (FunRich v3.1.3), which is designed to handle
   a variety of gene/protein datasets irrespective of the organism, to
   perform pathway enrichment analysis. FDR <0.05 was set as the cut-off
   criterion.

Construction of protein–protein interaction network for DEGs and
identification of hub proteins

   Protein–protein interaction (PPI) was used to analyze the
   interrelationship between differentially expressed proteins, and
   further elucidate the mechanisms of genes playing important roles in
   physiological and pathological conditions. The STRING database
   ([48]http://string-db.org) helps to provide a critical assessment and
   integration of protein–protein interactions, including direct
   (physical) and indirect (functional) associations. In this study, we
   used STRING database v10.5 to construct PPI network for DEGs. The
   parameters were set as follows: meaning of network edges: confidence;
   minimum required interaction score: medium confidence (0.400). Then,
   Cytoscape software v3.6.0 was used to visualize the PPI network. At
   last, Cytoscape software plug-in Molecular Complex Detection was used
   to screen hub proteins within the PPI network.

Kaplan–Meier (KM) plot analysis for hub proteins

   We ran a KM analysis for three hub genes to characterize the
   association of gene expression and corresponding clinical outcome in
   common malignant tumor. KM analysis is available on the website
   [49]http://www.kmplot.com/. The KM plotter is capable to assess the
   effect of any gene or gene combination on survival in breast, ovarian,
   lung, gastric, liver cancer patients using over 30,000 samples measured
   using gene chips or RNA-seq.

Results

Identification of DEGs from BMSCs-derived exosomes

   Based on GEO2R online platform, using FDR<0.05 and (|log2FC|≥1) as the
   cut-offs, 386 genes were identified to be differentially expressed in
   the tumor compared with the normal control samples, which included 150
   upregulated genes and 236 downregulated genes.

GO enrichment analysis

   The eight important terms are obtained by GO enrichment analysis in the
   software FunRich as shown in [50]Figures 1A–C.) The genes were enriched
   in biological processes of response to signal transduction, cell
   communication, cell growth and maintenance, and energy pathways
   metabolism. As for the molecular function, these genes showed
   enrichment in GTPase activity, chaperone activity, catalytic activity,
   calcium ion binding, and receptor signaling complex scaffold activity.
   Besides, the cell component indicated enrichment predominantly at
   exosomes, cytoplasm, lysosome, nucleus, and plasma membrane.

Figure 1.

   [51]Figure 1
   [52]Open in a new tab

   The eight important terms are obtained by GO enrichment analysis in the
   software FunRich.

   Notes: (A–C) The top eight terms of cell components, biological
   processes and molecular functions of GO enrichment analysis for DEGs,
   respectively. (D) The top eight terms of biological pathway of KEGG
   enrichment analysis for DEGs.

   Abbreviations: DEGs, differentially expressed genes; GO, Gene Ontology;
   KEGG, Kyoto Encyclopedia of Genes and Genomes; HES/HEY, hairy/enhancer
   of split.

KEGG pathway analysis

   KEGG pathway enrichment analysis suggested that the genes were
   significantly enriched in pathways including vascular endothelial
   growth factor (VEGF) and VEGF receptor (VEGFR) signaling network,
   thrombin/protease-activated receptor pathway, PAR1-mediated thrombin
   signaling events, signaling events mediated by VEGFR1 and VEGFR2,
   integrin family cell surface interactions, IL5-mediated signaling
   events, beta1 integrin cell surface interactions, syndecan-1-mediated
   signaling events, plasma membrane estrogen receptor signaling, and
   proteoglycan-mediated signaling events ([53]Figure 1D).

PPI network construction and hub module analysis

   We constructed a network for DEGs from the exosomes derived from BMSCs
   that included 579 interaction edges and 150 proteins nodes ([54]Figure
   2A). Next, we obtained four key modules by means of Cytoscape plug-in
   Molecular Complex Detection. At last, the three proteins that
   interacted with at least ten other proteins (P<0.05, FDR <0.05), which
   included PODN, ZNF521, and CFI, were identified as the hub proteins
   from four key modules ([55]Figure 2B–E).

Figure 2.

   [56]Figure 2
   [57]Open in a new tab

   Through PPi network construction and hub module analysis identifies
   four central modules of key proteins.

   Notes: (A) Protein–protein interaction network of the selected DEGs.
   (B–E) The four key gene modules calculated by MCODE software from the
   protein–protein interaction network. The red circle stands for the hub
   genes/proteins.

   Abbreviations: DEGs, differentially expressed genes; MCODE, Molecular
   Complex Detection.

Survival analysis of hub proteins

   We ran a KM plot analysis for three genes to visualize the association
   of hub genes expression and match the clinical outcome in breast
   cancer, ovarian cancer, lung cancer, and gastric cancer. The
   upregulation of PODN, ZNF521, and CFI indicated good survival in breast
   cancer ([58]Figure 3A–C). Downregulation of PODN and ZNF521 indicated a
   good outcome in ovarian cancer ([59]Figure 3D and E), while the
   upregulation of CFI indicated a good survival in ovarian cancer
   ([60]Figure 3F).The downregulation of PODN, ZNF521, and CFI indicated
   good survival in lung cancer [61]Figure 3G–I), while the upregulation
   of CFI indicated a good survival in gastric cancer ([62]Figure 3L).

Figure 3.

   [63]Figure 3
   [64]Open in a new tab

   The three hub of the survival analysis for cancer datasets by
   Kaplan–Meier.

   Notes: (A–C) Hub genes/proteins (PODN, ZNF521, CFI) of the survival
   analysis for breast cancer dataset by Kaplan–Meier. (D–F) Hub
   genes/proteins (PODN, ZNF521, CFI) of the survival analysis for ovarian
   cancer dataset by Kaplan–Meier. (G–I) Hub genes/proteins (PODN, ZNF521,
   CFI) of the survival analysis for lung cancer dataset by Kaplan–Meier.
   (J–L) Hub genes/proteins (PODN, ZNF521, CFI) of the survival analysis
   for gastric cancer dataset by Kaplan–Meier.

   Abbreviation: HR, hazard ratio.

Discussion

   Malignant tumor is a major public health problem that threatens human
   health and is second only to cardiovascular disease. Unfortunately,
   tumors lack early diagnostic and prognostic markers with a high
   sensitivity and specificity. Our study focuses on the critical role of
   mRNA of exosome derived from BMSCs. In our study, by extracting the
   gene expression information from GEO database, we obtained 386 DEGs,
   including 150 upregulated genes and 236 downregulated genes from the
   exosome derived from BMSCs. Next, we uncovered that those DEGs were
   mainly involved in proliferation and angiogenesis, such as VEGF and
   VEGFR signaling pathway. By constructing PPI network and identifying
   the hub genes, we obtained three hub proteins, including PODN, ZNF521,
   and CFI. These three genes are closely related to the clinical
   prognosis of common malignant tumors, including breast cancer, ovarian
   cancer, lung cancer, and gastric cancer, based on KM online platform.
   In addition, three hub proteins have been reported less in the context
   of BMSCs. Moreover, the multiple roles of the three genes in common
   malignant neoplasm were also less studied. As illustrated in [65]Figure
   3, the upregulation of PODN, ZNF521, and CFI indicated good survival in
   breast cancer ([66]Figure 3A–C) and lung cancer ([67]Figure 3G–I),
   compared with the downregulation of three hub genes. Downregulation of
   PODN and ZNF521 indicated a good outcome in ovarian cancer ([68]Figure
   3D and E), while the upregulation of CFI indicated a good survival in
   ovarian cancer ([69]Figure 3F). Besides, the downregulation of PODN and
   ZNF521 indicated a good outcome in gastric cancer ([70]Figure 3J and
   K), while the upregulation of CFI indicated a good survival in gastric
   cancer ([71]Figure 3L). As we know, malignant tumor is a group of
   diseases with high heterology. Moreover, a gene whether it is a tumor
   promoter or inhibitor depends on the microenvironment of the tumor.

   However, we have noticed that the upregulation of PODN, ZNF521, and CFI
   indicated good survival in breast cancer ([72]Figure 3A–C) and lung
   cancer ([73]Figure 3G–I), compared with the downregulation of three hub
   genes. We make a bold assumption: the three hub genes may be used in
   the clinic for breast and lung cancers.

   PODN was a seed gene in the first key modules. PODN, which is an
   important component of the extracellular matrix protein, belongs to the
   small leucine-rich repeat protein family.[74]^29 It was reported that
   PODN may bind type 1 collagen, thereby reducing the cell growth and
   migration in kidney. It is unfortunate that no literature has reported
   the role of PODN in tumors. ZNF521 was a seed gene in the second
   modules. ZNF521 is a zinc finger protein that consists almost entirely
   of 30 C2H2 Kruppel-like zinc fingers. Moreover, ZNF521 has been
   characterized as a potent inhibitor of EBF1 and is emerging as a
   potentially relevant contributor to the development of B-cell
   leukemias.[75]^31^,[76]^32 ZNF521 contributes to clonogenic growth,
   migration, and tumorigenicity in medulloblastoma cells through the
   recruitment of the NuRD complex.[77]^30 Furthermore, ZNF521 sustains
   the differentiation block in myeloid-lymphoid leukemia-rearranged acute
   myeloid leukemia. CFI was a seed gene in the fourth modules. CFI is a
   serum protease that inhibits all complement pathways. The common
   variant rs10033900 near the CFI gene is associated with age-related
   macular degeneration risk in Han Chinese population. Based on the gene
   expression profile, we found that upregulated CFI indicates good
   prognosis, while Okroj et al suggested that high expression of CFI is
   associated with poor prognosis and recurrence in breast cancer.[78]^33

   In our study, we systematically characterized DEGs from the exosomes
   derived from BMSCs and identified certain hub proteins and their PPI
   networks. Taken together, our findings suggested that three genes may
   be potential prognostic markers in common malignant tumors. However,
   the study results were obtained based on bioinformatics, which only
   previously considered datasets. However, further studies are required
   with large samples and for experimental verification.

Acknowledgments