Identication of Potential Biomarkers and Pathways in Neonatal Hypoxic-Ischemic Brain Injury: Based on Bioinformatics Technology

Neonatal hypoxic-ischemic brain damage (HIBD) is one of the most common serious diseases in newborns, with a high mortality and disability rate. This study aims to use the bioinformatics analysis to identify potential hematologic/immune systems tissue-specic genes and related signaling pathways neonatal HIBD. Methods Microarray datasets in HIBD were downloaded from the Gene Expression Omnibus database, and DEGs were identied by R software.Enrichment analyses were performed and protein–protein interaction networks were constructed to understand the functions and enriched pathways of DEGs and to identify central genes and key modules. The results The also showed similar results in biological and KEGG enrichment.and four hematologic/immune system tissues specically expressed potential biomarker genes were conrmed through a variety of methods, which were veried by GEO datasets published experimental signaling pathways, complement and pathways, hematopoietic cell neuroactive ligand receptor interaction, signaling pathway.


Introduction
Hypoxic ischemic brain injury (HIBD), also known as hypoxic-ischemic encephalopathy (HIE), is one of the most common serious diseases in newborns. In developed countries, HIBD occurs in 1-8 cases per 1,000 live births, and 10-26 cases per 1,000 in low and middle income countries [1][2]. In the long term, HIBD can cause cerebral palsy, epilepsy, visual impairment and motor cognitive impairment [3]. At present, neonatal HIBD is mainly based on the clinical manifestations of the nervous system, imaging and electrophysiological examinations to diagnose and predict prognosis [4]. These methods of diagnosis and prognostic judgment require a long time and cause delay in diagnosis. So that delay in diagnosis oftentimes leads to that many HIBD babies miss the optimal treatment time, which may cause different levels of neurological sequelae.With the rapid development of biomedicine, there have been many reports on biomarkers related to neonatal HIBD [5]. However, there is no blood-based marker that is stable enough to diagnose HIBD or predict prognosis [6]. Currently, there is still a lack of effective drugs to treat HIBD, and prevention is the key to this disease. In recent years, mild hypothermia neuroprotective therapy has become the standard for clinical treatment of HIBD, and other adjuvant therapies such as erythropoietin and neuroprotective agents are rapidly transitioning from basic science to clinical application [7][8].
However, pharmacological treatments were solely focused on neurons. Gradually, glial cells started to be considered as alternative targets for neuroprotection [9][10].
More and more evidences show that hypoxia-ischemia can induce in ammation in the brain parenchyma and peripheral immune system, which plays a vital role in mediating secondary neuronal death [11][12].The brain is considered a place of immune privilege because the blood-brain barrier restricts bloodborne cells and proteins from entering the central nervous system (CNS) [13].In the central nervous system, microglia, astrocytes, and oligodendrocytes play the roles of in ammatory cells and immune cells. There are currently studies that the sequelae of the nervous system are related to the activation of astrocytes [14].The in ammatory response of the brain can lead to the activation of astrocytes and the activation and aggregation of microglia to provide immune defense for nerve cells, which can lead to rapid changes in their gene expression [15].The research team of Dr. Ting Qi and Professor Jian Yang found that the blood has genetic targets that can be used to identify brain-related properties (such as IQ) and diseases (such as schizophrenia) [16].This discovery provides a new research idea for brain disease research.
Therefore, this study aims to apply the public gene expression comprehensive database (GEO) for bioinformatics analysis to determine the differentially expressed genes (DEGs) in neonatal cerebral cortex tissue and blood.Our research study focused on nding the potential hematologic/immune system tissue speci city gene and related signal pathways that are closely related to HIBD and revealing the underlying molecular mechanisms.

Microarray Data
The expression pro le datasets of HIBD were searched from the publicly available Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). The microarray datasets generated from the cerebral cortical in rats with HIBD and from the whole blood in human with HIBD and controls were used in the study. Four gene expression pro les include GSE37777, GSE18356, GSE121178 and GSE112137 were obtained.The details of the datasets are shown in Table 1.

Identi cation of differentially expressed genes DEGs
We analyzed the DEGs of healthy and HIBD samples in the cerebral cortex tissue and blood by using the limma software package, especially the raw data of datasets GSE121178 and GSE37777 were extracted with the affy package and individually normalized with the robust multi-array average (RMA) package, and batch effects were eliminated between experiments by applying the ComBat function in the SVA package. The DEGs cut-off thresholds of both are set as follows: (1) p<0.05; (2) log 2 (FoldChange, FC)> 0.585 or log 2 (FoldChange, FC) <-0.585.

GO Enrichment and KEGG Pathway Analysis
Gene ontology (GO) includes three types of data, namely, cellular components (CC), molecular functions (MF), and biological processes (BP) [17]. GO is often used to annotate genes according to a de ned set of structured words. The Kyoto Encyclopedia of Genes and Genomes (KEGG) is used to match the information of gene pathways [18]. We used the DAVID 6.8 (https://david.ncifcrf.gov) online tool [19] to perform GO biological process (GO-BP) analysis on DEGs in HIBD, and use KOBAS 3.0 (http://kobas.cbi.pku.edu.cn/) online tool for KEGG pathway analysis [20]. P<0.05 was selected as the threshold to identify the signi cant enrichment of GEGs in GO terms and signal pathways.

PPI network construction, central genes and key module Analysis
The tool for searching gene/protein interactions (STRING: http://string-db.org/) is an online biological database designed to evaluate the protein-protein interaction(PPI) network. Intersecting the DEGs of cerebral cortex tissue and whole blood to obtain a common DEGs. According to the information in the STRING database (PPI score> 0.4), we selected the protein related to the common DEGs (species: Homo), and used the Cytoscape software (http://cytoscape.org/ ) build a PPI network. Then we used ve methods(degree, Maximum Neighborhood Component (MNC), Radiality centrality, Stress centrality, Closeness centrality) to sequence and evaluate central genes . In addition, we used the MCODE algorithm to obtain key modules.

Validation of central genes and determination of tissue-speci c biomarker genes
We identi ed the central genes by using the Genecards database [21]and further determine the genes potentially related to the disease. The central genes was located for further determine the central genes expressed in the hematologic/immune system by using the Biogps database [22].The results obtained from the two databases were integrated to obtain HIBD hematologic/immune system tissue speci c biomarker genes. We applied the GSE121137 data set for external veri cation of tissue-speci c genes.Box-plot analysis and receiver operating characteristic(ROC) curve were performed to display tissue-speci c genes for external veri cation of the diagnostic potential of HIBD by using GraphPad Prism8.0 soft.

Prediction of transcription factors and miRNAs of key central genes
The CHEA3 database [23] are used to predict the relationship between key central genes and transcription factors (TF); The mirDIP database [24] was identi ed to predict the relationship between key central genes and miRNA.Network analysis was performed by running Cytoscape application,and the key TF and key miRNA was determined according to the Degree value.

Identi cation of Differentially Expressed Genes (DEGs)
In the cerebral cortex tissue, 2598 DEGs were screened, including 2362 up-regulated DEGs and 236 downregulated DEGs. In the blood, 1442 DEGs were screened, including 540 up-regulated DEGs and 902 downregulated DEGs. The volcano map in Figure 1A-B displays the general distribution of these genes, and expression of the top 30 up-regulation DEGs and the top 30 down-regulation DEGs of two kinds of samples were showed in the heatmap in Figure 1C-D.

GO Term Enrichment Analysis and KEGG Pathway Analysis
We applied the DAVID online tool and the KOBAS online tool to cluster the GO protein function and the KEGG signal pathway of the DEGs of HIBD. As shown in GO analysis results ( In general, the enrichment results of GO and KEGG of DEGs in the cerebral cortex tissue and DEGs in whole blood are related to hypoxic-ischemic brain injury, and both involve in ammation, immunity and apoptosis.

Common DEGs identi ed in the cerebral cortex tissue and blood of HIBD cases
There were 86 genes overlapping between DEGs in the cerebral cortex tissue with HIBD and DEGs in whole blood with HIBD. As shown in Figure 3, these 86 common DEGs included 67 up-regulation and 19 down-regulation.The biological processes that these common DEGs mainly involved in ammation and immunity ( Figure 4A, only showing the top 15 items in terms of gene enrichment), including participation in neutrophils, macrophage-mediated in ammation, platelet activation, in ammatory cells activation,participates in immune response, regulation of lymphocyte differentiation. Cross-examination of common DEGs and GO biological process terms showed ( Figure 4B) that some genes related to in ammation were also enriched in immune-related biological processes, indicating that these genes may be related to the in ammation-immune damage mechanism in HIBD.The enrichment results of cellular components and molecular functions ( Figure 4C) showed that the cellular components of common DEGs were mainly enriched in plasma membrane, extracellular body, cytoskeleton, extracellular zone, cell surface, extracellular space, and oscillating cilia; molecular functions mainly included binding to phosphotyrosine, phosphatase, interleukin-1, serine endopeptidase activity, phospholipase D activity, interleukin-1 receptor activity. Similarly, common DEGs were signi cantly enriched in the KEGG pathway ( Figure 4D, showing the signal pathway with FDR<0.05), mainly involving microbial infection, natural killer cell-mediated cytotoxicity, osteoclast differentiation, cancer pathway, FcεRI signaling pathways, complement and coagulation cascade pathways, hematopoietic cell spectrum, neuroactive ligand receptor interaction, Rap1 signaling pathway.  (Table 2.): The GO-BP of these two modules GO mainly involved integrin-mediated signaling pathways, positive regulation of chemokine biosynthesis, and response to hypoxia;The results of KEGG mainly involved HIF-1 signaling pathway, hepatocellular carcinoma, Rap1 signaling pathway, neomycin, kanamycin and gentamicin biosynthesis, galactose metabolism, etc.
3.5 Validation of key central genes and determination of hematologic/immune system tissue-speci c biomarker genes Using Genecards database to identify 10 central genes, it was found that 7 central genes may be related to HIBD, and no related reports were found for the other 3 genes.Intersecting these 7 genes with the genes of 2 modules to get 4 genes. It was con rmed in the Biogps database that all 4 genes were clearly expressed in the hematologic/immune system, and these genes were de ned as HIBD hematologic/immune system tissue speci city potential biomarkers (Table 3).In this study, 32 samples in the GSE12137 data set (hypoxia model of early cortical progenitor cells) were used to verify 4 potential biomarkers, and it was found that two potential biomarkers were signi cantly up-regulated.The box plots of the two genes ( Figure 6A-B) were clearly differentially expressed in the hypoxia model of cortical progenitors cells; the ROC curve ( Figure 6C-D) showed that the two genes have good performance in identifying cases and controls AUC EGR1 =0.8398,AUC HMOX1 =0.9063 .In addition, a literature search was conducted on ITGAM and TYROBP that have not been veri ed by the data set, and related studies (qt-PCR test) showed that ITGAM and TYROBP were signi cantly expressed in the brain tissue of the hypoxicischemic brain injury model [25].This study believed that these 4 genes may be speci c potential biomarker genes for the diagnosis of hematologic/immune system in HIBD.Especially, EGR1 and HMOX1 may have a diagnostic effect on early HIBD

Prediction of transcription factors and miRNAs of key central genes
According to the comprehensive ranking of multiple sub-databases in the CHEPA3 database, the top 10 potential transcription factors with higher scores are selected to construct an interaction network ( Figure  7A).In the mirDIP database, we set the credibility parameter to very high, and obtained at least 88 miRNAs predicted by 7 sub-databases ( Figure 7B).9 key transcription factors (TFEC, SPI1, CEBPE, HLX, ZNF438, ZNF641, CEBPB, ELF4, LYL1) and 3 key miRNAs (hsa-miR-218-5p, hsa-miR-217, hsa-miR-377-3p) were screened according to the topoisomeric method,and their interactive network may be a potential key regulatory pathway in the pathogenesis of HIBD.

Discussion
Neonatal hypoxic-ischemic brain injury is an important cause of neonatal death and disability [26]. The pathogenesis of HIBD is very complicated. At present, there are no stable diagnostic biomarkers and specialized treatment methods for neonatal HIBD. So far, many studies have reported that in ammatory response and immune response play a key role in the development and recovery of hypoxic-ischemic brain injury. The activation of immune cells and the release of cytokines in the central and peripheral immune systems were involved [27][28].
This research obtained mRNA expression data about HIBD from the Public Gene Expression Comprehensive (GEO) database for bioinformatics analysis,and DEGs were identi ed in normal and case from cerebral cortex tissue samples and blood samples. We performed GO function annotation and KEGG pathway enrichment analysis on the DEGs of the cerebral cortex tissue and blood respectively.Regardless of GO function annotation or KEGG analysis, the functions and signal pathways related to in ammation, immune response and apoptosis were screened in the DEGs of the cerebral cortex tissue and blood, in particular, PI3K-Akt signaling pathway, MAPK signaling pathway, Rap signaling pathway and other classic signaling pathways have been reported to be related to brain injury [29][30].This similarity provided a feasible basis for exploring genetic changes that re ect cerebral hypoxicischemic damage in the blood. 86 overlapping genes were obtained in their DEGs.The GO biological process of common DEGs also involved in ammation and immune response, such as participation in neutrophil, macrophage-mediated in ammation, platelet activation, in ammatory cell activation, immune response lymphocyte differentiation regulation,etc;The KEGG signaling pathways mainly included natural killer cell-mediated cytotoxicity, osteoclast differentiation, FcεRI signaling pathway, complement and coagulation cascade pathways.We believed that the path most related to the disease is the natural killer cell-mediated cytotoxicity pathway. It was found in the KEGG mapper database that its downstream can trigger the apoptosis pathway [31].Unfortunately, the correlation between the natural killer cell-mediated cytotoxicity pathway and HIBD had not been con rmed by relevant experimental studies, and further studies on animal experiments or clinical experiments are still needed.
The PPI network of Common DEGs was constructed through the String database and visualized through cytoscape. Ten central genes (ITGAM, SYK, PLD1, KLRD1, NCKAP1L, EGR1, MAPK3, TYROBP, HMOX1, IL1R1) were identi ed through ve sequencing methods in cytocentralba.In addition, we analyzed the PPI network of the common DEG and identi ed 2 key modules through MCODE (score ≥ 3).Then use DAVID and KOBAS to analyze the GO and KEGG pathway enrichment analysis of DEGs in the two modules.Our results show that these two modules are involved in biological processes such as integrin-mediated signaling pathways, positive regulation of chemokine biosynthesis processes, and responses to hypoxia.Integrin is a type I heterodimeric transmembrane protein composed of α and β subunits [32].Previous studies have shown that the regulation of integrin-mediated signaling pathways such as MAPK and PI3K signaling pathways can resist mitochondrial apoptosis in hypoxic cardiomyocytes [33].Chemokines are a family of small proteins (8-10 kDa).It has been discovered earlier that the expression of some chemokines such as macrophage in ammatory protein-1 precedes the in ltration of immune cells into the brain and participates in the in ammatory response [34].KEGG results mainly involved HIF-1 signaling pathway, hepatocellular carcinoma, Rap1 signaling pathway, neomycin, kanamycin and gentamicin biosynthesis, galactose metabolism and other signaling pathways.The HIF-1 signaling pathway is involved in the transcriptional regulation of many genes after hypoxia/ischemia, and it also plays many different roles in protecting the brain during hypoxia/ischemia preconditioning, including vascular regulation, apoptosis regulation, oxidation, excitement,sexual toxicity regulation, immune regulation and so on [35].
Combining the 10 Central genes with the genes of the two key modules, we identi ed 4 key central genes TYROBP, ITGAM, ERG1 and HMOX1.At the same time, the Biogps database was used to test their expression levels in the hematologic/immune system.TYROBP, ITGAM, ERG1 and HMOX1 are central proteins in the PPI network, and are also common DEGs in HIBD cortical tissue and whole blood.We de ne these genes as HIBD hematologic/immune system tissue speci c biomarkers.In this study, TYROBP participated in the natural killer cell-mediated cytotoxicity signaling pathway and hematopoietic cell spectrum signaling pathway.TYROBP is a transmembrane signal transduction adaptor protein, which has both pro-in ammatory and anti-in ammatory effects[36].Albertsson's study found that after HI injury, high expression of TYROBP was found in the damaged area, including the subcortical white matter area, hippocampus, periventricular area, and meningeal area,and may be involved in the occurrence of white matter damage in premature infants [37].In this study, ITGAM participated in the cell adhesion molecule pathway, the leukocyte transendothelial migration pathway, and the Rap1 signaling pathway.Microglial cells increase signi cantly in mice and human brain injury and neurodegenerative diseases, and ITGAM expression also shows a positive change [38].Previous studies reported that the protein CD11B encoded by ITGAM was signi cantly increased after HI injury, and ITGAM has a pro-in ammatory effect in the mechanism of HI injury [39]. ITGAM can be combined with TYROBP in microglia to control the production of superoxide ions in microglia, thereby promoting neuronal apoptosis [40]. In this study, HMOX1 participated in the HIF-1 signaling pathway and angiogenesis pathway. HMOX1 has anti-in ammatory, anti-oxidative stress and anti-proliferation effects, can promote the survival of neurons and oligodendrocytes, and can reduce the degree of HIBD nerve damage [41][42]. In this study, EGR1 participated in the interleukin-1 mediated signaling pathway and GnRH signaling pathway.Previous studies have shown that ERG1 induced by hypoxia-ischemia can reduce the expression of BDNF and aggravate brain damage [43].We applied the GEO data set to verify 4 hematologic/immune systemspeci c genes and indirectly con rmed the continuous differential expression of ERG1 and HMOX1 in the pathogenesis of HIBD (24h-8 weeks). ROC curve analysis shows that ERG1 and HMOX1 have high diagnostic value for early HIBD.However, TYROBP and ITGAM have not been validated in the data set, which may have a certain relationship with the number of in vitro cell models in the data set deviating from the hematologic/immune system regulation. Published experimental documents have con rmed that they are signi cantly highly expressed in the HI model.In general, these 4 potential blood/immune biomarker genes as candidate molecular targets may provide potential references for the diagnosis, treatment and prognosis of BIBD. ERG1 and HMOX1 may be the potential hematology/immune system speci c expression of early HIBD. However, further large-scale experiments are needed to verify and clarify their speci c mechanisms.
This study also constructed a mutual regulatory network of key central DEGs-TFs and key central DEGs-miRNAs to identify their transcriptional and post-transcriptional regulatory factors.And identi ed their key transcription factors (TFEC, SPI1, CEBPE, HLX, ZNF438, ZNF641, CEBPB, ELF4, LYL1) and key miRNAs (hsa-miR-218-5p, hsa-miR-217, hsa-miR-377 -3p). miRNAs are non-coding RNA molecules that play a vital role in regulating a series of basic cellular processes,which may induce RNA silencing or act as regulators after DNA transcription [44]. Looney's study con rmed that the expression pro le of miRNA in the cord blood of newborn HIE can be changed [45]. Currently, studies showed that hsa-miR−217 can target EGR1 to reduce the in ammatory damage of endothelial cells induced by oxidized LDL [46]. Moreover, it is shown that Overexpression of hsa-miR-217 aggravates hypoxia-induced cardiomyocyte damage by inhibiting the expression of Silent Information Regulator 1[47]. However, no relevant reports have been found for these key miRNAs in hypoxia-induced brain damage. In the regulatory network, the synergistic effect of TF and miRNA is largely unclear.The identi cation and analysis of miRNA and TF may also reveal the potential pathogenesis of HIBD at the molecular level, and help promote the diagnosis, prognosis and treatment of HIBD.

Conclusion
In summary, the DEGs of HIBD, including potential peripheral biomarkers TYROBP, ITGAM, EGR1 and HMOX1, may play a role in the pathogenesis of HIBD through in ammation and immune-mediated signaling pathways.These potential regulatory genes and models may provide enlightenment for the discovery of new molecular targets for the diagnosis, prognosis and treatment of HIBD, and new breakthroughs in anti-in ammatory and immunotherapy are expected in the future.In order to clarify the role of these new peripheral marker genes in HIBD, more in vitro and in vivo studies and veri cations are needed.

Declarations Data Availability
The data used to support the findings of this study are included within the article. Jiayu Huang, Jinfeng Meng, Weichen Yan, and Jie Wang performed the data extraction and the results visualization. Changjun Ren and Ling Hao designed the study and amended the paper. All authors are responsible for reviewing data. All authors read and approved the nal manuscript.

Funding
This study was funded by the government-funded clinical medicine outstanding talent training project leader (team) project in 2020, which was named "Study on the expression and mechanism of microRNA in neonatal bilirubin encephalopathy" [NO.LS202006].

Con icts of Interest
The authors declare that there is no con ict of interest.    Venn diagram of the DEGs in the cerebral cortex tissue and blood. A total of 86 common genes were obtained.