Analysis of Differentially Expressed Genes Associated With Alzheimer’s Disease Based on Bioinformatics Methods

Introduction

Alzheimer’s disease (AD) was first described ¹ by German neurologist Alois Alzheimerin in 1907. As one of the most common types of dementia, AD had characteristics such as progressive memory impairment, cognitive dysfunction, personality changes, language barriers, and other neuropsychiatric symptoms. ² With the current increasing disease rate of AD, previous studies indicate the total number of people with AD dementia in 2050 to be 13.8 million. ³ Nowadays, clinical diagnostic criteria for AD, especially early diagnosis, are not sufficient. Besides, studies on pathogenesis of AD are still unclear, especially the genetic mechanisms of AD.

Considering the huge damage of AD, preventive measures are urgently needed. In the past decades, research of AD mostly focused on the molecular mechanism. There are 5 main pathogenesis theories about AD including cerebral accumulation of amyloid β (Aβ) protein, ⁴ central nervous cholinergic damage, ⁵ excitatory amino acids, ⁶ abnormal phosphorylation of τ protein, ⁷ and dysregulation of intracellular calcium. ^8,9 Moreover, many genes have proved the association with AD. Amyloid protein precursor (APP), ¹⁰ presenilin 1 (ps-1), ¹¹ and presenilin-2 (ps-2) ¹² have been confirmed as virulence gene of familial AD, while apolipoprotein E (apoE) ¹³ has been proved tightly relevant to sporadic AD. Meanwhile, the mutation of APP is proved to cause an amino acid substitution of the carboxy terminus of the β-amyloid (Aβ) peptide. ¹⁰ Additionally, mutant PS1 influences APP processing both in vitro and in vivo, and the elevated extracellular concentrations of amyloidogenic Aβ1-42 peptides are precipitated disease in PS1-linked familial AD. ¹⁴ Furthermore, apoE is now considered to have best-established genetic association with AD, ¹⁵ especially the apoE ∊4 which is considered as the only well-verified susceptibility gene. The effect of apoE ∊4 on AD is influenced by age and ethnicity. ¹⁶ Although previous studies have identified several potential genes and proteins as determinants of AD, the needs for more research to elucidate the mechanism of AD are highlighted.

In this article, differentially expressed genes (DEGs) between AD samples and normal samples were identified based on the data downloaded from Gene Expression Omnibus (GEO) database. Bioinformatics methods were used to construct a protein–protein interaction (PPI) network of DEGs. Meanwhile, function and pathway annotation of DEGs were performed based on Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways databases. The present study hoped to provide a new view and evidence for the mechanism of AD.

Materials and Methods

Results

Analysis of DEGs

A total of 12 DEGs were identified (Figure 1). Clusters analysis of DEGs showed that the upregulated DEGs included dentin matrix protein 2 (DMP2), hormonally upregulated Neu-associated kinase (HUNK), neuropilin and tolloid-like 1 (NETO1), and solute carrier family 24, member 3 (SLC24A3), while the downregulated DEGs included complement component (3b/4b) receptor 1 (CR1), EF-hand calcium binding domain 6 (EFCAB6), glutathione S-transferase theta 1 (GSTT1), interleukin 1 receptor-like 1 (IL1RL1), olfactory receptor, family 7, subfamily A, member 5 (OR7A5), palladin, cytoskeletal associated protein (PALLD), tripartite motif containing 5 (TRIM5), and titin (TTN).

OPEN IN VIEWER

Figure 1.

Dendrogram of differentially expressed genes (DEGs) by cluster analysis. The below abscissa axis represents specimen: NDC represents normal sample, AD represents sample of Alzheimer’s disease. The above abscissa axis represents the clustering of specimens. The right longitudinal axis represents genes. Red shows upregulated genes while green shows downregulated genes.

Construction of PPI Network

The PPI network was constructed based on IPA database (Figure 2A). To further investigate the potential function of significant DEGs and their related proteins, the nodes and lines of top 5 DEGs were abscised (Figure 2B). The nodes of top 5 DEGs were ubiquitin C (UBC), β-estradiol, APP, interferon gamma (IFNG), and Ca²⁺. Moreover, 8 upregulated genes including TTN, CR1, IL1RL1, EFCAB6, PALLD, OR7A5, GSTT1, and TRIM5 as well as 4 downregulated genes including HUNK, DMP2, SLC24A3, and NETO1 were revealed in this PPI network.

OPEN IN VIEWER

Figure 2.

Protein–protein interaction (PPI) network of different expression genes (DEGs) based on Information-technology Promotion Agency Japan (IPA) database. A, PPI network of DEGs based on IPA database. B, PPI network after nodes and lines of top 5 were abscised. The red points stand for upregulated genes, the green points stand for downregulated genes, and the white points stand for genes which have PPI with DEGs. The size of node corresponds to the degree of this node. The thickness of the lines represents the ligation: the thicker the line is, the more important this connection is.

Function and Pathway Enrichment Analysis of DEGs

Gene Ontology functional enrichment indicated that DEGs were significantly enriched into several GO terms, such as α-actinin binding, interleukin 33 receptor activity, and telethonin binding. Top 10 of GO terms are listed in Table 1. Besides, the result of KEGG analysis revealed 3 enriched KEGG pathways: glycosylphosphatidylinositol (GPI)-anchor biosynthesis, N-glycan biosynthesis, and glutathione metabolism (Table 2).

OPEN IN VIEWER

Table 1.

Significantly Enriched Gene Ontology Terms of Differentially Expression Genes (Top 10).^a

GO ID	P Value	Count	Term
GOTERM_BP_ALL
GO:0031033	.002108	1	Myosin filament assembly or disassembly
GO:0018406	.002108	1	Protein amino acid C-linked glycosylation via 2'-α-mannosyl-L-tryptophan
GO:0018103	.002108	1	Protein amino acid C-linked glycosylation
GO:0018211	.002108	1	Peptidyl-tryptophan modification
GO:0035269	.002108	1	Protein amino acid O-linked mannosylation
GO:0030241	.002108	1	Muscle thick filament assembly
GO:0048739	.00281	1	Cardiac muscle fiber development
GO:0030240	.003511	1	Muscle thin filament assembly
GO:0055003	.004212	1	Cardiac myofibril assembly
GO:0045087	.005142	2	Innate immune response
GOTERM_MF_ALL
GO:0051393	.000721	1	α-actinin binding
GO:0002114	.000786	1	Interleukin 33 receptor activity
GO:0031433	.000786	1	Telethonin binding
GO:0008273	.001572	1	Calcium, potassium–sodium antiporter activity
GO:0051371	.001572	1	Muscle α-actinin binding
GO:0004875	.002357	1	Complement receptor activity
GO:0004582	.003141	1	Dolichyl-phosphate β-d-mannosyltransferase activity
GO:0004908	.005491	1	Interleukin 1 receptor activity
GO:0043621	.007836	1	Protein self-association
GO:0015491	.013288	1	Cation–cation antiporter activity
GOTERM_CC_ALL
GO:0031501	.002058	1	Mannosyltransferase complex
GO:0000506	.002058	1	Glycosylphosphatidylinositol-N-acetylglucosaminyltransferase (GPI-GnT) complex
GO:0000932	.015004	1	Cytoplasmic mRNA processing body
GO:0005884	.019735	1	Actin filament
GO:0030018	.023101	1	Z disk
GO:0000794	.025116	1	Condensed nuclear chromosome
GO:0030176	.027797	1	Integral to endoplasmic reticulum membrane

Abbreviations: GO, Gene Ontology; mRNA, messenger RNA.

^a P value less than .05 were consider to be significantly different.

OPEN IN VIEWER

Table 2.

The Enriched KEGG Pathways.

KEGG ID	P Value	Count	Term
563	.024489	1	Glycosylphosphatidylinositol (GPI)-anchor biosynthesis
510	.042779	1	N-Glycan biosynthesis
480	.048497	1	Glutathione metabolism

Abbreviation: KEGG, Kyoto Encyclopedia of Genes and Genomes.

Discussions

In the present study, a gene expression profile downloaded from GEO was used to explore the possible candidate genes of AD. A total of 12 DEGs (4 upregulated genes and 8 downregulated genes) were identified between the normal samples and samples with AD.

Protein–protein interaction network showed that TTN, CR1, IL1RL1, EFCAB6, PALLD, OR7A5, GSTT1, and TRIM5 were upregulated genes and HUNK, DMP2, SLC24A3, and NETO1 were downregulated genes. As a kind of protein degradation pathway, ²⁴ ubiquitin proteasome pathway (UPP) dysregulation could induce the overphosphorylation of τ and degradation dysfunction of Aβ in AD. ^25,26 The TRIM proteins consist of an N-terminal E3 ubiquitin ligase Really Interesting New Gene domain which has been demonstrated to possess E3 ubiquitin ligase activity in vitro allowing self-polyubiquitylation ²⁷ and played roles in the polyubiquitylation and turnover of the protein. ²⁸ The present study showed that TRIM5 was overexpressed in samples with AD group which inferred that the overexpression of TRIM5 might activate the UPP and increase decomposition of paraprotein. Besides, as TTN is reported to modulate Ca²⁺ homeostasis, ²⁹ and Ca²⁺ is an important factor in AD, ³⁰ there is no surprise that TTN is differently changed in AD. Moreover, previous study indicates that NETO1 is associated with neurological diseases by the effects on the maintenance of neural circuitry and synaptic plasticity. ³¹ Drugs that bind to or modulate the activity of protein encoded by PALLD gene are suggested for the treatment of AD. ³² Genetic dysregulation of the IL1RL1 axis appeared to be involved in conferring predisposition to AD, ³³ and the IL1RL1 pathway plays an important role in AD. ³⁴ Consistent with the above-mentioned studies, our results indicated that NETO1, PALLD, and IL1RL1 might be involved in AD. Furthermore, an association between AD risk and markers spanning CR1 has been observed, ³⁵ and human erythrocytes, which abundantly expressed CR1, was able to sequester Aβ ^36,37 and to favor its clearance via the C3b-mediated adherence to erythrocyte CR1. ³⁷ In a word, CR1 played a protective role via the generation and binding of C3b, which might contribute to Aβ clearance. ³⁸ Therefore, we speculated that CR1 might play a role in AD via effecting on Aβ. There was no literature drawn that EFCAB6 and OR7A5 were related to AD. Thus, the present study might add a new candidate to the list of genes involved in AD. The PPI network also showed that β-estradiol might enhance the expression of DMP2, HUNK, and scl24A3 and decrease the expression of GSTT1. Figure 1 shows DMP2, HUNK, and SLC24A3 are low expressed, and GSTT1 is overexpressed in samples with AD. Previous studies indicate that the expression of HUNK may be regulated by β-estradiol, ³⁹ and GSTT1 is involved in estrogen metabolism. ⁴⁰ The changed expression of DMP2, HUNK, scl24A3, GSTT1, and β-estradiol might result in the interactions between β-estradiol and these genes, which needed further studies. The PPI network result also showed that UBC, ⁴¹ β-estradiol, ¹⁰ APP, ⁴² IFNG, ⁴³ and Ca²⁺ were³⁰ important factors in AD, which have been reported to have important relevance with the pathogenesis of AD. These 5 factors might be involved in the 5 pathogenesis theories of AD which was mentioned earlier. ^{4 –9} Besides, β-estradiol has already been used as a potential therapeutic drug for AD, ⁴⁴ while APP has been considered as the criteria for the diagnosis of AD. ⁴⁵ This study indicated that DEGs that were related to UBC, IFNG, and Ca²⁺ pathways might be the important potential therapeutic targets for AD.

Additionally, the analysis of KEGG in the present study revealed 3 enriched pathways: GPI-anchor biosynthesis, N-glycan biosynthesis, and glutathione metabolism. Amyloid β, which is thought to be one of the causes of AD, is a cleaved fragment from N-glycans of amyloid precursor protein, and the number of N-glycans having a bisecting GlcNAc residue is proved in AD brains. ⁴⁶ Considering the above-mentioned information, we hypothesized that N-glycan biosynthesis might be involved in the pathogenesis of AD. Besides, research that focused on glutathione metabolism demonstrated the induction of glutathione reductase and glutathione peroxidase messages in the AD hippocampus ⁴⁷ and showed that decreased glutathione content might be involved in AD pathology in humans. ⁴⁸ However, there is no evidence to support that GPI-anchor biosynthesis is involved in AD.

In conclusion, the DEGs of AD were analyzed by computational bioinformatics approaches. The present study indicated that TTN, CR1, IL1RL1, EFCAB6, PALLD, OR7A5, and GSTT1 were upregulated genes and HUNK, DMP2, SLC24A3, NETO1 were downregulated genes that might play meaningful roles in AD. Meanwhile, the enriched pathway revealed several pathways including GPI-anchor biosynthesis pathway, N-glycan biosynthesis pathway, and glutathione metabolism pathway. The present study indicated several potential target molecules to treat the AD and might shed new light on the mechanism of AD.