Identification of Potential Key lncRNAs and Genes Associated with Aging Based on Microarray Data of Adipocytes from Mice

Objective. This study aimed to screen potential crucial lncRNAs and genes involved in aging. Methods. The data of 9 peripheral white adipocytes, respectively, taken from male C57BL/6J mice (6 months, 14 months, and 18 months of age) in GSE25905 were used in this study. Differentially time series expressed lncRNA genes (DE-lncRNAs) and mRNA genes (DEGs) were identified. After cluster analysis of lncRNAs expression pattern, target genes of DE-lncRNAs were predicted from the DEGs, and functional analysis for target genes was conducted. Results. A total of 8301 time series-related DEGs and 43 time series-related DE-lncRNAs were identified. Among them, 41 DE-lncRNAs targeted 1880 DEGs. The DEGs positively regulated by DE-lncRNAs were mainly related to the development of blood vessel and the pathways of cholesterol biosynthesis and elastic fibre formation. Furthermore, the DEGs negatively regulated by DE-lncRNAs were correlated with protein metabolism. Conclusion. These DE-lncRNAs and DEGs are potentially involved in the process of aging.


Introduction
Aging is an elevated risk of common diseases, including obesity, hypertension, atherosclerosis, and diabetes [1][2][3]. Currently, about 800 million people are at least 60 years old, which accounts for about 11% of the world's population, and aging population is estimated to increase to more than 2 billion by 2050 [4]. Aging is closely related to damaged adipogenesis in various fat depots in humans [5,6]. White adipose tissue (WAT) is considered as an important regulator for multiple physiological processes and highly linked to the development of multiple morbidities [7][8][9]. Therefore, it is significant and urgent to reveal the relationships of aging and adipose, which is very important for understanding the diseases in the elderly.
Previous studies have discovered a set of genes that are implicated in the aging process in an adipose depotdependent manner. For example, age-related increase in IL-6 (interleukin 6), which was related to stress responses and cellular senescence, was observed in a fat depot-dependent manner [5]. Sirt1 (sirtuin 1) and SOD2 (superoxide dismutase 2), which were correlated with mitochondrial aging, were significantly decreased in aging epididymal adipocytes [10].
Furthermore, the expression of MMP-3 (matrix metallopeptidase 3) was increased in mouse subcutaneous fat cells and human skin fibroblasts with aging [11,12]. Additionally, decreased expression in PPAR (peroxisome proliferatoractivated receptor gamma) through declining fat mass has been observed in monkey subcutaneous whole fat tissue [13]. In addition to these genes mentioned above, roles of long noncoding RNAs (lncRNAs) in age-related diseases have attracted more attention recently [14,15]. LncR-NAs are defined as the largest transcript class in human genome longer than 200 bp that lack protein-coding potential [16,17]. In aging murine aortas, mitochondrial lncRNA ASncmtRNA-2 is induced by replicative senescence [18]. Abnormal expression of the telomeric repeat-containing RNA lncRNA TERRA is responsible for premature senescence and aging through controlling telomere elongation [19]. In spite of much effort, the lncRNAs with known functions involved in aging remain rare.
Microarray technology has been widely used in molecular studies of human diseases [20,21]. Based on an age-related gene expression profile GSE25905, Liu et al. have found high expression of genes involved in inflammatory response and low adipose-specific gene expression in bone marrow 2 BioMed Research International adipocytes, and age has a greater influence on gene expression in epididymal adipocytes than bone marrow adipocytes [22]. However, the effect of aging on expression of lncRNAs is still elusive.
In the current study, to investigate the expression variation and functional roles of lncRNAs in aging, the microarray data deposited by Liu et al. [22] were used to identify the differentially time series expressed lncRNA genes (DE-lncRNAs) and differentially time series expressed mRNA genes (DEGs) in the process of aging. Additionally, DEGs targeted by DE-lncRNAs and their functions were analyzed. The results may provide new information for the molecular investigation of aging and a deeper insight into aging.

Methods and Materials
2.1. Tissue Samples and Data Acquisition. The gene expression profile GSE25905 [22] was downloaded from the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih .gov/geo/) database in National Center for Biotechnology Information (NCBI). The microarray data were produced on GPL6246 platform ([MoGene-1 0-st] Affymetrix Mouse Gene 1.0 ST Array, Affymetrix, CA, USA). This dataset contained 9 bone marrow adipocyte samples and 9 peripheral white adipocytes, respectively, taken from male C57BL/6J mice (6 months, 14 months, and 18 months of age), with three replicates at each age point. In this study, only gene microarray data of peripheral white adipocytes were used for further analysis.

Data Preprocessing.
The gene expression profile of GSE25905 was preprocessed by the Robust Microarray Analysis (RMA) algorithm [23]. The Affy package (available at http://master.bioconductor.org/packages/release/bioc/html/ affy.html) [24] of R. Probe IDs in CEL document was translated to corresponding gene symbols. If one gene symbol was matched by multiple probe IDs, the mean expression value was selected as the expression level of this gene.

Identification of DE-lncRNAs and DEGs.
Based on annotation information of lncRNAs in GENCODE (http://www .gencodegenes.org/) [25] and the array platform GPL6246, expression data of lncRNAs were obtained. Afterwards, the BETR (Bayesian Estimation of Temporal Regulation) algorithm in the BETR package (http://betterpackages.com/) [26] was applied to identify DE-lncRNAs and DEGs at the three time points, and this algorithm calculated the probability of differential expression for each gene. The probability >0.9 was set as the cut-off criterion.

Cluster Analysis of DE-lncRNAs Expression Pattern.
Hierarchical clustering is an analytical tool applied to discover the closest associations between gene profiles and specimens under evaluation [27,28]. In our study, the BHC (Bayesian Hierarchical Clustering) package (http://master.bioconductor.org/packages/release/bioc/html/BHC.html) [29] of R was utilized to perform clustering of DE-lncRNAs and construct the cluster heat map of DE-lncRNAs and samples.

Prediction of DE-lncRNA Target Genes.
Pearson correlation coefficient (PCC) [30] was used to calculate the expression similarity of DE-lncRNAs and DEGs at different time points. For each pair of DE-lncRNA and DEGs, significant correlation pairs with |PCC| > 0.95 and value < 0.05 were used to construct the DE-lncRNA/DEG regulatory network which was then visualized by Cytoscape (http://js.cytoscape.org/) [31].
Furthermore, DE-lncRNA target genes that were known to be associated with aging were identified based on the information in AGEMAP (Atlas of Gene Expression in Mouse Aging Project), which is a gene expression database for aging in mice (http://cmgm.stanford.edu/∼kimlab/aging mouse) [32]. Subsequently, the regulatory network of DE-lncRNAs and the known aging-related targets was visualized by Cytoscape.
2.6. Functional Analysis. Gene Ontology (GO) functional and pathway enrichment analyses for genes positively and negatively regulated by DE-lncRNAs were carried out using TargetMine (http://targetmine.mizuguchilab.org/) [33]. The value of each GO and pathway term was adjusted by the Holm-Bonferroni method [34], and adjusted value < 0.05 was considered statistically significant. Additionally, the pathway network was constructed using Cytoscape.

Identified DE-lncRNAs and DEGs.
Based on the annotation information in GENCODE and GPL6246 platform, a total of 203 probes were annotated as lncRNA genes, and 20564 probes were annotated as mRNA genes. With the cut-off of probability >0.9, 8301 time series DEGs and 43 DE-lncRNAs were identified in peripheral white adipocyte samples.

Clusters of DE-lncRNAs Expression Pattern.
To further explore the changes of the DE-lncRNAs expression levels at the three time points in peripheral white adipocytes, the cluster analysis was conducted. The samples at different time points were distinguished by DE-lncRNAs. The expression values of nearly half of DE-lncRNAs showed an uptrend in 6-14 months and a downtrend in 14-18 months (e.g., ENSMUSG00000086859 and ENSMUSG00000061510); a set of DE-lncRNAs were expressed in a decline trend (e.g., ENSMUSG00000087540 and ENSMUSG00000032048); and a small fraction of DE-lncRNAs were expressed in a rising trend (e.g., ENSMUSG00000066057) (Figure 1(a)).
According to the results of clustering analysis, 41 DE-lncRNAs were divided into 11 clusters (Table 1). It was clearly observed that, with the increase of age in mice, DE-lncRNAs in clusters 7 and 10 were expressed in a rising trend, whereas DE-lncRNAs in clusters 2, 4, and 8 were expressed in a decline trend. Clusters 5 and 11 showed an uptrend in 6-14 months and a downtrend in 14-18 months (Figure 1(b)).     and Fbln2. ENSMUSG00000066057 regulated the DEGs, such as CYP51, FDPS, and Eif2s1 ( Figure 2). The expression changes over time of ENSMUSG00000086859 and ENS-MUSG00000066057, as well as some targets of them, were shown in Figure 3.

DEGs
Based on the AGEMAP database, a total of 51 DE-lncRNA target genes had been discovered to be correlated with aging in mice. There were 16 DE-lncRNAs that regulated these genes, and 61 regulatory relationships were included in the network. Both ENSMUSG00000086859 and ENSMUSG00000061510 targeted Slc16a2 and Ifitm3; ENS-MUSG00000066057 regulated aging-related DEGs like Wdr1 (Figure 4(a)).

Enrichment Analysis of DE-lncRNA Targets.
To further reveal the potential functions mediated by DE-lncRNAs, the GO and pathway enrichment analyses of DE-lncRNA targets were performed, respectively. The DEGs positively regulated by DE-lncRNAs (e.g., Efna1 and Efnb2) were mainly enriched in a set of biology processes about the development of blood vessel, such as vasculature development and blood vessel morphogenesis (Table 2). A series of positively regulated target genes were significantly enriched in the pathways of cholesterol biosynthesis (e.g., Cyp51 and Fdps) and elastic fibre formation (e.g., Fbln1, Fbln2, and Fbln5) ( Table 2).
Furthermore, the negatively regulated target genes of DE-lncRNAs were mainly enriched in a set of biology processes, such as metabolic process (e.g., Abi3 and Acaca) and mitochondrion organization (e.g., Acaa2 and Bnip3), as well as pathways like metabolism of proteins (e.g., Eif2s1, Eif2s3x, and Eif3b) (Table 3).
Additionally, the pathway of metabolism of proteins was predicted to interact with five other pathways, such as posttranslational protein modification and asparagine N-linked glycosylation (Figure 4(b)).

Discussion
The increased occurrence of age-related diseases, such as cancers, chronic inflammatory, and neurodegenerative diseases, becomes a burden on health care provision in the developed  and developing countries [35]. In this study, gene expression profile GSE25905 was downloaded and analyzed using bioinformatics methods to explore the potential mechanisms of aging. A total of 8301 time series DEGs and 43 time series DE-lncRNAs were identified in peripheral white adipocyte samples. In the DE-lncRNAs/DEGs regulatory network, the DE-lncRNAs ENSMUSG00000066057, ENS-MUSG00000086859, and ENSMUSG00000061510 regulated multiple DEGs. The DEGs positively regulated by DE-lncRNAs were mainly enriched in the functions about the development of blood vessel (e.g., Efna1 and Efnb2), as well as the pathways of cholesterol biosynthesis (e.g., Cyp51 and Fdps) and elastic fibre formation (e.g., Fbln1, Fbln2, and Fbln5).
The function of blood vessel development was significantly enriched by a set of DE-lncRNA genes, such as Efna1 and Efnb2. During aging, angiogenesis is delayed, and capillary density as well as newly deposited collagen is decreased [36]. Cardiovascular structure and function are altered during aging, with elongated and stiffer aorta, as well as changed arterial baroreflex [37]. Both Efna1 and Efnb2 encode members of the ephrin family, which mediates developmental events [38]. It has been confirmed that the balance of alternatively expressed isoforms in Efna1 is disrupted in peripheral blood leukocytes of human population with advancing age [39]. Furthermore, the expression of Efnb2 is significantly decreased during the aging of the rat retina [40]. In this study, Efna1 was predicted to be regulated by the DE-lncRNA ENSMUSG00000086859 (gene name: 2810008D09Rik). There is no study that reports the role of ENSMUSG00000086859 in aging so far. Collectively, we speculate that ENSMUSG00000086859 may play key roles in aging through genes related to blood vessel development (e.g., Efna1).
In this study, several other DEGs positively regulated by DE-lncRNAs (e.g., Fbln1, Fbln2, and Fbln5) were significantly enriched in the pathway of elastic fibre formation. During cutaneous aging, elastic fibres exhibited disintegration and appeared to be loose [41]. All of Fbln1, Fbln2, and Fbln5 encode a secreted glycoprotein that is incorporated into a fibrillar extracellular matrix [42]. During aging, the balance between proteases and their inhibitors involved in extracellular matrix formation is destroyed [43]. In this study, Fbln1 and Fbln2 were discovered to be an age-regulated gene and regulated by the DE-lncRNA ENSMUSG00000086859 (gene name: 2810008D09Rik). The association of Fbln2   with human aging has also been discovered by previous studies [44,45]. Moreover, Fbln5 was predicted to be targeted by ENSMUSG00000061510. Therefore, ENS-MUSG00000086859 may also exert functions in aging via regulating the genes involved in elastic fibre formation (e.g., Fbln1 and Fbln2). ENSMUSG00000061510 may function in aging via regulating the expression of genes like Fbln5.
A previous study has demonstrated that aging is associated with altered cholesterol metabolism in T cells, causing increased cholesterol levels in lipid rafts [46]. Furthermore, cholesterol transport and lipid catabolism have been identified to be upregulated in normally aging rats [47,48]. In the present study, CYP51 and FDPS, the positively regulated target genes of the DE-lncRNA ENSMUSG00000066057 (gene name: Gm1976), were significantly enriched in cholesterol biosynthesis. CYP51, the most evolutionarily conserved member of cytochrome P450 gene superfamily, participates in the late portion of cholesterol biosynthesis [49]. In aging peripheral nervous system and liver, CYP51 is also detected to be involved in the deregulation of cholesterol biosynthesis [50,51]. FDPS encodes farnesyl diphosphate synthase, which is a key intermediate in cholesterol and sterol biosynthesis [52]. Previous studies have reported that FDPS is associated with bone mineral density of aging bone [53,54]. During aging, cholesterol synthesis is reduced in human hippocampus [55]. For example, the concentration of three cholesterol precursors (lathosterol, lanosterol, and desmosterol) is significantly decreased in the hippocampus [56]. Currently, there is no experimental evidence that ENSMUSG00000066057 is involved in aging. Therefore, this DE-lncRNA may play a role in aging via the genes related to cholesterol synthesis (e.g., Cyp51 and Fdps).
Furthermore, the DEGs that were negatively regulated by ENSMUSG00000066057 (e.g., Eif2s1) were mainly enriched in the pathways about protein metabolism along with Eif2s3x and Eif3b, and several pathways about protein metabolism interacted with each other. In aging humans, the balance between muscle protein synthesis and degradation is disrupted, which leads to the loss of skeletal muscle mass [57]. All of Eif2s1, Eif2s3x, and Eif3b encode subunits of eukaryotic translation initiation factors (EIFs), which regulate protein synthesis [58]. Decreased eIF2 phosphorylation has been detected in aged tissues and it is responsible for a higher level of protein phosphatase 1 and other proapoptotic proteins [59,60]. There is no evidence to prove the roles of Eif2s1, Eif2s3x, and Eif3b in aging so far. We speculate that the ENSMUSG00000066057 may also play critical roles in aging via regulating protein metabolism through Eif2s1. The DEGs Eif2s1, Eif2s3x, and Eif3b may also be involved in aging via protein metabolism.
Despite the aforementioned results, there were several limitations in this study. The predicted results should be confirmed by laboratory data. Furthermore, the included samples for analysis should be more. In our further studies, more samples of aging will be included to validate the expression levels and functions of the potential key lncRNAs and genes.
In conclusion, based on the gene expression data of peripheral white adipocytes taken from mice at different ages, a total of 8301 time series DEGs and 43 time series DE-lncRNAs were identified. Among them, 41 DE-lncRNAs targeted 1880 DEGs. The DE-lncRNAs ENSMUSG00000066057, ENS-MUSG00000086859, and ENSMUSG00000061510 regulated multiple DEGs. Furthermore, the DEGs positively regulated by DE-lncRNAs (e.g., ENSMUSG00000066057 and ENS-MUSG00000086859) were mainly related to the functions about the development of blood vessel (e.g., Efna1 and Efnb2), as well as the pathways of cholesterol biosynthesis (e.g., Cyp51 and Fdps) and elastic fibre formation (e.g., Fbln1, Fbln2, and Fbln5). Additionally, the DEGs (e.g., Eif2s1, Eif2s3x, and Eif3b) that were negatively regulated by DE-lncRNAs were correlated with the pathways about protein metabolism. These DE-lncRNAs and DEGs may be involved in aging, which provides novel information for the study of aging.