Comparative tail fat transcriptome analysis of key genes and pathways activated in response to fat deposition in two sheep breeds with extreme fat-tail phenotype differences

Fat tail in sheep presents a valuable energy reserve that has historically facilitated adaptation to harsh environments. However, in modern intensive and semi-intensive sheep industry systems, breeds with leaner tails are preferred. For ecient selection of lean sheep breeds, clarication of the regulatory mechanisms underlying tail fat deposition of sheep is crucial. Altay and Xinjiang Fine Wool (XFW) sheep, two important breeds with distinct tail fat deposition properties, are mainly distributed in the Xinjiang district of China, and serve as ideal models for investigating the mechanisms of tail fat deposition. In the present study, RNA-Seq was applied to determine the transcriptome proles of tail fat tissues in these two breeds, followed by analysis of differentially expressed genes (DEGs) and their sequence variations.


Abstract Background
Fat tail in sheep presents a valuable energy reserve that has historically facilitated adaptation to harsh environments. However, in modern intensive and semi-intensive sheep industry systems, breeds with leaner tails are preferred. For e cient selection of lean sheep breeds, clari cation of the regulatory mechanisms underlying tail fat deposition of sheep is crucial. Altay and Xinjiang Fine Wool (XFW) sheep, two important breeds with distinct tail fat deposition properties, are mainly distributed in the Xinjiang district of China, and serve as ideal models for investigating the mechanisms of tail fat deposition. In the present study, RNA-Seq was applied to determine the transcriptome pro les of tail fat tissues in these two breeds, followed by analysis of differentially expressed genes (DEGs) and their sequence variations.

Results
In total, 21,527 genes were detected, among which 3,965 displayed signi cant expression variations in tail fat tissues of the two sheep breeds, including 707 upregulated and 3,258 downregulated genes. Gene Ontology (GO) analysis disclosed that 198 DEGs were related to fat metabolism. In Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, the majority were signi cantly enriched in adipocytokine signaling, PPAR signaling, and metabolic pathways, with some genes being involved in multiple pathways. Among the 198 DEGs, 22 genes were markedly up or down regulated in tail fat tissue of Altay sheep, supporting their association with the fat tail trait of this breed. A total of 41,724 and 42,193 SNPs were detected in tail fat tissue transcriptomes of Altay and XFW sheep, respectively. The distribution of 7 SNPs in the coding regions of the 22 candidate genes was further investigated in three sheep populations with distinct tail types. In particular, the g.18167532 T/C mutation of ABCA1 and g.57036072 G/T mutation of SLC27A2 showed signi cantly different distributions and were closely associated with tail type.

Conclusions
The present study provides transcriptomic evidence explaining the differences in fat-and thin-tailed sheep breeds and reveals numerous DEGs and SNPs associated with tail type. Our data provide a valuable theoretical basis for selection of lean sheep breeds.

Background
Fat tail is a valuable trait that helps sheep (Ovis aries) adapt to harsh conditions, such as extremely cold winters, food shortages and drought seasons. Fat-tailed sheep characteristically deposit a mass of fatty tissue in the tail region during summer and autumn seasons when nutritious pastures are available. In the winter seasons when temperatures are extremely low and grasslands are covered with heavy snow for long periods of time, these animals obtain energy by decomposing the fat deposits in their tails to sustain life. According to archaeological ndings, modern sheep breeds were domesticated at the Fertile Crescent district of Iraq about 9,000 years ago [1][2][3]. Similar to their Asian mou on ancestor, modern breeds were thin-tailed at the early stages. Fat-tailed sheep appeared ~ 5,000 years ago through long-term arti cial and natural selection during the long evolution process to adapt to harsh local climate conditions [4][5][6][7]. However, in modern society, dietary habits and health concepts have undergone a profound revolution, and mutton with lower fat content is preferred for consumption. Furthermore, fat deposition requires more energy than deposition of lean tissue. Thus, the e ciency of meat production is higher than that of fat in intensive and semi-intensive sheep industry systems, and the tail fat of Altay sheep (comprising up to ~ 25% of carcass weight) markedly lowers their economic value [8]. For the above reasons, fat-tailed sheep breeds have been gradually rejected by producers and consumers, and leaner sheep breeds are more desirable [9][10][11][12][13][14]. Therefore, elucidation of the key genes that regulate the deposition and decomposition of tail fat in sheep and the molecular mechanisms controlling fat metabolism should greatly accelerate the breeding course of lean sheep and production of more healthy mutton for consumption.
Fat is not only used for energy storage but also an important endocrine tissue involved in regulating crucial physiological and biochemical reactions in organisms [15][16][17][18]. Traditionally, research on the molecular regulatory mechanisms of sheep tail adipose tissue development and deposition has involved investigation of the functions of single candidate genes. Earlier, Kumar and colleagues reported higher expression of the leptin gene in perirenal, backside and omental fat tissue of fat genotype than lean genotype Coopworth sheep [19]. Both activity and expression of lipoprotein lipase (LPL) were increased in fat tissue of the fat genotype sheep groups. Moreover, its expression was tissue-speci c, but not affected by the nutrition level [20]. In Guangling Large Tail sheep, the uncoupling protein 1 (UCP1) gene was expressed at signi cantly higher levels in perirenal fat than other tissues, but showed very low expression in subcutaneous fat [21]. The cell death-inducing DFFA-like effector c (CIDEC) gene was highly expressed in rump fat tissue of Altay sheep, which decreased signi cantly after a 4-week fasting period [22]. Expression of fatty acid binding protein 4 (FABP4) was signi cantly elevated in tail fat of Lori-Bakhtiari, a fat-tailed sheep breed, compared to Zel, a thin-tailed sheep breed [23]. Heart fatty acid-binding protein (H-FABP) gene in skeletal muscle of Lanzhou fat-tailed sheep was up or downregulated at different developmental stages and potentially related to the meat quality traits [24]. In Altay sheep, FABP4 was abundantly expressed in intestinal and rump fat tissues and showed no signi cant changes upon alteration of the nutritional status, suggesting a fundamental role in adipose metabolism [25]. The collective ndings highlight the involvement of several critical genes in regulation of tail fat deposition and mobilization in sheep. In general, traits of animals, including tail size, are modulated by multiple interconnected genes that form a re ned regulatory network to manage complex internal and external environments. Therefore, research speci cally focusing on one or several genes cannot completely understand elucidate such a network.
In recent years, due to the rapid development of next-generation sequencing and application of RNA-Seq technology, numerous genes expressed in tail fat tissue of sheep have been identi ed via analysis of transcriptome data in attempts to clarify the molecular network regulating adipose metabolism in fat tail of sheep. RNA-Seq has been used successfully to investigate the genes expressed in fat tissue of sheep. Wang and colleagues applied RNA sequencing technology to compare the transcriptome pro les of two sheep breeds, Kazak (fat-tailed) and Tibetan (short-tailed). Their study led to the identi cation of 646 differentially expressed genes between the two sheep breeds, including 280 upregulated and 366 downregulated genes. Moreover, the genes displaying the most signi cant fold changes, NELL1 and FMO3, were highly correlated with adipose deposition in tail [26]. Guangling Large-Tailed sheep and Small-Tailed Han sheep are two typical fat-tailed breeds in China. Upon application of RNA-Seq, a total of 4131 differentially expressed genes were determined in tail fat tissues of these two breeds, with FABP4, FABP5, ADIPOQ and CD36 identi ed as the four most highly transcribed genes [27]. Research on Small-Tailed Han and Dorset sheep revealed 602 differentially expressed genes, and GO analysis showed that several of these genes were enriched in the triglyceride biosynthetic process [28]. RNA-Seq analysis by Ma and colleagues to investigate the tail fat transcriptome of Lanzhou Fat-Tailed (long fat-tailed), Small-Tailed Han (short fat-tailed), and Tibetan (short thin-tailed) sheep identi ed several differentially expressed genes (DEGs) and long non-coding RNAs (lncRNAs). GO and pathway analysis of DEGs and target genes of differentially expressed lncRNAs revealed that the majority were enriched in fatty acid metabolism and fatty acid elongation-related pathways that contribute to fat deposition [29]. Previous studies clearly indicate that the tail fat deposition ability of sheep with different tail types is a complex quantitative trait regulated by multiple genes. The molecular mechanisms underlying tail fat deposition remain to be elucidated.
Altay and Xinjiang Fine Wool (XFW) sheep with distinct tail types are both distributed in the Xinjiang district of China. Altay is one of the most popular fat-tailed breeds. The tail and rump are fused together and these animals are characterized by their ability to deposit rump fat, which accounts for ~ 25% of carcass weight on average in this breed. In contrast, XFW, a typical long thin-tailed breed of sheep, stores almost no fat tissue in its tail. These two sheep breeds with distinct tail fat characteristics thus present ideal models for investigating DEGs involved in regulation of tail fat deposition (Fig. 1). Accordingly, we focused on rump and tail fat tissue of Altay and XFW sheep as model animals. RNA-Seq technology was applied to identify DEGs and the associated signaling pathways, with the aim of highlighting candidate genes and mechanisms that play critical roles in regulating adipose deposition in tail of sheep. Genetic variations of these DEGs were further investigated. Our collective data should provide fundamental information and theoretical guidelines for e cient breeding of lean sheep.

Summary of transcriptome sequencing data
Two cDNA libraries were constructed using mRNAs extracted from fat-rumped Altay sheep and thin-tailed XFW sheep, sequenced, and two sets of raw reads were obtained containing 51,943,518 and 51,770,440 raw reads, respectively. Low-quality raw reads and adapter sequences were then ltered, ultimately resulting in 46,614,192 and 46,646,110 clean reads. Approximately 84% and 81% clean reads could be mapped to the sheep reference genome Ovis aries v3.1. The clean reads were nally assembled into Unigenes, which were categorized to two classes, speci cally, clusters and singletons. Clusters were labeled by the pre x 'CL', followed by the cluster id. A single cluster included several Unigenes with > 70% sequence similarity. Singletons were indicated by the pre x 'Unigene' (Additional File 1, Table S1). In total, 153,914 and 117,254 clusters and 78,065 and 56,293 singletons were obtained from the two sample sets, respectively. The mean lengths of clusters were 335 nt and 317 nt, while mean lengths of singletons were 696 nt and 629 nt for Altay and XFW groups, respectively (Additional File 2, Figure S1). Clusters and singletons were further analyzed and ltered, resulting in a nal total of 48,894 Unigenes. Transcriptome sequencing data are summarized in Table 1. Annotation and expression analysis of Unigenes Comparison of the Unigenes obtained with known gene sequences of Bos taurus and Ovis aries revealed a total of 21,527 genes (E-value < 0.00001), which were subsequently matched to NR (RefSeq non-redundant proteins), Swiss-Prot, KEGG and COG (Cluster of Orthologous Groups of proteins) databases, leading to 57%, 53%, 61%, and 45% annotation, respectively (E-value < 0.00001). GO analysis was applied to clarify the biological functions of the above genes (Additional File 1, Table S1).
Calculation of gene coverage revealed that 65% (13,993/21,527) genes of Altay sheep and 68% (14,638/21,527) genes of XFW sheep had 90-100% coverage ( Fig. 2A, B). In total, 19,878 annotated genes with FPKM > 0 were detected in the two samples. The FPKM trends of the two samples were comparable, indicating similar expression patterns of the majority of genes in tail fat tissues of Altay and XFM sheep (Fig. 2C). The largest proportions of genes were expressed at low (1 < FPKM < 10) and moderate (10 < FPKM < 100) levels and only a small fraction expressed at high levels (FPKM > 100   Table S2).
To further clarify the functions of DEGs in tail fat metabolism of the two sheep breeds, we identi ed 198 DEGs (72 upregulated and 126 downregulated) closely related to adipose tissue development, deposition and mobilization.
Based on signi cant differences in expression of these genes between tail fat tissues of Altay and XFW breeds and their participation in regulation of fat metabolism, we speculate that the DEGs identi ed play potentially important roles in in uencing phenotypes of different sheep breeds. We further focused on 22 DEGs showing highly signi cant up-or down-regulation in tail fat tissue of Altay sheep as candidate genes.
qRT-PCR validation of RNA-seq data To further investigate expression patterns and validate the reliability of RNA-seq results, 22 candidate genes were selected and their relative expression levels in rump and tail fat tissue of Altay and XFW sheep, respectively, assessed via qRT-PCR using non-pooled RNA samples (n = 3 for each breed). qRT-PCR expression patterns of these genes were consistent with RNA-Seq data ( Fig. 4; Additional File 4, Figure S2), supporting the reliability of the expression pro le generated with RNA-Seq.
GO and KEGG analyses of DEGs between Altay and XFW sheep GO was applied for functional analysis of the 8,042 DEGs (Additional File 3, Table S2). GO terms with Q values ≤ 0.05 were considered signi cantly enriched and DEGs classi ed based on 'cellular component', 'molecular function', and 'biological process' categories. In total, 847 terms were enriched in cellular component, of which 24 were signi cantly enriched, such as 'membrane', 'membrane part', and 'cell periphery' (Fig. 5A). Overall, 7,014 biological process terms were enriched, 71 to a signi cant extent, including 'cell communication', 'response to stimulus', and 'multicellular organismal process' (Fig. 5B). Among the 1,903 terms enriched in molecular function, 32 were signi cantly enriched, including 'substrate-speci c transporter activity', 'insulin receptor binding', and 'protein kinase A regulatory subunit binding' (Fig. 5C).
To identify the biological pathways involved in fat deposition, DEGs were mapped to the KEGG pathway database.
Interaction network analysis of proteins encoded by DEGs related to adipose metabolism With the aid of STRING and Cytoscape software, an interaction network of proteins encoded by the 198 DEGs related to adipose metabolism was constructed (Fig. 8) roles of these pathways in adipose metabolism, we speculated that they may also in uence the tail types of different sheep breeds by regulating fat metabolism in tail tissue.
The interaction network of 22 proteins related to fat deposition was further analyzed, which revealed interactions among 19 of the proteins (Fig. 9A). The core nodes were identi ed as ACOX1, FASN, and ACAA1. ACOX1 interacted with ACADS, SLC27A2, CPT1A, FASN, and ACAA1. Interactions of FASN with CPT1A, ACACB, ACACA, ACLY, and ACOX1 were detected. ACAA1 showed interactions with PEX7, HADH, ACADS, ACOX1, and ACLY. GO analysis revealed that the majority of these proteins were related to the PPAR signaling pathway, Fatty acid metabolism, and  Table S3; Additional File 6, Table S4). We speci cally focused on the 22 candidate genes related to tail fat metabolism, which led to the identi cation of 13 SNPs in the coding regions of 9 genes, among which 12 induced amino acid alterations (Table 3).   Consequently, the speed of deposition and mobilization of tail fat in these breeds may be slower, compared to breeds in high-latitude areas.
Here, we used two highly suitable sheep breeds, fat-rumped Altay and thin-tailed XFW, for investigating transcriptome differences in tail fat tissue. These two breeds were selected for several reasons. Using RNA-seq, a total of 19,878 genes were identi ed in tail fat tissues, among which 8,042 were differentially expressed between the two sheep breeds (FDR ≤ 0.001 and |Log 2 Ratio| ≥ 1). Li and co-workers reported 5,395 DEGs between tail fat tissues of Guangling large-tailed and small-tailed Han sheep [27]. In addition, 646 DEGs between Kazak and Tibetan sheep were reported by Wang et al. [26], 390 DEGs between Lanzhou Fat-Tailed and Tibetan sheep by Ma et al. [29] and 602 DEGs between Small-Tailed Han and Dorset sheep by Miao et al. [28]. The numbers of total genes and DEGs identi ed in the current study were signi cantly higher than previously reported gures, which may be attributable to the higher suitability of our animal models.
Previous research indicates that the majority of these upregulated genes are critical for fat deposition while downregulated genes are related to fat mobilization. In bovine mammary glands, mRNA abundance at 60 d postpartum of FABP3 and ACSL1 was 80-and 7-fold greater relative to 15 d antenatal, with peak expression of SLC27A2 and SLC27A6 at 240 and 15 d relative to parturition respectively, which are signi cantly associated with milk fat synthesis [35]. ANGPTL4 promotes LPL protein intracellular degradation and triglyceride levels in adipocytes [36]. PLIN1, an Fsp27 activator, interacts with the CIDE-N domain of Fsp27 and markedly enhances lipid droplet growth by promoting lipid exchange and transfer [37]. ADIPOQ, an important adipocytokine, modulates glucose and fatty acid oxidation [38], and its polymorphisms are associated with adipose deposition in pig and cattle [39,40]. ACAA1 and ACADL play critical roles in beta-oxidation of fatty acids [41].
Upon further analysis of these DEGs and their pathways, we observed involvement of a number of DEGs in multiple pathways. For instance, ACSL1, ACSL3, ACSL4, and ACSL6 contribute to regulation of Adipocytokine signaling, PPAR signaling, Fatty acid degradation, and Metabolic pathways. MAPK8, AKT2, and AKT3 are involved in MAPK signaling, Adipocytokine signaling and Insulin signaling pathway while FASN and ACACA participate in regulation of Fatty acid biosynthesis, Insulin signaling and Metabolic pathways ( Table 2). The mechanisms underlying fat metabolism are complex. Fat is not only a tissue used to store energy in animals but also an important endocrine tissue involved in regulating crucial physiological and biochemical reactions [15,16]. Accordingly, fat metabolism is regulated by an elaborate network composed of numerous signaling pathways. We speculated that these genes involved in multiple pathways play bridging roles to connect these signaling mechanisms.
Among the 22 candidate genes, ABCA1, ACACA, and CIDEC were signi cantly upregulated in rump fat tissue of Altay sheep (P<0.01) with 5.37, 6.75, and 5.86 times higher expression, compared to tail fat tissue of XFM sheep. C1QTNF1 and HSL were signi cantly downregulated in rump fat tissue of Altay sheep with 0.15 and 0.13 times expression relative to tail fat tissue of XFM sheep (P<0.01). ABCA1 is a membrane transporter protein that plays an essential role in the e ux of cholesterol from peripheral tissues back to the liver for participating in lipid metabolism [46]. In sheep reared under intensive conditions and offered su cient feed, ACACA in muscle was signi cantly upregulated and the fat deposition accelerated [47]. CIDEC (FSP27) located on the surfaces of lipid droplets of adipocytes could promote enlargement or fusion of lipid droplets via clustering and lipid transfer [48][49][50]. Meanwhile, CIDEC (FSP27) suppressed HSL located on lipid droplet surfaces and inhibited lipolysis [51]. In Altay sheep fasted for 4 weeks, the CIDEC level in rump fat tissue was signi cantly downregulated [22]. In human liver samples of individuals with obesity and diabetes mellitus, CIDEC was signi cantly up-regulated [52,53]. The collective studies con rm essential roles of these candidate genes in fat metabolism.
In view of the signi cant differences in expression levels of these genes between tail fat tissue of Altay and XFW sheep, and their signaling pathways being closely related to fat metabolism, we speculated that the genes are key regulators of tail phenotype differences of sheep that require further investigation.

Relevance of gene variants and tail fat deposition of sheep
Previous research has con rmed that a number of gene variants are closely related to tail phenotypes of sheep.
Using Ovine SNP50k BeadChip, Moradi and colleagues investigated gene variants in two Iranian sheep breeds with fat tail and thin tail phenotypes, respectively. The group identi ed several mutations that were signi cantly different between the fat-tailed and thin-tailed sheep breeds in three regions located on chromosomes 5, 7 and X [54]. Our group additionally showed that polymorphisms of g.59571364, g.59912586, g.60149273, and g.59383635 loci on Chromosome X are markedly related to tail fat deposition ability of sheep breeds [55][56][57].
RNA-Seq offers novel opportunities for the e cient detection of transcriptome variants (SNPs and short indels) in different tissues and species [58,59]. Compared to whole-genome sequencing, RNA-Seq offers a more costeffective alternative for identifying variations and potentially causal mutations underlying the analyzed phenotypes [60,61]. Using RNA-Seq, Suárez-Vega and colleagues detected 57,795 variants in the regions harboring Quantitative Trait Loci (QTL) for mild yield, protein and fat percentages in sheep, among which 21.44% were novel [62]. In the current study, we detected 41,724 and 42,193 SNPs in tail fat tissue transcriptomes of Altay and XFW sheep, respectively, using RNA-Seq. We further focused on SNPs in 22 candidate genes related to tail fat metabolism, 12 of which altered the encoded amino acid ( Table 3).
The distributions of g. 18167532 T/C mutation of ABCA1 and g. 57036072 G/T mutation of SLC27A2 were signi cantly different in the three sheep breed populations with distinct tail phenotypes (Table 4). ABCA1 encodes a key protein regulating apolipoprotein-mediated e ux of cholesterol and phospholipid from peripheral cells to highdensity lipoprotein-cholesterol (HDL-C) [63,64]. Most previous studies have focused on the association of ABCA1 gene polymorphisms with human disease. The SNPs rs4149267, rs1800977, rs1800978 and rs2230806 of ABCA1 are associated with HDL-C concentrations in Caucasian, Sacramento and French populations [65,66]. A signi cant association was observed between the SNP rs3890182 and type 2 diabetes in patients of Han Chinese ancestry [67,68]. A rs2230806 genetic variation was signi cantly related to the development and severity of coronary artery disease (CAD) in an Iranian population. Moreover, the K allele of ABCA1 R219K polymorphism has been shown to exert a protective effect against CAD risk and is correlated with decreased severity of CAD, independently of plasma lipid levels [69,70]. Here, we detected a g.18167532 T/C mutation of ABCA1. Its distribution in different sheep breeds was signi cantly related to tail fat deposition ability, which should be further investigated.
The signi cance of mutations in SLC27A isoforms has been established in previous reports. Wang and colleagues identi ed a SNP in exon 7 leading to an amino acid alteration in Large White and Meishan pig breeds, which was signi cantly correlated with growth and carcass traits [71]. SNPs at SLC27A1 and SLC27A4 were associated with saturated fatty acid and stearic acid contents in longissimus dorsi muscle of pig [72]. Two SNPs in exon 3 and 3′UTR of bovine SLC27A1 exerted effects on milk production traits, such as milk protein and milk fat percentages, in Chinese Holstein cattle [73]. In the present investigation, the genotype of the g.57036072 G/T mutation of SLC27A2 was distinct in three sheep breeds with different tail phenotypes and the G allele was signi cantly related to rump tail type in Altay sheep ( Table 4).
The two mutations of ABCA1 and SLC27A2 identi ed in this study led to alterations in the encoded amino acids. In view of their signi cant association with tail phenotype of sheep, the issue of whether these mutations affect rump fat deposition in Altay sheep by in uencing the functions of the corresponding proteins requires further investigation. Here, we con rmed relationships between limited SNPs and tail fat deposition traits of sheep. Further research is warranted to ascertain the associations of several other detected genes with tail fat metabolism.

Conclusions
In this research, we examined the differences in transcriptome pro les and sequences of tail fat tissue from Altay and XFW sheep breeds that are distributed within the same district of China but display distinct tail fat deposition traits by applying RNA-seq followed by qRT-PCR validation to con rm the reliability of our ndings. DEGs were identi ed and their functions evaluated via GO and KEGG analyses. The genes associated with fat metabolism were ltered out for further analysis. Based on the data, 22 candidate genes and two SNPs were identi ed that potentially contribute to differences in tail fat deposition abilities of sheep. Determination of the speci c roles of these DEGs and candidate genes in tail fat deposition may aid in the selection of lean sheep breeds.

Methods
Tail fat collection and RNA extraction  Cary, NC, USA) and results expressed as means ± SD. Signi cance of differences was analyzed using one-way ANOVA. Differences were considered signi cant at P < 0.05 and highly signi cant P < 0.01.

Analysis of variations in candidate functional genes involved in fat deposition
Based on SNPs analysis of candidate genes, 10 SNPs existing in coding regions of the 22 candidate functional genes involved in tail fat deposition were selected for analysis. Sequences containing the selected SNP sites were exported from transcriptome data and mapped onto the sheep genome (ISGC Oar_v3.1/oviAri3, August 2012) applying Blat of the UCSC Genome Browser (http://genome.ucsc.edu/), and subsequently 1,000 bp sequences around the SNP sites were cut out to design primers using Oligo 6.0 (Additional File 8, Table S6). The PCR fragment sizes were 200 to 300 bp and SNP sites were located near the middle of the ampli ed fragments. Primers were synthesized by Sangon Biotech Co. Ltd (Shanghai, China).
Genomic DNA of Altay (n = 104), Hu (n = 104) and XFW (n = 104) sheep was ampli ed using the above primers (Additional File 8, Interaction analysis of candidate proteins involved in adipose metabolism STRING 11.0 (https://string-db.org/) and Cytoscape (https://cytoscape.org/) were applied to analyze the proteinprotein interactions of candidate genes involved in adipose metabolism, and the interaction network illustrated.        Protein-protein interaction analysis of 198 lipid metabolism-related differentially expressed genes.