Integrated transcriptome and proteome analysis reveals potential mechanisms for differential abdominal fat deposition between divergently selected chicken lines

Background: Genetic selection for meat production performance of broilers concomitantly causes excessive abdominal fat deposition, accompanied by several adverse effects, such as the reduction of feed conversion eciency and reproduction performance. Our previous studies have identied important genes regulating chicken fat deposition, using the Northeast Agricultural University broiler lines divergently selected for abdominal fat content (NEAUHLF) as an animal model. However, the molecular mechanism underlying fat deposition differences between fat and lean broilers remains largely unknown. Results: Here, we integrated the transcriptome (RNA-Seq) and quantitative proteome (isobaric tags for relative and absolute quantitation, iTRAQ) proling analyses on abdominal fat tissues from NEAUHLF chicken lines. Differentially expressed genes (466 DEGs) and proteins (231 DEPs) were identied, and enriched in pathways related to fatty acid metabolism, fatty acid biosynthesis, glycerophospholipid metabolism, and PPAR signaling, and in pathways mainly involved in protein processing, endocytosis and lipid metabolism, respectively. Moreover, several key DEGs and DEPs involved in long-chain fatty acid uptake, in situ lipogenesis (fatty acid and cholesterol synthesis), and lipid droplets accumulation were discovered, and most of them were up-regulated in the fat line, after integrated transcriptome and proteome analysis. Conclusions: Together, our ndings provided a novel insight into abdominal fat content discrepancy between the fat and lean chicken lines.

molecular mechanism for abdominal fat deposition differences between fat and lean broiler lines remains unclear.
Recently, with the development of high-throughput sequencing technology, integration of transcriptome and proteome technologies has become an important means and routine to analyze the molecular mechanism of agricultural complex traits in farm animals [8][9][10].In the present study, we examined the differences of transcriptome and quantitative proteome pro ling on abdominal adipose tissues between the two broiler lines at 7 weeks of age. We identi ed several key DEGs and DEPs potentially involved in long-chain fatty acid uptake, in situ lipogenesis (fatty acid and cholesterol synthesis), and lipid droplets accumulation, facilitating our understanding of abdominal fat content differences between chicken lines under divergent selection. fat content as previously described [3]. In total, ten male birds (lean line, n = 5, and fat line, n = 5) at 7 weeks of age from the 19th generation of NEAUHLF were used for RNA-seq and iTRAQ analysis, and these birds were kept under the same environmental conditions and had free access to feed and water.

Methods
Abdominal fat tissues were collected right after these birds were euthanized by intramuscular injection of pentobarbital (Sigma, St. Louis, MO, USA) (40mg/kg) under deep anesthesia, and then immediately frozen in liquid nitrogen and stored at -80℃. The detailed information of selected chickens' body weights, abdominal fat weights (AFW) and abdominal fat percentages (AFP) were listed in Table S1.

Transcriptomic data collection and analysis
Total RNA from abdominal fat tissues was extracted using the TRIzol reagent (Invitrogen, New Jersey, NJ, USA), and genomic DNA was removed by DnaseI treatment. RNA purity, concentration and integrity were checked by NanoPhotometer® spectrophotometer (IMPLEN, CA, USA), Qubit® RNA Assay Kit in Qubit® 2.0 Flurometer (Life Technologies, CA, USA), and RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA), respectively. After removal of ribosomal RNA and cleaning-up of rRNA free residue by a Ribo-Zero™ rRNA Removal Kit (Epicentre, USA), the sequencing libraries were generated using the NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's recommendations. cDNA fragments of 150-200 bp in length were selected and puri ed with the AMPure XP system (Beckman Coulter, Beverly, USA). Then, library quality was assessed by the Agilent Bioanalyzer 2100 system. Finally, after cluster generation (cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS, Illumina), the libraries were sequenced on an Illumina Hiseq 4000 platform, and 150 bp paired-end reads were produced.
After demultiplex and quality ltering of raw data, clean reads were obtained and aligned to the G. gallus 5.0 reference genome assembly using HISAT2 (v.2.0.4) [11]. The mapped reads of each sample were assembled by StringTie (v1.3.1) [12] in a reference-based approach. The software Cuffdiff (v2.1.1) [13], which provides statistical routines for determining gene expression data using a model based on the negative binomial distribution, was used to calculate FPKMs (fragments per kilobase of exon per million fragments mapped). Transcripts with a p-value ≤ 0.01 and fold changes ≥ 1.5 or ≤ 0.67 were assigned as signi cantly differentially expressed.
The MS/MS data were processed with ProteinPilot Software v. 5.0 (AB SCIEX, USA) against Gallus gallus database using the Paragon algorithm [16]. The experimental data from tandem mass spectrometry (MS) were utilized to match the theoretical data to identify proteins, which was performed by the search option (with an emphasis on biological modi cations). An automatic decoy database search strategy was used to estimate the false discovery rate (FDR) calculated as the false positive matches divided by total matches, using the PSPEP (Proteomics System Performance Evaluation Pipeline Software, integrated into the ProteinPilot Software). The signi cantly differentially expressed proteins were identi ed using the following criteria: 1) peptide groups considered for quanti cation required at least 2 peptides, and a global FDR less than 1% was used; and 2) a fold change ≥ 1.5 or ≤ 0.67 and with p-value ≤ 0.05 (t-test).

RT-qPCR analysis
To validate RNA-Seq results, 20 DEGs with higher expression levels and larger fold changes were validated by RT-qPCR. Ten male birds (n=5 for each line) from the 19 th generation of NEAUHLF were used. Total RNA from abdominal fat tissue was reversely transcribed into cDNA using a PrimeScript™ RT Reagent Kit (Takara, Dalian, China). FastStart Universal SYBR Green Master kit (Roche) and the ABI 7900 PCR detection system were used to perform RT-qPCR. The program began at 95 ℃ for 30 s for activation, followed by 40 cycles of ampli cation at 95 ℃ for 5 s and 58 ℃ for 30 s. An additional 15 s at 95 ℃, 1 min at 60 ℃, and 15 s at 95 ℃ were performed for the melt curve stage. The housekeeping gene TATA-Box binding protein (TBP) was used as the control. RT-qPCR primer pairs were designed by Primer Premier 6.0 and the detailed information were listed in Table S2. The comparative 2 -ΔΔCt method was used to determine the statistical signi cance.

PRM-MS analysis
To verify the protein expression level obtained by iTRAQ analysis, 10 DEPs with higher expression levels and larger fold changes were selected for validation. Signature peptides for the target proteins were de ned according to the iTRAQ data, and only were unique peptide sequences selected for the PRM-MS analysis (Table S3). We randomly selected 6 male birds from the 19 th generation (n=3 for each line) of NEAUHLF, and extracted the proteins from abdominal fat tissues, which were prepared following the iTRAQ protocol. Tryptic peptides were loaded on C18 stage tips for desalting prior to reversed-phase chromatography on an Easy nLC-1200 system (Thermo Scienti c).
One-hour liquid chromatography gradients with acetonitrile ranging from 5 to 35% were used, and PRM was performed on a Q-Exactive Plus mass spectrometer (Thermo Scienti c). Methods were optimized for collision energy, charge state, and retention times for the most signi cantly regulated peptides experimentally, using unique peptides of high intensity and con dence for each target protein. For PRM MS data, each sample's average base peak intensity was extracted from the full scan acquisition using RawMeat (version 2.1, VAST Scienti c). The normalization factor for sample N was calculated as f N = the average base peak intensity of sample N divided by the median of average base peak intensities of all samples. The area under curve (AUC) of each transition from sample N was multiplied by this factor. After normalization, the AUC of each transition was summed to obtain AUCs at the peptide level. Relative protein abundance was de ned as the intensity of a certain peptide.
The thresholds for signi cant enrichment was set at p-value < 0.05.

Statistical analysis
Chicken body measurement data were shown as mean ± SD. Student's t-test was used to compare the differences between two groups, and the threshold of signi cance was set at p < 0.05.

Results
Transcriptome pro ling RNA-Seq generated 77,841,982 to 95,203,262 raw reads for each library (Table S4). After ltering the lowquality reads, the average number of clean reads was 89,456,223 (99.93%) and 85,002,558 (99.90%) for the lean line (LL) and fat line (FL), respectively (Table S5). Furthermore, the distribution of mapping rates for clean reads to the reference genome ranged from 81.7% to 86.99%. FPKM was used to estimate the level of gene expression, and a total of 12501 mRNA transcripts (corresponding to 9603 genes) were identi ed after quality control for each sample. The correlation analysis of the gene expression level between 10 samples was also performed (Fig.S2). Finally, 483 differentially expressed transcripts were found, corresponding to 466 genes (DEGs) , of which 154 were up-regulated and 312 down-regulated in LL compared to FL (Fig 1a. and Table S6). To verify the accuracy of RNA-Seq data, 20 DEGs were chosen and their expression levels were assayed by RT-qPCR. Except for MAPK6, other genes showed consistent results for both RNA-Seq and RT-qPCR (Fig.1b). Then, GO analysis showed that GO terms, such as "regulation of endocytosis", "cellular response to prostaglandin E stimulus", " lamentous actin" and "nucleolus" were signi cantly enriched (p < 0.05) ( Fig.1c and Table S7), and the KEGG pathway analysis also revealed that these 466 DEGs signi cantly enriched in 6 pathways, including "fatty acid metabolism", "PPAR signaling pathway", "biosynthesis of unsaturated fatty acids", "metabolic pathways" and "glycerophospholipid metabolism", "fatty acid biosynthesis" (p < 0.05) ( Fig.2a and Table S8). Thus, transcriptome sequencing found that pathways related to fatty acid metabolism could be important in fat deposition differences in our divergently selected chicken lines.

iTRAQ-based Proteomics
To better understand the potential mechanisms underlying differential abdominal fat deposition between fat and lean broilers, iTRAQ-based proteomics was also performed. Eight-plex iTRAQ generated 61,498 spectra and 18,672 unique peptides (matching to 2,424 proteins); four-plex iTRAQ generated 45,203 spectra and 14,300 unique peptides (matching to 2,185 proteins). In addition, after stringent selection of unique peptides (95% con dence limit and global FDR < 1%), we identi ed 2,137 and 1,727 proteins in the two iTRAQ experiments, respectively. There were 1,470 overlapped proteins, which were considered as credible proteins. Finally, 231 differentially expressed proteins (DEPs) were identi ed between the two chicken lines, of which 139 were up-regulated and 92 down-regulated in LL compare to FL (Fig.3a). The heat map of hierarchical clustering of DEPs was shown in Figure S3, and detailed information about every DEP was listed in Table S9. To validate the iTRAQ data, we selected 10 DEPs for the PRM analysis, and 9 of 10 proteins were successfully quanti ed. The PRM results of these nine proteins were consistent with our iTRAQ data (Fig.3b).
Next, GO enrichment analysis was performed to determine the function of the DEPs. For the biological process (BP) category, "fatty acid biosynthetic process", "cell redox homeostasis" and "oxidationreduction" were signi cantly enriched. In addition, "extracellular exosome" and "calcium ion binding" were the most representative GO term for the cellular component (CC) and the molecular function (MF), respectively ( Fig.3c and Table S10). KEGG pathway analysis showed that the DEPs were mainly enriched in metabolic pathways, protein processing, endocytosis and lipid metabolism-associated pathways, such as PPAR signaling pathway, fatty acid degradation, fatty acid biosynthesis and fatty acid metabolism ( Fig.2a and Table S11). In support of our transcriptome analysis, proteome analysis further discovered that pathways related to fatty acid metabolism are likely important in fat deposition differences between our fat and lean chicken lines.

Integrated transcriptome and proteome analysis
In order to further distinguish the critical DEGs and DEPs that may affect chicken abdominal fat deposition, integrated transcriptome and proteome analysis was conducted by combined analysis on our RNA-Seq and iTRAQ data.. First, we compared the DEGs and DEPs and found 22 genes were overlapped (Fig.4a). It was worth noting that theses 22 genes show the same expression tendency ( Fig.4b and Table  1), suggesting these genes could be the key genes involved in the regulation of abdominal fat deposition.
Second, by comparing the pathways obtained after DEGs and DEPs enrichment analysis, we found that there were ve overlapped pathways, four of which were related to lipid metabolism, such as fatty acid metabolism, PPAR signaling pathway, fatty acid biosynthesis and biosynthesis of unsaturated fatty acids ( Fig.2a and 2b), including 15 genes that may important for fat deposition. Together, 35 key DEGs/DEPs [22 signi cantly differentially expressed at both mRNA and protein levels, and 15 genes in overlapped pathways with acetyl-CoA carboxylase alpha (ACACA) and stearoyl-CoA desaturase (SCD) appear twice] that may affect chicken abdominal fat deposition were discovered through our integrated transcriptome and proteome analysis (Table 1).

Discussion
Adipose tissue is no longer viewed as a passive repository for triacylglycerol storage and a source of free fatty acids but as an active endocrine and paracrine organ secreting an ever increasing number of cytokines that participate in diverse metabolic processes including food intake, regulation of energy balance, insulin action, glucose and lipid metabolism, angiogenesis and vascular remodeling, regulation of blood pressure, and coagulation [18][19]. Excess adipose tissue leads to obesity and metabolic syndrome, such as insulin resistance, type 2 diabetes, heart disease, atherosclerosis and hypertension [20]. Exploring the molecular mechanism of adipose development and fat deposition is helpful for the therapy of obesity and related metabolic diseases. Our fat and lean broilers have similar body weight but acquire a divergent abdominal fat content. So, it is the ideal animal model to study the molecular basis of fat deposition. In the current study, we combined RNA-Seq and iTRAQ techniques on abdominal adipose tissues from 7-week-old FL and LL broilers, and identi ed a number of key DEGs and DEPs that may affect fat deposition (Table 1). These genes are mainly involved in lipid metabolism associated processes, such as long-chain fatty acids uptake, in situ lipogenesis (fatty acid and cholesterol synthesis), and lipid droplets accumulation.

Long-chain fatty acid uptake
In poultry, fatty acids are taken up by the adipose tissue, which mainly come from triglycerides in plasma lipoproteins (such as VLDL) synthesized and packaged by the liver, and also from triglycerides in portomicrons (PM) assembled by long-chain fatty acids in dietary fat [21]. The triglycerides contained in VLDL and PM are hydrolyzed by lipoprotein lipase (LPL) located in adipose tissue-lined endothelial cells to produce free fatty acids, which can be taken up by adipocytes and then re-esteri ed and stored in lipid droplet as triglycerides [22]. Previous studies suggested that increased uptake of fatty acids in abdominal adipose tissue is a major cause of obesity in chickens [23]. In general, most cells show less ability in longchain fatty acid uptake, whereas adipocytes and cardiomyocytes can e ciently and speci cally absorb long-chain fatty acids [24]. In the present study, DEG [caveolin 1 (CAV1)] and DEPs [LPL, CAV1, and acyl-CoA synthetase long-chain family member 1 (ACSL1)] were implicated in long-chain fatty acid uptake. So, we speculate that long-chain fatty acid uptake may play an important role in chicken adiposity.
LPL is considered to be a rate-limiting enzyme in fat accretion in chicken adipose tissue [25], responsible for decomposing triglycerides in VLDL or PM to release free fatty acids. CAV1 was identi ed as the main plasma membrane fatty acid binding protein in adipocytes that can bind long-chain fatty acids with high a nity [26]. Lack of CAV1 results in the loss of caveolae and defects in long-chain fatty acid uptake in adipocytes [27]. In addition, CAV1 can bind to the long chain fatty acids on the inner lea et of the lipid bilayer, and transport fatty acids to the subcellular membrane compartment through vesicle-mediated transport [28]. ACSL1 is an acyl-CoA synthetase, and functions as long-chain fatty acid transport protein in adipocyte [29]. The rst step in using long-chain fatty acids in cells is their esteri cation reaction with CoA, and this reaction is catalyzed by acyl-CoA synthetase. In humans, there are two related long-chain fatty acid activation-related protein families: fatty acid transporters (FATP) and long-chain acyl-CoA synthetase (ACSLs). ACSL1 was found to co-localize with FATP1 in a small number of 3T3-L1 cells [30]. Furthermore, ACSL1 can promote fatty acid uptake into cells depending on their expression levels [31][32]. In the present study, the expression levels of LPL, ACSL1, and CAV1 were signi cantly higher in the FL adipose tissue, indicating the adipose tissues of the fat broilers have stronger long-chain fatty acid uptake ability to synthesize more triglycerides.

Fatty acids synthesis
The liver is widely considered to be the main site of de novo lipid synthesis in avian species, with more than 70% of de novo fatty acid synthesis occurring in liver tissue [33], contradictory to the ndings in the present study that a large number of lipogenic genes expressed in chicken abdominal fat tissue. Recent studies have also shown that the lipid synthesis ability of avian adipose tissue may be underestimated. Resnyk et al. [34] performed microarray analysis on 9-week-old chicken abdominal fat tissue and found many genes associated with lipogenesis were highly expressed in fat chicken. Similarly, one RNA-Seq analysis on 7-week-old broilers showed a large number of lipogenic genes were also up-regulated in abdominal adipose tissues from fat chicken [35]. Another RNA sequencing analysis showed that the 7week-old fast growth chickens (fatter than slow growth chickens) over-express numerous lipogenic genes in adipose tissue, which should enhance in situ lipogenesis and ultimately adiposity [36]. Intriguingly, in the present study, we also found several key genes associated with fatty acids synthesis, including DEGs [ACACA, 3-oxoacyl-ACP synthase, mitochondrial (OXSM), fatty acid desaturase 2 (FADS2), SCD, peroxisomal trans-2-enoyl-CoA reductase (PECR), and 3-hydroxyacyl-CoA dehydratase 2 (HACD2)] and DEPs [ACACA, fatty acid synthase (FASN), SCD, ACSL1, and acyl-CoA oxidase 1 (ACOX1)]. KEGG analysis showed that ACACA, OXSM, FASN, and ACSL1 were enriched in fatty acid biosynthesis pathway, and FADS2, SCD, PECR, HACD2, and ACOX1 were enriched in the biosynthesis of unsaturated fatty acids pathway (Fig.2b). It is worth noting that two proteins (ACACA and SCD) can work as critical enzymes to synthesize fatty acids. ACACA is the rate-limiting enzyme in fatty acid biosynthesis, which can catalyze the synthesis of malonyl-CoA from two molecules of acetyl-CoA, and produce fatty acids under the action of fatty acid synthase [37]. SCD is a rate-limiting enzyme that catalyzes the formation of monounsaturated fatty acids from saturated fatty acids [38].
Thus, we found that the expression levels of genes related to fatty acid synthesis were signi cantly higher in the fat line, suggesting the adipose tissues in fat birds have stronger ability of triglycerides synthesis in adipocytes.

Cholesterol synthesis
At the cellular level, the deposition of adipose tissue is the result of the increase of the number of adipocytes (adipogenesis) and the size of single fat cells (triglyceride and cholesterol accumulation in lipid droplets) [39][40].
Adipose tissue is the major site for the storage of cholesterol, containing both free and esteri ed forms of cholesterol [41]. In the current study, some critical DEGs [epoxide hydrolase 2, cytoplasmic (EPHX2) and cytochrome p450 oxidoreductase (POR)] and DEPs [acetyl-CoA acetyltransferase 1 (ACAT1), EPHX2, and POR] were related to cholesterol synthesis. ACAT1 is an acetyl-coenzyme A acetyltransferase, which can catalyze the formation of cholesteryl esters from cholesterol and long-chain fatty acyl-CoAs [42]. EPHX2 is a member of the epoxide hydrolase family, and the N-terminal activity of EPHX2 can increase the cell's cholesterol level [43][44]. POR is a microsomal membrane-associated protein of two types: type I and type II, of which type II is responsible for cholesterol synthesis [45]. Lanosterol-14α-demethylase and squalene monooxygenase can participate in cholesterol biosynthesis and require POR as the electron donor [46][47]. EPHX2, POR and ACAT1 were all up-regulated in abdominal adipose tissue of fat line in the current study, suggesting the fat broilers could accumulate more cholesterol to expand the size of adipocytes.

Lipid droplet accumulation
Lipid droplets are dynamic organelles involved in intracellular lipid metabolism in almost all eukaryotic cells, and in white adipocytes, the large unique lipid droplet occupies most of the cell space and volume [48]. Perilipin1 (PLIN1), PLIN2, PLIN4 and CAV1 associated with lipid droplet accumulation were also discovered in the present study. PLINs are proteins that coat lipid droplets in adipocytes, which control the lipolysis of stored neutral lipids by cytosolic lipases. PLIN1 is the most abundant lipid droplet coat protein, and plays a crucial role in restricting adipose lipolysis under basal or fed conditions [49].
However, PLIN2 has minimal control over lipolysis, and may affect lipid droplets accumulation by a different mode. PLIN2 de cient mice can increase triglycerides accumulation in the heart by altered lipophagy [50]. PLIN4 mainly exists in white adipose tissue and is associated with tiny nascent lipid droplets. As a lipid droplet coat protein, PLIN4 can quickly package newly synthesized triacylglycerol, and store energy to the greatest extent during excessive nutrition [51]. Another lipid droplet coat protein is CAV1, which is an essential component for the assembly of caveola organelles in highly differentiated cells, such as adipocytes. CAV1 usually plays a key structural role in the accumulation of lipid droplets in adipocytes, since the deletion of CAV1 can reduce lipid accumulation, which leads to progressive atrophy of white adipose tissue [52]. PLIN1, PLIN4 and CAV1 were up-regulated and PLIN2 was down-regulated in the adipose tissue of fat line, indicating that fat birds may accumulate more lipid droplets in adipocytes than the lean birds.
Herein, through the joint analysis of transcriptome and proteome, we found many key genes that may affect chicken fat deposition ( Table 1). The differential expression and molecular function of these genes likely lead to the differential accumulation of abdominal fat content, although some of them have not been reported to be directly related to adiposity, such as amino acid metabolism-related genes (ADP ribosylation factor like GTPase 6 interacting protein 5 and glutathione S-transferase theta 1-like 1) and oxidation-reduction-related genes (steroid 5 alpha-reductase 3, glycerol-3-phosphate dehydrogenase 2, mitochondrial, retinol saturase and vesicle amine transport 1). Functions of these genes in adipose tissue development and fat deposition awaits further investigation.

Conclusion
In summary, molecular differences related to long-chain fatty acid uptake, in situ lipogenesis (fatty acid and cholesterol synthesis), and lipid droplets accumulation were discovered to exist between the fat and lean chicken lines, which may contribute to the striking differences of abdominal fat deposition.