Background

Ducks have been widely reared worldwide for their meat, egg, down, and feathers. Pekin duck is a famous breed all over the world for its extensive adaptability, rapid growth, and superior taste owing to the superior adipose deposition [1]. Pekin duck also is an important raw material source for Pekin roast duck which needs high subcutaneous fat content to ensure its fine flavor [2].

Fat deposition has a crucial role in animal meat flavor. The coordinated processes of fat synthesis and deposition were observed in the liver and fat tissue, respectively [3, 4]. Avian has been utilized as a good animal model for exploring basic adipogenesis mechanisms [5]. Adipogenesis involves a temporally regulated set of gene-expression events [6], and a set of studies regarding subcutaneous and abdominal fat have been previously investigated in avian [7,8,9]. In a genome-wide association study (GWAS) analysis for fat traits across poultry breeds, significant associations were identified on 21 QTLs for skin fat traits and 396 QTLs for abdominal fat traits in a variety of chicken chromosomal regions [10]. Examining currently available meat quality studies, hundreds of QTLs have been curated in pigs, cattle, and chickens [11]. However, owing to a lack of microarrays and cheap genotyping methods in ducks, few studies have been reported using whole genome-sequencing data [12, 13]. Recently, we applied for the first time the genotyping-by-sequencing (GBS) method in ducks [14,15,16], which makes it possible to detect meat quality-related QTLs. Further, some newly identified candidate genes were discovered for fat-deposition and meat-quality traits, which reported fat-deposition and meat-quality traits in Pekin ducks [17]. Avian fat-related traits are typical complex phenotypes, which need a large amount of data to analyze the genetic loci. In this study, we collected large-scale duck fat traits and targeted genome-wide variation using resequencing methods, aiming to identify loci with pleiotropy for subcutaneous and abdominal fat traits by GWAS.

In this study, we first phenotyped fat traits and genotyped whole-genome variations for 1,246 ducks, combined with the previous GWAS data, we obtained a group of new loci that affect the traits related to avian fat deposition. Among them, CLN8 is a new gene with great potential in regulating fat deposition. The functional basis of the most intriguing significant loci was elucidated in the present study. The significant signal was discovered to be driven by 5 SNPs, which were the variants located in 1,920 bp upstream of the CLN8 gene that alters the binding affinity of several transcript factors and may lead to differential CLN8 gene expression. Further, this study analyzed the upstream regulatory region of CLN8 by a dual-luciferase reporter system and found that the important haplotype mutation of the regulatory region composed of 5 SNPs obtained by GWAS analysis, which lead to the decreased transcriptional activity of CLN8. In further, we performed gain-of-function experiments on the candidate gene CLN8, showing that CLN8 is a promoter of adipocyte differentiation. Our findings provide a theoretical basis for the functional analysis and regulatory mechanism of CLN8 in avian adipogenesis and propose a novel function of CLN8 in avian adipocyte differentiation.

Materials and methods

Phenotype collection

Two batches of ducks were reared and phenotypes were collected. This study reared about 1,400 ducks from the day-old ducklings which are inherited from randomly mated Pekin ducks. Each batch included about 700 ducks and was reared in a large population for 21 d, and then all ducks were recorded feed consumption individually to 42 d. All ducks were fed the same diet and maintained the same lighting conditions as described previously [18]. At 41 d, blood was collected from the brachial vein of the duck by venipuncture using citrated syringes during a routine health inspection. Ducks were weighed and killed by stunning and exsanguination, after withholding feeding for 6 h at 42 d. After the carcass composition traits were determined, meat quality traits were measured using methods previously described in detail [17], including subcutaneous fat weight (SFW), subcutaneous fat percentage (SFP), abdominal fat weight (AFW), abdominal fat percentage (AFP) and the body weight of 42 d (BW42). Finally, this study collected 1,246 ducks with full phenotypes for further whole-genome genotyping. We also combined with previously published data, and finally, a total of 1,880 (males, 932; females, 948) Pekin ducks were used in this study.

DNA extraction and whole genome sequencing

Populations from different years were genotyped using two types of genotyping strategies. Genomics DNA from Population 2014 (Pop2014) with 634 individuals was extracted from blood using the standard phenol/chloroform method, and genotyping by sequencing (GBS) was used for genotyping as described previously [14]. Genomic DNA from Population 2019 (Pop2019) and Population 2020 (Pop2020) with total of 1,246 ducks was extracted using the Qiagen kit (QIAGEN, Valencia, CA, USA), and samples were subjected to next-generation sequencing (NGS) using Illumina paired-end cycles. The whole genome sequencing of all 1,880 duck samples was available publicly. The data were deposited in the NCBI sequence read archive (PRJNA506902, SRP068685 and PRJNA921894).

Variant calling and annotation

Adaptors and low-quality sequences were removed using the Picard software (http://picard.sourceforge.net/) (version 2.28). High-quality reads were aligned to the Mallard reference genome (ASM874695v1, CAU_wild_1.0) using the BWA-MEM algorithm of the BWA software package (v0.7.17) [19]. Duplicate reads were excluded using the Picard tool MarkDuplicates. Variant calling was performed with the GATK HaplotypeCaller (v4.1.8) and using joint genotyping across all sequenced samples [20]. PLINK (v1.922) was used to exclude SNPs and individuals with more than 5% missing data, markers with minor allele frequency < 1%, and sex chromosome related SNPs. Variant data were imputed using Beagle (v5.1) [21]. After applying these filters, 8,448,069 SNPs and 1,880 individuals were retained for further analysis. Known variants from the Ensemble variation database (release 10) were used for variant annotation with VEP (v88.9) software [22].

Statistical genetic analyses and GWAS analysis

The correlation between traits and covariates was examined by calculating Pearson's correlation coefficients. We used Univariate Linear Mixed Models from GEMMA (v0.98.3) to perform association analysis with the Wald test [23, 24], an additive genetic model incorporating birthday, sex, batch, and 5 genetic principal components (PCs) as covariates. PCs were obtained from an LD-pruned dataset of 573,457 SNPs. The number of PCs to be included in the regression was determined by inspecting the proportion of variance explained and by checking PC scatter plots.

ATAC-seq and analysis of CLN8 promoter binding region

Fifty thousand nuclei from Pekin duck subcutaneous preadipocytes (n = 2) before and after oleic acid-induced preadipocyte differentiation were transposed using Tn5 transposase as previously described [25]. The PE150 ATAC-seq reads were mapped to the Mallard reference genome (GenBank: ASM874695v1) using Burrows-Wheeler Aligner (BWA) with default parameters [19]. SAM files were converted to the BAM format using Samtools (v0.1.19) [26]. The peaks were called using MACS2 in each sample and using the filtered BAM files with the parameters (-g 9.6e8 -q 0.01 -B –SPMR –nomodel -shift -100 -extsize 200) [27]. All the alignment files were scaled to RPKM-normalized read coverage files using DeepTools [28]. The library size factors were estimated to compare binding profiles between different samples in an unbiased manner by the DESeq2 [29]. Differentially accessible regions were detected using DESeq2 with a fold change of more than 1.5 and adjusted P value (Q-value) below 0.05. The HOMER tool (http://homer.salk.edu/homer/motif/) was used to detect the motifs.

Plasmid construction

CLN8 knock-in plasmids: In order to construct the CLN8-over-expression vector, the full-length coding sequence of chCLN8 (NCBI reference sequence: NM_204214.2) was amplified from chicken subcutaneous adipose cDNA by PCR using the primers listed in Additional file 1, and cloned into the CMV promoter-driven piggyBac and an EF1α promoter-driven cherry plasmid by replacing cherry using NheI and AccI (New England Biolabs, Ipswich, MA, USA). pGL3.1 dual-luciferase reporters: Different length fragments of the upstream area of CLN8 (NCBI reference sequence: NC_052534.1) were amplified by PCR using the primers listed in Additional file 1 from Pekin duck subcutaneous adipose DNA, and then, cloned into pGL3.1-basic vector. The important candidate mutant fragments were synthesized by Tsingke Biotechnology Co., Ltd., and cloned into pGL3.1-promoter vector.

Gene over-expression vector: The C/EBPα over-expression vector was constructed according to the user manual of Easy Ligation Kit (Sidansai, Shanghai, China). C/EBPα coding sequence (NCBI reference sequence: NM_001031459.2) was amplified from chicken subcutaneous adipose cDNA by PCR. The PCR product was cloned into the pcDNA3.1 vector. The successful C/EBPα over-expression vector, was confirmed by DNA sequencing.

Cell culture and transfections

A cell line of immortalized chicken preadipocytes (ICPs) [30] was cultured in DMEM/F12 (Gibco, Gaithersburg, MD, USA) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT, USA), and 0.1% penicillin/streptomycin (Invitrogen, Carlsbad, CA, USA). To induce ICPs differentiation, we added 160 µmol/L sodium oleate (Sigma Life Science, St. Louis, MO, USA) to the medium [31].

The pGL3.1 dual-luciferase reporters' transfections were performed with Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s direction. Nucleic acids were diluted in OPTI-MEM Medium (Gibco, Gaithersburg, MD, USA). All experiments were carried out at least three times independently.

For CLN8OE cell selection, ICPs were seeded in 6-well plates for further transfection using FuGENE® HD Transfection Reagent (Promega, Madison, WI, USA). After a 48-h recovery period, the cells were supplemented with 1.5 μg/mL of puromycin (Sigma-Aldrich, MO, USA) in the culture medium for 3 d until wild cells dead. Cells were harvested using 0.25% trypsin/EDTA (Gibco, Gaithersburg, MD, USA).

Oil red O staining and quantification

The cells were washed with PBS and fixed in 4% formaldehyde for 20 min. Then the cells were stained with Modified Oil red O Staining Kit (Beyotime, Jiangsu, China) according to the manufacturer’s manual. Morphological changes were observed and photographed under an inverted fluorescent microscope (Nikon). The Oil red O dyes were then extracted in an isopropanol solution containing 4% Nonidet P-40 and quantified by a Model 680 Microplate Reader (Bio-Rad) at 510 nm.

Dual-luciferase reporter assay

For the promoter activity assays, ICPs were transfected with the luciferase reporter gene vector which contained different deletion fragments of the CLN8 promoter or control vector, and the TK-Renilla reporter was also co-transfected to each sample as an internal control using the Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) in 24-well plates. After 48 h transfection, cells were washed by PBS twice, and the activities of Firefly and Renilla luciferase were measured according to the manual of Dual-Luciferase Reporter Assay Kit (Vazyme, Nanjing, China), and the relative fluorescence intensity was detected by a microplate reader. All the data were acquired by averaging the results from three independent repeats.

RNA extraction, cDNA synthesis, and quantitative real-time PCR

Total RNA was isolated from the cells using RNAiso reagent (Takara, Otsu, Japan) according to the manufacturer’s instructions. According to the manufacturer's manual, the reverse transcription reaction for mRNA was performed with the PrimeScript RT reagent Kit (Perfect Real-Time; Takara, Otsu, Japan). Using total RNA as a template and oligo (DT) primer, the first strand of cDNA was inverted. The specific RT-qPCR primers of mRNA were designed using Primer 3 software (version 0.4.0, Howard Hughes Medical Institute). Primer sets are listed in Additional file 1. The amplification efficiency and specificity of all the primers were tested before the formal quantitative experiment. Real-Time fluorescence quantification PCR (RT-qPCR) was performed using TB green premix Ex Taq™ fluorescence quantitative kit (Takara, Otsu, Japan), and the qPCR program was carried out in an ABI-7500 PCR machine (Applied Biosystems, MA, USA). The housekeeping gene GAPDH was used as the internal control, and the method of 2-ΔΔCt was used to quantification the results as described [32]. All reactions were run in triplicate.

Western blot

Cultured cells were washed with PBS and homogenized with RIPA buffer (Beyotime, Jiangsu, China) containing a protease inhibitor cocktail (Beyotime, Jiangsu, China). Protein concentrations were determined using the BCA Protein Assay Kit (Beyotime, Jiangsu, China). Proteins were denatured and subjected to 12% polyacrylamide gel and transferred to methanol‐activated PVDF membranes. Blots were probed using the primary antibodies: Rabbit polyclonal anti-FLAG antibody for Western blots (1:500; Sigma, Cat. No./F7425), Rabbit polyclonal anti-beta actin antibody, (1:3,000; Proteintech, Cat. No./20536–1-AP), overnight at 4 °C. After 1 h incubation with HRP-conjugated Affinipure goat anti-rabbit IgG (H + L) as a secondary antibody (1:5,000, Proteintech, Cat. No./SA00001-2), immunodetection was performed using enhanced chemiluminescence (ECL) Western blotting substrate (Beyotime, Jiangsu, China) and detected with FluoChem R imaging system (ProteinSimple, CA, USA).

RNA-Seq analysis

Raw reads were trimmed to remove adapters and low-quality reads, with Trimmomatic (version 0.39) [33]. Read counts for each gene were calculated using Salmon (v1.8.0) [34] and GRCg7b (GalGal1.mat.broiler.GRCg7b) as the reference genome. We used the DESeq2 (v.1.28.1) package [29] to identify differentially expressed genes (DEGs) between CLN8NC, and CLN8OE cells at different days (d 0 and 3). Therefore, samples were excluded from further analysis due to its low global Pearson correlation with the other repeat samples (R2 < 0.95). Genes with |log2FC| ≥ 2 and the Benjamini & Hochberg (BH) adjusted P-value (adjusted-P value) < 0.05 were considered as DEGs. Transcription factor prediction using an online website (http://gene-regulation.com/pub/programs/alibaba2/index.html) [35].

Functional annotation and enrichment

Genes were annotated with gene Symbols from the UniProt database for functional annotation. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of the enriched genes was performed using the web-based Metascape (http://metascape.org/gp/index.html#/main) [36].

Results

Summary of the phenotypes and genotypes

The basic statistics of 42-d body weight, subcutaneous fat weight, subcutaneous fat percentage, abdominal fat weight, and abdominal fat percentage for each population were shown in Additional file 2. The heritability, genetic correlations, and phenotypic correlations of each trait were shown in Table 1. The estimates of heritability range from 0.42 to 0.61. Trait SFW presented the highest heritability (0.61) estimates, while the AFP was the lowest (0.42). The highest phenotypic correlation (0.95) and genetic correlation (0.98) were shown between AFW and AFP, and the lowest phenotypic correlation (0.33) and genetic correlation (0.26) were calculated between BW42 and AFP. The phenotypic and genetic correlations among these fat-related traits (SFW, SFP, AFW, and AFP) were all greater than or equal to 0.62.

Table 1 Genetic parameters*
Full size table

The population of Pop2014, Pop2019, and Pop2020 were obtained 1,778 Gb, 4,799.83 Gb, and 4,709.25 Gb clean data, respectively. The average reads mapping alignment rate was 96.57%. The variant calling pipeline identified 14,220,859 SNPs within the duck genomes. After quality filtration and exclusion of variants of uncharacterized chromosomes, 8,448,069 SNPs and 573,457 independent SNPs passed our filters. A principal component analysis (PCA) was constructed to examine the relatedness among three populations (Additional file 3). The PCs showed that all samples were divided into two subpopulations, which were Pop2014 and the other two.

Genome-wide association study identified the candidate variants and genes

A total of 648 chromosome-wide significant SNPs (1/573457, P < 1.74E-06) and 242 genome-wide significant SNPs (0.05/573457, P < 8.72E-08) across 40 chromosomes were identified by loci-based analysis. We defined candidate genes as genes surrounding significant SNPs in 5 kb extend region. A list of the genome-level significantly associated variants is shown in Additional file 4 and a list of the chromosome-level significantly associated variants is shown in Additional file 5. There were 71, 47, 155, 64, and 14 genome-wide significant SNPs identified for traits BW42, SFW, SFP, AFW, and AFP, respectively. In total, 39 candidate genes were annotated from 68 genome-level significant SNPs. The estimated genomic inflation factor λ ranged from 1.027 to 1.066 among 5 traits (Additional file 6), suggesting no population stratification in the studied population after specifying sex, age, and batch effects as covariates. Manhattan and Q-Q plot of association results from genome-wide association analysis of each trait are shown in Additional file 6. We identified a 7-Mb region between nucleotide position 54 Mb and 61 Mb in chromosome 4 that showed a significant association with the 42-day body weight phenotype with the most significant of 3.86E−10 (rs458603786, T > C, intergenic variant, P = 3.86E−10). As for fat traits, two genome-wide significant SNPs (rs117362452, G > A, synonymous variant, and rs291299411, G > C, intron variant) were detected to be correlated with SFW, SFP, AFW, and AFP, which were located at 7.36 Mb on Chr11 and 1.29 Mb on Chr 29. There was a high genetic correlation between subcutaneous fat and abdominal fat, 11 genes (Additional file 5) were common genes between subcutaneous and abdominal fat were identified among a total of 35 significant genes from SFW, SFP, AFW, and AFP. By querying the functional annotations of these genes, we found that CLN8 is a potential gene affecting lipid synthesis and transport, which was related to SFW, SFP, and AFW. The joint Manhattan plot of SFW and SFP showed in Fig. 1A.

Fig. 1
figure 1

Identification of positional candidate genes for the skin fat phenotype in ducks. A Manhattan plot of association results from genome-wide association analysis. Y axis shows −log10 (P-value) of the association result for each SNP. Each SNP is indicated by a colored dot. SNPs are plotted based on the reference duck genome chromosome (x-axis). The horizontal solid line is the threshold for the Bonferroni level of significance (P < 8.72E−08). The horizontal dashed line is the threshold for the Bonferroni level of significance (P < 1.74E−06). B Haplotype block at linkage disequilibrium (LD) with the skin fat based on 5 candidate variations. The diagram above the haplotype block indicates genes around candidate variations. C The skin fat weight (SFW) of the ducks with different Genotypes. The WT represents the genotype with identical alleles (-T-C-A-A-C) at corresponding chromosomal loci, the Homozygous represents the genotype with identical alleles (-A-T-C–C-G-) at corresponding chromosomal loci, and the Heterozygous represents the genotype with two different alleles (-T-C-A-A-C and -A-T-C–C-G-) at the given locus. The level of significance was presented as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 

Full size image

ATAC-seq analysis of mutations in the upstream of CLN8

The five SNPs (rs322493594, rs322493619, rs322493641, rs322493648, rs322493651) were located in 1,920 bp upstream of the CLN8 gene on chromosome 3 (Additional file 7). The rs322493651 was the most significant SNP accounting for 1.46% of the genetic variance, and 11.18% of the phenotypic variance. Linkage analysis revealed that they were in high linkage disequilibrium (LD, r2 > 0.9; Fig. 1B). These results prompted us to speculate that the causative SNPs should be the 5 variations. To confirm the association of the 5 SNPs with the subcutaneous fat phenotype, 616 ducks were genotyped for the CLN8 heterozygous mutation and 3 for the homozygous, the wild type, and heterozygous mutation of haplotype with 5 SNPs significantly decreased the subcutaneous fat weight (P < 0.0001) (Fig. 1C). We used the ATAC-seq to investigate the chromatin accessibility landscape of subcutaneous preadipocytes. Fragments from the nucleosome-free regions (NFR) are expected to be enriched around the transcription start site (TSS) of genes (Additional file 8). Transcription factor binding sites were identified from 22,493,347 bp to 22,494,147 bp, which is located in the upstream of gene CLN8 (Fig. 2A). To exploit motif information, the JASPAR database was used to predict the motif binding site [37]. Different motif binding site was predicted as shown in Fig. 2B and C between wild-type sequence (up) and mutation sequence (down). ATAC-seq analysis showed that transcription factor binding sites were identified in a region close to the haplotype.

Fig. 2
figure 2

ATAC-seq predicts transcription factor binding sites in the upstream of CLN8. A Gene structure and nucleosome-free regions fragments in the upstream of CLN8. It showed the levels of transposase-accessible chromatin in the region from 22.492 Mb to 22.493 Mb. The P0 and P3 represent subcutaneous preadipocytes (n = 2) before (d 0) and after (d 3) oleic acid-induced differentiation respectively. The F0 and F3 represent abdominal preadipocytes (n = 2) before (d 0) and F3 after (d 3) oleic acid-induced differentiation, respectively. B The Motif binding site in the annotated promoter region. Motif predicted in wild type sequence (up, -T-C-A-A-C-) and in mutant type (down, -A-T-C–C-G-). C Sequence logo map of the promoter region, the box area indicated the position of the mutation

Full size image

Active region screening of the CLN8 promoter

Based on Pekin duck genomic DNA, we designed primers in the proposed promoter region of CLN8 and amplified seven deletion fragments of different lengths: 3,009 bp, 2,519 bp, 2,026 bp, 1,349 bp, 811 bp, 501 bp, and 285 bp (Fig. 3A). To explore the active region of the CLN8 promoter, seven truncated fragments of the CLN8 promoter were amplified and inserted into the pGL3.1-Basic vector (Promega, USA) using restriction enzymes Kpn I and Xho I (Fig. 3B), and were co-transfected into ICP1 cells with the pRL-TK vector for luciferase activity detection. The results showed that the luciferase activity of pGL3.1-F3, pGL3.1-F2, and pGL3.1-F1 were significantly increased compared to that of pGL3.1-Basic, which was the control group (P < 0.05), the promoter activity of pGL3.1-F4 was significantly reduced (P < 0.05), and the activity of pGL3.1-F3 was significantly increased compared to that of pGL3.1-F4 (P < 0.05) (Fig. 3C). These results demonstrated that the region from −1,884 to −1,207 bp upstream of the transcription start site of CLN8 was the promoter core transcription active region.

Fig. 3
figure 3

Active region screening of the CLN8 promoter. A Different lengths fragment of CLN8 promoter cloned by PCR. B Plasmid digested of pGL3.1-Basic Vector by NheI and XhoI. C Relative luciferase activity of different promoter fragments. F1-F7: Fluorescence activity values detected after transfection of ICP1 cells with fluorescence vectors containing promoter truncated fragments of different lengths, using the pRL-TK as reference. Data are shown as mean ± SD of three biological replicates. An independent sample t-test was used to analyze the statistical differences between groups. a–cDifferent lowercase letters showed significant differences (P < 0.05) in relative luciferase activity between groups

Full size image

The functional polymorphism SNPs decreased the transcriptional activity of CLN8

We next investigated the effect of the 5 SNPs (wild-type: -A-T-C–C-G- and Mutant-type: -T-C-A-A-C-) variant on the transcriptional activity of the CLN8 gene. Wild and Mutant haplotype CLN8 plasmids were verified before transfection by direct sequencing (Fig. 4A). To assess the effects of the polymorphisms, we generated two luciferase reporter gene constructs that share identical backbone sequences except for the polymorphisms, as shown in Fig. 4B, reporter gene expression driven by the A-T-C–C-G-containing CLN8 promoter (wild-type) was greater than that driven by the T-C-A-A-C-containing counterpart (mutant-type) (P < 0.001). This result demonstrated that significantly lower transcriptional activity of the mutant haplotype was observed when compared with the wild types.

Fig. 4
figure 4

The functional polymorphism SNPs decreased the transcriptional activity of CLN8. A Wild-type and mutant-type CLN8 plasmids were verified by direct sequencing. B Three luciferase reporter gene constructs were generated. They share identical backbone sequences except for the polymorphisms as left. Significantly lower luciferase activity of the mutant type (-A-T-C–C-G-) haplotype was observed when compared with the wild type (-T-C-A-A-C-) haplotype vectors in the ICP1 cell line. Data are shown as mean ± SD of three biological replicates. An independent sample t-test was used to analyze the statistical differences between groups. The level of significance was presented as *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

CLN8 promotes the differentiation of avian adipocytes

To further determine the roles of CLN8 in avian adipocyte differentiation, we first performed gain-of-function experiments by using piggyback delivery of CLN8 into ICPs, and we detected its expression in the overexpression group and control group cells. As shown in Fig. 5A and B, a transfected CLN8 clone can significantly increase CLN8 expression at the level of transcription and translation by qPCR and western blot. Further, we detected the expression of adipocyte markers, PPARγ, and FABP4, and the results showed that they were significantly increased in CLN8OE cells (Fig. 5C and D). As shown by Oil red O staining of neutral lipids and detection of the relative lipid content, over-expression of CLN8 in these cells (CLN8OE) significantly facilitated lipid formation (Fig. 5E and F). These results indicate that CLN8 is a positive regulator for the avian adipogenesis.

Fig. 5
figure 5

Over-expression with CLN8 promotes the differentiation of adipocytes. A and B CLN8OE cell significantly promoted CLN8 mRNA and protein expression in ICPs. C and D mRNA levels of adipocyte genes PPARγ and FABP4 were analyzed with RT-PCR. E Representative images of CLN8OE cells increased the lipid droplet formation by Oil red O staining on d 3. F Comparison of the lipid droplet content of CLN8OE and CLN8NC cells obtained by Oil red O staining and extraction methods. Data are shown as mean ± SD of three biological replicates. An independent sample t-test was used to analyze the statistical differences between groups. The level of significance was presented as *P < 0.05,  **P < 0.01, ***P < 0.001

Full size image

CLN8 is involved in the gene regulation network of preadipocyte differentiation

To better understand the effect of CLN8 on adipogenesis, we performed mRNA-Seq experiments in CLN8OE and CLN8NC cells prior to differentiation (d 0) and d 3 after differentiation (Additional file 9, 10). To confirm the results from mRNA-Seq, RT-PCR was performed in the present study firstly. Our RT-PCR results were highly consistent with the mRNA-Seq results (Additional files 11, 12, 13). In total, 2,488 and 2,159 DEGs were up-regulated and down-regulated in CLN8OE cells compared to CLN8NC cells on d 3, respectively (Additional file 14). GO analysis for DEGs up-regulated in CLN8OE cells are enriched for cell differentiation processes, while DEGs up-regulated in CLN8NC cells are enriched for the terms related to cell cycle (Table 2; Additional file 15). These results indicate that the knock-in of CLN8 alone has initiated the preadipocyte differentiation early. Notably, we focused on the DEGs enriched in entries directly related to fat cell differentiation in CLN8OE cells compared to the CLN8NC group. There are 9 up-regulated DEGs were enriched in the positive regulation of fat cell differentiation (PTGS2, PPARG, MAPK14, ASXL2, PPARD, STK4, KLF5, BMP2, ZC3H12A), and 9 down-regulated DEGs were enriched in the negative regulation of fat cell differentiation (ARNTL, RUNX1T1, GATA2, CCN4, YAP1, ANKRD26, ZFPM1, BBS12, ASXL1) (Additional file 15). These data suggest that CLN8 is involved in the gene regulation network of avian preadipocyte differentiation.

Table 2 The representative GO enrichment analysis terms of DEGs in CLN8OE and CLN8NC cells
Full size table

Discussion

Numerous studies have shown that SNP variation in coding regions, which affects the sequence of amino acids or the folding of peptide chains by changing codons, can alter certain phenotypic traits of organisms [38,39,40]. However, the effect of regulatory regions on animal phenotypes may be underestimated, and a growing number of studies have found the effects of regulatory regions on phenotypes [41]. In recent studies, regulatory regions have also been found to influence individual phenotypes by regulating gene expression. In humans, ICOS SNPs in 3′ UTR are able to regulate ICOS gene expression by influencing miRNA binding sites [42]. In addition, when the intron region is used as an enhancer, the associated SNP variant affects the expression of genes and may affect obesity in humans [43]. In livestock, variation in the promoter zone may affect the number of litter births in sheep [44], breed differentiation, and fat deposition of pigs [45, 46]. Therefore, as an important regulatory region of gene expression, the promoter region can also change the gene expression level by affecting the binding efficiency of a variety of transcription factors, thereby affecting gene function polymorphisms and animal body development process. So, the study of the variation of the promoter region is of great significance for the future field of animal breeding.

In this study, by querying the functional annotations of all significant genes and identification of transcription factor binding sites in the upstream of gene CLN8 by ATAC-seq, we identified that CLN8 is a positive regulator of avian adipocyte differentiation. Moreover, we investigated the effect of these variants on the transcriptional activity of the CLN8 gene. Our data showed that 5 SNPs found on chromosome 3 are functional polymorphisms and that the mutant haplotype leads to the lower transcriptional activity of the CLN8 gene than the wild-type, which is consistent with observations from our phenotype-profiling experiments that the wild-type, and heterozygous mutation of haplotype with 5 SNPs significantly decreased the subcutaneous fat weight. We analyzed the transcriptional activity of the upstream regulatory region of the CLN8 gene through experiments, and further through Alibaba2 website predicting, it was found that there are many adipogenesis-related transcription factors binding sites on its core transcriptional region, including C/EBPα, C/EBPβ, C/EBPδ, Oct-1, Sp1, GATA1, c-Jun, NF-1, MYOD, AP1, Myc, etc. (Additional file 16), which provide important clues for further exploration of the regulatory mechanism of CLN8 on avian fat deposition. Further, we found that the mutation can lead to a different binding of transcript factors and affect gene expression in adipocytes. These variant alleles may up-regulate CLN8 expression through overexpression C/EBPα (Additional file 17), which may function as a regulatory factor to combine the upstream gene region to affect expression of CLN8. C/EBPα is one of the most studied transcription factors in hematopoiesis is the leucine zipper CCAAT-enhancer binding protein, which is mainly involved in cell fate decisions for myeloid differentiation [47]. As a transcription factor, C/EBPα can translocate into the nucleus and further regulate a variety of genes directly or indirectly, which are all key factors for cell differentiation [48]. The currently available data are persuasive evidence suggesting that C/EBPα expression contributes to the development of fat deposition, and CLN8 expression as well as its function represents a promising target for gene C/EBPα. Therefore, we can further explore the specific mechanism of C/EBPα regulating the expression of adipose-promoting factor CLN8 by combining with candidate causal SNPs through experimental techniques, such as EMSA, ChIP, and their corresponding sequencing analysis techniques.

Previous and current transcriptomic data have shown that a large number of genes are significantly changed during avian adipocyte differentiation [49,50,51], but their regulatory role in adipocyte differentiation is still poorly understood. In this study, we successfully established an avian CLN8OE cell line with the Piggyback technique and discovered that CLN8 is an important positive regulator of avian adipocyte differentiation. Among these differentially expressed genes, most genes have been reported to have well-defined regulatory functions during adipogenesis in different species. Several potential regulatory pathways could be inferred from the differentially expressed genes and previous studies. For example, KLF5, in turn, acts in concert with C/EBPβ/δ to activate the PPARγ2 promoter, and functions as a key component of the transcription factor network controlling adipocyte differentiation [52]. Murine GATA2 has specifically been expressed in white adipocyte precursors and their down-regulation sets the stage for terminal differentiation. GATA2 and GATA3 regulate adipocyte differentiation through molecular control of the preadipocyte-adipocyte transition [53]. In addition to its previously recognized function in suppressing PPARγ transcriptional activity, the interaction of GATA factors with C/EBP is necessary for their ability to negatively regulate adipogenesis [54]. The RNA-Seq results suggest that CLN8 may positively regulate avian adipogenesis by modulating the expression of a set of genes related to fat cell differentiation, including PPARG, which is a core regulator of adipocyte differentiation. Overexpression of the CLN8 gene could affect these gene expressions directly or indirectly and then effects adipogenesis in avians.

Conclusions

Avian fat deposition is an important genetic trait that has been related to numerous interesting biological functions, such as energy storage, stress resistance, and immunity. Large-scale genome-wide association analysis identified 35 genes related to duck adipose traits. Among these candidate genes, we demonstrated that the complex transcriptional of CLN8 is regulated by the 5 linkage SNPs in the upstream, leading to a significantly decreased subcutaneous fat weight in Pekin duck. SNPs in the promoter of CLN8 can lead to an altered transcriptional activity, and consequently, modulate the adipogenesis in avians. This study presented that promoter mutation of a novel CLN8 could cause phenotypes changes, and is useful for avian adipose biology and breeding.