Selection signature analysis in chickens divergently selected for growth rate reveals novel candidate genes regulating fat deposition

Abstract

Background

Decades of intensive breeding for rapid growth rate has resulted in increased abdominal fat content in commercial broilers, which also led to significant economic loss in this industry. In the present study, we integrated RNA-Seq datasets of 44 samples, including 22 fat- and 22 lean-line, to identify the selection signatures linked to abdominal fat content in chickens.

Results

In total, 68 selection signature regions in the top 0.1% were screened out via Fst analysis on chromosomes 1, 2, 4, 5, 9, 11, 13, 15, and 18 were identified under differential selection between fat- and lean-lines, which harbored 1,140 SNPs and 44 genes. Functional annotation analysis highlighted key biological processes and KEGG pathways related to fat metabolism, such as “Fatty Acid Metabolic Process”, “Lipid Import Into Cell”, “Triglyceride Biosynthetic Process” and “Long-Chain Fatty Acid Transport”. The results confirmed several previously reported candidate genes involved in fat deposition, such as ACSL3, ACSF2, MOGAT1, TBXAS1, and NDST4. Notably, the NDST4 and YIPF7 genes were of particular interest, as they are closely linked with the QTLs associated with fatness traits, which makes them well suited for future applications in poultry breeding programs. Moreover, some novel candidate genes associated with fat metabolism were identified, including GUF1, GNPDA2, SLC25A48, RBFOX3, FRMD4A and KCNE4. While the exact mechanisms by which novel candidate genes contribute to abdominal fat deposition are not fully understood yet, but they appear to play relevant roles in fat metabolism that make them promising candidates for further investigations. Furthermore, the identified candidate regions harbored two miRNAs, of which mir-205b is of particular interest and can be considered as a potential candidate involved in the genetic control of abdominal fat deposition, as mainly participate in pathways associated with lipid metabolism. These genes can pave the way for the optimization of the breeding programs associated with fatness to promote chicken broilers performance.

Conclusions

Overall, these findings enrich our understanding of the genetic mechanisms underlying abdominal fat deposition in chickens. Of note, our approach for selection signature analysis can be applied to other traits as well as species with available RNA-Seq data to identify the genomic signals associated with the divergence of different phenotypes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-025-12360-9.

Background

Nowadays, broiler chickens (Gallus gallus) industry plays a crucial role in supplying animal protein to meet the nutritional needs of a world’s growing human population. Throughout a long history of genetic breeding, chickens have been selected for rapid growth and higher body weight, which has always been accompanied by increased abdominal fat deposition. While the intensive artificial selection has dramatically increased poultry productivity, excessive abdominal fat content in broilers poses challenges, causing skeletal abnormalities, metabolic disorders and decreasing feed efficiency [1]. Additionally, abdominal fat accumulation needs more energy than accretion of an equivalent amount of lean tissue, which sharply increase feed cost [2]. On the other hand, the achievements of conventional breeding methods to decrease abdominal fat deposition have been substantially disrupted due to the positive genetic and phenotypic correlations between abdominal fat content and body weight [3]. Therefore, a comprehensive understanding of the molecular mechanisms and pathways governing abdominal fat deposition may be helpful for advancing breeding programs in broiler chickens.

Selected genomic regions due to intensive selection leave footprints across the genome over time, called selection signatures, which often harbor functionally important genes directly associated to the phenotypic trait targeted by the selection process. Therefore, a comprehensive search for the selection signatures and positional candidate genes associated with fatness traits, may provide a direct insight into the molecular mechanisms contributing to the regulation of this process [4]. Different genomic approaches have been applied to identify genes or genetics factors associated with abdominal fat related traits in chicken [5, 6]. In this context, more than 300 quantitative trait loci (QTLs) have been reported for various fat-related traits of chicken in AnimalQTLdb database [7], including abdominal fat weight, intramuscular fat percentage, carcass fat content and intramuscular fat percentage [8–10]. However, most of these QTLs have been detected using low density of SNP arrays and the genomic locations of them have large confidence intervals [11]. Furthermore, some studies applied high-density SNP arrays to perform selection signature analysis and identify genomic signals and potential causal genes related to fatness traits in chickens. The results of these studies have revealed some genomic regions and genes associated with various traits, such as abdominal fat content [12, 13], fat deposition and muscle development [14], body composition and meat quality [15, 16]; and fatness [17].

Despite these studies, the selection signature analysis based on SNP arrays suffer from some limitations, including 1) arrays have constraints in SNP coverage and depth, as do not consider genome-wide genetic diversity and rare SNPs are missed (MAF < 0.1%), 2) only known SNPs are genotyped and unknown population-specific SNPs are neglected, and 3) due to lack of probe design options, some SNPs in duplicated regions of the genome cannot be genotyped [18]. Currently, with the development of next generation sequencing (NGS) technology, whole-genome resequencing (WGS) has become a powerful approach to identify genome-wide SNPs comprehensively and efficiently. Zhu et al., used WGS to investigate the genetic basis of abdominal fat in chickens by GWAS and reported a SNP, rs312715211, in ZNF652 as the key variant associated with this trait [19]. So far, to the best of our knowledge, only a few studies have been conducted to use WGS data to identify selection signature regions associated with fatness traits in chickens [20]. In contrast, there are many studies that used RNA sequencing (RNA-Seq) method to investigate expression profile between different conditions in chickens. In our previous studies, we demonstrated that RNA-Seq datasets can be a cost-effective alternative to WGS for SNP calling [21–24]. Furthermore, in a recent study we showed the great potential of the combined use of diverse RNA-Seq datasets to study genome-wide signatures of selection [18]. Hence, while the genetic basis of fat-deposition in broiler chickens has not been fully understood and remain an active subject of research, integrating RNA-Seq datasets to perform selection signature analysis may be helpful in this regard [18]. Following our earlier work in sheep, we aimed to identify the selection signatures associated with fat deposition in broiler chickens, through comparison between fat- and lean-line chickens. To do this, four RNA-Seq studies were selected, which have investigated gene expression changes in abdominal fat tissue between two divergent chicken lines (fat- vs lean). According to a stringent bioinformatic pipeline, selection signature regions underlying variation of fat deposition were detected. Also, the identified regions were subjected to functional enrichment analysis to elucidate the significance of the candidate genes under selection to pathways related to fat metabolism. Findings of the present study are anticipated to enhance the understanding of the genetic mechanisms regulating fat deposition in chickens and may provide a promising avenue for future genetic improvement in poultry production.

Methods

RNA-Seq datasets

A detailed review of the literature identified four cohorts of RNA-Seq studies that have compared chicken lines with extreme abdominal fat content (fat- vs lean-line). All samples were collected from abdominal fat tissue of male broiler chickens and subjected to paired-end sequencing. All datasets, including 22 fat- and 22 lean-line samples (Table 1), were retrieved from NCBI GEO database and processed using the same bioinformatics pipeline (Fig. 1). A similar method to our previous study was used to detect signatures of selection, explained as follow [18].

Table 1.

List of datasets used in the present study

Accession number	Registration date	Source	Fat-line	Lean-line	Reads Length	Reference
PRJNA657377	16-Aug-2020	Northeast Agricultural University	Arbor Acres (3)a	Arbor Acres (3)	150 bp	[25]
PRJNA184019	18-Dec-2012	Avian Functional Genomics, Animal and Food Sciences, University of Delaware	Leghorn (12)	Leghorn (12)	101–105 bp	[26]
PRJNA354990	26-Nov-2016	Northeast Agricultural University	Arbor Acres (5)	Arbor Acres (5)	150 bp	[27]
PRJNA248570	26-May-2014	French National Institute for Agronomical Research (INRA)	Fayoumi (2)	Fayoumi (2)	100 bp	[28]

^aNumbers in parentheses indicate biological replicates

Fig. 1.

Analysis pipeline used for detecting selection signature regions based on RNA-Seq datasets

Quality control and selection sweep mapping

The quality control of the raw and clean reads were performed using FastQC (version 0.11.5) [29]. Trimmomatic (version 0.38) [30] was applied to trim low-quality bases/reads and adaptor sequences (TRAILING: 20, MAXINFO: 120:0.90, and MINLEN: 120). Then, clean reads were mapped to the most current chicken genome (broiler GRCg7b available at ensembl (https://ftp.ensembl.org/pub/release-114/fasta/gallus_gallus/dna/Gallus_gallus.bGalGal1.mat.broiler.GRCg7b.dna.toplevel.fa.gz.)) using the STAR (version a2.7.9) software with a two-pass alignment approach, incorporating known and novel splice junctions. The accuracy of alignment is enhanced by two-pass alignment approach, as splicing junctions are collected from each sample separately during the first pass, and integrated into the genome to provide consistent junctions among samples during the second pass [31]. The known splice site information was extracted from the available GTF annotation file of the chicken genome (GRCg7b, release 110, ENSEMBL database). Only uniquely and concordantly paired-end aligned reads were retained, and PCR duplicates were marked and excluded using GATK’s MarkDuplicates function [32]. Base quality score recalibration was performed using GATK tool (version 4.2.6.1) based on Ensembl chicken SNP database to enhance SNP calling accuracy.

SNP calling

The HaplotypeCaller module of GATK was used to identify SNPs across all the samples (stand_call_conf and stand_emit_conf value of 30 and mbq of 25). This process was followed by SNP quality filtering using standard parameters, including HomopolymerRun > 5, total depth of coverage < 10, QualitybyDepth < 2, RMSMappingQuality < 40, MappingQualityRankSum < −12.5, and ReadPosRankSum < −8, to ensure high-quality data [32]. The selection criteria for including variants were those that supported by a minimum of three reads and exhibited biallelic polymorphism. Also, the variants located in problematic regions (including simple sequence repeats regions ± 3 bases and splice signal regions ± 5 bases) were excluded [18]. Only variants that passed these pre-established filters and were listed as known SNPs in the Ensembl chicken SNP database were retained for further analysis, thereby enhancing the confidence of the detected SNPs.

Imputation

To enhance the completeness of genotypic data, genotype imputation was conducted using a reference SNP file from the Chicken Genotype-Tissue Expression (ChickenGTEx) project. This reference panel has been constructed based on 2,869 WGS data from 152 chicken breeds, which provide a high-quality multiple-breed genotype imputation panel (https://ngdc.cncb.ac.cn/chickengtex/) [33, 34]. Given that the genomic coordinates in this reference panel were based on the previous reference genome of chicken (GRCg6a), LiftOver tool (version 46e) was employed to convert the genomic coordinates to the current chicken reference genome (GRCg7b) [35]. Imputation was performed using Beagle software (version 5.4) according to the haplotype frequency model. Only the imputed SNPs with a prediction probability above 80% (according to DR2 value, squared correlation between imputed and real genotypes) and a minor allele frequency (MAF) greater than 0.05 were kept for further analysis [36].

Screening for genomic regions under selection

Based on the SNPs remaining after imputation, genomic regions that may be under selection between the fat- and lean-lines were investigated using Fst analysis (Weir and Cockerham) [37]. Fst measures the level of genetic differentiation between two populations (here, fat- vs lean-lines chicken), based on differences in allele frequencies. The Fst values for the SNPs were calculated using VCFtools tool and a sliding window approach, by setting a scanning window of 30 kb and a step size of 10 kb within each sliding window [14]. Regions with extreme Fst values (top 0.1%) were considered potential signatures of selection regions [38]. R software was employed to perform principal component analysis (PCA) based on the identified SNPs in the candidate regions and the first two principal components (PC) were plotted.

Functional annotation analysis

According to the genome annotation file (GRCg7b, release 110, ENSEMBL database), a gene was assumed to be under selection if it located within a selection signature region. Hence, the candidate regions were annotated using SNPeff tool (version 5.1) and the overlapped genes were detected. To explore the potential biological significance of the genes, functional enrichment analysis was conducted using the EnrichR tool [39]. Biological processes terms, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome pathways with a false discovery rate (FDR) below 0.05 were considered significant. It is worth noting that we designed a new Circos plot, named Circos functional annotation plot, to visualize the results of functional enrichment analysis. This plot shows the terms, their p-values and connections among the genes and related terms.

Overlap with known QTLs

The overlap of the identified candidate genes with the relevant QTLs to the fattens-related traits was determined using the information available at the Chicken QTLdb database. The current version of this database (release 55, accessed in March 2025) contains 26,309 QTLs representing 206 different traits. First, 263 QTLs associated with fattens, such as intramuscular fat percentage, abdominal fat weight and carcass fat content, were retrieved from the database. Then, genomic coordinates of the genes with these QTLs were compared and the overlapped regions were detected.

Results

RNA-Seq dataset analysis

A total number of 44 RNA-Seq samples (including 22 fat- and 22 lean-lines) related to four studies were integrated. The total number of raw reads was ~ 1,824 million, with an average of 41 million reads per sample. Of these, 1,572 million reads were retained after trimming, with an average of 36 million reads per sample. On average, 88% of the reads were aligned to the reference genome. Of the aligned reads, only uniquely aligned reads were retained for further analysis (~ 1,221 million reads). Details of the number of reads and mapping analysis is provided (Supplementary File S1).

SNP calling

After variant calling and imputation, 2,883,748 unique biallelic SNPs were identified across all samples, which have also been reported in chicken dbSNP database. It is worth emphasizing that only known SNPs were used to increase the reliability of the results. The complete list of the identified SNPs per sample is provided (Supplementary File S2).

Detection of selection signature regions

To explore the genomic evidence related to selection, Fst method was employed and distinctions between the fat- and lean-lines of chickens were investigated. According to the window-based approach, 68,070 genomic regions of 30 kb with a step of 10 kb were tested by computing an average Fst based on the SNPs overlapping the window (Supplementary File S3). The results revealed that most regions (> 92%) exhibited moderate genetic variance (Fig. 2, Fst < 0.15). Mean Fst values for all the investigated regions was 0.06, while the highest F_ST value was 0.67. To minimize false-positive results, the regions that exhibited high Fst values (in the top 0.1% of the right tail, where Fst > 0.45) were considered as the candidate selective regions. This strategy led to the identification of 68 candidate regions, with a mean Fst value of 0.53, on nine chromosomes (1, 2, 4, 5, 9, 11, 13, 15 and 18). These regions contained 1,140 SNPs and 44 genes (23 lncRNA, 19 mRNA and two miRNA). A circos plot was employed to visualize the genome-wide distribution of SNPs in the candidate regions along with the genes within the regions (Fig. 2). More information about these regions can be found (Supplementary File S4). Following the identification of the SNPs within the selection signature regions, a PCA analysis was performed. The first three PCs (PC1, PC2 and PC3) effectively divided the 44 samples into their respective lines (fat or lean). Collectively, all three PCs accounted for > 65% of the total variance. The analysis revealed that the 22 samples related to lean-lines clustered together due to their positive PC1 values, while 22 samples related to fat-lines formed the other group with negative PC1 values (Fig. 3).

Fig. 2.

Circos plot of the distribution of SNPs within the candidate signature of selection regions. From the outer layer to the inner layer, the outermost layer represents the chromosomes, QTLs associated with fatness traits are presented in the second layer, genes identified in the regions of interest are displayed in the next layer (genes overlapped with the QTLs and miRNAs are highlighted in red and green colors, respectively); the following layer displays the SNPs located in the top 0.1% of regions with the highest genetic differentiation; innermost layer shows all investigated SNPs through the analysis

Fig. 3.

Graphical 3D PCA plot of the first three PCs for the 44 chicken samples analyzed in the present study. Samples from fat-lines clustered together (green color) and samples related to lean-lines formed the other group (red color). Fat- and lean-lines samples of each study are tagged with different color points. The labels include “St,” which refers to the study number (e.g., “St1” for Study 1), while “F” and “L” denote fat-and lean-line, respectively. Additionally, “S” indicates the sample number associated with each study

Functional annotation analysis

Functional association of the candidate mRNAs was further investigated using the EnrichR tool. In total, 50 biological processes terms, 21 KEGG pathways and one Reactome pathway were significantly enriched after FDR correction test (Supplementary File S5). In terms of biological processes, several terms were found to be related to fat metabolism including “Fatty Acid Metabolic Process”, “Lipid Import Into Cell”, “Triglyceride Biosynthetic Process”, “Long-Chain fatty-acyl-CoA Biosynthetic Process”, “Long-Chain Fatty Acid Transport” and “acyl-CoA Metabolic Process”. From the KEGG analysis, several pathways related to fatty acid metabolism were enriched, such as, “Long-Chain Fatty acid-CoA Ligase Activity” and “arachidonate-CoA Ligase Activity”. The only significant pathway identified by the Reactome analysis was “Fatty Acid Metabolism”, and the other pathways were also mainly focused on metabolism of lipids. These findings highlight the importance of the detected regions in the phenotypic divergence between the fat- and lean-lines of chickens (Fig. 4) represent Circos functional annotation plot to show some of the important enriched terms.

Fig. 4.

Circos functional annotation plot of the candidate genes detected in the present study. From the outer layer to the inner layer, the outermost layer represents the GO ids (biological process (blue), Reactome pathways (yellow) and KEGG pathways (green)), adjusted p-values (FDR) of the terms presented in the second layer, heatmap of the p-values are displayed in the next; innermost layer and links connect the genes to the related functional terms

Overlap with known QTLs

To investigate if the identified candidate genes are co-localized with QTLs associated with fattens traits, genomic coordinates of the genes and QTLs were compared. Genome-wide distribution of these QTLs is displayed (Fig. 2). Out of the candidate genes, two genes, NDST4 and YIPF7, co-localized with some of those QTLs (shown in red color in Fig. 2). Both QTLs located on chromosome 4 and associated with abdominal fat weight (19,470) and abdominal fat percentage (14,487).

Discussion

Fat metabolism in chickens is a polygenic trait regulated by many genes, transcription factors, proteins and metabolites [40]. The past decades have witnessed the development of RNA-Seq method as a powerful tool to identify key pathways/genes associated with various traits [18, 24, 41–43]. A further potential advantage of RNA-Seq data is that it allows for variant discovery, as it has been demonstrated in our previous studies [21–24] as well as the works of others [44, 45]. Hence, in the present study, we have taken advantage of RNA-Seq data to identify genomic regions under divergent selection between fat- and lean-lines of chickens. Investigating the genetic footprints left on the genome during intensive selection for growth rate, may provide valuable insights into phenotypic variability caused by different fat deposition patterns. To do this, the information is combined from four independent RNA-Seq studies to identify signatures of selection that is not possible to be detected in the individual studies. It is worth noting that, in all studies, abdominal fat tissue of fat-line chicken samples had been compared with lean-lines. Our choice of an allele frequencies-based approach to identify the signatures of selection, Fst method, is motivated by the fact that this approach is recommended for the traits that are result of completed selection, not an ongoing selection, such as abdominal fat content in chicken [18].

PCA analysis clearly separated all samples of fat-lines from those of lean-lines, which can be attributed to the genetic differentiation between them due to growth rate variation as well as abdominal fat content. Furthermore, as it can be seen (Fig. 3), samples belong to the fat- or lean-lines of each study were obviously formed sub-groups and clusterd together. It was expected, because the different chicken breeds were used by each study. In this regard, fat-line samples of study 1 and 3 were gathered closer althogether and way from the other studies, as in both studies a same breed (Arbor-Acres) were used and tend to get more relative to each other. This finding indicated that the genetic divergence between fat- and lean-lines is consistent across studies, reinforcing the robustness of the identified selection signals.

In total, 68 regions containing 1,140 SNPs and 44 genes detected as signatures of selection. Of the candidate genes, 17 protein-coding genes with gene symbol and two miRNAs were considered for more detailed analysis. Here, to reduce the number of false-positive regions, a very conservative threshold (regions with extreme Fst values, top 0.1%) was used, which is lower than in other studies [38]. Therefore, it is likely that some functional selection signatures be missed due to such stringent criteria. However, this approach favors specificity over sensitivity and increases confidence in the identified regions, as they are less likely to result from neutral selection or technical issues. On the other hand, it is important to note that the breed diversity of the samples of both lines enabled us to identify the most remarkable selection signatures related to fatness traits. Stated differently, different chicken breeds in fat-lines are selected for growth traits and thus the same genomic regions of these breeds were targeted. This might be explained by the well-known hypothesis that organ physiology/biological process is conserved across animals as well as breeds [18, 46]. Therefore, the identified genomic signals are those conserved regions and makes it is unlikely that these loci be irrelevant to fat deposition. This claim was further confirmed by the result of functional annotation analysis of the candidate genes in these regions. These genes were significantly enriched in fatty acid metabolism related pathways and thereby potentially contribute to the biological mechanisms controlling fat deposition processes. Notably, most of the terms associated with fat metabolism, were significantly enriched by some genes, including ACSL3, ACSF2, MOGAT1 and TBXAS1. Additionally, 399 important SNPs were found according to PCA analysis results, as these SNPs explained the most variance based on PC1. In other words, these SNPs were more important to differentiate the samples than the other ones and the regions containing these SNPs were under strong positive selection. These SNPs located in ACSL3, ACSF2, MOGAT1, TBXAS1, GUF1, SIM2 and RGS22 genes. Furthermore, 19 important SNPs were found based on PC2, which were mostly located in ACSF2 gene. Interestingly, most of the genes were directly involved in fat metabolism, such as ACSL3, ACSF2, MOGAT1, TBXAS1 and GUF1.

Adipose tissue development in chickens involves adipocyte formation (adipogenesis) and triglyceride (TG) accumulation in lipid droplets, driven by hyperplasia (proliferation of new adipocytes) and hypertrophy (lipid accumulation in adipocytes). Fat accumulation primarily results from plasma VLDL or dietary fats. Key transcription factors, such as long-chain acyl-CoA synthetases (ACSLs) family, are involved in fatty acid activation and transport. ACSLs family (including ACSL1, ACSL3, ACSL4, ACSL5, and ACSL6) play critical roles in de novo lipid synthesis and β-oxidation [47, 48]. Interestingly, various members of ACSLs family have been reported to be involved in fat deposition in different animals such as ACLS1 in sheep [23] and chicken [49], ACSL3 in cattle [50], ACSL4 in pig [51] and ACSL5 in goat [52]. ACSL3, plays a role in fatty acid activation and encodes key enzymes in fatty acid and TG synthesis pathways [53]. ACSL3 knockdown led to decrease in lipid synthesis [50, 54] and change the accumulation of lipid droplets in intramuscular preadipocytes [55]. Furthermore, ACSL3 has been found in selection signature regions linked to fatness traits in mammals and birds [18, 45, 56, 57]. Our findings in agreement with the previous studies reinforced the importance of ACSL3 as a crucial candidate involved in the regulation of the fat deposition processes in chickens. Therefore, it is reasonable to considers ACSL3 as a high potential candidate associated with abdominal fat content of chickens. Acyl-CoA synthetase family member 2 (ACSF2) is involved in long-chain fatty acid metabolic processes and plays a role in adipocyte differentiation [58]. In a meta-analysis of RNA-Seq datasets, ACSF2 was differentially expressed in adipose tissue of sheep (fat- vs thin-tail breeds) [41]. Transcriptome analysis of abdominal fat tissue in chickens divergently selected for a large difference in growth rate and abdominal fatness, showed that ACSF2 is involved in a network controlling lipid metabolism [59]. In light of this, ACSF2 is considered a key gene involved in fatty acid metabolism in both mammals and chickens [59–61]. The MOGAT family members (MOGAT1, MOGAT2, and MOGAT3) participate in fat metabolism through triglyceride synthesis, dietary fat absorption and fat deposition. MOGAT1 catalyze the synthesis of diacylglycerol from monoacylglycerol and plays an important role in energy storage and lipid homeostasis [62]. Association between MOGAT family members (MOGAT1 and MOGAT3) and growth traits have been demonstrated in Holstein calves [63]. Knockdown of MOGAT1 resulted in glucose intolerance, insulin sensitivity and obesity in mice in response to high fat diet [64]. It has been reported that MOGAT1 promotes the lipogenesis in the high fat diet-induced obese chickens [65]. Adipocyte dysfunction in obesity can result from the downstream prostanoids levels, which is associated with aberrant arachidonic acid metabolism. In this regard, TBXAS1 encodes an enzyme in the arachidonic acid cascade (thromboxane synthase). Role of TBXAS1 in obesity in mice has been documented [66]. This gene is reported as a core biomarker of fatty acid metabolism-related gene to predict the prognosis and immunotherapy response of cutaneous melanoma patients [67]. In a recent study, the critical role of TBXAS1 in amount of lipids in the muscle tissues and the lightness and yellowness qualities of chicken meat has been described [68]. These findings lead us to conclude that these candidate genes can be served as effective markers to improve breeding efforts to mitigate abdominal fat content in chickens.

Among the selection signatures, regions containing NDST4 and YIPF7 genes, may be the most biologically relevant ones, owing to their overlap with the QTLs associated with fatness traits. Of these, NDST4 plays important role in fat metabolism related pathways. As a potential regulator of abdominal fat content, NDST4 modulates fat deposition through pathways related to extracellular matrix remodeling and signaling processes [5]. By performing a genome scan using whole genome sequence data from mouse lines divergently selected for fatness and leanness, Šimon et al. (2024) reported that NDST4 is associated with fat deposition [69]. Notably, this gene was also reported as candidate gene controlling fat deposition in chickens [5]. Yip1 domain family (YIPF) includes nine members that appear to have overlapping functions [70]. In this context, role of YIPF6 in the development of obesity in mice is documented, which highlight the potential role of the other members of this family in fat metabolism [71]. Therefore, based on the evidence from the previous related studies described above, it is possible to suggest both genes, NDST4 and YIPF7, as genetic factors involved in controlling abdominal fat content in chickens. Of these, YIPF7 can be considered as a novel and promising candidate associated with fat deposition.

Furthermore, some of the other genes were also known to be related to fat metabolism, such as GNPDA2, SLC25A48, RBFOX3, FRMD4A and KCNE4. As a member of glucosamine-6-phosphate deaminase subfamily, GNPDA2 is reported to be associated with fat deposition and fatness traits in human [72] and chickens [73, 74]. SLC25A48 encodes a member of slc25 family that transports fatty acids across the inner mitochondrial membrane. It is reported that overexpression of this gene could increase fatty acid transport to the mitochondria for β-oxidation. Moreover, it is suggested that SLC25A48 modulates PPAR signaling pathway, which is closely related to fatty acid metabolism [75, 76]. RBFOX3 belongs to RNA-binding Fox family (RBFOX), an ancient family of splicing factors, which includes three paralogues in mammals (RBFOX1, RBFOX2 and RBFOX3). It is reported that RBFOX2 plays an important role in maintaining cholesterol homeostasis in the liver of mice, by regulating alternative splicing a set of genes involved in lipid homeostasis [77]. The other paralogue of RBFOX3, RBFOX, has been reported as an obesity gene [78]. Of note, RBFOX3 gene has been previously reported to be associated with HDL-cholesterol levels [79]. Overall, these findings seem to suggest a potential role of RBFOX3 in fat deposition in broiler chickens.

Previous studies underscore the critical roles of miRNAs in lipid metabolism and fat deposition. Consistent with these studies, two miRNAs (mir-205b and mir-7443) were found in the selection signatures regions. It is reported that mir-205 targets glycogen synthase kinase 3β to promotes the differentiation of 3T3-L1 preadipocytes [80]. Importantly, it is well documented that mir-205 is regulated by PPARγ, which is mainly involved in fatty acid storage [81]. Adi et al., showed that miR-205 inhibition in 3T3-L1 preadipocytes increases cell proliferation and lipid accumulation after adipogenic induction in a mouse model of polygenic susceptibility for obesity and type 2 diabetes [82]. Interestingly, in a recent study mir-205b proposed as a promising candidate involved in abdominal fat deposition in duck [83]. Therefore, the reported roles of mir-205 in the related pathways to fat deposition along with the presence of this miRNA in a selection signature region, make it a potential candidate involved in abdominal fat deposition in chicken.

A particular result that deserves to be highlighted is that some of the candidate genes (such as ACSL3, ACSF2, MOGAT1, TBXAS1, GNPDA2, NDST4, RBFOX3) have been also found in the selection signatures related to fat deposition in the other species like cattle, sheep, goat and mouse [5, 84–88]. As discussed above, based on the hypothesis of the conservation of organ physiology/biological process across animals [18, 46], it is expected to lead to some common candidate genes among studies aiming to detect selection signatures between fat- and lean breeds/lines. This finding also emphasizes the reliability of our findings. Altogether, our findings indicated that the identified genomic signals as well as genes (including ACSL3, ACSF2, MOGAT1, TBXAS1, NDST4, GUF1, YIPF7, GNPDA2, SLC25A48, RBFOX3, FRMD4A and KCNE4) can be considered as interesting candidates associated with fat deposition, possibly under divergent selection between fat- and lean-lines of chickens. Notably, some of the genes were previously known to be directly related to fat metabolism, such as ACSL3, ACSF2, MOGAT1, TBXAS1, and NDST4, and our findings reinforced their potential roles in abdominal fat deposition. What is worth emphasizing is that some of the other genes, such as GUF1, YIPF7, GNPDA2, SLC25A48, RBFOX3, FRMD4A and KCNE4, are novel candidates that can be further validated for their potential to serve as effective markers in breeding improvement of broiler chickens. Our findings highlighted the potential biological roles of these candidate genes in regulating fat deposition, which might be a key point that contributed in morphological diversity between fat- and lean-lines of chickens. However, additional studies are needed to further elucidate the functional roles of the candidate genes in fat metabolism, especially in abdominal fat content in chickens, possibly through increasing the sample size, sequencing depth and conducting experimental validation approaches.

While we integrated four datasets to increase sample size, as this comes as a limitation to our study, higher sample size, particularly the higher number of samples for each breed pair, may improve the statistical power of the findings as well as the generalizability of the results. We used imputation approach along with stringent filtering processes to improve accuracy of genotyping, however, it is clear that RNA-Seq data only captures SNPs within expressed gene regions, inherently excludes regulatory variants, introns, and other non-coding regions, which we acknowledge as another limitation of this study. Also, selection signature analysis using RNA-Seq data can be considered preliminary, providing valuable initial insights into candidate genes regulating fat deposition. Hence, although direct experimental evidence is lacking, the candidate genes we have proposed warrant further functional validation.

Conclusion

To identify genomic regions putatively under differential selection in fat- vs lean-lines of chickens, a selection signature analysis using four RNA-Seq datasets (based on Fst method) performed. In total, 68 genomic regions were identified under differential selection between the two lines of chickens, which harbored 1,140 SNPs and 44 genes. Various functional evidences were found supporting the idea that the candidate regions/genes contribute to the phenotypic differences and genetic mechanisms underlying abdominal fat content in broiler chickens. Our findings confirm previously reported candidate genes known to be involved in abdominal fat deposition in chickens, such as ACSL3, ACSF2, MOGAT1, TBXAS1, and NDST4. Moreover, the results give novel insights into some promising candidate genes associated with abdominal fat content, including GUF1, YIPF7, GNPDA2, SLC25A48, RBFOX3, FRMD4A, KCNE4 and mir-205b, as their roles in fat metabolism related pathways are documented. By pinpointing genes and pathways linked to abdominal fat content, this study lays the groundwork for narrowing down the list of candidate genes required for exploring the molecular mechanisms behind the abdominal fat deposition in chickens, which also provide a theoretical basis for optimizing selective breeding strategies.

Abbreviations

RNA-Seq

RNA Sequencing

GTF:

Gene transfer format

GWAS

Genome-wide association study

miRNA

MicroRNA

mRNA

Messenger RNA

PCA

Principal component analysis

PCR

Polymerase chain reaction

QTL

Quantitative trait locus

SNP

Single nucleotide polymorphism

WGS

Whole genome sequencing

Authors’ contributions

HA conducted project administration, developed methodology, contributed to visualization, performed investigation, and drafted the manuscript. MRB supervised the study, conceptualized the research, developed methodology, contributed to visualization, drafted the manuscript, and performed formal analysis. HM carried out investigation, contributed to methodology and visualization, and reviewed and edited the manuscript. BG-G assisted in methodology and reviewed and the manuscript. All authors read and approved the final version of the manuscript.

Funding

No application.

Data availability

The data that support the findings of this study are available from the tables and supplementary materials. The raw data of sequencing are available in GEO database in NCBI (PRJNA657377, PRJNA184019, PRJNA354990 and PRJNA248570).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.