Identification of selection signatures involved in performance traits in a paternal broiler line

Background Natural and artificial selection leads to changes in certain regions of the genome resulting in selection signatures that can reveal genes associated with the selected traits. Selection signatures may be identified using different methodologies, of which some are based on detecting contiguous sequences of homozygous identical-by-descent haplotypes, called runs of homozygosity (ROH), or estimating fixation index (FST) of genomic windows that indicates genetic differentiation. This study aimed to identify selection signatures in a paternal broiler TT line at generations 7th and 16th of selection and to investigate the genes annotated in these regions as well as the biological pathways involved. For such purpose, ROH and FST-based analysis were performed using whole genome sequence of twenty-eight chickens from two different generations. Results ROH analysis identified homozygous regions of short and moderate size. Analysis of ROH patterns revealed regions commonly shared among animals and changes in ROH abundance and size between the two generations. Results also suggest that whole genome sequencing (WGS) outperforms SNPchip data avoiding overestimation of ROH size and underestimation of ROH number; however, sequencing costs can limited the number of animals analyzed. FST-based analysis revealed genetic differentiation in several genomic windows. Annotation of the consensus regions of ROH and FST windows revealed new and previously identified genes associated with traits of economic interest, such as APOB, IGF1, IGFBP2, POMC, PPARG, and ZNF423. Over-representation analysis of the genes resulted in biological terms of skeletal muscle, matrilin proteins, adipose tissue, hyperglycemia, diabetes, Salmonella infections and tyrosine. Conclusions Identification of ROH and FST-based analyses revealed selection signatures in TT line and genes that have important role in traits of economic interest. Changes in the genome of the chickens were observed between the 7th and 16th generations showing that ancient and recent selection in TT line may have acted over genomic regions affecting diseases and performance traits. Electronic supplementary material The online version of this article (10.1186/s12864-019-5811-1) contains supplementary material, which is available to authorized users.


Background
Artificial selection of animals lead to changes on particular genomic regions that affect traits of economic interest, as well as traits involved in adaptation to climatic and stress conditions, immune response, and disease resistance [1]. Thus, selection signature regions are printed along the genome as a result of selection pressure. Detecting selection signatures is important for a better understanding of population history and genetic mechanisms affecting phenotypic differentiation in humans, livestock and wild animals [2]. Understanding how selection acts on livestock populations may also benefit breeding programs in order to improve traits of economic interest in these animals, such as chicken breeds which have been intensively selected for fast growth and muscle development [3]. Detection approaches rely on scanning the genome for regions of homozygosity, as well as on estimating allele or haplotype frequency differences between populations or generations within a population. There are several statistical methods for these analyses, such as extended haplotype homozygosity (EHH) [4], integrated haplotype score (iHS) [5], runs of homozygosity (ROH) [2], and F ST statistics [6].
Runs of homozygosity are regions in the genome containing contiguous homozygous genotypes identical by descent (IBD), i.e. regions where the pairs of alleles are most likely inherited from a common ancestor [2]. Recent studies used this approach to better understand human diseases [7][8][9], human ancestry [10], and population structure and traits of interest in livestock species, such as cattle [11][12][13][14], swine [15], poultry [16,17], and sheep [18]. The fixation index (F ST ), first defined by Wright [19], is a measure that exploits differences in allele frequencies to infer the genetic differentiation between populations or generations [20]. A certain locus under selection pressure changes its frequency over the generations. Thus, high values of F ST indicate candidate selection signatures due to differences in locus frequency among populations or across generations. Previous studies have reported important selection signatures in Virginia [21] and Brazilian broiler and layer chicken lines [22] using this method.
Embrapa Swine and Poultry, a Brazilian National Research Center, has been raising experimental chicken populations under selection since the 1970's. One of these lines is the paternal broiler line called TT, which has been under multi-trait selection since 1992 [23]. Identification of selection signatures in chicken lines can help understand which regions underwent selection pressure over time and how their biological mechanisms act to express the traits of interest, such as muscle growth and fat deposition. In this sense, we aimed to investigate selection signatures in TT broiler line by detecting ROH in the 7th and 16th generations, raised in the years of 1998 and 2007, respectively, and estimating F ST statistic between these two generations. The identification of those regions will provide better understanding of artificial selection effects on broiler lines, and may point out candidate genes and biological mechanisms underlying performance traits.

Ethics statement
This study followed experimental protocols pertinent to animal experimentation with the approval of the Embrapa Swine and Poultry Ethics Committee on Animal Utilization (CEUA) in Concordia, Santa Catarina State, Brazil, on resolution number 011/2010. It followed the rules of National Council of Animal Experimentation Control (CONCEA) in accordance with international guidelines to guarantee animal welfare.

Chicken population
Chickens used in this study were from a broiler line developed by the Embrapa Swine and Poultry National Research Center. This line, called TT, was originated from Cornish and White Plymouth Rock breeds, that has been under a multi-trait selection process focused on body weight, feed conversion, cut yields, breast weight, abdominal fat, and other traits, since 1992 [17,23,24]. The chickens were raised in open sided poultry houses, receiving commercial broiler diet and water ad libitum [17,24]. Chickens were euthanized by cervical dislocation at 42 days of age. Two groups of animals from this line were analyzed, 14 chickens (half male and half female) from the seventh-generation (7th) raised in the year of 1998 and 14 male chickens from the sixteenth-generation (16th) raised in 2007.
The performance of birds from the 7th generation, as hatched average live weight at 35 days of age was 2272 g; the breast area in the live bird was 96,1 cm 2 and the individual feed conversion rate (FCR) for males, from 36 to 43 days of age was 2268 g. The performance of birds from 16th, as hatched average live weight at 42 days of age was 2457 g; the breast area was 112 cm 2 and the individual FCR from 43 to 49 days of age was 2798 g [17]. Note that the age of selection for BW and BA has changed from 35 to 42 days of age in the described period, as well as the FCR evaluation period, which has changed from 36 to 43 days in 1998 to 43-49 days of age in 2007.

Sequencing and quality control
Whole genome sequencing (WGS) data of 28 chickens were used in this study. Animals were individually sequenced to a minimum coverage of 11.4x using the HiSeq2500 (Illumina) platform, and the alignment of reads was done against the chicken genome assembly (Gallus_ gallus-5.0, UCSC) chicken reference genome using Bow-tie2 [25]. Detailed information about library preparation, sequencing, quality control of reads, alignment and SNP and INDEL identification are fully described in Boschiero et al. [22] and Moreira et al. [26]. Variants identified in sexual, mitochondrial, random or unplaced chromosomes were removed from our analysis.

Principal component analysis
Genetic relationship between the 28 animals was assessed with a principal component analysis (PCA) using the SNP dataset (n = 9,914,904). The analysis was performed using the SNPRelate package of Bioconductor by means of an in-house script in R.

Identification of runs of homozygosity
The identification of ROH was chosen to obtain information about selection signatures and how they are shared between animals in both generations. Analyses were performed using PLINK v1.9 software [27,28], which uses a sliding window approach: a window, with a minimum size, slides across the genome, calling a segment if it is in accordance with the parameters established and the threshold of calculated proportion of homozygous windows overlapping each SNP in that segment. The parameters used in the analysis were set based on Ceballos et al. [29] and they are listed in Table 1.
Dataset of the 28 animals comprised 9,914,904 SNP, and all INDEL were excluded. The parameter -homozyggroup was also used to obtain information of the overlapping ROH (pools), i.e., ROH that appeared in at least two animals in the same region of the genome. The output plink.hom.overlap shows each ROH of each animal overlapping and their respective union (uROH) and consensus sequences (cROH), besides their genome position, size and number of SNP. The consensus ROH (cROH, i.e. a consensus segment of ROH that appeared in a common region in at least two animals) of the pools were used for annotation and enrichment analysis, to avoid randomly assigned ROH and to represent what changed and what is conserved between the animals [30,31]. In addition, we used an in-house script in R to check the overlap between the regions of all cROH and the ROH previously identified in the TT Reference population (originated from TT broiler line) by Marchesi et al. [17], given the positional coordinates (chromosome, start and end) of these regions and considering at least one overlapped base pair.

Genomic inbreeding coefficients
Individual genomic inbreeding coefficients were calculated based on ROH data (F ROH ), as defined by McQuillan et al. [32], to know if there was a difference of inbreeding between the 7th and 16th generations. F ROH was calculated as: where L ROH is the total size of ROH in the genome and L aut is the total size of autosomal genome covered by SNP of an individual (933.071 Mb, Gallus_gallus-5.0 chicken reference genome -UCSC).

F ST analysis
This method was applied to compare the two generations, i.e. to identify selection signatures by estimating the differences in allele frequency between the 7th and 16th generations. The fixation index was calculated according to Weir  signatures. The software BEDTools [35] was used to check if there were equivalent regions identified in both datasets.

Functional analysis
Functional analysis was performed to identify genes annotated within the candidate selection signature regions identified and, consequently, the biological mechanisms that may be involved with traits of adaptation and performance. Such information was obtained assessing the position (start and end coordinates) of the candidate selection signatures (cROH and FST windows ≥0.3) in the chicken genome available at BioMart Ensembl genome browser platform (Ensembl Genes release 94, Gallus_gallus-5.0 assembly) [36]. We also assessed the genes annotated in the candidate selection signatures under different perspectives in order to understand the effects of selection on TT line in different periods. First, we investigated changes that occurred between the 7th and 16th generations: (i) cROH of regions shared exclusively among animals of the 7th; (ii) cROH of regions shared exclusively among animals of the 16th; (iii) cROH of regions that were shared among at least four animals of the 16th more than animals of 7th; (iv) FST SNP windows (≥ 0.3); and (v) FST INDEL windows (≥ 0.3). In addition, we looked for genes annotated in (vi) cROH of regions shared with 12 or more animals (among the 28), to identify regions probably related to chicken domestication or even, specialization into broilers.
Functional enrichment was performed using MeSH Enrichment and Semantic Analysis, Bioconductor's package [37,38], in R software [39] to investigate if there was overrepresentation of any biological processes and components. For such purpose, datasets of genes annotated in the specific candidate selection signatures previously mentioned (i-vi) were analyzed separately. The p-value was adjusted using the Benjamin-Hochberg false discovery rate (FDR) method [40].

Overlapping selection signatures with QTL
In order to confirm the role of the selection signatures detected herein in the regulation of important phenotypes in chickens, we investigated the overlap with QTL associated with traits of economic interest. The analysis was performed using an in-house script in R to overlap the regions of all cROH, F ST SNP and INDEL windows against the QTLs available at the Chicken QTL database [41], given the positions (chromosome, start and end) of these regions and considering at least one overlapped base pair. Particularly, we also analyzed if there was overlap of candidate selection signatures with QTL associated with fat deposition previously identified by Moreira et al. [42] in the TT Reference population, originated from an expansion of TT line in 2007 for genomic studies purpose [17].

Principal component analysis
Principal component analysis using genomic data revealed a cluster separation between animals of the 7th and the 16th generations (Fig. 1). The distinct clustering demonstrated that genome data successfully separate these animals accordingly to their generation.

Runs of homozygosity
Analysis of the whole genome sequence data of the 28 animals (14 of the 7th and 14 of the 16th generation) with PLINK's sliding window approach identified 5721 ROH (1944 in the 7th and 3777 in the 16th generation animals) (Additional file 1). The average number of segments per animal was lower in the 7th (138.9 ROH/ animal) than in the 16th generation (269.8 ROH/animal) ( Table 2). The ROH presented small and moderate sizes, ranging between 300 Kb and 4.9 Mb, and most of them had sizes smaller than 1.0 Mb in both generations (1821 ROH in the 7th and 3120 in the 16th generation animals). A change in the distribution of ROH sizes was also observed between both generations (Fig. 2). The proportion of ROH smaller than 1.0 Mb decreased (93.7% in the 7th to 82.6% in 16th generation) while the proportion of ROH with sizes between 1.0 and 2.0 Mb increased (6.1% in the 7th to 14.7% in the 16 th generation) as well as ROH bigger than 2.0 Mb (0.3% in the 7th to 2.7% in 16th generation). The average total size of ROH per animal was 73.2 Mb in the 7th and 188.6 Mb in the 16th generation (Table 2).
ROH were identified in all chromosomes, except on GGA16 and GGA30-32 (Table 3). Regions commonly shared among the animals become more frequent in the 16th generation. We observed a greater number of ROH pools shared among animals from 16th generation than animals from 7th generation ( Fig. 4-a, b and c).

Genomic inbreeding coefficients
Individual genomic inbreeding coefficients based on ROH (F ROH ) were calculated for both generations (Table  2). Mean, maximum and minimum individual F ROH for animals of the 7th generation were 0.0784, 0.1340, and 0.0215, respectively, with a coefficient of variation (CV) of 52.8%. For animals of the 16th, the mean, maximum and minimum individual F ROH were 0.2021, 0.2213, and 0.1761 (CV = 7.4%). Most of these windows were in the macrochromosomes (Table 3) and approximately 87% of the INDEL windows overlapped with SNP windows (Fig. 4 c and d).

Genes in consensus selection signature regions
Annotation analysis using the Ensembl genome browser revealed 5681 genes annotated in the 1941 cROH pools shared among at least two of the 28 animals (Additional file 7). Annotation of F ST windows (Fst ≥0.3) identified 56 and 60 genes for SNP and INDEL datasets, respectively (Additional files 8 and 9). Since a great part of SNP windows overlapped with INDEL windows, 37 of these genes were common for both datasets (Table 4). In addition, about 46.1% of SNP F ST windows and 37.7% of INDEL F ST windows overlapped with cROH. Thus, we found 34 genes annotated in cROH in common with genes annotated in FST (SNP and/or INDEL) windows (Table 4).
Based on Biomart Ensembl database, some of the genes commonly annotated in Fst and cROH regions, plays a role in biological processes involved in traits of economic interest in chicken or in other model animals; all the biological processes related to the genes mentioned in Table  4, are available at Additional file 10. There were genes involved in lipid metabolic processes, glucose metabolism and homeostasis and adipose tissue development. Other genes were described to be involved in muscle cell differentiation, muscle tissue development, and constituents of skeletal muscle. Moreover, there were a group of genes related to different types of behavior, such as grooming, locomotion, fear response, feeding behavior, aggressiveness, and social, exploration and maternal behaviors. Genes involved in the immune humoral system, differentiation, proliferation, homeostasis and chemotaxis of B cells, and regulation of cytokines production were also annotated in regions of the candidate selection signatures. In order to complement the Fst analysis we identified genes in cROH regions that were either exclusive between the 7th and 16th generations. There were 71 genes annotated in the regions that were in homozygosity only in the 7th generation and 1881 genes annotated in regions of cROH shared only among animals of the 16th generation. We also identified genes in cROH regions that had changed between the 7th and 16th generations. For that, we considered regions that either increased or decreased by at least four animals in cROH regions between generations. There were 1318 genes annotated on these regions. For example, the gene IGF-I (GGA1 55,335,204 -55,383, 631) was annotated in a cROH region (GGA1 55,149,208 -55,359,089) shared between seven animals of the 7th and 13 animals of the 16th. Additional file 11 presents genes annotated in these regions and that were previously associated with traits of economic interest in chickens..   These different gene lists were used to perform MeSH overrepresentation analysis, with the purpose of having an integrated knowledge of biological processes may be involved in the selection of TT line. MeSH analysis indicates if there is an overrepresentation of a particular group of genes in a biological category, such as anatomy, diseases or phenomena and processes. In this sense, overrepresentation analysis resulted in eight different biological terms: 'matrilin proteins' , 'skeletal muscle' , 'Salmonella infections in animals' , 'adipose tissue' , 'cystatins' , 'tyrosine' , 'pregnancy in diabetics' , and 'hyperglycemia'. Table 5 presents each one of these terms and their respective gene counts and p-values.

Selection signatures in overlap with QTL in the TT line
To identify selection signatures associated with quantitative traits, we investigated the overlap between the regions of cROH, F ST SNP and INDEL windows (≥ 0.3) with QTL regions previously associated with traits of economic interest in chickens available at the Chicken QTL database (release 37). About 72.8% of the 1941 cROH overlapped with QTL regions (n = 2617). There were also overlap of 60.1% of the 178 F ST SNP windows and 68.2% of the 154 F ST INDEL windows with QTL regions (n = 107 and 105, respectively). These QTL regions were associated with 143 different traits (Additional file 12) of which some are very important for broilers' breeding program goals such as feed conversion rate, feed intake, average daily gain, body weight, breast muscle weight, and others (Figs. 7, 8 and 9).
We also identified cROH regions that overlapped with 10 QTL previously mapped for fatness traits on GGA5, GGA9, GGA10, GGA13, GGA15, and GGA27 in the same population utilized herein (TT Reference Population) ( Table 6) [42]. There was no overlap between the F ST windows (≥ 0.3) and these QTLs mapped for fatness.

Runs of homozygosity
ROH studies have addressed human evolution and diseases, conservation and evolution of wild species, and genomic features of livestock animals [2,7,13,17]. The investigation of ROH as selection signatures using different generations of the TT line provided two new insights. First, the identification of important selection signatures that may refer to periods preceding the generations under study. Second, it allowed the comparison of how these selection signatures were shared among the individuals and how they have changed over the generations. Since shared ROH is an indication of regions under selection [18], the consensus regions of ROH (cROH) were used in this study to understand the possible biological consequences of selection in this broiler line.
The ROH identified in animals of TT line presented small to moderate sizes, ranging from 300 Kb to 4.9 Mb. Short ROH are most probable to be IBD genomic regions inherited from ancient ancestors indicating long term selection [43]. Over the generations, IBD segments tends to break down due to recombination events by repeated meiosis. Thus, ROH size is associated with the degree of shared parental ancestry and for how long it was passed across generations [32] and, as expected, short ROH regions shared between animals of the 7th and 16th generations encompassed genes associated with traits of interest (Additional file 10). It is important to mention that not all short ROH are IBD and a proportion of them may be identical-by-state (IBS) due to genetic drift, as well as population bottlenecks, and therefore some authors recommend caution in attributing these regions as candidate selection signatures [18,44,45]. The minimum size of 300 Kb was set in our analysis to detect ROH, considering that the use of WGS data calls ROH with smaller sizes compared to SNPchip data. A limitation in our study is that we were not able to determine which proportion of ROH is attributed to genetic drift and may lead to false positives. However, strategies were adopted to improve the chances to detect true selection signatures associated with the selection program. They were (i) using ROH regions in common with at least two animals, (ii) overlapping these regions with the Chicken QTL database, and (iii) performing MeSH overrepresentation analysis. These combined strategies reduce the chance to detect candidate selection signatures due to genetic drift. In a study with offspring from animals of the 16th generation, Marchesi et al. [17] identified ROH in 1279 chickens using a high-density 600 K genotyping array data. They adopted the same parameters used in the present study, except for the minimum size of ROH and number of heterozygous SNP allowed. As expected, regions of ROH were commonly shared between both populations. More than 98% of the cROH identified herein in animals of the 16th generations, overlapped with ROH identified in the study of Marchesi et al. [17], corroborating our findings (Additional file 4). However, a higher number of ROH per animal and ROH with smaller sizes were identified in the 28 animals compared to those of Marchesi et al. [17]. We suggest that the reasons for identifying shorter ROH relies on the higher resolution of WGS data in comparison with SNPchip data. This difference in ROH calling was also observed in another study with feline that used both types of dataset for the same individuals [46].
It is also important to highlight that low coverage WGS data may present higher error rate of variant calling in comparison with SNPchip data, and this may lead to inaccuracy of ROH calling [29]. Thus, we followed parameters based on Ceballos et al. [29], which demonstrated equivalent results to SNPchip data's results, when dealing with low coverage WGS. Moreover, in order to extend the chances of detecting accurate ROH, we opted to investigate consensus regions of ROH, i.e. regions of ROH in common with at least two animals, that might indicate regions under selection [18,30]. All these observations corroborate with our suggestion that the smaller size and higher number of ROH possibly relies on the better resolution of WGS.
In our study, an increase in ROH abundance was observed, i.e. between the 7th and 16th generations there was an increase on the average number of ROH segments per animal and in the average size of segment per animal ( Fig. 2 and Table 2). Investigation of the history of the breeding program indicate that these differences may have occurred due to a bottleneck effect. Marchesi et al. [17] estimated the N e in TT line backing to 200 generations ago and reported a decay in N e , especially in the last five generations, ranging from 157 to 113 chickens (N e of TT Reference Population). Thus, when a population size is reduced, the average of heterozygosity in a certain locus is expected to decline, depending on the N e [47,48]. The occurrence of a bottleneck effect is supported by the observed increase in the mean genomic inbreeding coefficient from the 7th to the 16th generation (7th F ROH = 0.078 and 16th F ROH = 0.202). It  is worth to mention that, even if mating between close related individuals is avoided some level of inbreeding is unavoidable, because TT line is a closed population [17].
Furthermore, the percentage of animals sharing a ROH region increased from the 7th to the 16th generation (Fig. 4). Mastrangelo et al. [18] reported similar observations of an increase in the abundance of ROH in a sheep breed it and suggested that a decrease in the effective population size (N e ) had occurred resulting in recent and historical autozygosity events. Thus, we suggest that the increase in homozygous regions across generations in TT line is consequence of selection pressure over genomic regions that are important to the breeding program's goals jointly to reduction on the N e and inbreeding. In fact, genes associated with traits of economic interest, such as the APOB, POMC, PPARG and other genes (Additional file 10), were annotated in regions shared with more animals of the 16th than with 7th generation, supporting that the regions containing these genes were under selection pressure in the respective period.

F ST windows
An alternative approach applied in this study for identification of selection signatures was the genetic differentiation method based on allele frequency differences called F ST statistics. Previous studies have used this method for detecting selection signature in livestock species, such as broilers [14,21,22,49,50]. Here we compared two groups of animals of TT broiler line, 10 generations distant from each other. The parameters were the same used by Boschiero et al. [22], in the comparison of the TT line against a layer line. The authors states that windows of 20 Kb allow a finer resolution of the regions in addition to windows with sufficient number of markers, considering that the amount of variants in a window is essential for increasing the power of the analysis [22,51]. This intent was achieved in our results since we obtained an average of 216 SNP/window and 18 INDEL/window, which were similar to the results obtained by Boschiero et al. [22] with averages of 268 SNP/ window and 26 INDEL/window. Furthermore, there was a considerable number of windows in common between SNP and INDEL datasets with 87% of overlapping, a fact also observed by Boschiero et al. [22].
Estimates of F ST range from zero, meaning no genetic difference between the subpopulations, up to 1.0, meaning complete genetic differentiation [52]. Although there is not a determined threshold to capture regions that indicate genetic differentiation as a candidate selection signature, some authors use a threshold for the top 0.1% values of F ST [21,22]. Here we established a threshold value of 0.3 in order to obtain regions that might be in a differentiation process in the TT line. Only 0.002% of the windows had F ST values above this threshold, for SNP and INDEL datasets, and the highest estimated values were 0.598 and 0.555 for SNP and INDEL datasets, respectively. In addition, as discussed by Boschiero et al. [22], combining strategies to detect selection signatures minimizes the occurrence of false positives.

Selection signatures of broiler performance and adaptation
Besides identifying regions in the chicken genome under selection pressure, knowing the genes annotated in these regions and how they biologically act is essential for understanding how the selection signatures contributed to the current phenotype of the evaluated animals. Since TT broiler line is under multi-trait selection since 1992 aiming to improve body weight, feed conversion, cut yields, breast weight, viability, fertility, and hatchability and to reduce abdominal fat [17,23], it is expected that genes influencing the performance of these traits are under selection pressure.
Therefore, investigating which genes were annotated in the candidate selection signatures regions identified in both F ST and ROH analysis helps to understand the biological mechanisms that affected the construction and   TOX thymocyte selection associated high mobility evolution of the phenotype of TT line. In this sense, genes involved with traits of economic interest were identified in these regions. The genes IGFB2, TGFB2, HOXD9, HOXD10, POMC SPP1, SPP2, and IGF1 were some of the genes annotated in the candidate selection signatures of TT line and that were previously found in other selection signatures and associated with traits such as growth, body weight and composition, abdominal fat, organogenesis and feed intake and consumption [22,50,[53][54][55][56][57][58][59][60][61][62][63][64][65][66]. Furthermore, we identified a group of genes annotated in the selection signatures that are involved with structural constituents, cell differentiation, and development of muscle tissue: ACTC1, AKAP6, ATP2A2, KCNMA1, MYO1B, MYO1C, MYO1E, MYO1F, MYO6, MYO7A, MYO10, MYO16, TPM4, VCL, and V1PR1 [22,67]. Selection signatures identified in our analysis also indicate regions involved in lipid metabolism and adipose tissue development, encompassing the ADCY2, AKAP6, APOB, ATPR2, IGFBP2, PLA2R1, PPARG, SCARB1 and ZNF423 genes [22,[68][69][70][71][72][73][74][75][76][77]. Chickens raised in production systems are under several stressful conditions that can affect performance and   [82]. Adipose tissue was in overrepresentation among the genes of these regions, what is expected since selection for growth in broilers, could lead to elevated fat deposition [83]. In addition, overlaps of cROH regions with QTL associated with fat deposition in TT Reference population support that regions affecting these traits were indirectly selected across generations in TT line. Hyperglycemia was also overrepresented and it is possibly a consequence of fat deposition in these animals. The excess of adipose tissue in chickens may lead to a condition similar to the early stage of type 2 diabetes in humans, manifesting hyperglycemia and exogenous insulin resistance [84,85]. Another overrepresentation was 'pregnancy in diabetics' , and, as it is known, pregnancy is not a biological mechanism of birds. However, Nadaf et al. [83] discuss that some QTL associated with chicken fatness have genes playing a role in obesity and diabetes in humans, and since MeSH is a tool that comprises animals in general, this association may have be done due to genes with similar functions. Cystatins was among genes annotated in regions that underwent allele frequency changes during the 7th and 16th generations. Cystatin is a superfamily of reversible competitive inhibitors of cysteine proteases such as calpains, cathepsins, and ficins, and the cystatin system have important roles in protein turnover, antigen presentation and disease immunity [86,87]. As well, tyrosine was overrepresented among the genes of F ST SNP windows, which is considered a nonessential amino acid in animals [88].
The ROH analysis also provided information about regions that may indicate selection in a period that precedes the studied generations. In these regions there were overrepresentation of genes involved in skeletal muscle and the matrilin proteins, both important for the growth of chickens. Matrilin is a four-member family of proteins composing extracellular matrix of some tissues as cartilage, a connective tissue. They bind to collagencontaining fiber and other matrix constituents and can form oligomers [89].
Furthermore, the abundant amount of overlaps between cROH and F ST windows and QTL regions associated with traits of economic interest in broilers, such as feed conversion, feed intake, growth, and abdominal fat, enforces the results of candidate selection signatures involved in performance traits. This fact together corroborates that TT line have been selected for growth and muscle deposition for a long period with a possible consequence of increased fat deposition, and for some traits such as proteic turnover and metabolism of tyrosine.