Genetics of the phenotypic evolution in sheep: a molecular look at diversity-driving genes

After domestication, the evolution of phenotypically-varied sheep breeds has generated rich biodiversity. This wide phenotypic variation arises as a result of hidden genomic changes that range from a single nucleotide to several thousands of nucleotides. Thus, it is of interest and significance to reveal and understand the genomic changes underlying the phenotypic variation of sheep breeds in order to drive selection towards economically important traits. Various traits contribute to the emergence of variation in sheep phenotypic characteristics, including coat color, horns, tail, wool, ears, udder, vertebrae, among others. The genes that determine most of these phenotypic traits have been investigated, which has generated knowledge regarding the genetic determinism of several agriculturally-relevant traits in sheep. In this review, we discuss the genomic knowledge that has emerged in the past few decades regarding the phenotypic traits in sheep, and our ultimate aim is to encourage its practical application in sheep breeding. In addition, in order to expand the current understanding of the sheep genome, we shed light on research gaps that require further investigation. Although significant research efforts have been conducted in the past few decades, several aspects of the sheep genome remain unexplored. For the full utilization of the current knowledge of the sheep genome, a wide practical application is still required in order to boost sheep productive performance and contribute to the generation of improved sheep breeds. The accumulated knowledge on the sheep genome will help advance and strengthen sheep breeding programs to face future challenges in the sector, such as climate change, global human population growth, and the increasing demand for products of animal origin.

predators and as sexual weaponry. In modern breeding practices, the hornless (or polled) phenotype has become increasingly desirable as a way to avoid the need for dehorning [5], which is usually performed to protect animals and producers from accidental injuries. In addition, coat color serves in the adaptation to environmental conditions and avoiding predators. Currently, selection for specific sheep coat color patterns is performed based on the production system, e.g., white color is desirable in several wool-producing sheep breeds [6]. Combined with other phenotypic traits, coat color is considered a determinant of sheep breeds. Several phenotypic traits are significantly associated with production and animal welfare parameters. Hence, understanding the genetic background of these phenotypic traits is of great importance for a rapid and direct selection and improvement of sheep breeds. The sequencing of the sheep genome and the elucidation of its content and variations have offered an unprecedented opportunity for a better understanding of the underlying genetic traits controlling sheep phenotype [7][8][9][10].
Recent advances in sheep genomics have helped in identifying genes that are implicated in determining major phenotypic traits in sheep. For instance, it has  8) can be found in China, and over 21 improved and crossed breeds were generated by breeding local and imported breeds to yield multipurpose breeds for improved meat and wool production, as well as other productive traits. Images used herein were adapted with permission from Animal Genetic Resources in China: Sheep and Goats [217] been reported that the PDGFD, BMP2, HOXB13, and TBXT genes regulate sheep tail phenotype [11][12][13][14], RXFP2 and HOXD1 regulate horn and multi-horn traits [5,15], whereas ASIP, MC1R, and TYRP1 regulate coat color patterns [16][17][18]. However, the causative genomic variations within or around some of these genes are currently unknown. Given the importance of the genes controlling sheep phenotypic traits that are linked with adaption, production, and welfare traits, in this review, we collectively discuss the genomic history of these traits and describe genes and variants that are currently known for these traits and highlight traits which need further investigation to elucidate their genetic determinism. We discuss potential genes and their variants that control main phenotypic traits in sheep, including tail, horns, coat color, bone color, fat color, wool length, fleece type, ear size, number of nipples, number of thoracic and lumbar vertebrae, body height, and stature. We anticipate that exploring genes that control significant phenotypic traits in sheep will foment selective breeding programs and genetic-based strategies for the rapid improvement of productivity and welfare of sheep populations.

Main text
The long-standing conundrum of sheep tails: a genomic perspective Sheep tail phenotypic patterns vary among different sheep breeds in terms of length, directionality, and fat accumulation level [19] (Fig. 2). The genetic basis of sheep tail phenotypic variations has been elucidated based on genomic studies. Moradi et al. [4] first applied a genomic scan approach to identify genomic loci that underlie the phenotypic differences between fat-and thin-tailed Iranian sheep breeds, and revealed seven genomic regions that overlap with quantitative trait loci (QTL) shown to affect fat and carcass yield. In addition, three genomic regions located on Ovis aries chromosome 5 (OAR5), OAR7, and OARX were suggested to be associated with the sheep tail phenotype [4]. A subsequent genome-wide analysis by Wei et al. [11] using ten representatives of Chinese sheep breeds revealed the association of PPP1CC and PDGFD with significant differences between fat-and thin-tailed animals. Specifically, PDGFD was positively selected in fat-tailed sheep breeds. In another study, BMP2 and VRTN were identified as the genes that are most probably involved in the determination of sheep tail phenotype [12].
Subsequently, Yuan et al. [20] proposed that the HOXA11, BMP2, PPP1CC, SP3, SP9, WDR92, PROKR1, and ETAA1 genes potentially play an important role in determining the fat-tailed phenotype. An integrated analysis of genome-wide association study (GWAS) and X-chromosome-wide association analysis (XWAS) in small-tailed and large-tailed Han sheep revealed a set of genes (namely, CREB1, STEAP4, CTBP1, and NRIP1) that could be potentially associated with lipid metabolism or adipocyte development [21]. A candidate genomic region, i.e.,  Mb, was also identified, which was shown to contain significant single nucleotide polymorphisms (SNPs) that could be associated with tail phenotype [21]. In addition, by analyzing genotypic data of 11 Ethiopian indigenous sheep breeds and two populations of thin-tailed sheep from Sudan, a set of potential genes associated with sheep tail phenotype was proposed [22]. This includes putative candidate regions that spanned genes affecting growth traits and fat deposition (NPR2, HINT2, SPAG8, and INSR), tail formation (ALX4, HOXB13, and BMP4), embryonic development of tendons, bones, and cartilages (EYA2 and SULF2), lipogenesis and intracellular long-chain fatty acids transport (FABP3) [22]. In another study, genomic data of three Iranian sheep breeds (i.e., Baluchi, Lori-Bakhtiari, and Zel) were also analyzed, which revealed that the RPS6KA3, HOXB13, HOXB9, ACAN, and CHRDL1 genes as potentially influencing the skeletal system and tail traits when comparing fat-tailed Baluchi sheep and thin-tailed Zel sheep; conversely, NPR2 and EN1 were suggested as potentially influencing energy metabolism, skeletal system, and tail traits when comparing fat-tailed Lori-Bakhtiari sheep and thin-tailed Zel sheep [23].
Pan et al. [24] analyzed the whole-genome sequences of 89 sheep and identified two significant signals on OAR13 and OAR15. The sweep signal on OAR13 was consistent with previous studies [11,12]. In particular, Moioli et al. [12] suggested that the BMP2 gene could be potentially causative for the sheep fat tail formation, whereas Wei et al. [11] suggested the PPP1CC gene, which is a paralogue of the original PPP1CC gene on OAR13 (NCBI gene identifier: LOC101117953). The other signal on OAR15 contained the PDGFD gene, which was previously highlighted by Wei et al. [11]. The newly characterized LOC101117953 has been suggested to originate from a retro-copy (r-PPP1CC) of the PPP1CC gene via clear intron losses [24]. Since r-PPP1CC is a promoterless non-functional pseudo-gene, it was assumed to be less likely the causative gene. However, the locus can potentially serve as a regulatory element, which changes the expressional patterns of the distant BMP2 gene [24].
Mastrangelo et al. [25] compared the genomic data of two fat-and thin-tailed sheep breeds, and again revealed previously reported genes, i.e., VRTN and BMP2, with a potential role in determining variations in the number of vertebrae and tail formation, respectively, and PDG-FRA and PDGFD, which seem to be involved in preadipocyte differentiation. Moreover, Mastrangelo et al. [26] also reported a set of genes, namely, ANAPC1, MSRB3, CXXC5, PSD2, SLIT2, EPHA5, CHP1, OIP5, PCDH9, CDS2, BMP2, and OAS2, which could be associated with fat deposition in sheep breeds from Africa and Eurasia.
Recently, a number of large-scale genomic studies have been performed to elucidate the phenotypic differences in sheep tail formation. Li et al. [27] investigated the genetic background of the configuration of sheep tails by performing separate pairwise-population selection tests and analyzing genotype patterns located in the promoter region; in this study, the PDGFD gene was consistently selected by multiple comparisons and its expression varied between fat-and thin-tailed sheep breeds. Dong et al. [28] highlighted a 6.8-kb region within the first intron of the PDGFD gene that contained the most differentiated SNPs between fat-and thin-tailed sheep breeds and that likely harbors regulatory mutation(s) determining the fat-tail configuration. Furthermore, it has been demonstrated that expression of PDGFD was high on days 60 and 70 of embryonic development in the tail adipose tissue, which suggests a potential role of PDGFD in the initiation and/or commitment of preadipocytes; in contrast, expression of PDGFD decreased between days 80 and 90 of embryonic development when fat started to accumulate in the tail, thus suggesting a negative association with fat maturation [28].
Baazaoui et al. [29] employed a combination of runs of homozygosity (ROH) hotpots analysis and an F ST -outlier approach to identify genomic regions under selection in Tunisian sheep breeds. An outstanding selection signal harbored several relevant genes linked with lipid metabolic processes (i.e., CDS2, PROKR1, and BMP2). The study also emphasized the potential involvement of BMP2 in the sheep tail formation [29]. Genes linked with the skeletal tail phenotype, including HOXB13, NR5A1, NR6A1, KDM6B, and NR3C1, have also been proposed to play a major role in the adaptation of African sheep [30]. In addition, allele-specific expression (ASE) in thin-tailed Texel sheep and fat-rumped Kazakh sheep was investigated [31]. In this study, 2405 ASE genes were described from adipose tissue; among these, PDGFD and LYPLA1 were highlighted as possibly linked with lipid-related biological processes. In addition, seven ASE SNPs were found in the PDGFD gene, and a distinct genotype pattern located in the upstream region of PDGFD was detected by selective sweep analysis between mouflon, fat-tailed, and thin-tailed sheep. Moreover, potential exon splicing events within the PDGFD gene in thintailed sheep have been described [31]. More recently, Li et al. [13] emphasized the significance of both PDGFD and BMP2 signals in sheep tail formation, and described a new variant based on a 169-bp insertion close to the 5′-UTR region of HOXB13, which is likely to be associated with the long tail phenotype. This variant was further confirmed as a 167-bp segment with a linked missense mutation (c.23C>G) in Merinolandschaf sheep [32]. It is likely that multiple genes may contribute to the configuration of sheep tails since tail phenotype and related fat deposition levels vary greatly between different sheep breeds from different geographical regions.
Although the level of fat deposition mainly determines sheep tail phenotype, the length and number of caudal vertebrae have also been considered as important factors that influence tail morphology (Fig. 3). Zhi et al. [14] analyzed genomic differences between two lines of Chinese Hulunbuir sheep of short tail and normal tail, which revealed the TBXT gene (also known as T; Brachyury) as a candidate for determining the number of caudal vertebrae in sheep. Subsequently, Han et al. [33] confirmed the previously reported missense mutation (c.334G>T) with a completely linked synonymous variant (c.333G>C) to be associated with the characteristic short caudal vertebrae in fat-rumped sheep breeds. Interestingly, the functional role of the TBXT gene in forming the sheep tail length was validated using the CRISPR/Cas9 system [34]. Collectively, the abovementioned findings provide significant information obtained in research studies from the past few decades to understand the genetic causes of phenotypic variations in the configuration of sheep tails. These genomic studies identified sets of genes that could potentially be associated with phenotypic variations in sheep tails (see Additional file 1: Fig. S1a), and narrowed down the candidate genes and genomic regions involved in sheep tail formation (e.g., BMP2, LOC101117953, PDGFD, HOXB13, and TBXT). Further genomic investigations of sheep tails are shown in Additional file 2: Table S1. Moreover, sheep tail formation has been investigated using transcriptomic [mRNAs (see Additional file 2: Table S2 and Additional file 1: Fig. S1b), miRNAs (see Additional file 2: Table S3), and lncRNAs (see Additional file 2: Table S4)] and proteomic (see Additional file 2: Table S5) approaches, as well as direct gene expression and potential SNP associations (see Additional file 2: Table S6). These studies further identified potential genes, regulatory factors, and molecular pathways that might contribute to the formation and phenotype variation of sheep tails.

Development of sheep horns: the remainders of their wild past
The RXFP2 gene and its role in sheep horn development Horns (cranial appendages or headgear) [35] are typically used by male animals in mating conflicts for dominance [18]. Sheep can be horned, polled (hornless), or scurred (i.e., vestigial and deformed horns) (Fig. 4). Hornedness and polledness phenotypes have varying patterns among different sheep breeds. In certain cases, horns are found in both sexes, but their size is larger in males; males can be horned but females polled; and both sexes can be polled [36]. In other cases, horned and polled male and female representatives can be found within the same breed. Early attempts to understand the genetics of sheep horn development date from over a century ago [37][38][39][40][41]. Since then, several efforts have been made to understand the mode of inheritance of this unique phenotypic trait in sheep [42][43][44][45][46][47][48]. With the emergence of genomic-based approaches, the underlying genetic basis of this phenotypic trait in sheep has been progressively revealed. In 1996, Montgomery et al. [49] refuted the existence of a genetic locus determining horn traits in sheep on OAR1 as seen in cattle. However, a horn-determining locus was mapped on OAR10 in sheep that originate from Merino and Romney crosses. Subsequently, using a free-living population of Soay sheep (Fig. 5) in St. Kilda, Beraldi et al. [18] also mapped the horn locus on OAR10 at a location similar to that previously described in modern domestic sheep. In addition, Pickering et al. [50] localized the horn locus within a 14-Mb region using six genetic markers. Subsequently, by using fine-mapping with additional genetic markers, the horn locus was localized within a 50-kb region. This study also reported a large (approximately 3 kb) retrotransposed insertion in the non-coding 3′-UTR region of a then unnamed candidate gene, which was found exclusively in polled individuals [50].
Johnston et al. [51] further narrowed the horn locus down to a ~ 7.4-cM interval on OAR10 in a Soay sheep population. Subsequently, Johnston et al. [5] performed a GWAS using ~ 36,000 SNPs and determined that the RXFP2 gene was the main candidate for determining horn phenotype in sheep. Consistently, when comparing the genomes of global sheep breeds using the OvineSNP50K BeadChip, the strongest SNP under selection was determined to be located immediately adjacent to RXFP2, which suggested that RXFP2 can be potentially targeted by breeding for horn removal, being likely one of the oldest morphological changes occurring during sheep domestication [8]. Johnston et al. [52] later described two RXFP2 alleles in Soay sheep, namely, Ho + and Ho P , with the Ho + allele conferring larger, normal horns and the Ho P allele conferring smaller horns. Ho + Ho + , Ho + Ho P , and Ho P Ho P males were horned, and approximately half of the Ho P Ho P males were scurred. In contrast, the Ho + Ho + genotype led to horned, Ho + Ho P to scurred, and Ho P Ho P to polled females. In addition, it has been proposed that the Ho + allele that confers larger horns is associated with higher reproductive success, whereas the smaller horn Ho P allele is associated with increased survival. Collectively, these pieces of evidence suggest a trade-off between reproduction and survival for this allele in male Soay sheep. Several subsequent genomic studies highlighted the role of RXFP2 in determining horn phenotype when comparing different regional sheep populations. Pan et al. [53] described RXFP2 as a consistently strong marker of positive selection based on a GWAS in two semi-feral sheep populations from the Qinghai-Tibetan Plateau. In addition, the expression of RXFP2 in sheep has been studied in 13 tissues, and expression was exclusively in horn-related tissues [53]. Moreover, an introgression of genomic regions into the RXFP2 locus in Tibetan sheep from Argali sheep has been suggested, in which two missense variants (c.1999C>T and c.1957C>T) were also detected [54]. Recently, the same region was suggested to be introgressed from Iranian mouflon into Tibetan sheep [55]. Nevertheless, in this region, three major highly divergent haplogroups were found in haplotype patterns across sheep breeds with different horn phenotypes [55].
Wiedemar et al. [56] described a 1833-bp genomic insertion (namely, an EEF1A1-like insertion) in the flanking 3′-region of the sheep RXFP2 gene that is potentially associated with polledness. In this study, this region was genotyped in seven Swiss sheep breeds. It was shown that horned individuals were homozygous for the absence of the insertion, whereas polled individuals were homozygous for its presence. It has been speculated that when the EEF1A1-like insertion is present, EEF1A1 transcripts bind to the 3′-UTR of the RXFP2 mRNA, leading to post-transcriptional downregulation [56,57]. In Bündner Oberländer sheep with both polled and horned phenotypes, horned individuals were homozygous for the absence of the insertion; however, two polled individuals were found to be heterozygous for the insertion. Later, Lühken et al. [58] reported that the actual length of the EEF1A1-like insertion was 1780 bp. By genotyping this insertion in a large number of sheep breeds and individuals with different horn statuses, it was concluded that this insertion cannot be considered as the exclusive determinant of polledness in sheep, and thus could not be considered as a universal marker for determining the genetic causality of horns in sheep. The insertion was shown to  Table S14 segregate with horn status (such as polled or horned) only to a certain extent and within certain sheep breeds. In other breeds with variable horn statuses in both sexes or with a sex-dependent horn status, the segregation of this potential insertion is weaker. It has been suggested that this insertion is among the potential various causative factors for polledness in sheep [59]. He et al. [60] performed PCR amplification of the DNA of 49 Chinese  Table S14 sheep with three horn statuses (polled, two-horned, and four-horned), and found the EEF1A1-like insertion in 68.4% (13/19) of the polled individuals and 43.3% (13/30) of the horned individuals. These findings suggest a yet unverified association between the EEF1A1-like insertion and the polledness phenotype in sheep.
Large-scale investigations have also been performed to elucidate the origins of mammalian horns [61,62]. Wang et al. [61] identified 624 horn-related genes that are highly co-expressed in bone, skin, testis, and brain tissues. In addition, 201 genes were commonly found in both horn and antler tissues, and these genes were enriched in bone and skin development, and neurogenesis pathways. This study provided important insights into the origin of horns and further proposed a number of genes potentially associated with horn development regulation. Among these genes, SOX10, SNAI1, and TFAP2A were highly expressed in fetal sheep horn buds but not in adjacent skin tissues. In addition, this study also suggested a single evolutionary origin for the ruminant headgear, as well as proposed that the horn-stimulating gene RXFP2 is under convergent pseudogenization in secondarily headgearless lineages [61]. Furthermore, using transcriptomic data, seven highly expressed keratin genes, namely, KRT1, KRT2, KRT3, KRT5, KRT10, KRT14, and KRT84, were found to be associated with the formation of keratinous sheath [62]. Type II a-keratin proteins expressed by the above-mentioned genes (except KRT10 and KRT14) have been suggested to play an essential role in the development of the horn keratinous sheath [62]. Further investigations on the sheep horn phenotype are shown in Additional file 2: Table S7. Collectively, these findings highlight the contribution of a gene set to the determination of sheep horn status; in particular, the prominent role of RXFP2 with the presence/absence of horns in sheep has been suggested. Although a significant association between RXFP2 and horn development has been indicated, the exact variants and the underlying mechanisms by which RXFP2 contributes to phenotypic differences in horn status in sheep have not yet been elucidated.

Beyond the horned or polled dichotomy: polyceraty in sheep
Polyceraty is related to the condition where the animal carries multiple horns (Fig. 6), and which is considered rare in sheep [63]. Among native sheep populations, only few breeds carry multiple horns (Fig. 7), and polycerate sheep usually carry four or six horns. Ren et al. [64] performed a GWAS comparing 24 two-horned and 22 four-horned Chinese Sishui Fur sheep, which enabled  Table S14 the identification of a potential genomic region (132.0-133.1 Mb) in OAR2 that contained the top ten SNPs associated with the polycerate phenotype. In humans and mice, this genomic region includes the HOXD gene cluster, which comprises nine genes, including HOXD1, HOXD3, HOXD4, and HOXD8-HOXD13, as well as other functional genes, such as EVX2 and KIAA171, which was shown to have a potential association with the formation of limbs and genital buds [64]. Subsequently, Kijas et al. [65] performed a GWAS mapping 125 two-and fourhorned individuals from Jacob and Navajo-Churro sheep breeds to investigate the genetic cause of four-hornedness as well as examine whether it was also related with eyelid abnormality at various degrees of severity. This study confirmed that OAR2 harbors a strong marker linked with polyceraty (131.9-132.6 Mb). The MTX2 gene and the HOXD gene cluster (especially HOXD1) were the closest to the genetic marker that was most strongly associated with polyceraty. In Jacob and Navajo-Churro sheep breeds, eyelid abnormality was strictly observed in polled individuals and in polycerate individuals, which suggests an absence of eyelid abnormality in two-horned animals and a potential developmental link between eyelid abnormality and polyceraty [65]; thus, this study proposed a pleiotropic effect for the development of the horn trait in sheep. Greyvenstein et al. [66] conducted a GWAS on 43 Damara sheep (19 fourhorned, 7 three-horned, 16 two-horned, and one polled), and confirmed that the proposed genomic region (128-135 Mb) spanning nine embryonic development HOXD genes is causal for the polyceraty phenotype.
He et al. [60] performed a GWAS on 34 two-horned and 32 four-horned individuals from three Chinese breeds (namely, Altay, Mongolian, and Sishui Fur). The potential genomic region (132.6-132.7 Mb) on OAR2 was similar to that previously identified in polycerate sheep breeds, i.e., Sishui Fur, Jacob, Navajo-Churro, and Damara, which suggests a common genomic cause for the polyceraty phenotype among different sheep breeds. In a subsequent study, the same research group performed another GWAS using 72 polycerate and 24 two-horned Tibetan sheep, to map the genomic regions of typical (four horns) and atypical polycerate (three, five, and six horns) phenotypes [67]. The same candidate genomic region (131.990-133.525 Mb) was identified, thus suggesting that this region affects both typical and atypical polycerate phenotypes. In a more recent study, Allais-Bonnet et al. [15] revisited the Illumina OvineHD Beadchip genotyping data of 111 cases and 87 control sheep previously reported [65,66], and identified the polycerate locus between the positions 132,717,593 and 133,151,166 bp on OAR2. Subsequently, by comparing whole-genome sequences of 11 polycerate specimens and 1179 control sheep from various populations, a four-nucleotide deletion variant (g.132,832,249_132,832,252del) was revealed at position +4 to +7 bp after the first exon of the HOXD1 gene. Moreover, this potential variant further showed a perfect association when 236 animals from various populations, containing polycerate specimens, were investigated [15]. This variant was proposed to influence the splicing of HOXD1 precursor RNAs, thus generating the multi-horn phenotype in sheep.

Sheep coat color: a major breed determinant The ASIP gene
Sheep coat color has been investigated since a long time [68][69][70][71][72][73] and its variations have sparked the curiosity of a wide spectrum of researchers and specialists to understand its biological and genetic basis [6,74,75]. Coat colors in sheep vary widely among and within breeds (see Additional file 1: Fig. S2). In Australian Merino sheep, a uniform white fleece is a desirable trait, and the small proportion of colored-fleece lambs born annually in these flocks leads to significant financial loss for sheep breeders and wool processors. To reveal genes linked to pigmentation in the sheep genome, Parsons et al. [16] used comparative mapping information to investigate the ovine homologous position of the Agouti gene. The study showed that the locus responsible for recessive pigmentation in Australian Merino sheep is positioned within OAR13 in a region homologous to murine chromosome 2 and human chromosome 20 regions that contain the Agouti gene locus. In a subsequent report [76], the same authors used the available sequencing data of human and mouse homologs to isolate the coding region of the ovine Agouti gene, which revealed that the isolated ovine sequence has an authentic agouti homolog at the nucleotide and amino acid levels to other previously reported mammalian species. In addition, by developing a linkage map for loci responsible for certain phenotypic traits in Soay sheep, the 'coat pattern' (i.e., intensified coat color in the 'self ' individuals vs. the contrast in color between the belly and the rest of the body in 'wild' individuals) mapped to OAR13 and close to the candidate Agouti locus [18].
Smit et al. [77] amplified a 142-bp segment in exon 2 of the Agouti gene (ASIP) using 361 samples obtained from sheep. Bands of 142 bp were present in all animals and bands of 137 bp (agouti-like sequence) were shown in 173 of the investigated sheep. The sequences of the 142bp amplicons were identical to the previously published ovine agouti sequence [76] and a 5-bp deletion (D 5 ; positions 98-102 from the start codon in the normal agouti transcript) was identified in the 137-bp amplicon. Subsequently, D 5 was surveyed in ten sheep breeds, and the presence of D 5 was observed in white-faced sheep breeds but not in black-faced sheep breeds [77]. Royo et al. [77] reanalyzed the potential of D 5 by sequencing the ASIP exon 2 and intron 2 using samples obtained from 188 individuals of the rare Xalda sheep breed, and the deletion (c.100_104delAGGAA ) was verified in the surveyed individuals. All Xalda individuals carrying the deletion in a homozygous mode were phenotypically black; however, other black-coated Xalda individuals (91%) were not homozygous, which suggests that the homozygous deletion could not be the only determinant of black coat color in sheep. In addition, mRNA levels of ASIP was evaluated in 24 (14 black-coated and 10 white-coated) Xalda individuals, which revealed that ASIP expression in black-coated individuals was significantly lower than in white-coated individuals, and that D 5 did not influence ASIP expression [78].
The role of ASIP in determining coat color pattern was primarily proposed by classical genetics. The dominant white or tan (A Wt ) ASIP allele is responsible for the yellow/red phenotype (caused by phaeomelanin), whereas the recessive allele non-agouti (A a ) is responsible for the black/brown phenotype (caused by eumelanin). In addition to other patterns such as badgerface (A b ) where the dorsal region is pale and the ventral region is darker. To reveal the genomic basis of these patterns, Norris et al. [79] performed sequence analysis of genomic DNA, bacterial artificial chromosome clones, and RT-PCR products from a group of sheep breeds with different coat colors. Using these molecular tools, the study proposed a model for the evolution of the ovine ASIP locus. This includes (i) a 190-kb tandem duplication encompassing the ASIP and AHCY genes and the ITCH promoter region, which was identified as the genetic determinant of the white coat color of dominant white/tan (A Wt ) agouti sheep; (ii) a single copy of ASIP with a silenced ASIP promoter was identified as the genetic cause of the recessive black sheep; and (iii) a single copy of wild-type ASIP (A + ) with a functional promoter was identified as the genetic cause of the ancient Barbary sheep coat color (tan body and pale belly) [79]. Notably, a 9-bp deletion (D 9 ) was detected in this study in exon 2 of ASIP. Collectively, the findings of this study showed the complexity of the mechanism underlying ASIP regulation and provided novel insights into its role in determining sheep coat color [79]. In addition, the copy number variation (CNV) in the Agouti locus was also used as a model for a segregation analysis to estimate allelic copy number [80].
Li et al. [81] performed a GWAS to identify the genetic determinants of the white vs. non-white coat color variation in a Finnsheep population. Among the 35 significant SNPs identified, 25 were located in or adjacent to five previously reported genes (ASIP, TYRP1, KIT, MC1R, and MITF) known to be involved in sheep coat color development. In addition, the s66432.1 SNP in the ASIP gene showed the strongest association signal with the development of coat color patterns between white and non-white individuals. In particular, the white coat color pattern was associated with the A allele, whereas the non-white coat color pattern was associated with the G allele. Additional studies on the sheep ASIP gene are reported in Additional file 2: Table S8. Interestingly, with the emergence of gene-editing endonucleases, the ASIP gene was targeted using the CRISPR/Cas9-based knockout approach [82,83]. Among the generated Merino sheep, three individuals harboring a 4-bp deletion in the ASIP gene showed (i) a badger-face with black body coat color (n = 2) and (ii) brown coat color with light ventral pigmentation (n = 1), whereas (iii) two individuals with a 20-bp deletion showed a black-white spotted coat color [83] (see Additional file 1: Fig. S3). Collectively, these findings led to the conclusion that the CRISPR/Cas9mediated ASIP disruption together with D 9 /D 5 mutations and ASIP duplications contributed to the emergence of coat color in ASIP-targeted white Merino sheep. Thus, the manipulation of the ASIP gene using gene-editing tools enabled the determination of its functional role in sheep coat pigmentation.

The MC1R gene
It was expected that various loci or genes would epistatically interact to generate a wide variety of sheep coat colors. In addition to the Agouti locus that encodes ASIP, the Extension locus encoding the MC1R gene was shown to play a role in the formation of coat color in sheep [84]. It plays the opposite role to ASIP, i.e., its high functional expression is responsible for the dark coat color phenotype. Våge et al. [17] analyzed the ovine MC1R gene in Norwegian Dala sheep, which comprise individuals with completely white or black coat colors. The study identified the complete co-segregation of c.218 T>A (p.73Met>Lys) and c.361G>A (p.121Asp > Asn) mutations with dominant black coat color (the dominant E D allele indicates the black coat pattern and the wild-type E + indicates the white coat pattern). When these two mutations are induced in the mice MC1 receptor gene, the recapitulated p.73Met>Lys mutation alone was able to constitutively activate the receptor, whereas the p.121Asp >Asn mutation lacked this functionality, being thus proposedly required for high-affinity ligand binding [17]. In addition, Våge et al. [85] confirmed the cosegregation of these two mutations (the dominant E D allele) with the black coat color pattern in Damara, Black Merino, and Black Corriedale. Moreover, Royo et al. [78] detected the presence of the dominant black E D allele in Karakul, Black Merino, and Black Castellana sheep.
In addition, a GWAS was performed using the OvineSNP50K BeadChip to determine the major pigmentation gene using 42 samples obtained from Manchega and Rasa Aragonesa sheep, in which the solid black and solid white coat colors segregate [86]. In this study, the SNP s26449 was the most strongly associated with coat pigmentation and the closest to MC1R [86]. In Brazilian Morada Nova sheep, a GWAS also revealed the occurrence of MC1R in the highly significant SNPcontaining region [87]. In this study, white-coated and red-coated animals were homozygous for the E + allele, whereas black-coated animals carried at least one copy of the E D allele [87]. In a study performed to identify pigmentation-specific genes associated with skin photosensitization, MC1R was also the most significant genomic region under selection, and a number of other genes were also suggested to be potentially associated with this trait, including KIT, PDGFRA, IRF4, SOX10, PICK1, and EDN3 [88]. Moreover, a GWAS using the OvineSNP600K Beadchip was performed on 75 Chinese Tan sheep individuals (29 black-headed and 46 white-coated; see Additional file 1: Fig. S4), which revealed two significant SNPs (rs409651063 and rs408511664) in MC1R that were associated with coat color development [89]. Further information from previous studies on the sheep MC1R gene is shown in Additional file 2: Table S9. Collectively, these results highlight the role of MC1R and its variants in the development of coat color pattern in sheep.

The TYRP1 gene
Considering the complexity of the coat color trait, it has been suggested that more than one gene could determine the phenotypic variation of coat colors. In Soay sheep, coat color density can be classified into dark and light patterns (in a ratio of approximately 3:1). Beraldi et al. [18] developed a complete linkage map to identify the loci responsible for determining several phenotypic traits, including coat color, which revealed TYRP1 as a strong candidate gene influencing sheep coat color. Subsequently, Gratten et al. [90] analyzed TYRP1 in Soay sheep, and showed that a nonsynonymous substitution (c.869G>T) in exon 4 was highly associated with coat color pattern. For this SNP, light-colored sheep were homozygous for T, whereas dark-colored sheep were either homozygous for G or heterozygous. In addition, no significant difference was found when analyzing TYRP1 expression in dark-and light-coated Soay sheep, which hindered the assessment of the potential role of this SNP in determining sheep coat color [90]. Furthermore, an association between the TYRP1 c.869G>T mutation and body weight in Soay sheep has been reported [91]. It has been shown that dark-coated sheep with GG and GT genotypes were heavier at birth than light-coated sheep with the TT genotype. Moreover, dark-coated sheep with the GG genotype showed a fitness disadvantage compared to dark-coated sheep with the GT genotype. In this case, it was also difficult to predict how TYPR1 influences the body weight or fitness [91]. It has been proposed that other genetic factors in the vicinity of TYRP1 are likely responsible for microevolutionary constraints that influence these traits. It has also been suggested that the changing ratio between dark-and light-coated Soay sheep is a result of climate change [92,93]; however, such a claim is not well evidenced [94].
In addition, Raadsma et al. [95] performed an extended QTL study on 13 skin and fiber pigmentation traits in Awassi × Merino × Merino backcross sheep. Significant QTL signals were identified on OAR2 where TYRP1 is located. Notably, the nonsynonymous TYRP1 SNP proposed by Gratten et al. [90] in Soay sheep was not confirmed in the examined flock. However, this study proposed an influence of the nonsynonymous SNP (c.2240C>G) in exon 2 of TYRP1 on coat pigmentation [95]. In addition, two mutations, g.80608128G>T and g.80611700C>T within the TYRP1 gene have been reportedly associated with brown coat color development in American sheep populations [96]. Furthermore, a GWAS was performed using Valais Red sheep with the aim to compare brown and black individuals, which showed a strong signal in the TYRP1 locus, and three variants were shown to be associated with the recessively inherited brown coat color, namely (i) the previously reported missense variant (c.869G>T), (ii) a frame-shift variant (c.86_87delGA), and (iii) a nonsense variant (c.1066C>T) [97]. Taken together, these studies provide evidence for the potential involvement of TYRP1 in the development of coat pigmentation in sheep. Additional information on sheep coat color variation is shown in Additional file 2: Table S10, which highlights the potential contribution of a set of genes to the formation of coat color in sheep.

Black bone color or hyperpigmentation
Black bone color or hyperpigmentation has been observed in few sheep breeds. In this condition, the pigment is found not only in the coat but also in the skin, muscles, inner organs, and bones. In the indigenous Chinese Nanping black-boned sheep, two synonymous mutations (c.192G>C and c.462C>T) in exon 1 of the TYR gene, as well as two synonymous mutations (c.12A>G and c.144G>C) in the MC1R gene were detected [98]. In addition, four combinations of a synonymous (c.90C>T) mutation and a missense (c.203C>T) mutation were found in the TYRP1 gene in both black-boned and normal sheep [98]. Two other SNPs, a synonymous (c.102C>T) and a missense (c.215T>C) were also indicated in exon 2 of TYRP1 in black-boned sheep [98]. Patterns of these SNPs showed elevated plasma tyrosinase activity [99][100][101]. In general, black-boned sheep show higher levels of melanin and related pigments in tissues and organs compared to the other examined breeds with white bone color [101]. Another gene, EDN3, known to cause hyperpigmentation in black-boned chickens (the silky fowl), was also analyzed in sheep [102]. However, the results of this study showed non-significant differences in genotype and allele frequencies of the identified SNPs and copy number of EDN3 gene among blackboned and non-black-boned sheep. Interestingly, a different bone pigmentation, e.g., red pigmentation, has also been described in goats [103][104][105]. Such pigmentation characteristics might influence carcass quality and standards and, subsequently, consumer preference.

Fat color
Yellow fat color is an inherited recessive trait caused by the accumulation of carotenoids in adipose tissues [106][107][108][109][110]. This trait is known to influence carcass quality; thus it is of great interest to understand its genetic background. Våge et al. [111] analyzed two genes (BCMO1 and BCO2) known to be important for carotenoid degradation in Norwegian sheep, and highlighted the potential of a nonsense mutation (c.196C>T) to be associated with the emergence of the yellow fat phenotype. However, individuals who lacked this mutation had yellow fat, which suggested that another mechanism might be involved in the emergence of this trait. To further explore this, a more recent investigation was conducted in Norwegian spaelsau sheep with aberrations in the BCO2 mRNA, which revealed the presence of a 7.9-kb insertion of an endogenous Jaagsiekte Sheep retrovirus (enJSRV) sequence into the first intron of BCO2 [112]. This insertion led to the inactivation of BCO2 due to the generation of a truncated product of 58 amino acids (29 from BCO2 and 29 from enJSRV), whereas the functional BCO2 product is known to contain 576 amino acids [112]. Interestingly, the ovine BCO2 gene was previously targeted using the CRISPR/Cas9 gene-editing system in a study aimed at generating sheep with multiplex editing in three genes (MSTN, ASIP, and BCO2) [82,113]. The BCO2 gene was targeted to functionally validate its biological role, and was used as a gene with an obvious phenotype to validate the efficiency of CRISPR/Cas9 in sheep. The study revealed that the biallelic modification of BCO2 resulted in yellow fat compared to the monoallelic and wild-type individuals, confirming its functional role in fat color formation in sheep [113].

Genetic aspects of other phenotypic traits Wool length
Several genes are known to influence sheep wool quality and quantity, e.g., those included in the KAP families [114][115][116]. In the following sections, we will focus on genes related to the wool phenotype rather than genes that influence quantitatively wool aspects. In addition to the KAP families, a number of genes were shown to greatly influence fiber traits, e.g., FGF5. The inactivation of the FGF5 gene was first proven in mice to cause abnormally long hair [117]. Natural variants in the sheep FGF5 gene causing the long hair phenotype have not yet been described [118]. However, the potential of FGF5 has been investigated using gene modification tools. In order to induce wool growth by antagonizing the FGF5 function, transgenic Chinese Merino sheep with ectopic expression of short alternative FGF5 transcripts (FGF5s) were generated by zygotic injection of recombinant lentivirus; wool length was significantly longer in the transgenic group, and greasy wool weight was increased in yearling [119]. In addition, direct knockout of the FGF5 was also achieved using the CRISPR/Cas9 system, and FGF5-disrupted sheep showed the desired long wool phenotype [120][121][122][123]. Interestingly, similar findings were also found in goats [124][125][126]. Using the FGF5disrupted sheep model, a crosstalk between androgen and the Wnt/β-catenin signaling pathway in influencing wool increase and active hair-follicle density was revealed [123]. This interplay has been linked to the activation of the c-MYC and KRT genes associated with the inner root sheath. Furthermore, the Shh pathway being located downstream of the Wnt/β-catenin signaling pathway, it was also shown to be involved in wool length [123]. Moreover, a recent study on putative SNPs within the FGF5 full-length gene in fine-wool sheep revealed potential molecular markers associated with wool length and greasy wool weight [127]. Based on these results, FGF5 has been highlighted as a potential gene that could lead to improving yield in wool-producing sheep breeds, thus promoting the development of the wool industry.

Curly fleece
Chinese Tan sheep have curly fleece during the first months following birth (Fig. 8g), and this phenotype gradually disappears with aging. To understand this phenotypic change, a transcriptomic study involving sheep at two different developmental stages was performed, which enabled the identification of 600 differentially-expressed genes (DEG) [128]. The top 20 highly expressed transcripts belonged to genes included in KAP families as well as other genes, namely, KAP3.2, KAP2.3,  KRT25, KAP1.1, KRT5, KRT71, KRT14, KRT2.11, KRT83,  KRT31, KRT34, KRT33, KRT27, KRT33B, TCHH, RPS8, PRR9, CST6, ASIP, and CO3. Furthermore, isoform analysis revealed that the MT3 isoform was upregulated in lamb skin compared to adult individuals, suggesting its potential involvement in the development of curly fleece phenotype [128]. In addition, using an integrated miRNA-mRNA analysis of samples obtained from sheep at these two physiological stages, 49 differentially-expressed (DE) miRNAs were identified [129]. In addition, 36 potential miRNA-mRNA target pairs were identified (including 25 DE miRNAs and 10 DEG). Among DE miRNAs, it was confirmed that miR-432 could inhibit KRT83 expression [129]. Moreover, using suppression subtractive hybridization analysis, 67 DEG were identified in the samples of sheep at two developmental stages, and KRT71 was significantly highly expressed in individuals within the curly fleece group compared to the non-curling group [130]. The MFZ1 transcription factor has also been suggested to regulate the expression of KRT71 [130]. It has also been suggested that KRT83 expression was higher in Chinese Tan lambs with curly fleece compared to adults, and that the expression of the CAP1 transcription factor was low [131]. When CAP1 was overexpressed, KRT83 expression was repressed, which suggests that CAP1 has a role in the regulation of KRT83 expression [131].
In a genome-wide detection of CNV on 48 Chinese Tan sheep, a CNV region overlapping the homeobox transcription factor DLX3 was suggested as a candidate CNV for the curly fleece phenotype [132]. SNPs (c.118T>C,  c.228T>C, c.688A>G, and c.1,038_1,039 insC) within the 3′-UTR of DLX3 were also previously described as associated with wool crimp in sheep [133]. In Chinese Merino sheep, a set of genes (namely, CDC6, RARA , IGFBP4, and GHR) was identified as candidates for crimp number, in particular IGFBP4 [134]. In addition, the levels of PAPPA2 expression in dermal papilla cells (DPC) of straight-wool Hu lambs was significantly higher than in DPC of curved wool Hu lambs [135]. It has been speculated that PAPPA2 affects the proliferation of DPC by regulating IGFBP5 expression, which influences coat phenotypic patterns [135]. In Hu sheep, miR-143 has been speculated to influence wool curvature by targeting CUX1 thereby inhibiting the proliferation of DPC and KRT71, which ultimately regulates the growth and development of hair follicles [136]. The higher expression of CUX1 and KRT71 was proposed to be related to wool curvature [136]. Furthermore, an analysis of DNA methylation levels in Chinese Tan sheep at three growth stages (1-, 24-, and 48-month old) revealed 51 differentially methylated genes (DMG) associated with age growth and curly fleece formation [137]. Among these DMG, KRT71 and CD44 were highly methylated in one-month-old lambs, whereas ROR2 and ZDHHC13 were highly methylated in 48-month-old sheep. Interestingly, a similar phenotype of curly fibers was also found in goats [138][139][140][141]. Further information on the curly fleece phenotype in sheep is provided in Additional file 2: Table S11. Several genes and small regulatory RNAs were proposed to be involved in the formation of the curly fleece phenotype. Understanding the genetic basis of wool-related Kalds Table S14 traits in sheep is of significant importance for the genetic customization of wool characteristics and quality for economic purposes.

Fine vs. coarse fleece
Wool fiber diameter varies widely among sheep breeds. Fine wool fiber is characterized by a small diameter, whereas coarse wool fiber has a larger diameter compared to fine fiber. Fine wool is employed in manufacturing soft clothes, whereas coarse wool is employed in carpet weaving. As it is significantly determinant for wool characteristics, early research efforts have concentrated on understanding the genetic determinants leading to differences in wool fiber diameter [142][143][144][145]. In a GWAS using 765 Chinese Merino sheep, nine significant SNPs were suggested to be associated with wool fiber diameter; among these, four SNPs were located within previously known or predicted genes, including TSPEAR, PIK3R4, KRTCAP3, and YWHAZ [146]. These genes were found to be directly or indirectly associated with skin development, and their role in wool follicle development and fiber diameter has been proposed. To determine the wool fiber diameter coefficient of variation, three significant SNPs were detected, among which one was found to be located within the KIF16B gene [146]. In addition, when 18 markers under selection in fine-wool Australian Merino and coarse-wool Spanish Churra sheep breeds were analyzed, divergent variants potentially involved in phenotypic variation of wool type, growth, and meat production/quality traits were identified [147]. However, in this study, the investigated two breeds had varied phenotypic traits (wool, color, carcass, milk production, parasite resistance, among others), thus the identified markers under selection are likely not exclusively related to wool traits [147]. In addition, in another study, BNC1 has been suggested as significantly associated with the mean fiber diameter in Chinese Merino sheep [134].
To further understand wool fiber diameter at the transcriptome level, a study was conducted using Gansu Alpine Merino sheep, which have wool of two fiber diameters (super-fine and fine), were analyzed, and 40 DEG were identified [148]. Also, when analyzing natural antisense transcripts, three of 40 DEG (AQP8, Bos_d2, and SPRR) had significant natural antisense transcripts and were all significantly downregulated in the investigated super-fine wool group, thus indicating the correlation between gene transcripts and natural antisense transcripts in gene expression regulation [148]. Additional information on wool diameter is provided in Additional file 2: Table S12. Collectively, the findings discussed above point towards the existence of sets of potential genes that regulate wool fiber diameter, but further studies are required to elucidate their underlying genetic mechanism.

Hairy vs. woolly fleece
Several modern domestic sheep breeds have been selected for short and woolly fleece coat phenotype, which differs substantially from the ancestral long and hairy fleece coat phenotype. To identify the molecular basis of this evolutionary phenotypic variation, Demars et al. [149] analyzed the Romane sheep breed in which fleece pattern segregates, to determine the causal mutation that controls fleece variation. An insertion of an antisense EIF2S2 retrogene (asEIF2S2; ~ 1.5 kb) into the 3′-UTR of IRF2BP2 was identified as leading to the production of an abnormal IRF2BP2/asEIF2S2 transcript. This chimeric messenger targets the genuine sense of EIF2S2 RNA and alters the expression of both EIF2S2 and IRF2BP2 mRNA by a retroposition mechanism-led RNA-RNA hybridization. It was shown that woolly sheep carried an insertion (IRF2BP2 asEIF2S2 allele), whereas hairy sheep carried the ancestral IRF2BP2 wt allele [149]. This study provided the keys to understand the unique mechanism that underlies a causal variant leading to phenotype variation related to fiber folliculogenesis in sheep [150]. Lv et al. [151] identified a novel mutation in OAR25 (g.7068586T>C) in the 3′-UTR of IRF2BP2 as a causal variant influencing fleece fiber diameter. In addition, the expression of a bindable miRNA (oar-miR-20a-3p) was detected exclusively in the skin tissues of coarse-wool sheep carrying the ancestral T allele. Furthermore, a strong selection marker in the secondary wool follicle-associated VEGFA gene was detected, which was shown to be co-activated by IRF2BP2. It has been proposed that oar-miR-20a-3p binds to the wild-type 3′-UTR of IRF2BP2, regulating its expression and subsequently inhibiting VEGFA expression in the skin tissues of coarse-wool sheep [151].
Interestingly, when genotyping the hairy/woolly locus in a ~ 1600-year-old naturally mummified sheep found in a salt mine (Chehrābād) in northwestern Iran, the ancestral "hairy" allele (likely IRF2BP2 wt/wt or IRF2BP2 wt/ asEIF2S2 ) was identified [152]. In addition, using scanning electron microscopic observations of the mummified hair fibers showed characteristics that are specific to sheep hair fibers of mouflon and medium-wool breeds. This suggested that, during that ancient period, the herd was maintained for meat or milk production rather than wool extraction, and ovicaprids were destined to be used as food source for workers in that location [152]. This study highlighted the potentiality of using a locus for the phenotypic characterization of ancestral breeds using ancient DNA. Fleece pattern in sheep influences adaptation to environmental conditions, thus it is expected that the discovered variants related to different fleece patterns will be used in selection-or molecular genetics-based strategies to customize sheep breeds in order to better adapt to future environmental changes and sustainable sheep production.

Ear size
Sheep ear morphology varies widely in terms of length (large, medium, small or "microtia", and earless or "anotia") and erectness (pricked or "erect" and dropped or "floppy"), and is considered a sheep breed determinant (Fig. 8b-e). In sheep breeds with varying ear length, such as Awassi, sheep farmers prefer to breed individuals with the normal large ear phenotype owing to their better physical constitution. In addition, it is indicated that earless Awassi sheep tend to show a more flighty behavior, which likely results from impaired hearing ability [153]. Ear phenotype has also been associated with environmental adaptation, which highlights the importance of investigating the molecular basis of the ear phenotype in sheep. A SNP within the MSRB3 gene was to strongly segregate within the Chinese Duolang sheep with large ear phenotype and not in other breeds with normal or smaller ears, thus suggesting a role for MSRB3 in determining the ear phenotype in sheep [11]. In a GWAS performed to investigate microtia in Jordanian Awassi sheep (in which 12 normal-eared and 8 earless individuals were included), a SNP adjacent to GATA6 was identified [153]. GATA6 was proposed based on its location and function, which is related to developmental processes associated with outer ear shaping (e.g., chondrogenesis, the process formation of cartilage) [153].
To further investigate genes that influence the ear phenotype in Duolang sheep, a GWAS was carried out using 115 adult individuals with different ear sizes [154]. The most significant SNP was located within the DCC gene which is known to be associated with ear development; two other significant SNPs (PTPRD and SOX5) showed a physical association with ear development. In this study, no association with GATA6 was found [154]. When genotyping 40 individuals (20 with microtia and 20 normal) of the Valle del Belice sheep breed, a strong marker within intron 3 of CLRN1 was identified [155]. CLRN1 is associated with Usher syndrome type 3A, which is a progressive hearing and vision loss, and was thus proposed as a novel candidate gene for determining microtia in sheep [155]. In another GWAS using Chinese Altay sheep (26 microtia individuals and 29 normal-ear size individuals) combined with direct genotyping of a larger number of individuals, a 76-bp duplication in the evolutionarily conserved region (ECR) of HMX1 was determined [156]. This gene is known to be associated with a congenital deformity of the outer ear in species such as cattle and rats, and was proposed to determine microtia in sheep [156]. Furthermore, by analyzing previously studied genes in 84 Awassi sheep (16 earless, 41 short-eared, and 27 normaleared), the ECR 76-bp duplication of HMX1 and a missense SNP (g.34498242A>G) in exon 7 of GATA6 were shown to be significantly associated with microtia [157]. In addition, using a pruned dataset of 515 sheep from 17 breeds with different ear phenotypes, the most relevant marker for determining sheep ear size was found within the previously described MSRB3 gene [158]. The strongest markers for ear erectness (dropped vs. pricked ears) were also located within the 3'-region of MSRB3 [158]. Potential introgressive haplotypes of MSRB3 were found to be associated with ear morphology [55]. Xu et al. [159] reported potential novel genes associated with ear size, which included PTCH1 and EMX2. Information on genetic variants known to determine the ear phenotype can be used in genetic screening to improve the selection against undesirable ear phenotypes (e.g., microtia) in certain sheep breeds.
Interestingly, other regions (including e.g., the SUPT3H, STXBP5L, DENND1A, and VPS13B genes) that have been introgressed from wild Ovis species into domestic sheep were suggested to be potentially associated with facial shape traits (e.g., nose bridge breadth, nose morphology, chin dimples, and forehead protrusion) [55].

Number of nipples
The supernumerary nipple/teat phenotype (also known as polythelia; see Additional file 1: Fig. S5a-c) was observed in sheep and documented over a century ago [160][161][162][163][164][165][166]. Ewes with supernumerary nipples usually carry three, four, five, or six nipples, and extra nipples are extremely small and undeveloped. In the past, extranippled ewes were believed to be more fertile, and selective breeding for supernumerary nipples could have been developed to ultimately lead them from their rudimentary condition to fully functional milk-producing nipples [160]. However, it has been suggested that this trait has no practical value in sheep production [165], and now it is considered a morphological defect that might impede milking efficiency by increasing milking time and the risk of injuries [167], as well as negative implications for overall udder health [168]. A GWAS was performed using 144 ewes of the Chinese Wadi breed (75 cases with four teats and 69 control cases with two teats), which revealed 63 significant SNPs and a set of functional genes (BBX, SEMA3D, IRF2BP2, SETBP1, NAV3, CD47, LRP1B, CERS5, CASK, and CADM2); a significant SNP near the BBX and CD47 genes was determined to be highly associated with the supernumerary nipple phenotype [169]. In addition, Gene Ontology (GO) enrichment analysis revealed a set of functional genes that are potentially involved in determining nipple number and development, including CASK, a gene that was characterized as a risk factor associated with breast cancer development [169].
In addition to the five genes associated with breast cancer (LRP1B, GRM3, MACROD2, SETBP1, and GPC3) [169], seven novel genes (PHGDH, KDM3A, GLIS3, FSHR, CSNV2, CSN1S1, and ROBO2) previously reported in mice were associated with mammary and nipple development in sheep [27]. Furthermore, by performing GWAS using CNV data, the associated functional gene, GPC5, was found to be associated with nipple number [27]. In Hu sheep, LHFP, DPYSL2, and TDP-43 genes were also found to be related to nipple number [170]. The identification of genes and variants that control nipple number trait will likely enable the molecular characterization of sheep breeds for the desired nipple phenotype.

Number of thoracic and lumbar vertebrae
Vertebrae are the interlocking bones that constitute the spinal cord. Sheep, as a tetrapod, have five types of vertebrae, namely cervical, thoracic, lumbar, sacral, and coccygeal or caudal vertebrae (see Additional file 1: Fig. S6). In addition to studies performed to identify genes associated with caudal vertebrae in sheep [14,33], it has been suggested that the number of thoracolumbar vertebrae is associated with carcass length and weight [171,172]. Thus, it is of interest to understand the genetic basis of vertebrae-related traits. Six hundred Chinese Kazakh sheep (400 with a larger number of thoracic vertebrae and 200 with an average number of thoracic vertebrae) were investigated by DNA sequencing. Eleven putatively selected genomic regions related to the indicated phenotype were revealed, including the vertebral developmentrelated VRTN gene and the HOXA gene family [173]. Further investigation showed that VRTN is the major promising gene involved in the determination of the number of thoracic vertebrae in sheep. In addition, during fetal development, the expression level of VRTN was shown to be significantly higher in sheep with more thoracic vertebrae compared to sheep in the control group [173].
A recent study identified genes associated with the number of ribs in 36 Hu sheep with 14 ribs (supernumerary condition) and 36 Hu sheep with 13 ribs (normal condition), and a combined effect of CPOX (g.178730623T>G; Oar_rambouillet_v1.0), KCNH1 (g.75716237C>G), and CPQ (g.88323841A>G) was proposed to be associated with rib number [174]. Furthermore, using whole-genome resequencing of Chinese Duolang sheep with a different number of lumber vertebrate (6 vs. 7) revealed a significant region under selection (49.68-49.74 Mb; Oar_v4.0) that is contained within the SFRP4 gene [175]. In addition, two SNPs (rs600370085C>T and rs415133338A>G) within this gene were significantly associated with multi-lumbar vertebrae [175]. Additional file 2: Tables S13 and S14 contains further information on SNPs associated with the number of thoracic and lumbar vertebrae in sheep. Increasing the understanding of genes and variants controlling this phenotype in sheep will enable the design of selective breeding programs for increasing carcass length and weight.

Body height and stature
Dwarfism, whether proportionate or disproportionate, is included among the conditions that influence sheep phenotype [176,177]. Ancon sheep is one of the breeds known to be affected by dwarfism, as a result of a putative dwarf mutation, leading to individuals with elongated bodies and very short legs [178]. This phenotype was subjected to selection by breeders since such affected sheep are unable to jump fences. Ancon sheep were raised for a brief period of time, but went extinct as their phenotype was no longer desired [179,180]. Another physical condition including chondrodysplasia, characterized by dwarfism and angular deformities of forelimbs, was reported in Texel as a result of a 1-bp deletion of T (g.25513delT) at position 107 in exon 3 of the SLC13A1 gene [181]. Another condition of dysplasia was also reported in the Brazilian Cabugi sheep but its potential genetic basis has not been elucidated; in this context, dwarf lambs are characterized by a small size, short legs, domed heads, and other abnormalities [182]. The spider lamb syndrome that is associated with severe skeletal deformities (e.g., abnormal spines and long splayed legs) is characterized by a loss-of-function nonsynonymous T>A transversion (p.700Val>Glu) in the FGFR3 gene [183]. Subsequently, it has been proposed that the decreased expression of FGFR3 increased the proliferation and growth of plate chondrocytes by increasing telomerase reverse transcriptase (TERT) expression levels in vitro, suggesting a potential mechanism for skeletal overgrowth in vivo [184]. In another report, it has been proposed that heterozygosity of the inactivated FGFR3 gene was used to improve skeletal growth [185]. Furthermore, several genomic analyses identified a selective sweep on OAR6 harboring the LCORL locus that is implicated in determining height and stature in sheep [186][187][188][189][190][191][192][193].

Generic notes on other phenotypic traits
Other phenotypic traits in sheep are not yet well understood at the genetic level. Among these are included belly bareness (referring to the naturally bare area of skin on the belly [194,195]), breech bareness (referring to the area of naturally bare skin around the perineum [196][197][198]), skin wrinkles (the folds or creases in the skin [199][200][201]), albinism (the congenital absence of any pigmentation [202,203]), dewlap (a fold of loose skin which hangs from the neck or the throat as found in Persian sheep; Fig. 8a), manes (the growth of long hair on the neck as found in Barbados Blackbelly sheep; Fig. 8h), wattles (skin appendages found in the cervical region [204,205] as found in Madras Red sheep; Fig. 8i), udder conformation [206], and other wool-related traits [207,208].
Other significant traits for which the genes involved are well known have a great influence on sheep phenotype. These include, e.g., MSTN that causes increased body weight by generating the double-muscle phenotype [209][210][211], callipyge that causes the muscular rump phenotype [212][213][214], and SOCS2 that is associated with body weight and size [215,216]. Since these traits are production traits rather than phenotypic traits, we have not presented them in detail in this review. Collectively, it can be stated that all the above-mentioned, well-validated genes and their variants function in an orchestrated manner to give rise to the phenotypic variation among sheep breeds.

Concluding remarks
In the present review, several genes known to be or potentially involved in determining sheep phenotype variation have been discussed. Understanding the genetic basis of phenotypic trait variation in sheep is of great importance, since these are breed determinants and highly linked with production and welfare traits. In spite of the number of genes discussed here, the genetic link and expression patterns of several of these genes are not yet well understood. Further studies are required to identify novel genomic loci and genes for traits for which the genetic determinism has not yet been described. The identification of genes, their causative variants, and their expression characteristics will help to improve sheep breeding through genomic-based selection and modern molecular genetic tools.
Additional file 1: Figure S1. High-throughput sequencing outcomes for candidate genes linked with the sheep tail phenotype pattern. Outcomes of genomic a and transcriptomic b approaches. Figure S2. An example of coat color variation in sheep. Coat colors and markings in Shetland sheep breed. Poster copyright and published with the consent of the Shetland Sheep Society (www. shetl and-sheep. org. uk). Figure S3. CRISPR/ Cas9-generated ASIP knockout Chinese Merino sheep. a Badger-face with black body coat color. b Brown coat color with light ventral pigmentation. c Black-white spotted coat color. d Wild-type phenotype. Figure adapted from Zhang et al. [83]. Figure S4. Phenotypic variations of face coat color in sheep. A representative example from Chinese Tan sheep with varied face coat colors, including white, light brown spots, dark brown spots, brown, black spots, black marks, and black. Photos were taken by P.K. at Ningxia Tianyuan Tan Sheep Farm, Hongsibu, China. Figure S5. Phenotypic variations of nipple (teat) number in sheep. A representative example from Chinese Tan sheep, including a normal nipple number (a pair of nipples) and supernumerary nipples [b with three nipples and c four nipples]. Photos were taken by P.K. Figure S6. Location of vertebrae in the sheep skeleton. This includes cervical, thoracic, lumbar, sacral, and coccygeal vertebrae (item source: PNGITEM).