The mammary gland fulfills the essential role of providing all the nutrients needed to sustain the life and growth of the newborn under the form of milk, a white fluid composed primarily by water, carbohydrates, lipids, proteins and minerals. Domestication of cow, sheep and goats in the Near East 9,000 YBP and the subsequent creation of breeds specialized in milk production allowed humans to take profit of this rich source of proteins and minerals, becoming an important component of their diet either in the form of fresh milk or derived products such as cheese, yogurt, kefir, butter and many others (Figure 1). Because of their adaptability to a harsh climate and scarce vegetation, dairy goats occupy an important niche in the economy of tropical countries such as India, Bangladesh and Sudan, which happen to be the three main goat milk producers at a worldwide scale (FAOSTAT 2009). In Europe, France, Spain and Greece are the largest goat milk producers and, in comparison with Asian and African countries, have a much more intensified production system (FAOSTAT 2009).
Proteins and lipids are essential components of milk and they can have a very strong impact on milk nutritional and technological properties (Bauman et al. 2006). Milk casein content, for instance, is one of the main determinants of cheese yield and both traits are positively correlated (Remeuf and Hurtaud 1991). Similarly, fat content and composition are key factors determining milk and cheese attributes. In this way, milk with a low fat percentage is associated with a reduced cheese yield and firmness as well as with negative effects on flavor and color (Lamberet et al. 2001). Moreover, short-chain fatty acids (FA), such as C4:0-C12:0, have been implicated in the appearance of a rancid soapy flavor in milk, whilst the hardness and melting point of fat is largely determined by its unsaturated FA content (Fox and Sweeney 2003). Importantly, a relevant fraction (around 70%) of goat milk fat is composed by saturated FA that have detrimental effects on human health because they are associated with an increased risk of suffering cardiovascular diseases (Pfeuffer and Schrezenmeir 2000). It is also worth to mention that hydrolysis of milk proteins releases a wide array of short bioactive peptides which might have many beneficial effects on human health, such as (i) hypotensive, antithrombotic, antioxidative, antimicrobial, and immunomodulatory activities (Fitzgerald and Meisel 2004, Korhonen and Pihlanto 2001), (ii) enhancement of mineral absorption (Meisel 1998), and (iii) antitumoral properties (Matar et al. 2003). Milk protein and fat contents vary from breed to breed due to both environmental and genetic factors (Figure 2)
The major protein milk fraction (around 80%) is constituted by caseins, i.e. insoluble phosphoproteins organized in complex multi-molecular aggregates, named micelles, that at certain conditions of temperature and acidity (T ~ 20 ºC, pH = 4.6), precipitate from skim milk (Thompson et al. 1965). Besides caseins, micelles contain inorganic materials such as calcium phosphate and calcium citrate (Smiddy et al. 2006). In ruminants, four casein types have been distinguished so far (Martin et al. 2002): αS1-casein (CSN1S1), αS2-casein (CSN1S2), β-casein (CSN2) and κ-casein (CSN3). CSN1S1, CSN1S2 and CSN2 are considered to be calcium-sensitive caseins because they have a high phosphate content and precipitate in the presence of calcium (Grosclaude 1991). Loci encoding these three caseins have been shown to descend, through successive gene duplication events, from an ancestral locus encoding the secretory calcium-binding phosphoprotein proline-glutamine rich 1 (SCPPPQ1) molecule, that plays a key role in dental enamel mineralization (Kawasaki and Weiss 2003). In contrast, CSN3 is a calcium-insensitive casein responsible of micelle stabilization and whose molecular ancestry is completely different, since it evolved from the follicular dendritic cell secreted peptide (FDCSP) gene, which is expressed in fibroblasts producing periodontal ligament (Grosclaude 1991, Kawasaki and Weiss 2003). The enzymatic hydrolysis of CSN3 by chymosin involves the destabilization of micelles and the subsequent coagulation of milk, which is the initial step in the manufacturing of cheese (Jollès 1975, Remeuf et al. 1991).
Whey proteins represent around 20% of the total protein content of milk and their common denominator is that they remain soluble after milk coagulation (Madureira et al. 2007). Examples of whey proteins are β-lactoglobulin (BLG), α-lactalbumin (LALBA), immunoglobulins, serum albumin, lactoferrin, and lactoperoxidase. Functions of these proteins are very heterogeneous (Madureira et al. 2007), affecting processes as different as lactose synthesis (LALBA), transport of hydrophobic molecules (BLG) and microbial immunity (lactoferrin and lactoperoxidase).
Protein synthesis in the mammary gland depends on the uptake of amino acids from the circulatory system and it is controlled by lactogenic hormones (insulin, prolactin, and glucocorticoids) as well as by the blood concentrations of circulating amino acids (Weekes et al. 2006, Rhoads and Grudzien-Nogalska 2007). Milk protein gene expression is strongly affected by the reproductive cycle, being activated at mid-pregnancy, peaking during lactation and declining after weaning. Generally, the conversion efficiency of dietary nitrogen to milk proteins is relatively poor in the mammary epithelial cells of goats (25-30%) for reasons that remain to be defined (Bequette et al. 1998). The transport of amino acids, and even short peptides, through the membrane of secretory cells is facilitated by a variety of molecular systems that have been broadly reviewed by Shennan and Peaker (2000).
With regard to milk lipids, the most abundant fraction (around 95%) is constituted by triglycerides, whilst the remaining 5% encompasses free FA, phospholipids, cholesterol esters, diglycerides and monoglycerides (Harvatine et al. 2009). A distinctive feature of goat milk is that medium chain FA (e.g. C8, C10 and C12) and C18:1 unsaturated FA are particularly abundant (Fontecha et al. 2000). Milk FA are either synthesized by mammary epithelial cells through the lipogenic pathway or uptaken from the circulating plasma (Bauman and Griinari, 2003). In ruminants, lipogenesis contributes short and medium-chain FA (C4:0 to C16:0) encompassing 50% of the milk FA pool (Barber et al. 1997). Ruminal fermentation provides the main precursors (i.e. acetate and β-hydroxybutyrate) for the de novo synthesis of FA (Van Soest, 1994). In contrast, long-chain FA are mostly obtained through the hydrolysis of blood lipoproteins by lipoprotein lipase (Bauman and Davis 1974). These circulating lipids, in turn, come from the mobilization of adipose tissue stores as well as from dietary lipid absorption in the digestive tract (Bauman and Griinari, 2003).
Functional genomic studies have allowed to dissect the complex network of genes that drive forward fat synthesis in the bovine mammary gland (Bionaz and Loor 2008). This network is composed by genes that participate in a wide variety of metabolic pathways related with FA uptake and transport (e.g. LPL, VLDLR, ACSL1, CD36, FABP3), triglyceride synthesis (e.g. LPIN1, DGAT1, AGPAT6, GPAM), lipid droplet formation (e.g. XDH, BTN1A1), lipogenesis (e.g. ACACA, FASN), FA desaturation (SCD1, FADS1) and activation (ACSL1, ACSS2) and membrane-associated transport of metabolites (ABCG2). The multiple components of the milk fat synthesis machinery defined above are coordinated by diverse transcription factors with a well-known role on lipid metabolism such as SREBF1, SREBF2, PPARG, INSIG1, and PPARGC1A (Bionaz and Loor 2008).
2. Genetic parameters of milk protein and lipid traits in goats
Although casein content is a crucial determinant of cheese yield, its utilization as a selection criterion has been hindered by the fact that this phenotype cannot be easily measured with routine analytical methods. Advances in infrared spectroscopy techniques, however, may overcome this difficulty by providing accurate measurements of the clotting protein fraction at a reasonable cost and time expense (Díaz-Carrillo et al. 1993). Fat content, another important factor determining cheese yield, is routinely recorded and it has been included as an important selection objective in most of breeding programs (Barillet 2007).
Genetic parameters of milk traits under selection need to be accurately defined in order to implement breeding strategies (selection vs. crossbreeding), estimate breeding values and predict selection responses. Many of the heritability and genetic correlation estimates that have been reported in the scientific literature for milk protein and fat contents of dairy goats are listed in Tables 1 and 2. Large parameter ranges can be observed, a feature that might be likely explained by the fact that multiple breeds and methods have been used to estimate them. Another important difference amongst studies is the lactation time point at which phenotypes are obtained, an environmental factor that can have dramatic effects on genetic parameter estimation. With regard to the heritability of total casein and casein fraction contents, very few estimates have been reported so far. Worth to mention those obtained for casein content in the Alpine (h2 = 0.65-0.66), Murciano-Granadina (h2 = 0.13-0.19) and Malagueña (h2 = 0.29) breeds (Ricordeau and Bouillon 1971, Sigwald et al. 1981, Benradi et al. 2007 and 2009). It is also necessary to highlight the study of Benradi et al. (2007), where heritabilities of CSN1S1 (h2 = 0.25) and CSN1S2 (h2 = 0.09) contents were estimated in the Murciano-Granadina breed.
|Saanen (Norway)||-||0.28||Ronningen (1965)|
|Saanen (Norway)||-||0.40-0.59||Ronningen (1967)|
|Alpine (France)||0.59||0.62||Ricordeau et al. (1979)|
|-||0.54||Kennedy et al. (1982)|
|Alpine (France)||0.52||0.50||Boichard et al. (1988)|
|Alpine (France)||0.49-0.53||0.46||Bouloc (1987)|
|Alpine (France)||0.67||0.56||Bouillon & Ricordeau (1975)|
|La Mancha (USA)||-||0.63||Iloeje et al. (1981)|
|Nubian (USA)||-||0.66||Iloeje et al. (1981)|
|Toggenburg (USA)||-||0.54||Iloeje et al. (1981)|
|Saanen (France)||0.41||0.47||Boichard et al. (1988)|
|Saanen (France)||0.42||0.42||Bouloc (1987)|
|Saanen (México)||0.38-0.63||0.32-0.64||Torres-Vázquez et al. (2009)|
|Saanen (South Africa)||0.44||0.21||Muller et al. (2002)|
|Alpine (France)||0.58||0.58||Bélichon et al. (1998)|
|Saanen (France)||0.50||0.60||Bélichon et al. (1998)|
|Local breeds (Greece)||0.51||0.38||Zygoyiannis (1994)|
|Murciano-Granadina (Spain)||0.25-0.47||-||Analla et al (1996)|
|Alpine (France)||0.66-0.85||0.58-0.77||Barbieri et al. (1995)|
|Verata (Spain)||0.14-0.42||0.30-0.32||Rabasco et al. (1993)|
|Murciano-Granadina (Spain)||0.30||-||Benradi et al. (2009)|
|Murciano-Granadina (Spain)||0.50||0.15||Benradi (2007)|
|Malagueña (Spain)||0.30||-||Benradi (2007)|
The main trend that emerges from these analyses is that protein, casein and fat contents have moderate heritabilities, so they can be improved at a reasonable pace by using classical selection. These traits show, in general, positive medium or high genetic correlations among themselves; being, consequently, also positive their respective correlated responses to selection. However, they display negative genetic correlations with milk yield, a circumstance that is quite unfavourable given the high economic impact of this phenotype. In spite of this, protein and fat contents are important selection criteria and they are frequently included in selection indexes (Manfredi et al. 2000; Montaldo and Manfredi 2002)
Another interesting issue that deserves to be discussed is the effect that the highly polymorphic CSN1S1 gene has on the estimation of genetic parameters of milk traits in goats. In this way, Barbieri et al. (1995) reported that heritability of protein content was reduced from 0.66 to 0.34 and that the genetic correlation between protein yield and content dropped from 0.09 to -0.22 when the CSN1S1 genotype was taken into account as a fixed effect. This would mean that a substantial part of the genetic variance of goat protein content is explained by the polymorphism of the CSN1S1 gene.
An important methodological advance in the estimation of genetic parameters of dairy traits has consisted in the use of random regression models applied to test day records (Schaeffer, 2004)). This approach has allowed to obtain estimates of heritabilities at, and genetic correlations among, different timepoints throughout the lactation curve in a wide variety of populations such as Norwegian dairy goats (Andonov et al. 1998), Spanish Payoya and Murciano-Granadina goats (Menéndez-Buxadera et al. 2008, 2010) and Canadian dairy breeds of Alpine origine (Bishop et al. 1994). To illustrate this concept, evolution of variance components for milk yield, fat, protein and dry matter contents in Murciano-Granadina goats is shown in Figure 3. From these data it can be inferred that estimates of heritability of yields and contents of milk components do not have stable values but, on the contrary, they vary throughout the lactation curve, being more variable near parturition and drying off. Moreover, genetic correlations between adjacent records are much higher (0.70-0.99) than those between records far apart (0.00-0.40). With no doubt, this statistical methodology allows a better genetic evaluation of dairy goats anf facilitates the selection of improving genotypes for lactation persistency (the ability of a goat to maintain as high as possible milk daily yield during lactation).
Genetic correlations between milk components and rheological traits as well as cheese yield have been also studied by Benradi et al. (2009). These authors reported high positive genetic correlations of protein, total caseins and CSN1S1 contents with time to curdling onset, curd firmness and curdling speed and a moderate positive correlation of the aforementioned milk components with cheese yield, so confirming the important influence of protein, and especially of total caseins and CSN1S1 contents, on the efficiency of cheese manufacture and quality.
3. Genomic architecture of protein and lipid phenotypes in goats and other ruminants
The lack of appropiate molecular tools in goats, such as large microsatellite panels uniformly covering the goat genome, microarrays and high throughput genotyping SNP chips, has indered the fine mapping of genes related with milk protein and lipid traits at a genomic scale. While dense quantitative trait loci (QTL) maps of dairy traits have been obtained in cattle and sheep, a single partial genome scan for milk yield and fat and protein contents has been performed so far (Roldán et al. 2008). However, many of the findings obtained in cattle can be probably extrapolated to goats, so they will be commented in the following paragraphs. First of all, it should be highlighted that relative chromosomal contributions to protein and fat genetic variance are quite uneven. In cattle, strong evidences of protein content QTL have been obtained on chromosomes 3, 6 and 20, whilst protein yield QTL map to bovine chromosomes (BTA) 1, 3, 6, 9, 14 and 20 (Khatkar et al. 2004). Khatkar et al. (2004) have also reported QTL hotspots for milk fat content and yield (BTA6 and BTA 14). Second, many of the reported protein and fat content QTL just segregate in specific populations or families, meaning that they have a very restricted distribution (Khatkar et al. 2004). Third, and with a few exceptions, the identity of the genes and mutations explaining these QTL remain unsolved. The BTA14 QTL is one of the few cases where the underlying causal polymorphism has been identified i.e. a non-conservative amino acid substitution (K232A) in the DGAT1 enzyme that explains 51% of the daughter yield deviation variance of fat percentage (Winter et al. 2002, Grisart et al. 2002). The lack of success in finding causal mutations underlying milk trait QTL might be partly explained by the low resolution of microsatellite-based QTL mapping (confidence intervals usually encompass 20 cM or more).
Low resolution QTL detection methods have been (cattle, sheep) or will be (goats) progressively replaced by high throughput single nucleotide polymorphism (SNP) genotyping platforms. In this way, the recent advent of a bovine 50K SNP BeadChip has allowed to perform genome-wide association analyses of dairy traits at an unprecedented resolution (Hayes and Goddard 2010). As an example of the power of this approach, Schopen et al. (2011) analysed the segregation of 50,228 SNP in 1,713 Dutch Holstein cows with records for milk CSN1S1, CSN1S2, CSN2, CSN3, LALBA and BLG contents. Their results showed highly significant associations between polymorphisms mapping to BTA5, 6, 11 and 14 and protein percentage. With regard to specific protein fractions, the number of associated regions varied from three (CSN2 and BLG) to 12 (CSN1S2) and the percentage of additive genetic variance explained by the joint sets of significant SNP oscillated between 25% and 35%. Strong associations were observed between SNP mapping to casein genes and the corresponding casein fraction contents, but evidences of SNP acting in trans were also obtained e.g. SNP with effects on CSN1S1 content were detected in BTA11 and BTA14 (whilst the CSN1S1 gene resides on BTA6). As a general conclusion, valid not only for cattle but also for sheep or goats, we can state that casein content is regulated not only by polymorphisms located within the casein genes but also by mutations mapping to other loci and chromosomes.
In goats, the precise position of genes with effects on milk lipid or protein composition or their contributions to phenotypic variance are mostly unknown. One of the few exceptions is the caprine casein gene cluster, which is known to have strong effects on milk composition and that has been finely dissected at the genomic and sequence levels (Martin et al. 2002). The four caprine casein genes are tightly clustered in a 250 kb region mapping to chromosome 6, in the order CSN1S1-CSN2-CSN1S2-CSN3 (Rijnkels et al. 2002). Genomic organization and gene structure are highly conserved in all mammals (Rijnkels et al. 2002), with the only exception of the CSN1S2 gene that might contain from 11 to 19 exons depending on the species under consideration (Rijnkels et al. 2002). Moreover, CSN1S2 is duplicated in human and rodents but not in ruminants.
The goat CSN1S1 gene encodes a 199 amino acid protein and it was completely sequenced by Ramunno et al. (2004). This gene is 16.7 kb long and contains 19 exons with sizes that go from 24 to 385 bp. The signal peptide and the first two amino acids of the mature protein are encoded by exon 2, whilst the stop codon is located at the boundary of exons 17 and 18. Leroux and Martin (1996) demonstrated that the goat CSN2 gene is located 12 kb apart from the CSN1S1 gene and that both are convergently transcribed (Leroux and Martin 1996). The transcription unit of CSN2 contains 9 exons and encodes a 1.09 kb mRNA. The translation initiation site is located at exon 2, which encodes the signal peptide plus two amino acids of the mature protein (Roberts et al. 1992). The caprine CSN1S2 and CSN3 genes have been also characterized at the molecular level by Bouniol et al. (1993) and Coll et al. (1993).
The whey protein genes LALBA and BLG have been mapped to goat chromosomes 5 (Hayes et al. 1993) and 11 (Folch et al. 1994), respectively. The LALBA gene contains four exons and encodes a 123 amino acid protein. Structural characterization of the goat BLG gene has shown that it contains seven exons as previously described in other ruminant species (Folch et al. 1994). Two short interspersed nucleotide elements were found in the 3’end of the gene Moreover, a duplicated goat BLG pseudogene with a genomic organization very similar to BLG and also mapping to caprine chromosome 11 has been identified by Folch et al. (1996).
4. Candidate genes and their association with milk protein traits
From a structural and functional point of view, the casein cluster is one of the best studied genetic systems in goats. As outlined above, casein genes have been sequenced and their variability has been characterized in depth (Martin et al. 2002, Moioli et al. 2007). Even more, in several cases consistent association between this variability and milk composition traits has been firmly established (Table 3). In this regard, the most paradigmatic case is represented by the CSN1S1 gene, where causal relationships between regulatory polymorphisms and CSN1S1 synthesis rate have been found. In the following sections, we will describe the variability of the casein and whey protein genes and its impact on the phenotypic variation of dairy and rheological traits.
4.1. The caprine αS1-casein gene
This locus is highly polymorphic in goats (reviewed in Martin et al. 2002 and Moioli et al. 2007), with 17 alleles (Table 3) that can be classified as strong (A, B1, B2, B3, B4, C, H, L and M ), medium (E and I), low (F, D and G) and null (01, 02 and N). The existence of this remarkable level of variability was firstly outlined by Boulanger et al. (1984). By means of starch gel protein electrophoresis, these authors evidenced the existence of one variant A, two B variants (associated with different electrophoretic band intensities) and one variant C. These results suggested the existence of a polymorphism with quantitative effects on CSN1S1 synthesis. This interpretation was subsequently confirmed by Grosclaude et al. (1987), who identified two additional alleles (F and 0) and provided a first estimate of quantitative differences amongst CSN1S1 genotypes based on rocket immunoelectrophoresis i.e. strong, medium, low and null alleles were distinguished (Table 3). These two studies have become classics in goats genetics because they pionereed the discovery of genetic variants with effects on milk traits in this ruminant species.
The advent of DNA-based methods allowed to characterize the specific mutations causing this variability as well as to identify new variants not detectable through electrophoresis techniques. Strong variants A, B and C were shown to differ by several amino acid substitutions i.e. B vs A: P16L and E77N and B vs C: H8I, R100K and T195A, but none of these polymorphisms seemed to have quantitative effects. Protein variants H and L were identified and characterized by Chianese et al. (1997) by using a variety of proteomic techniques, whilst variant M was reported by Bevilacqua et al. (2002). The main feature of the M variant is the loss of two phosphate residues in the multiple phosphorylation site consecutively to a S66L substitution. Likely, this allele emerged as a result of an interallelic recombination event between A and B2 alleles followed by a C to T transition at exon 9 (Bevilacqua et al. 2002).
(g casein /L/allele)
|CSN1S1||A, B1, B2, B3, B4, C, H, L, M||3.5|
|D, F, G||0.45|
|01, 02, N||0|
|CSN1S2||A, B, C, E, F||2.5|
|CSN2||A, A1, B, C, D, E||5|
Variants E and I have been associated with an intermediate level of CSN1S1 synthesis (Boulanger 1984, Pérez et al. 1994, Chianese et al. 1997). Extensive sequencing of the E-allele revealed the existence of a 457 bp insertion at exon 19 caused by a truncated long interspersed nucleotide element (Pérez et al. 1994). This insertion might destabilize the corresponding CSN1S1 mRNA diminishing 3-fold the synthesis of the corresponding protein (Pérez et al. 1994). Interestingly, the bovine CSN1S1 allele G, that also contains a retrotransposon insertion at exon 19, has been associated with a lower milk CSN1S1 concentration (Rando et al. 1998). These results clearly indicate that the retrotransposon insertion is the causal mutation explaining the reduced levels of CSN1S1 in milk. However, the exact molecular mechanism by which the retrotranposon insertion represses CSN1S1 synthesis has not been elucidated yet, although a RNA interference mechanism might be suspected.
Low synthesis of CSN1S1 is explained by a defective processing of the corresponding transcript due to mutations that promote exon-skipping events and, in consequence, result in internally deleted CSN1S1 proteins (Martin et al. 2002). As much as nine different transcripts seem to be associated with allele F, the most abundant of which lacks exons 9, 10 and 11 and provokes a 37 amino acid deletion encompassing the multiple phosphorylation site (Leroux et al. 1992). A single nucleotide frameshift deletion in exon 9 (that induces the appearance of a premature stop codon at exon 12) and two 11 and 3 bp insertions at intron 9 might explain the defective processing of allele F. Similarly, the D allele is characterized by skipping of exon 9, while the G allele displays a G to A mutation in the 5’ splice site consensus sequence of intron 4 that causes the skipping of exon 4 and the synthesis of a protein lacking amino acids 14 to 26 (Martin et al. 1999).
The complete absence of CSN1S1 in milk is explained by a couple of genetic mechanisms. In the case of the 01 allele, a genomic deletion of at least 8.5 kb, that encompasses intron 12 to exon 19 of the CSN1S1 gene, abrogates the synthesis of the corresponding protein (Cosenza et al. 2003). In close similarity with the F variant, the N allele contains a 1 bp deletion at the 23rd site of exon 9 determining a premature stop codon at exon 19. Sequencing of RT-PCR clones revealed that this variant is represented by at least 12 different transcripts lacking combinations of exons 9, 10, 11, 16 and 17 as a result of the defective processing of the mRNA. The abundances of transcripts carrying a premature stop codon are 30% and 14% for the N and F alleles, respectively (Ramunno et al. 2005). This finding might explain why the synthesis rate of the N-allele is 3-fold lower than that of the F-allele
Genotyping techniques have been developed to characterize the polymorphism of the CSN1S1 gene in diverse goat populations. Initially, CSN1S1 variants were typed through the analysis of milk samples by SDS-polyacrylamide gel electrophoresis combined with isoelectric focusing (Grosclaude et al. 1987). The main inconvenient of this approach was that only lactating females could be typed, whilst in breeding schemes the main contributors to genetic improvement are bucks. To circumvent this problem, molecular techniques were developed. The first one was based on Southern blotting and restriction fragment length polymorphims analysis (Leroux et al. 1990). Although useful, this approach was very time consuming and not applicable to the large throughput genotyping required in breeding schemes. Later on, PCR-based techniques were published (Pérez et al. 1994, Ramunno et al. 2000) allowing the fast genotyping of the most abundant CSN1S1 alleles.
With the aid of these molecular tools, the segregation of CSN1S1 alleles has been studied in a wide array of breeds. Estimation of allelic frequencies in the Spanish Murciano-Granadina and Malagueña breeds showed that the E-allele was the most frequent one, followed by the B variant (Jordana et al. 1996). In contrast, in the French and Italian Saanen and Alpine breeds as well as in the French Corse breed the low content F-allele was predominant, while the E-allele would rank second (reviewed in Trujillo et al. 1998). It should be taken into account, however, that these estimates are quite outdated and that selection for CSN1S1 variants might have changed their frequencies dramatically (at least in French breeds). In African, Canarian, Maltese and Garganica breeds, strong CSN1S1 content alleles are the most frequent ones (reviewed in Trujillo et al. 1998). A recent survey of American goat breeds highlighted the coexistence of different allelic CSN1S1 frequency patterns, with breeds in which the F (e.g. Alpine), E (e.g. Saanen and Oberhasli) and A+B alleles (e.g. LaMancha, Nigerian dwarf and Nubian) were predominant (Maga et al. 2009). These differences in the frequencies of CSN1S1 variants might be explained by a combination of effects produced by genetic drift, selection and other evolutionary and demographic factors.
There is substantial evidence that the aforementioned polymorphisms not only affect the synthesis rate of CSN1S1 but also a wide array of production traits. A within-sire analysis of the progeny of five Alpine bucks revealed significant effects of CSN1S1 genotype on milk protein and fat content, with the A-allele showing a clear superiority over E and F (Mahé et al. 1994). A similar trend was observed by Manfredi et al. (1995) when surveying 184 Alpine and 96 Saanen bucks. Moreover, Barbieri et al. (1995) demonstrated that the A-allele is associated with higher protein (AA > AE, AF > EE, EF > FF) and fat (AA, AE, AF > EE, EF) contents but also with a lower milk yield. Differences in protein content between genotypes might be in the order of 4 g/l (AA vs EE) to 6 g/l (AA vs FF). Chanat et al. (1999) offered a biological explanation to these findings by demonstrating that CSN1S1 genotype has relevant effects on casein transport from the endoplasmic reticulum to the Golgi compartment (in goats with low or null CSN1S1 genotypes this transport is severely impaired). An important question is if results obtained in French breeds can be safely extrapolated to breeds from other countries. In fact, milk composition is affected by many genetic and environmental factors that might differentially modulate the effects of goat CSN1S1 genotype depending on the population under consideration. Results obtained in Spanish goat breeds are consistent with this hypothesis. Whilst significant differences were observed in the Malagueña breed when comparing milk CSN1S1 concentrations in BB (6.94 ± 0.38), BF (5.36 ± 0.22), EE (4.58 ± 0.13) and FF (3.98 ± 0.27 g/l) goats, in the case of the Murciano-Granadina breed only the BB genotype (8.50 ± 0.60 g/l) was significantly associated with increased levels of CSN1S1, whereas BF, EE and EF genotypes displayed non-significant differences when compared with each other (Caravaca et al. 2008). Even more, the CSN1S1 genotype did not display any significant association with protein, casein or fat content (Caravaca et al. 2009). These results suggest that CSN1S1 genotype has significant and consistent effects on the synthesis rate of the corresponding protein, but associations with other milk components might vary from breed to breed, likely due to differences in their genetic backgrounds and production systems. Interestingly, recent data suggest a certain level of dominance of strong alleles over the weak ones (Berget et al. 2010). In this way, CSN1S1 expression of goats carrying one strong and one weak allele at the CSN1S1 locus is much more similar to those with a strong homozygous genotype than to goats with a weak homozygous genotype (Berget et al. 2010). Noteworthy, in cattle most of genetic correlations between casein fractions are negative or null (Schopen et al. 2009). As a whole, these findings suggest the existence of complex inter-loci and intra-locus interactions between casein genes that might impact their relative contributions to phenotypic variance of milk composition.
Not only fat content but milk FA composition has been reported to be affected by CSN1S1 genotype. In this way, low-content CSN1S1 alleles have been associated with less C8-C12 saturated FA, less stearic acid and more palmitic, linoleic and rumenic acids than their high-content counterparts (Chilliard et al. 2006). Low-content alleles have also been linked to an increased mammary desaturase activity (Chilliard et al. 2006). There is a certain controversy about the influence of CSN1S1 polymorphism on the mRNA expression of lipid metabolism genes, with studies that support a regulatory effect (Badaoui 2008) and others that do not (Leroux et al. 2003). Since lipids and proteins are synthesized in the endoplasmic reticulum and their transport is, to a certain extent, coupled and co-regulated, it has been hypothesized that the perturbation of casein transport induced by the CSN1S1 genotype might also alter lipid trafficking resulting in a reduced fat secretion (Ollivier-Bousquet et al. 2002).
Another trait influenced by CSN1S1 genotype is micelle size, that happens to be lower in AA (221 nm) than in EE (265 nm) or FF (268 nm) milks (Remeuf 1993, Pirisi et al. 1994). This feature together with an augmented global protein content might explain the better coagulation properties and increased cheese yield of the AA milk (Ambrosoli et al. 1988, Vassal et al. 1994). In this way, AA milk produces a firmer curd and displays a slower coagulation time than FF milk (Ambrosoli et al. 1988). Moreover, corrected cheese yield (kg of cheese obtained from 100 kg of milk) was around 21-23 kg for AA, 20 kg for EE and 18 kg for FF goats (Vassal et al. 1994). These associations, however, may change depending on the breed under consideration. In this regard, Caravaca et al. (2011) were unable to find significant differences between the cheese yields of milks from BB, EE and FF Murciano-Granadina goats, whilst EE milk had a significantly higher curdling rate than its BB counterpart. As mentioned above, these differences amongst studies might be explained by a complex mixture of biological and technical factors.
From a sensorial point of view (Vassal et al. 1994), the AA cheese has been reported to display a higher hardness than the FF one (score of 3.23/5 vs 2.85/5), but a weaker goat flavor intensity (score of 2.10/5 vs 2.02/5). This means that the AA milk has better technological properties than the FF one in order to produce cheese but, unfortunately, the resulting product has a less intense taste and odour. It can be speculated that CSN1S1 effects on cheese flavor might be caused by differences in the FA content and composition (goat flavor is mostly explained by the presence of volatile branched-chain 4-methyl and 4-ethyl octanoic FA) as well as in the lipolysis rate of AA vs FF milks (Chilliard et al. 2003, 2006),
4.2. The caprine αS2-casein gene
Currently, five variants encoding “normal” levels of CSN1S2 have been found (Table 3), i.e. A and B (Boulanger et al. 1984), C (Bouniol et al. 1994), E (Veltri et al. 2000, Lagonigro et al. 2001) and F (Ramunno et al. 2001a). Two other variants D and 0 linked to reduced concentrations of CSN1S2 have also been detected (Ramunno et al. 2001a,b). The main feature of the D allele is a 106-nucleotide deletion, starting from the last 11 nucleotides of exon 11 and including 95 bp of intron 11, which causes the loss of three amino acid residues i.e. Pro122, Thr123, and Val124. Ramunno et al. (2001a) proposed that this variant might be associated with an intermediate level of CSN1S2 synthesis, but evidence is still preliminar and needs to be confirmed. The 0 allele contains a non-sense G>A mutation at exon 11 that changes the codon TGG (coding for Trp110) into a TAG stop codon (Ramunno et al. 2001b), thus hindering the synthesis of the corresponding protein. PCR-RFLP and PCR-SSCP protocols have been implemented in order to characterize CSN1S2 variation (Ramunno et al. 2001a,b, Chessa et al. 2008) but, to the best of our knowledge, association studies with milk and cheese traits are still lacking.
4.3. The caprine β-casein gene
Alleles at the CSN2 locus can be classified in two main categories depending on the synthesis rate they are associated with (Table 3) i.e. alleles that are associated with “normal” concentrations of CSN2 in milk (A, A1, B, C, D, E) and the null ones (0 and 01) in which CSN2 expression is completely abrogated. Cosenza et al. (2007) have also reported a single nucleotide polymorphism at the CSN2 promoter but this variant has not been named yet. With regard to the A, B, C, D and E variants, they have been identified by means of proteomic techniques such as isoelectric focusing (Mahé and Grosclaude 1993), peptide mass fingerprinting and tandem mass spectrometry (Neveu et al. 2002), reversed phase HPLC/electrospray ionization mass spectrophotometry (Galliano et al. 2004) and immunoelectrophoretic analysis (Chianese et al. 2007). Molecular characterization of the null variants revealed that the 0 allele contains a 1 bp deletion in the 5’end of exon 7 that induces the appearance of a premature stop codon resulting in a 20-fold reduction in mRNA synthesis and the generation of a much shorter protein (72 vs 223 amino acids, Persuy et al. 1999). This variant has been found in Creole and Pyrenean goats (Persuy et al. 1999). Similarly, Ramunno et al (1995) identified a null 01 variant with a substitution at position 373 of exon 7 that converts a triplet encoding glutamine into a stop codon. This event results in a 10-fold reduction in CSN2 mRNA synthesis and yields a non-functional protein truncated at position 181 (Ramunno et al. 1995). PCR-SSCP (Chessa et al. 2005, Caroli et al. 2006), allele-specific-PCR (Ramunno et al. 1995), and PCR-RFLP (Cosenza et al. 2005) methods have been developed to identify genetic variants A, A1, C, E, and 0 at the DNA level.
4.4. The caprine κ-casein gene
The caprine CSN3 gene is extraordinarily polymorphic with 16 alleles (A, B, B′, B″, C, C′, D, E, F, G, H, I, J, K, L, and M) identified to date (Yahyaoui et al. 2003, Jann et al. 2004, Prinzenberg et al. 2005). Interestingly, most of the detected genetic variation is non-synonymous and preliminar evidence of positive selection acting on this locus has been found (Prinzenberg et al. 2005, Clop et al. 2009). This high genetic variation can be characterized with diverse molecular techniques, although multiplexed primer-extension analysis is particularly suitable in order to detect all alleles (Yahyaoui et al. 2003). The two most abundant CSN3 alleles are A and B, while the remaining ones can be considered as low-frequency variants in the majority of caprine breeds. Interestingly, Caravaca et al. (2010) have reported that the CSN3 genotype has significant effects on casein and protein contents (BB, AB > AA). Similar results have been obtained in the Orobica breed (Chiatti et al. 2007), where isolectric focusing variant B was associated with higher protein and casein contents than its A counterpart. The molecular basis of these associations is unclear because, as far as we know, the A and B variants only differ by a I119V substitution (Yahyaoui et al. 2003). In silico analyses with the Polyphen software have shown that in most mammalian species, Ile119 is a highly conserved residue, suggesting that it might have an important functional role (Caravaca et al. 2010). It is also worth to mention that associations between CSN3 genotype and rennet coagulation time (BB>AB) have been recently reported (Caravaca et al. 2011). This finding is consistent with the key role of CSN3 in the initial phase of milk rennet coagulation, where 87–90% of CSN3 is enzymatically degraded before micellar aggregation takes place. However, it should be emphasized that CSN3 genotype did not affect cheese yield, one of the main factors determining the economic income of goat farmers (at least in Spain and other Mediterranean countries). In consequence, using CSN3 genotype in marker-assisted selection schemes might not be advisable if cheese yield is a major breeding goal.
4.5. Casein haplotypes and their association with milk traits
The most meaningful approach to investigate the effect of casein genes on milk composition implies the genotyping of haplotypes rather than individual locus-specific alleles. However, this methodological strategy has been rarely carried out, probably because of the technical challenge of simultaneously typing such highly polymorphic loci in a fast and reliable way. Casein haplotypes have been characterized in a number of breeds from Italy (Sacchi et al. 2005, Caroli et al. 2006, Finocchiaro et al. 2008, Gigli et al. 2008), Norway (Hayes et al. 2006, Finocchiaro et al. 2008, Berget et al. 2010), Germany (Küpper et al. 2010), Czech Republic (Sztankóová et al. 2009), West Africa (Caroli et al. 2007) and India (Rout et al. 2010). However, most of these studies just report the variability of the casein cluster in selected populations rather than analysing its impact on milk quality phenotypes. An exception to this general statement is the work performed by Hayes et al. (1996). These authors genotyped 436 goats for 39 SNP distributed in the casein loci. They found higher levels of linkage disequilibrium between SNP pairs within casein loci than between casein loci, meaning that levels of intragenic recombination in casein genes are somewhat low. Moreover, they found significant associations between CSN1S1 haplotypes and protein percentage and fat yield, as well as between CSN3 haplotypes and fat and protein percentages (Hayes et al. 1996). In the next future, extensive characterization of the variability of the casein cluster region with next generation sequencing techniques, construction of large casein SNP panels and genotyping with high throughput platforms will be instrumental to elucidate the influence of casein haplotypes on dairy traits.
4.6. The caprine α-lactalbumin and β-lactoglobulin genes
The LALBA gene has been poorly characterized in goats with a few polymorphisms described to date. Cosenza et al. (2003) reported a silent SNP at exon 3 that can be analysed by PCR-RFLP (MvaI). Another synonymous mutation has ben found at exon 1 (Ma et al. 2010). However, to the best of our knowledge none of these polymorphisms has been associated with milk traits in goats. With regard to the BLG locus, most of the polymorphisms that have been found so far lie at the promoter region. In this regard, Yahyaoui et al. (2000) reported a C>T change at position -60, while Ballester et al. (2005) found 9 SNP at the promoter region and 6 silent SNP at the coding region (exons 1,2,3 and 6). Pena et al. (2000) also found two SNP at exon 7 encoding the 3’UTR. In close resemblance with results contributed by Ballester et al. (2005), Sardina et al. (2011) reported extensive variability at the BLG promoter, with a total of 36 SNP identified in a panel of Sicilian goats. Several of the SNP identified by Ballester et al. (2005) and Sardina et al. (2011) have been mapped to potential transcription factor binding sites, so they might be good candidates to regulate BLG mRNA expression. Pending tasks are to investigate the existence of quantitative differences (in terms of BLG mRNA) between promoter alleles as well as to find out if they have a detectable influence on milk composition.
5. Candidate genes and their association with milk fat content and composition traits
The detection of electrophoretic variants of goat milk protein genes three decades ago gave a strong impetus to the identification of the underlying mutations and the performance of association analyses with milk traits. This has resulted in the establishment of a wide catalog of polymorphisms located in the casein and whey protein genes, several of which have well demonstrated causal effects on milk composition. In comparison, the study of the genetic basis of milk fat traits is much less advanced. So far, a reduced number of candidate genes have been characterized at the molecular level and associations with milk fat content and/or composition have been reported (Table 4). However, causality has not been demonstrated for any of these associations, that most likely are produced by the existence of linkage disequilibrium between the analysed SNP and the true causal mutation. This contrasts with results obtained in cattle, where causal effects have been proposed (and sometimes convincingly demonstrated) for polymorphisms located at the DGAT1 (Grisart et al. 2002, Winter et al. 2002), ABCG2 (Cohen-Zinder et al. 2005) and PPARGC1A (Weikard et al. 2005) genes. An important difference between studies performed in cattle and goats is that in the latter species candidate genes were exclusively selected on the basis of physiological criteria, since positional information (e.g. QTL landscape of traits under study) was not available. This is an important limitation that has severely hindered progress in goat genetics research
The acetyl-CoA carboxylase α (ACACA) enzyme catalyses the carboxylation of acetyl-CoA to form malonyl-CoA, that can be used by fatty acid synthase as a substrate. This is a key rate-limiting step in the synthesis of FA (Abu-Elheiga et al. 1997). Sequence analysis of 5.5 kb of the coding region of the caprine ACACA gene revealed a silent SNP at exon 45 that was suggestively associated with fat yield and other traits (Badaoui et al. 2007a). Moreover, Federica et al. (2009) found 3 SNP at promoter III of the caprine ACACA gene that map to putative transcription factor binding sites but none of them displayed significant associations with lipid traits. Lipoprotein lipase is another fundamental enzyme involved in FA release and absorption through the hydrolysis of triglycerides from chylomicrons and other lipoprotein particles (Olivecrona and Olivecrona 1998). One missense polymorphism involving a S17T change has been suggestively associated with milk fat content (Badaoui et al. 2007b). This polymorphism is located in the signal peptide so it has been hypothesized that it might alter protein localization or expression. Variability at the goat milk fat globule epidermal growth factor and butyrophilin genes has also been associated with milk fat yield (Qu et al. 2011).
Zidi and coworkers (2010a,b,c,d) pionereed the study of the genetic basis of milk FA composition in goats through the analysis of several candidate genes and the performance of association analyses. As said in previous sections, milk FA composition constitutes an important set of traits with a key influence on the nutritional and technological properties of milk. Results obtained by Zidi and colleagues are summarized in Table 4. Worth to mention the identification of a 3-bp indel in the 3’UTR of caprine stearoyl-CoA desaturase (SCD1) gene, previously reported by Bernard et al. (2001), that was suggestively associated with conjugated linoleic acid and polyunsaturated FA content. This deletion is predicted to cause a dramatic change in the secondary structure of the 3’UTR so it has been hypothesized that it might exert its effect by influencing mRNA stability. In cattle, polymorphism at the SCD1 gene has been associated with milk FA composition by several authors (Taniguchi et al. 2004, Schennink et al. 2008, Kgwatalala et al. 2009). A significant advancement in the dissection of the genetic factors that regulate milk fat content and composition in goats will necessarily involve the use of high throughput genotyping tools, such as the Illumina BeadChip that will be soon available, to type large goat populations with multiple records for these traits (throughout the lactation and/or successive lactations). This approach would allow to identify the genomic regions influencing milk lipid phenotypes, and then the daunting task of finding the causal mutations might begin with reasonable prospects of being successful.
|Acetyl coenzyme A carboxylase α (ACACA)||C5493T in exon 45||fat yield, lactose content, and somatic cell count||Badaoui et al. (2007a)|
|1206 pb C/T at promoter III (locus AJ292286)||fat and protein percentages||Federica et al. (2009)|
|1322 pb T/C at promoter III (locus AJ292286)||percentage, and fat and protein yields|
|Growth hormone (GH)||SSCP patterns in exons 2, 4 and 5||milk, fat, and protein yields||Malveiro et al. (2001)|
|Lipoprotein lipase (LPL)||G50C (Ser17Thr)||milk fat content||Badaoui et al. (2007b)|
|Stearoyl Co-A desaturase 1 (SCD1)||c.*1902_1904delTGT|
|trans-10, cis-12 CLA, PUFA, and total CLA||Zidi et al. (2010a)|
|Malic enzyme 1 (ME1)||c.483C"/>T||C16:0 C17:0, C18:1n-9c, C20:1, SFA, MUFA and performed fatty acids.||Zidi et al. (2010b)|
|c.667G"/>A||C18:1n-9t, and C17:0, C18:0, C18:1n-9c, trans-10, cis-12 CLA, C20:1 and total CLA.|
|c.1200G"/>A||total CLA, and C17:0, C17:1, C18:0, C18:1n-9t, C18:1n-9c, cis-9, trans-11 CLA, and trans-10, cis-12 CLA.|
|Hormone sensitive lipase (LIPE)||c.327C"/>A"/>T||C12:0 FA, C15:0 and de novo FA||Zidi et al. (2010c)|
|c.558C"/>T||fat content, trans-10, cis-12 CLA|
|Prolactin receptor (PRLR)||c. 1201G"/>A (R401G)||C16:1 FA and C16:1, C18:2n6c FA and PUFA||Zidi et al. (2010d)|
|c.1355C"/>T (T452I)||C16:1, C18:1n9t FA, SFA, MUFA, and omega 3|
Milk protein and fat content and composition are key determinants of the nutritional and technological quality of goat dairy products as fresh milk, cheese and yogurt. Classical quantitative genetic studies have demonstrated that there is a remarkable amount of additive genetic variance for these dairy phenotypes, prompting the search of the causal mutations that explain the observed variability. These investigations have been particularly succesful when studying the genetic basis of casein concentrations in milk, since causal mutations have been identified in the goat CSN1S1 and relevant associations have been found for CSN3. These findings have allowed the implementation of marker assisted selection schemes to improve milk quality in goats. Milk fat related phenotypes have been much less studied, although we can anticipate that the development and application of high throughput genotyping and sequencing methods will revolutionize the field in the near future.