Milk protein polymorphisms and casein haplotypes in Blanco Orejinegro cattle of Colombia

The aim was to determine the genetic variation in the CSN1S1, CSN2, CSN1S2, CSN3 and LGB genes in Blanco Orejinegro cattle. 419 animals from 15 herds were genotyped with GGP Bovine 150 K (n= 70) and 50 K (n= 349) chips. Information was obtained from 43 SNPs in the mentioned genes and protein variants *B, *C and *D of α S1 -CN ; *A 1 , *A 2 , *B , *H 2 and *F of β-CN ; *A and *D of α S2 -CN, *A , *A 1 , *B , *I and *H of κ-CN and *A , *B , *C , *D , *E , *F and *H of β-LG were reconstructed. Allele and genotypic frequencies were estimated for SNPs and for protein variants; Hardy-Weinberg equilibrium and F ST values were evaluated for each of the SNPs under different structuring criteria. LD values and haplotypic frequencies were estimated for caseins. The most frequent variants were CSN1S1*B (0.804), CSN2*A 2 (0.509), CSN1S2*A (0.997), CSN3*A (0.679) and β-LG*B (0.657). None of the variants showed deviations from HWE, but the CSN2*A 2 allele showed a slight increasing trend over time. The F ST values were low (0.035) regardless of the structuring criteria. Twenty-eight CSN1S1-CSN2-CSN1S2-CSN3 haplotypes were found, 22 of them with frequencies <5%; the three most frequent were BB-A 1 A 2 -AA-AA-AA-AA (16.6%), BB-A 1 A 2 -AA-AA-AA- AB (14.1%) and BB-A 2 A 2 -AA-AA-AA (10.1%). A good potential of BON cattle to produce high quality milk with functional value was reported.

Caseins are important proteins because they are associated with high dairy yield, milk quality and technological performance.For example, a higher amount of caseins and whey proteins has been associated with the CSN1S1 gene (MOHAN et al. 2021); high milk production has been associated with the CSN1S2 gene (ARDICLI et al. 2018); the CSN2 gene is not only associated with high production and quality (MOHAN et al. 2021, VIGOLO et al. 2022), but also with a healthier product, specifically the animal milk with the *A 2 variant, as the *A 1 variant during its digestion digests the bioactive pet β-casomorphin-7 properties have opioid properties, and has been postulated as a risk factor in the development of diseases such as atherosclerosis, type I diabetes, sudden infant death syndrome, neurological diseases and milk allergy (CHESSA et al. 2020, THIRUVENGADAM et al. 2021, VILLALOBOS-CORTÉS et al. 2023).
On the other hand, the CSN3 gene has been linked to the structural stability of casein micelles, which may affect cheese quality by reducing coagulation time and improving curd firmness (POULSEN et al. 2017, ZEPEDA-BATISTA et al. 2017).Similarly, the BB-A 2 A 2 -BB haplotype of CSN1S1-CSN2-CSN3 has been associated with higher milk protein content, better coagulation, better ability to retain fat and protein in the curd, and consequently higher cheese yield compared to the BB-A 1 A 1 -AA haplotype (PERNA et al. 2016a, b).Likewise, variations in the LAA and LGB genes, which account for about 90% of whey proteins, can influence milk protein composition, curd formation and cheese yield (MACEDO et al. 2020).
The Blanco Orejinegro (BON) breed of cattle is one of the most important Colombian Creole breeds, adapted to high tropical conditions.It is mainly used in dual-purpose production systems (milk and meat), with a birth weight of 28.39±3.77kg for males and 26.68±3.59kg for females.Adjusted weaning weight at 240 days was 165±29.16kg and adjusted weight at 480 days was 215.38±30.57kg.Mean age at first calving 1104±141 days and calving interval 487±147 days.In terms of milk production, it averages 3.6 l/cow per day and 1439 l/lactation corrected to 270 days, with compositional values of fat 4.39%, protein 3.92%, solids non-fat 9.69%, solids total 14.08% and lactose 5.04% (MARTÍNEZ et al. 2012c).
Unlike the other American Creole breeds (brown in color), it has pigmented epidermis and mucous membranes, with a predominantly white, grey, and sarda-colored coat (MARTÍNEZ et al. 2012c).It has Iberian and British ancestry (MARTÍNEZ et al. 2012a) and more than 500 years of adaptation to tropical conditions (CAIVIO-NASNER et al. 2021b), which has allowed it to develop traits such as hardiness (ARRIETA et al. 2021), resistance to some diseases (LÓPEZ-HERRERA et al. 2001) and parasites (ROCHA et al. 2019), adaptation to climate (LEÓN et al. 2019), ability to digest low quality forage (MARTÍNEZ et al. 2012b) and good productive (CAIVIO-NASNER et al. 2021c) and reproductive efficiency (RIVEROS et al. 2022).This performance, together with its resistance and adaptive traits, confirms its value as an important genetic resource for the tropics, not only in Colombia.
Genomic studies on genetic diversity (CAIVIO-NASNER et al. 2021a), allele frequencies associated with genetic diseases (CAIVIO-NASNER et al. 2021b), selection indices (AMAYA et al. 2022) and genomewide association studies on growth traits (LONDOÑO-GIL et al. 2021) have also been carried out.However, the study of genes related to milk production and quality has not been advanced in this breed.In Colombia, studies have only been carried out in Holsteins from the departments of Antioquia (PADILLA-DOVAL et al. 2021) and Nariño (SOLARTE et al. 2009), while only one study reports on the variation of the LAA and LGB genes in local breeds (ROSERO-ALPALA et al. 2011) and one on the CSN3 gene in the Hartón del Valle breed (NARANJO et al. 2007).Therefore, the aim of this study was to determine the genetic variation in the CSN1S1, CSN2, CSN1S2, CSN3 and LGB genes in Blanco Orejinegro cattle.
Using the GeneJET Genomic DNA Purification Kit (Thermo Scientific™), DNA was extracted from 404 whole blood samples obtained by venipuncture using the BD Vacutainer® system with 21-gauge needles and tubes containing EDTA (7.2 mg) as anticoagulant, and from 15 semen samples stored at -196 °C in thermos flasks containing liquid nitrogen (CAIVIO-NASNER et al. 2021b).

Analysis of data
Allele and genotypic frequencies for the 43 SNPs were estimated using R software (R core team 2020).Intragenic haplotypes or protein variants (CHESSA et al. 2020) were then reconstructed for all genes, and genotypic and allele frequencies were estimated and plotted for each gene, taking into account sex, year of birth (2004 to 2018) and herd (1 to 15), using R (http://www.r-project.org/) and PLINK v1.9 (PURCELL et al. 2007).
Using PLINK v1.9 software, Hardy-Weinberg equilibrium (HWE) and FST values were estimated for each of the evaluated SNPs, using herd, sex and year as population structuring factors (only for casein genes due to their economic importance).Then, using the LD function of the Gaston package v1.5.9 (DANDINE-ROULLAND & PERDRY 2018) for R, the data were trimmed using sliding windows and a criterion r 2 = 0.2, with which the global FST was estimated from the informative markers.Then, the LD values obtained among all variants were plotted using the LD.plot function of the same package.Finally, the haplotype frequencies for the caseins (CSN1S1-CSN2-CSN1S2-CSN3) were estimated from the reconstructed intragenic haplotypes (CHESSA et al. 2020), using the pegas package v1.0-1 (PARADIS 2010) with R software.

RESULTS
In the CSN1S1 gene, only the rs132656458 variant was monomorphic, in the others the reference genotype and allele were most frequent (Table 1).The allele and genotypic frequencies of the SNPs present in this gene did not vary by the effect of sex, year and herd, so they were in HWE.The four variables of the CSN1S1 gene allowed the reconstruction of the CSN1S1*B, *C and *D variants of the αS1-CN protein, of which only the last one has a low frequency (0.006).In the BON population studied, 4 genotypes were found, of which the homozygous *B/*B (0.646) and heterozygous *B/*C (0.311) were the most frequent (Table 1).
Of 10 SNPs reconstructing the CSN2*A1 variant of the β-CN protein, six were fixed in the population, and the frequency of the alternative allele in the polymorphic variants did not exceed 0.003 (Table 2).Thus, the frequency of the CSN2*A 1 allele was 0.470.The SNP rs43703011, which discriminates the CSN2*A 2 allele from the β-CN protein, was the most frequent (0.509) in the BON cattle studied.Although the variant did not deviate from the theoretical HWE, the frequency of this allele varied according to sex, herd and year.Thus, this allele was higher in males (Figure 1A), with frequencies ranging from 0.20 in herd 11 (n=11) to fixed in herd 13 (n=48) (Figure 1B), and frequencies above 50% were found in at least nine of the 13 years, with a slightly increasing trend over time (Figure 1C).The genotyping chips used in this study only assessed two variants in the CSN1S2 gene (rs441966828 and rs463985801).For both SNPs, the reference genotype and allele were the most common (Table 3), with no significant variation for sex, year of birth and herd effects, and no significant deviations from HWE.The SNP rs441966828 discriminates the CSN1S2*A from the αS2-CN protein, which was the most common in BON cattle, whereas the SNP rs463985801 discriminates the CSN1S2*D variant, which was the least common in the study population.Only two of the variants evaluated in the CSN3 gene were fixed in the BON (Table 4), both associated with the κ-CN variant A, which has the highest frequency (0.679).Similarly, the variant rs43703015, which determines the κ-CN*B variant, had the second highest frequency (0.293).This variant was significantly higher in females (Figure 2A), in flocks 2, 5 and 7 (0.5, 0.55 and 0.51, respectively) (Figure 2B), and with a slight tendency to decrease over time (Figure 2C), since the frequency was higher than 40% between the years 2005 and 2010, whereas this frequency did not exceed 35% in the last three years studied.
On the other hand, seven of the 11 variants analyzed in the LGB gene were fixed in the study population (Table 5).All variants were in HWE and did not vary by sex, year or herd effects.Reconstruction of the β-LG variants identified 7 alleles and 10 genotypes, with the β-LG*B allele (0.657) and the *B/*B genotype (0.389) being the most common, followed by the β-LG*A allele and the heterozygous *A/*B genotype.Of the other alleles, only the β-LG*H allele had a frequency greater than 5%.
After pruning, 14 of the 32 variants identified in the CSN1S1, CSN2, CSN1S2 and CSN3 genes were considered informative.There were no differences in the FST values before and after pruning.The coefficient of population structure considering all SNPs was higher when herd (FST= 0.035) was considered as structuring factor compared to sex (FST= 0.003) and year (FST= 0.001).The population structure value increased when only CSN3 gene and herd were considered (FST= 0.054) compared to CSN2 gene and herd (FST= 0.035).Clustering by sex was higher with CSN2 gene (FST= 0.003) compared to CSN3 gene (FST= 0.001).Clustering by year was similar for CSN2 and CSN3 genes (FST= 0.001).Missen se GG GC CC G C 1.000 0.000 0.000 1.000 0.000 Figure 3A, shows the LD values among the 14 SNPs after pruning in the genes determining milk proteins.The highest LD value was found between two variants determining the κ-CN variant A (CSN3_AY380228_13104_1 and CSN3_AY380228_13165), also the linkage between the SNP CSN3_AY380228_13068 determining the protein variant κ-CN*B with those mentioned above in κ-CN*A showed values of LD=0.99.Other high LD values were found among the SNPs classifying the CSN1S1*B and CSN1S1*C variants of the αS1-CN, followed by the linkage between the *A 1 variants of the CSN2 and CSN3 genes (LD= 0.66).In particular, the CSN2_7 SNP, which determines the CSN2*A 2 variant, showed a higher linkage with the variants analyzed in the CSN1S1 gene than with other SNPs in the CSN2 gene and only 0.06 with the CSN3*B variant.
On the other hand, after data pruning, only 4 of the 11 analyzed β-LG variants were considered informative.The LD values of these are shown in Figure 3B.The variant LGB_X14710_3080 (β-LG*F allele) showed no linkage with the other SNPs.While the LGB_X14710_3982 and LGB_X14710_5174 variants, which reconstruct the β-LG*B allele, showed high linkage with the LGB_X14710_5263 variant, which determines the *A allele of β-LG.

Haplotype CSN1S1-CSN2-CSN1S2-CSN3 Frequency Haplotype CSN1S1-CSN2-CSN1S2-CSN3
Frequency  (GALLINAT et al. 2013).This could be related to the breeds used in the chip design (AL KALALDEH et al. 2023) or to the ease of reconstructing variants from DNA sequencing, as all possible polymorphisms are available.

DISCUSSION
The CSN1S1*B allele was the most common in 13 breeds of Bos taurus studied, as in the present investigation, with values ranging from 0.53 in the Sarabi breed to being fixed in the Angler, German Yellow, Highland Cattle, Limpurger and Shorthorn breeds (GALLINAT et al. 2013).In Turkish Holstein, the *B allele frequency was 0.95 (ARDICLI et al. 2018) and in Czech Simmental it was 0.90 (ČÍTEK et al. 2023).On the other hand, the CSN1S1*C allele was most frequent in Bos indicus Butana (0.66, n= 50), Gir (0.50, n= 7), Golpayegani (0.48,n= 20) and Sistani (0.88, n= 21) (GALLINAT et al. 2013).Several authors report that animals with *B/*B genotype have higher milk production and higher milk protein and fat concentration (ARDICLI et al. 2018, MOHAN et al. 2021, ČÍTEK et al. 2023), the frequency of this genotype in the studied BON cattle population was 0.646 (Table 1).
From the 14 SNPs present in the genotyping chips used in this work (GGP Bovine 150 K chip, n= 70 and GGP Bovine 50 K chip, n= 349), located in the CSN2 gene, it is possible to reconstruct five of the 12 variants of the β-CN protein reported (CAROLI et al. 2009 (MOHAN et al. 2021), the CSN2*A2 allele was the most frequent, with values above 0.5 in all breeds, values similar to those reported in this work.Similarly, this variant was even more common in the Bos indicus breeds evaluated by AHMED et al. (2017) and GALLINAT et al. (2013).Notably, the *H 2 variant was not reported by the above authors.Whereas the *B allele was the most common allele in Colombian Holstein (PADILLA-DOVAL et al. 2021).
As mentioned above, intestinal digestion of β-CN*A 1 protein releases the bioactive peptide βcasomorphin-7, which is considered a risk factor for the development of several diseases (CHESSA et al. 2020, THIRUVENGADAM et al. 2021, VILLALOBOS-CORTÉS et al. 2023), in particular milk allergy and lactose intolerance (MUNTEAN et al. 2022), compared to β-CN*A 2 protein.Lactose intolerance and cow's milk allergy can be confused.The former is essentially due to lactase deficiency in the intestine (JAISWAL & WORKU 2022), whereas the latter may be a humoral, cellular, or mixed process mediated by B and T lymphocytes to epitopes resulting from intestinal digestion of milk, which bind mainly to immunoglobulin E (MUNTEAN et al. 2022).In this regard, at least five IgE-binding epitopes, three common to the β-CN*A 1 , *A 2 and *B variants and two to the β-CN*A 1 and *A 2 variants, have been reported following in vitro gastrointestinal digestion of cow's milk (LISSON et al. 2013).
Several studies have reported the possible effect of genetic variation in the CSN2 gene on the immunomodulatory, antihypertensive, antimicrobial and antithrombotic function of bioactive peptides derived from the digestion (in vitro, in vivo and in situ) of the β-CN protein (FITZGERALD et al. 2020, JAISWAL & WORKU 2022, MUNTEAN et al. 2022), with particular focus on the β-CN*A 2 variant for the production of A2type milk.Therefore, selection in favor of this allele has started in several countries (CHESSA et al. 2020).For example, in the Italian Holstein breed, the frequency of this allele has increased by 10% between 1990 and 2017, to the detriment of the frequency of the β-CN*A 1 allele (CHESSA et al. 2020).The latter is similar to the results presented in this study (Figure 1C, the red dotted line indicates the trend over time), possibly due to the response of a selection process not necessarily related to the farmers' knowledge of the nutritional quality of A2-type milk, but rather to the indirect positive effects of the *A 2 /*A 2 genotype and to the increase in milk yield, protein and fat content or technological performance for the production of curd (MOHAN et al. 2021, VIGOLO et al. 2022).However, there is a strong interest among producers in the new market for specialty milks, which has led them to use male genotyping as a criterion for sire selection, which may partly explain the observed trend.On the other hand, it must be emphasized that A2-type milk can also be obtained from the *A 3 , *D, *E and *H 2 variants, as all of them do not release β-casomorphin-7 after digestion compared to the other variants (*B, *C, *F, *G and *H 1 ).
The five κ-CN variants that can be reconstructed from the 12 SNPs present in the genotyping chips used in this study are *A, *A1, *B, *J and *I of the 14 κ-CN variants (*A, *A 1 , *B, *B 2 , *C, *D, *E, *F 1 , *F 2 ,*G 1 , *G 2 , *H, *I and *J) reported (CAROLI et al. 2009).The frequency of the *A variant reported here was higher than that reported in nine B. taurus (GALLINAT et al. 2013) (NARANJO et al. 2007) and Karan Fries (India) (MOHAN et al. 2021), but lower than reported in Colombian Holstein for the regions of Antioquia and Nariño (SOLARTE et al. 2009, PADILLA-DOVAL et al. 2021).In contrast, the *B allele has a high frequency of 0.9 in Italian Jersey (n=622) (CHESSA et al. 2020), values far from those reported in the BON in the present work.On the other hand, in B. indicus breeds, the *H variant is more frequent (f(*H) > 0.5) compared to what was found here (*H= 0.024).
Regarding the effect of the CSN3 gene on traits related to milk production and milk quality, a metaanalysis (MAHMOUDI et al. 2020) including 48 articles with genotypes of 30471 animals for the *A and *B variants, in specialized and local breeds, taking into consideration four different inheritance models, reports a higher daily milk yield in animals with the *B/*B and *A/*B genotypes (p< 0.05) in the recessive and additive models.In the dominant and additive models, the *B/*B genotype improves milk fat yield.In a dominant and recessive model, the *A/*A genotype decreases milk protein percentage (MAHMOUDI et al. 2020).These results are confirmed by MOHAN et al. (2021) who report that the *B/*B genotype is associated with higher milk yield and higher protein, calcium and phosphorus content.The same genotype is also reported to produce milk with a better fatty acid profile (HEWA NADUGALA et al. 2022).In addition, the *B/*B genotype has been associated with shorter coagulation time, higher stability of micelle structure, higher gel strength and higher cheese yield (POULSEN et al. 2017, ZEPEDA-BATISTA et al. 2017, KRUCHININ et al. 2023).
However, the possible beneficial health effects of the bioactive peptides described for κ-CN include antithrombotic activity, reduction of inflammatory processes and oxidative stress, and improvement of the immune response (VARGAS- BELLO-PÉREZ et al. 2019).These effects are certainly related in some way to the polymorphism in the CSN3 gene.For example, after in vitro digestion of κ-CN, two common IgE-binding epitopes are reported between the *A and *E alleles and only one in the *B variant (LISSON et al. 2013), which can be partly explained by the low homology (58%) between human and bovine κ-CN (JAISWAL & WORKU 2022), which could imply a lower allergenicity of the κ-CN*B variant.
The above reports justify the targeting of breeding programs towards the selection of the CSN3*B allele.This is demonstrated by CHESSA et al. (2020) in Italian Holstein, where the frequency of this allele increased from 0.186 before 1990 to 0.355 in 2017.On the contrary, in the BON breed there is a slight decrease over time (Figure 2C, red dotted line) of this allele, and although the frequency of the homozygous *B/*B genotype was low (0.077), the allele frequency has an important magnitude (Table 4), with high variation in its frequency between herds (Figure 2B), which would facilitate and justify a selection scheme in favor of this allele.
The *B allele of the β-LG protein (Ensembl ID: ENSBTAT00000019538) is used as a reference to reconstruct the 11 reported variants (*A, *B, *C, *D, *E, *F, *G, *W, *I, *J and *W) in this protein (CAROLI et al. 2009).The 11 SNPs included in the genotyping chips allow the identification of the *A, *B, *C, *D, *E, *F and *H variants.In a previous study in BON cattle, the *A variant was found to be the most frequent (0.51) (ROSERO-ALPALA et al. 2011), a value that contrasts with this study, but which can be explained by the limited sample size used (n= 30) and the origin of the samples (a single farm in one department vs. 15 herds in six departments).The *B allele was the most frequent allele in several cattle populations (AHMED et al. 2017), as well as in the Holstein and Simmental populations of the Czech Republic (ČÍTEK et al. 2023), which is consistent with the results of this work.In contrast, the *A variant is more common in the Gyr and Jersey breeds (BARBOSA et al. 2019, GAI et al. 2021).
Several authors have reported that the β-LG*B/*B genotype correlates with lower milk yield, but with higher protein, fat and casein content, additionally with shorter rennet formation time and lower somatic cell score (BARBOSA et al. 2019, GAI et al. 2021, ČÍTEK et al. 2023), this genotype was the most frequent in the BON population studied (Table 5).Different bioactive peptides with antihypertensive, antimicrobial, antilipemic, antioxidant and immunomodulatory activity have been reported from in vitro and in silico digestion of β-LG (DULLIUS et al. 2018, KIM et al. 2019, ČÍTEK et al. 2023) but reports on their bioactivity according to genetic variation are null.
There is extensive knowledge about the unique effect of casein variants on the properties and functionality of milk.However, some authors suggest that due to their proximity, it is possible that they are inherited as a linkage group rather than independently (VILLALOBOS-CORTÉS et al. 2023) and therefore the joint effect, i.e. the haplotypes, needs to be investigated (HEWA NADUGALA et al. 2022).The results of this investigation show that linkage between the CSN1S1-CSN2-CSN1S2-CSN3 genes is low (Figure 3A), except for the *A1 variants (LOD= 0.66) of the CSN2 (BCN_8491) and CSN3 (CSN3_AY380228_13111) genes, which may partly explain the 28 haplotypes found and the low frequencies (<5%) of most of them (22 haplotypes).In contrast, linkage between variants of the same gene was logically higher, such as between variants of the CSN3 or LGB genes (Figure 3B).AHMED et al. (2017) found 84 haplotypes for the CSN1S1-CSN2-CSN1S2-CSN3 genes.However, unlike this study, AHMED et al. 2017 constructed the haplotypes from the protein variants rather than the genotypes, which would explain the higher number of haplotypes, nevertheless the haplotypes C-A 2 -A-A-A (0.156), B-A 2 -A-A-A (0.119) and B-A 2 -A-A-A-A (0.085) were the most common in their report.In another study using only β-casein and κ-casein, the most frequent haplotypes in the Italian Holstein breed were A 2 -A (0.474) and A 1 -A (0.218), whereas in the Italian Jersey breed they were A 2 -B (0.646) and B-B (0.171) (CHESSA et al. 2020).
Finally, among some effects of haplotypic variants on milk production, quality and technological performance, the haplotype BB-A 2 A 2 -BB of the CSN1S1-CSN2-CSN3 genes (with a frequency of 3.5% in BON) has been associated with higher milk protein content, better coagulation, better ability to retain fat and protein in milk (PERNA et al. 2016a,b).The BB-A 2 A 2 -AA combination of CSN1S1-CSN2-CSN3 genes correlated with low or no rennet-induced coagulation, this haplotype was found with a frequency of 10.1% in BON, however, this same haplotype was associated with better but acid-induced coagulation characteristics (KETTO et al. 2017), which may be of interest for cheese producers with milk of BON origin.Additionally, the haplotypes CC-A 2 A 2 -BB (0.3% in BON), BC-A2A2-BB (0.29% in BON) and BB-A 1 A 2 -AA (19.6% in BON) with a cumulative frequency of 20.19% in BON, are associated with good rennet formation, coagulation time and cheese yield (KETTO et al. 2017).
The BB-BB haplotype of the CSN2-CSN3 genes (only found as heterozygous in BON with a cumulative frequency of 0.6%) with higher cheese yield (CHESSA et al. 2020).The B-A 1 -B variant of αS1-CN, β-CN and κ-CN proteins was associated with higher milk yield and protein content, found in heterozygous state in BON with cumulative frequency close to 40% (HEWA NADUGALA et al. 2022).Haplotype B-A 1 -A proteins αS1-CN, β-CN and κ-CN were associated with higher milk fat content (HEWA NADUGALA et al. 2022).In the Italian Reggiana breed the C-A2-B haplotype was correlated with higher fat and protein content (CAROLI et al. 2009, HEWA NADUGALA et al. 2022), with a frequency of 0.3% in BON.Finally, the possible effects on human health in terms of allergenicity and intolerance to milk and milk products of haplotypic variants or composite genotypes have not been studied, probably due to the existence of synergistic combinations between the variants and the complexity of analysing the results.

CONCLUSION
The BON cattle population showed a high variation in the SNPs evaluated, as 62.79% of the SNPs were polymorphic.The SNPs present in the genotyping chips allow the reconstruction of the most important casein and β-lactoglobulin variants from the point of view of their relationship with milk production, quality and technological performance.The high frequency of the protein variants of interest (CSN1S1*B, β-CN*A2, κ-CN*B and β-LG*B) in BON cattle suggests a potential in terms of production, quality, yield and functionality of the milk produced by this breed, without ignoring the non-genetic factors that can modify it.The high frequency in both sexes and the variation between farms of the β-CN*A 2 and κ-CN*B variants could be used as a basis for a future genetic improvement program in the breed, with the aim of producing quality milk with a function in human health.These results indicate that there is no linkage in the casein genes, but there is linkage between some intragenic variants useful for indirect genotyping.It will be necessary to extend this study to the complete sequencing (including UTRs) of the CSN1S1, CSN2, CSN1S2, CSN1S2, CSN3, LAA and LGB genes in order to detect new SNPs that characterize the breed or have a potential effect on protein quality, as well as to replicate it in the other Criollo breeds present in Colombia.Also, to study not only the individual effect of the variants, but also the effect of the caseins/α-LA/β-LG as a whole, to understand the possible synergistic or opposite effect on the production, quality, yield and functionality of cow's milk.

Figure 1 .
Figure 1.Frequency of the CSN2*A 2 allelic variant according to sex (A), herd (B) and year (C) in the BON breed.Different letters indicate statistically significant differences (p< 0.05) between allele frequencies by sex, herd and year.

Figure 2 .
Figure 2. Frequency of the CSN3*B allelic variant according to sex (A), herd (B) and year (C) in the BON breed.Different letters indicate statistically significant differences (p< 0.05) between allele frequencies by sex, herd and year.

Figure 3 .
Figure 3. Linkage disequilibrium heat map of informative variants in the κ-Casein (A) and β-Lactoglobulin (B) genes in BON breed.
In the Butana breed (Bos indicus), 17 SNPs in the CSN1S1 gene were found by sequencing using the *B variant (ENSBTAT00000010119.3) as a reference (AHMED et al. 2017), the genotyping chips used in this research contain only four SNPs in this gene, which allow distinguishing the *B, *C and *D variants from the nine variants (*A, *B, *C, *D, *E, *F, *G, *H and *I) of the αS1-CN protein reported ) (*A1, *A2, *A3, *B, *C, *D, *E, *F, *G, *H1, *H2 and *I).Consistent with what GALLINAT et al. (2013) reported in 13 breeds of Bos taurus, in Holstein and Jersey from Italy (CHESSA et al. 2020) and in the synthetic breed Karan Fries from India (B.taurus x B. indicus) ). Forty-three SNPs were identified in the genotyping chips, four in the CSN1S1 gene, 14 in the CSN2 gene, two in the CSN1S2 gene, 12 in the CSN3 gene and 11 in the LGB gene.According to the method proposed by SANCHEZ et al. (2020), these polymorphisms were used to reconstruct the protein variants *B, *C and *D of αS1-CN; *A 1 , *A 2 , *B, *H 2 and *F of β-CN; *A and *D of αS2-CN, *A, *A 1 , *B, *I and *H of κ-CN and *A, *B, *C, *D, *E, *F and *H of β-LG.

Table 1 .
Description of the variants present in the genotyping chip located in the CSN1S1 gene and their frequencies.Allelic and genotypic frequencies of the reconstructed variants in the αS1-CN protein found in the BON breed.

Table 2 .
Description of the variants present in the genotyping chip located in the CSN2 gene and their frequencies.Allelic and genotypic frequencies of the reconstructed variants in the β-CN protein found in the BON breed.

Table 3
. Description of the variants present in the genotyping chip located in the CSN1S2 gene and their frequencies.Allelic and genotypic frequencies of the reconstructed variants in the αS2-CN protein found in the BON breed.

Table 4 .
Description of the variants present in the genotyping chip located in the CSN3 gene and their frequencies.Allelic and genotypic frequencies of the reconstructed variants in the κ-CN protein found in the BON breed.

Table 5 .
Description of the variants present in the genotyping chip located in the LGB gene and their frequencies.Allelic and genotypic frequencies of the reconstructed variants in the β-LG protein found in the BON breed.