Casein polymorphism and relation between milk production traits

Abstract

Introduction The trait relation between milk yield, fat yield and protein yield on one hand and protein percentage and fat percentage on the other is partly characterized by negative genetic correlation coefficients. Nevertheless, selection of high yielding cows in breeding of Holstein Friesian has been accompanied by approximately +1% of fat and nearly no gain in protein percentage during the last 30 years. The index selection involving yield and percentage of fat and protein simultaneously results in lower selection responses for each trait compared with single trait selection. Investigations of linkage of quantitative trait loci (QTL) for milk production traits to genetic markers could be useful for examining the genetic background for the specific trait relationship. The main focus is the detection of loci which are linked to QTL for a specific trait. The question arises whether identical or closely linked genes are responsible for the negative correlation between yield and percentage traits. It is rare for cows to be superior in both yield traits and percentage traits. The results of different studies regarding the casein locus on chromosome 6 suggest that this super locus has an effect on the performance ( Velmala et al. 1995 ; Vilkki 1996). The casein locus can thus be marked as was done by Graml et al. (1985) because almost no recombination occurs between the single loci. The location of the casein genes (αS1-CN, β-CN, κ-CN) within the casein locus on chromosome 6 is known (e.g. Eggen & Fries 1995). The statistical analysis may be carried out in different ways. It ranges from analyses of variance ( anova), regressive models, the examination of differences in daughter performance of carriers of two paternal alleles (e.g. Kühn et al. 1996 ), maximum likelihood approaches to Gibbs sampling (e.g. Wild 1997). Liu (1994) reported significant QTL effects for milk yield and protein yield using the mixed model of inheritance. Simianer (1993) found a significant QTL effect for fat yield near the casein locus. The evidence of genetic linkage between genetic markers and putative QTL for single milk production traits was given in some family analyses. A significant linkage between QTL for milk yield, fat percentage and protein percentage and markers on chromosome 6 in one out of 14 families, using a major gene model, has been published by Georges et al. (1995) . Spelman et al. (1996) obtained evidence of a putative QTL for protein percentage in a loose linkage to the casein locus. Graml et al. (1985 , Graml et al. (1985) investigated the influence of the milk protein loci on milk traits using a mixed model analysis of variance. Results of segregation and linkage analysis may contribute to clarifying the genetic background of marker effects and the relation between yield traits and percentages of milk fat and milk protein in Black and White cattle. The present study presents results of estimated direct marker effects of the loci for β-CN and κ-CN on milk traits, together with results of a segregation and linkage analysis with respect to the negative relationship of yield traits and percentage traits. Material The examinations were based on performances of cows originating from different groups of Black and White cattle ( Table 1). The sample SMR/HF involves both SMR (crossbred from German Black and White, 25% Jersey and 50% Holstein Friesian genes) and pure Holstein Friesian (HF). A number of SMR cows resulted from the mating scheme HF × SMR. So the amount of HF genes was 75% in this case. The German Black and White (DSR) were the second sample. For the estimation of direct marker effects two samples were used to obtain results from independent data sets. As genetic values were being used the linkage analysis could be carried out in one analysis using a half sib design. Not all traits showed a normal distribution of genetic values whereas the absolute milk performances were normally distributed. The data were reduced considerably for linkage analysis according to model requirements for 1258 daughters from 191 sires. Eight of the sires had more than 20 daughters, 170 sires between five and 10 daughters. Table 1. . Number of cows, averages and standard deviations of milk production traits milk yield (kg) fat yield (kg) protein yield (kg) fat (%) Sample/group Number Mean SD Mean protein (%) namest="col7" nameend="col8">SD Mean SD Mean SD Mean SD Sample SMR/HF SMR 796 5206 925 226 39 182 31 4.37 0.40 3.51 0.22 HF 280 5810 992 244 40 195 31 4.23 0.36 3.37 0.17 Sample DSR SR 645 4647 674 184 28 161 23 3.97 0.31 3.47 0.19 The data information for individual performances was supplied by VIT Paretz/Verden. Methods The genotyping for milk protein genes was carried out by isoelectric focusing ( Erhardt 1989, 1993). The heterozygosity of the αS1-casein, β-casein and κ-casein was computed to be 0.07, 0.50 and 0.41, respectively. Linkage analysis with αS1-casein was limited for this reason and therefore the results are not shown. Estimation of direct marker effects The estimation of direct marker effects was carried out by using the PEST programme ( Groeneveld et al. 1992 ) involving four milk protein markers simultaneously to avoid confounding effects between the closely linked loci. They are marked as direct marker effects to distinguish them from QTL effects estimated by a segregation analysis. The fixed effects of herd, year and season of calving and lactation number were considered in the model. An additional effect of the group of Black and White population (Ra) was included to get marker effects unbiased by breed effects. The reason for this was the significant difference in milk yield between SMR and HF. The model is: where yi is the phenotypic value of cow I;μ is the general mean; HYS is the fixed effect of herd, year and season of calving; Ra is the fixed effect of population group (SMR/HF); L is the fixed effect of age (lactation number); GT is the effect of the genotype (j, k, l, m) of the markers αS1-CN, κ-CN, β-CN and β-LG; Ai is the effect of animal i (considering the relationship between the cows); and e is the residual. Each milk protein marker was involved in the model simultaneously in order to estimate the specific genotypic effects on milk production traits without confounding. Results of a model including interaction effects between κ-casein and β-casein (not shown here) were comparable with the results obtained from this model. Comparison of correlation coefficients The Pearson correlation coefficient between yield traits and percentages of fat and protein was calculated from the genetic values using the computer program package SAS. It was compared with the correlation coefficients after subtracting the effects estimated for the casein genotypes from the individual genetic value of each cow. The amount of change was considered to be due to casein polymorphism on the relation between yield and percentage traits. Linkage analysis Both the estimation of the effects of QTL and the analysis of genetic linkage were carried out with the approach of Liu (1994) and Simianer 1993) for the underlying daughter design. The genetic model is where yij is the genetic value of daughter i of sire j;ui the polygenic value of daughter i and vj the effect of the QTL. The parameters of QTL (a, additive effect of QTL, d, dominance effect at QTL and p frequency of the desired QTL-allele) and the recombination rate (θ) were estimated by the likelihood function (L). It may be written in a simplified form ( Stricker 1997) as follows: where: L(y,M) is the joint likelihood of all phenotypes y and genotypes of markers M;P is the probability; g is the genotype of the marked QTL; Φ(x,v) is the density in x of a normal distribution with average 0 and variance v ; uj is the polygenic value for sire j, integration variable; vQi is the effect of daughter’s QTL; l is the daughter’s index within the sire; h2, is the polygenic heritability; and σp2 is the phenotypic variance. The indices: m, mother; f, sire, o (offspring), ni (number of families) and no (number of offspring). Five different genetic models were adapted for testing of hypotheses: (i) mixed model of inheritance; θ estimated versus θ = 0.5 (considers one QTL and polygenic component, the results of this model are reported in this paper because it fits best); (ii) single locus model of inheritance; θ estimated versus θ = 0.5 (major gene model); (iii) polygenic model of inheritance (no QTL with large effect is considered) The experiment-wise LOD score was used as a linkage test criterion. The sires were considered to be unrelated. The analyses were carried out for one marker and QTL (two-point analyses). The model is identical to Guo & Thompson (1992) involving one QTL assuming two alleles, fixed effects and polygenic component. The numerical methods are different. Both methods are maximum likelihood approaches. The method used integrates by means of Gauss–Hermite quadrature with 20 starting points whereas G uo and T hompson applied the Monte Carlo method. The likelihood was maximized repeatedly by the Powell method from different starting points. Results and discussion An association study as well as a linkage analysis was carried out using the data and genotypic information of Black and White cows in two samples SMR/HF and DSR. The relatively high degree of heterozygosity found in sample SMR/HF promised that the linkage analysis might lead to meaningful results. The crossbred SMR contains a maximum 25% Jersey genes and a minimum 50% HF genes. The allelic frequency of Jersey cattle is different from those of Black and White cattle as may be shown by κ-CNB ( Table 2). This explains the increased frequency of κ-CNB (+0.15), lower frequency of β-CNA2 (−0.11) and αS1-CNB (−0.01) in SMR compared with purebred HF cows. Cows of both groups SMR and HF were managed simultaneously in the same farms. The analysis of cows with different percentages of HF genes was possible by allowing for a fixed effect of group SMR and HF in the association study and analysing single half sib families within the segregation and linkage analysis. Table 2. . Allelic frequencies of κ-CN B in dairy breeds Breed Number of cows κ-CNB Author Holstein Friesian 3346 0.13 Oloffs (1991) DSR 481 0.20 Panicke et al. (1991) SMR 804 0.28 Panicke et al. 1991 ) Montbiliarde 646 0.37 Grosclaude (1988) Jersey 437 0.56 Erhardt (1989) Braunvieh 1742 0.59 Ortner et al. (1995) The estimated direct marker effects of β-CN and κ-CN on milk production traits from two independent samples are presented for each of the three most frequent genotypes. Genotypic effects of rare allelic combinations (including β-CNA3, β-CNB, κ-CNE) could only be estimated with large standard errors ( Freyer 1997) and should not be discussed. The genotypic marker effects and the differences between the homozygous genotypes most frequent in both casein loci are considered together with the results of the linkage analysis. Comparable genotypic effects of casein markers on yield traits and percentage traits were obtained by Ortner et al. (1995) . The estimated parameters of putative QTL are from the analyses with β-CN and κ-CN ( Tables 8 and 9). The model involving a QTL and polygenic inheritance fitted the data best compared with the usual polygenic model and the single locus model (α < 0.001). Table 8. Estimated parameters for yield traits from linkage analysis to casein markers Estimated parameters, marker Milk yield Fat yield Protein yield Additive effect of QTL (2a) 728 ± 14 33 ± 0.5 20 ± 0.5 Dominance effect of QTL d −150 ± 23 −7.8 ± 1.1 −5.9 ± 0.8 Allele frequency p 0.23 ± 0.02 0.17 ± 0.2 0.23 ± 0.02 β-CN Recombination rate 0.02 ± 0.11 0 ± 0 0 ± 0 LOD score 0.40 1.89 1.11 κc-CN Recombination rate 0.21 ± 0 0.5 ± 0 0.5 ± 0 LOD score 0.65 0 0 Table 9. Estimated parameters of percentage traits from linkage analysis with casein markers Parameter/marker Fat-(%) Protein-(%) EFQ Additive effect of QTL (2a) 0.66 ± 0.001 0.09 ± 0.003 0.003 ± 0 Dominance effect of QTL d −0.05 ± 0.02 −0.06 ± 0.007 −0.001 ± 0 Allelic frequency p 0.02 ± 0.05 0.12 ± 0.02 0.13 ± 0 β-Casein Recombination rate 0 ± 0 0.10 ± 0.43 0.5 ± 0 LOD score 1.09 0.04 0 κ-Casein Recombination rate 0.26 ± 0.18 0.5 ± 0 0.19 ± 0.7 LOD score 3.03 0 0.006 Marker ββ-casein The allele β-CNA2 affects the milk yield positively. The difference of effects between the homozygous genotypes β-CNA2A2 and β-CNA1A1 amounts to 207 kg milk (sample SMR/HF) and 144 kg milk (sample DSR). This marker genotype is superior in fat yield and protein yield. The total difference of these two traits is more than 10 kg fat and protein ( Tables 3 and 4). The favourable yield genotype β-CNA2A2 also simultaneously affects the percentage of fat and protein significantly in a negative manner. According to the results one may assume that the allele β-CNA2 and the QTL allele for the yield traits are inherited jointly. In the case of percentage traits the opposite direction may be noted. The β-CNA1 is associated with the favourable allele for fat and protein percentage and simultaneously with the negative allele for yield traits. The specific genotypic effects of β-CN on yield traits on one hand and percentage traits on the other reflect the negative correlation between them. Table 3. . Effects of milk protein genotypes on yield traits in SMR/HF and differences between effects of homozygous genotypes (|A1 A1-A2 A2|, |AA-BB|) Marker Genotype Frequency Milk-yield Fat-yield Protein yield β-Casein A1 A1 0.24 −83a −1.83 −2.15 A1 A2 0.47 2b −0.07 −0.01 A2 A2 0.19 124b 3.81 3.68 A1 A3 <0.01 −159 −3.02 −8.60 A2 A3 <0.01 368 7.47 12.59 A1B 0.05 −47 −1.14 −0.62 A2B 0.04 −71 −4.68 −3.15 |difference| 207 5.64 5.83 κ-Casein AA 0.52 −15 −1.31 −1.25 AB 0.37 33 1.81 1.82 BB 0.06 −88 −2.81 0.74 AE 0.03 −117 −0.59 −4.74 BE 0.01 323 11.04 7.10 EE <0.01 197 4.98 −4.70 |difference| 73 1.50 1.99 Unequal suffixes indicate significant differences (p < 0.05) Table 4. Effects of milk protein genotypes on yield traits in DSR and differences between effects of homozygous genotypes (|A1 A1-A2 A2|, |AA-BB|) Marker Genotype Frequency Milk-yield Fat-yield Protein yield β-Casein A1 A1 0.61 −93 −2.49 −1.31 A1 A2 0.32 −57 −2.64 −0.55 A2 A2 0.04 51 1.33 1.01 A1B 0.03 −100 −2.11 −7.96 A2B <0.01 −613 −34.07 −28.18 |difference| 144 3.82 2.32 κ-Casein AA 0.54 −10 −0.53 −1.31 AB 0.27 −5 0.25 −2.45 BB 0.02 −33 −2.26 −4.49 AE 0.13 91 4.23 2.31 BE 0.03 −143 −6.69 −1.99 EE <0.01 −96 −11.83 5.42 |difference| 23 1.73 3.18 The results of linkage analysis suggest a close linkage of marker β-CN to QTL for fat yield, protein yield and fat percentage indicated by an estimated recombination rate θ = 0 ( Tables 8 and 9). The LOD scores range from 1.09 to 1.89. Marker κκ-casein The estimated direct effects of marker κ-CN on fat yield neither lead to clear results ( Tables 3 and 4). The differences of the direct effects of marker genotypes κ-CNAA and κ-CNBB exceed the differences which were found for the homozygous genotypes of β-CN: −0.073% of fat and –0.058% of protein in sample SMR/HF (α < 0.05). The effects of κ-CN genotypes on yield traits are difficult to interpret and may suggest some heterotic effects, since genotype κ-CNAB has a slightly higher effect than both homozygous genotypes. Some evidence for linkage was found for the fat percentage. The recombination rate was estimated to be 0.26 (LOD score 3.03, Table 9). The standardized effects of homozygous genotypes on yield and percentages show the trait specific influence ( Fig. 1) very clearly. Positive effects of β-CNA2A2 on yield traits (10–16% of standard deviation) and negative effects on percentages (8 to 9%) and also the mirror image effects of β-CNA1A1 are evident. We may notice stepwise changes for κ-CNBB from a negative effect on milk yield to high positive effects on percentages, especially on protein. Figure 1Open in figure viewerPowerPoint Standardized effects of homozygous β-CN and κ-CN genotypes on milk production traits (SMR/HF) Although the casein loci are closely linked together, each of them might have a specific effect on yield and on percentage traits. The genetic background is not completely clear. It might be assumed that κ-CN is an indicator for percentage of fat and protein and β-CN for yield traits. Correlation coefficients between yield traits and percentage of fat and protein A significant decrease for the correlation coefficient between milk yield on one hand and fat and protein percentage on the other (difference 0.05 and 0.07 shown in Table 7) can be seen when comparing the negative correlation coefficients between genetic values (i) before and (ii) after subtraction of estimated effects of casein genotypes from the individual genetic values. This result demonstrates that the polymorphism is involved in the negative relation between milk yield and percentage of fat and protein. Table 7. Correlation coefficients of milk traits (genetic values above diagonal, genetic values subtracted for estimated effects of Casein genotypes under diagonal) Milk yield Fat yield Protein yield Fat (%) Protein (%) EFQ Milk yield 1 0.84 0.91 −0.36 −0.33 0.21 Fat yield 0.69 1 0.86 0.18 (−0.02) −0.19 Protein yield 0.76 0.86 1 −0.16 (−0.01) 0.22 Fat (%) −0.31 0.19 −0.16 1 0.56 −0.73 Protein (%) −0.26 (−0.02) (−0.01) 0.56 1 (0.05) EFQ 0.20 −0.19 0.56 −0.72 (0.05) 1 Significant correlations at α = 0 . . . 0.0001 (brackets indicate nonsignificance) A considerable decrease of the positive correlation coefficient occurred between milk yield on one hand and fat and protein yield on the other. The decrease amounted to 0.15 each. The correlation coefficients between fat and protein yield and the percentages are nearly unchanged after adjusting for casein genotypic effects. Graml (1989) also reported decreased correlation coefficients when milk protein markers were involved. Compared with their results we found a much higher decrease of correlation coefficients amounting up to 21% between milk yield and content of protein was found in the present study. Graml et al. (1989) involved CN haplotypes and β-Lactoglobulin genotypes as well. Another fact is that the allelic frequencies have an influence on the results. Different statistical methods may also be the reason for different changes of correlation coefficients by using markers in both studies. Graml et al. (1985 , 1986) reported marker associated and polygenic correlations using a quite extensive amount of material of Simmental and Brown cattle. The results between the two breeds agreed and led to the conclusion that all milk protein genotypes affect all milk traits although to a small extent because the variance components were relatively small. The authors found similar effects in both breeds and partly significant antagonistic effects between yield and percentages which led to a compensation in fat and protein yield. This agrees with the present results. During recent years a large number of investigations aimed at the estimation of such effects. This paper does not intend a complete evaluation of them. Some results correspond with the present results (e.g. Bovenhuis & Weller 1994; Ortner et al. 1995 ) despite different breeds with different allelic frequencies. Other results ( Graml et al. 1985 , 1986; Van Eenenam & Medrano 1991) are different probably due to the different genetic models and statistical methods used. Trait-specific different directions of genotypic effects might also be due to an existing linkage disequilibrium between the casein loci and the putative-linked QTL. The linkage phase could be different between the breeds. Intensive selection and breed specific selection schemes may be the reason for this. Hypothesis of two existing QTL The highest LOD score (3.03) was found for fat percentage and κ-Casein ( Tables 8 and 9). Map specific LOD scores for different locus orders show some indication of close linkage for fat yield and protein yield ( Freyer et al. 1996 ). For fat percentage, evidence for linkage is obtained with estimated intermediate recombination fractions when κ-CN is the marker. The linkage analysis with β-casein led to estimated recombination rates θ = 0 for fat yield, protein yield and fat percentage (LOD scores 1.89, 1.11, 1.09, respectively). Bovenhuis & Weller (1994) and Spelman et al. (1996) have found similar results showing more than one peak in their analyses of milk production traits and microsatellite markers on chromosome 6. Mosig et al. (1998) and Gomez-Raya et al. (1998) reported further results suggesting at least two QTL on chromosome 6 at a distance to the casein locus up to 30 cM. Conclusions The results presented clearly show an association between casein polymorphism and relation between the two groups of milk traits (i) milk yield and (ii) percentage of fat and protein. The main differences between genotypic effects on yield traits are caused by β-CN genotypes in favour of genotype β-CNA2A2. β-CNA1A1 is superior regarding fat and protein percentage. The homozygous genotype κ-CNBB is superior in fat and protein percentage in SMR/HF whereas the heterozygote genotype κ-CNAB is best for yield traits suggesting heterotic effects. These results are underlined by changes of correlation coefficients and indicate that the casein locus affects the relation between yield and percentage traits. The results of linkage analysis indicate that these trait-specific genotypic effects of casein loci are due to genetic linkage in case of fat and protein yield and fat percentage. Evidence was given for a medium linkage between κ-casein and QTL for fat percentage (LOD score 3.03). The hypothesis of the putative existence of more than one QTL near to the casein locus might be examined using a model of segregation analysis allowing for effects for at least two QTL. Acknowledgements Support from the foundation ‘Wilhelm Schaumann’ is thankfully acknowledged. The authors would like to thank C hristian S tricker for numerous discussions. Summary The relation between milk yield and fat yield on one side and fat percentage and protein percentage on the other is partly described by strong negative correlation coefficients. An influence of casein polymorphism on this trait relation could be proved. The estimated direct effects of markers and the results of segregation and linkage analysis are considered in this connection. One or more than one QTL exist on chromosome 6 linked to the casein super locus that affects yield traits as well as percentage traits. It has been shown that the casein genotypes (CN) contribute to the correlation between yield traits and percentages of protein and fat as well. There are clear suggestions to different directions of casein genotypic influence for fat yield, protein yield and percentages of protein and fat according to the results from estimated direct marker effects and partly also from linkage analysis. Significant evidence for linkage was obtained regarding fat percentage and κ-casein (LOD score 3.03). The estimated genotypic effects show a positive influence of β-CNA2 and κ-CNA on milk yield, fat yield and protein yield and a simultaneous negative effect on fat percentage and protein percentage.