Genotype by Environment Interaction and Genetic Correlations Among Parities for Somatic Cell Count and Milk Yield 1 G. BANOS 2 and G. E. SHOOK Dairy Science Department University of Wisconsin Madison 53706 ABSTRACT
traits. Between second and third parity genetic correlation estimates were around unity for all traits. Records from all parities should be used for sire evaluation. (Key words: genotype environment interaction, genetic correlation, somatic cell count)
Lactation measures of somatic cell concentration and total sec production were developed. Data were separated into three parity groups. Within parity, five data sets were created: four subsets by herd-year average sec, and one with all records. Records on lactation sec, total sec production, and 305- 3.02. 3Second parity Ql: ~ 2.37; Q2: 2.38 to 2.81; Q3: 2.82 to 3.28; Q4: > 3.28. ~d parity QI: ~ 2.64; Q2: 2.65 to 3.14; Q3: 3.15 to 3.64; Q4: > 3.64. Journal of Dairy Science Vol. 73,
No.9, 1990
2570
BANOS AND SHOOK
TABLE 10. Product-moment (PM) correlation coefficients between sire effects for 305-d milk (ML) estimated in different levels of herd avenge somatic cell concentrations (HAVSC), 95% confidence intervals (CI), approximate limits of expectations (EXP) of PM, and genetic correlations (fg) given by the ratio PM:EXP. ML
First pari Q1/Q2 1'y Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4 Secondjarlty Q1/Q2 Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4 Third parity Q1/Q24 Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4
PM
CI
.55 .57 .54 .60 .54 .55
.49 .52 .49 .55 .49 .49
-
.59 .62 .59 .65 .59 .59
.56 .54 .53 .54 .54 .52
-
.59 .57 .57 .58 .58 .56
.93 1.00 .95 1.03 .93 .98
-
.98 1.06 1.02 1.11 1.00 1.06
.53
.48 .53 .54 .44
.47 .39 .42 .47 .48 .37
-
.58 .52 .54 .58 .59 .50
.44 .41 .43 .43 .44 .42
-
.49 .47 .49 .48 .50 .48
1.08 .98 .98 1.10 1.08 .92
-
1.20 1.12 1.12 1.23 .122 1.05
.50 .43 .45 .47 .46 .39
.43 .36 .38 .40 .39 .32
-
.56 .50 .51 .54 .53 .46
.37 .39 .37 .38 .37 .40
-
.46 .47 .46 .46 .47 .46
1.09 .91 .98 1.02 1.02 .85
-
1.35 1.10 1.22 1.24 1.24 .98
.46
EXP
f
g
IThe HAVSC stratification from low (Ql) to high (Q4). 2p irst parity Ql: S 2.30; Q2: 2.31 to 2.64; Q3: 2.65 to 3.02; Q4: > 3.02. 3Second parity Ql: S 2.37; Q2: 2.38 to 2.81; Q3: 2.82 to 328; Q4: > 3.28.
~d parity Ql: S 2.64; Q2: 2.65 to 3.14; Q3: 3.15 to 3.64; Q4: > 3.64.
though there were differences in c2 estimates by level of HAYSC, they showed no trend and they were nonsignificant due to large standard errors associated with them. For ML, c 2 estimates over all records within parity were a little higher than for SC. Within parity by level of HAYSC, they ranged from 1.31 and 7.33% and were in agreement with estimates of others (17, 26, 27). Product-moment correlations between levels of HAYSC for sire effects under Model [1], and approximate limits of expected correlations, are in Tables 9 and 10 for se and ML. Confidence intervals for the observed correlations using Fisher's log transformation are included. Estimates of genetic correlation (rg) for expression of the same genotype in two environments were obtained by the ratio of observed to expected correlations and are also presented in Tables 9 and 10. Lower limits of rg between SCC performance in different environments were above .80, indicating little GxE. For SC, lowest values of rg were between the Journal of Dairy Science Vol. 73,
No.9. 1990
lowest and highest HAYSC subset (Ql and 04) in parities 1 and 2 (Table 9). High estimates of r g for ML (Table 10) demonstrate the similarity of sire effects for ML between herds with different average sec. These estimates of r¥ are sensitive to the effective number of daugnters per sire. In the present study, the average effective number of daughters in each subset ranged from 22.3 to 24.5 in frrst, from 15.9 to 17.8 in second, and from 10.7 to 12.7 in third parity. Sampling error due to the small average effective number of daughters may have caused some estimates of r g to be larger than unity, especially in second and third parities. Phenotypic and Genetic Correlation Between Traits
Estimates of phenotypic (rp ) and genetic correlations (r~) between SC and ML by level of HAYSC WIthin parity are in Table 11. Estimates of rp were always negative, more so in second and third parity than in frrst. This was an expected result, because several reports have
2571
GENOTYPE-ENVIRONMENT INTERACTION
TABLE 11. Phenotypic and genetic correlations between somatic cell concentration (SC) and 305-d milk (ML), by level of herd average somatic cell concentration (HAVSC) within parity. Phenotypic correlation
HAVScl First parity Ql ~ 2.30 Q2 2.31 to 2.64 Q3 2.65 to 3.02 Q4 > 3.02 Total Second parity Ql ~ 2.37 Q2 2.38 to 2.81 Q3 2.82 to 3.28 Q4 > 3.28 Total Third parity Ql ~ 2.64 Q2 2.65 to 3.14 Q3 3.15 to 3.64 Q4 > 3.64 Total
-.04 -.04 -.05 -.08 -.05
Genetic correlation .20 .31 .31 .24
(.10)2 (.09) (.10) (.11)
.24 (.06)
-.06 (.14)
-.14 -.16 -.17 -.17 -.16
-.17 -.11 .12 -.17
(.13) (.14) (.14) (.07)
-.16 -.15 -.16 -.19 -.16
-.29 -.12 -.08 -.17 -.12
(.16) (.17) (.17) (.20)
different genetic factors may influence milk and see in first and later parities. Also culling in frrst parity based on milk yield, mastitis, or both may influence the correlation in later parities, Culling practices would remove low milk producers or potentially high milk producers with mastitis infection and high see. eonsequently, high milk producers with low see would be favored to have second and later parities. Within parity, there were not substantial differences between rg estimates obtained in different HAvse subsets. Genetic Correlation Among Parities
(.09)
I The HAVSC stratification from low (Ql) to high (Q4). 2Approximate standard errors in parentheses.
already shown a similar decline in milk production with increasing see (9, 18). Estimates of rg were positive in first and negative in later parities. Positive r g reflect an antagonistic relationship between milk yield and see, meaning that genetically high milk producers have a tendency toward higher see and greater susceptibility to mastitis. Similar results in first parity have been reported (7, 9, 11). Negative estimates of r between ML and se in later parities have k n observed (9, 19, 21). A possible explanation for this change is that
Product-moment correlations between sire effects for se and ML in different parities and their 95% confidence intervals are in Table 12, Approximate limits of expected values of the correlation coefficients and of estimates of rg between parities are also given. Genetic correlations of frrst with later parities were moderately high for se ranging from .71 to .81, whereas between second and third parity were around unity, eomparative results from the literature for see measures are quite contradictory. Shook et al. (23) estimated rg between measures of lactation see in pairs of the first five parities by simultaneously obtaining estimates of variance and covariance components. They reported rg between adjacent parities ranging between .44 and .77 and averaging .55. Monardes and Hayes (20), however, estimated r g between measures of lactation see in pairs of the first three parities between .90 and .97. For ML, rg estimates of first with later parities were between .77 and .86. Between second and
TABLE 12. Product-moment (PM) correlation coefficients between sire effects for somatic cell concentrations (SC) and 305-d milk (ML) estimated in different parities, 95% confidence intervals (CI), approximate limits of expectations (EXP) of PM, and genetic correlations (rg) given by the ratio PM:EXP.
SC Pl/P21 PIIP3
P2/P3
PM
a
EXP
.39 .39 .52
.33 - .45 .32 - .45 .47 - .57
.50 - .55 .48 - .54 .44 - .50
.78 .72 - .81 1.04 - 1.18
.51 .46 .56
.46 - .56 .40 - .51 .51 - .60
.59 - .62 .55 - .60 .51 - .56
.82 - .86 .77 - .84 1.00 - 1.10
.71 -
ML
PI/P2 PIIP3
nIP3 1PI
= First parity;
P2
= second
parity; P3
= third
parity. Journal of Dairy Science Vol. 73.
No.9. 1990
2572
BANOS AND SHOOK
third parity, correlations were around unity. Similarly, Maijala and Hanna (14), in a review of the literature, reported genetic correlations for milk yield between first and later parities of .80 to .85 and between second and third of .91 to 1.00. They concluded that milk yield may be a somewhat different trait in later parities than in first. A possible explanation for the less than perfect rg of first with later parities observed in the present study is culling based on first parity records, on both milk yield and mastitis. It has been shown that culling reduces correlations between sire evaluations for milk on first and second parity records, under mixed models (12), and modified contemporary comparisons (3). Another explanation is that different sets of genes may influence a trait in first and later parities. H sec in first and later parities are two correlated but different traits, sire evaluation and selection based only on first parity records may not be the most effective scheme in reducing see and mastitis in later parities, although such a strategy would decrease generation interval. Selecting on first parity sec records, however, to reduce overall see is expected to be as effective as selecting on first parity milk yield to increase overall milk yield. Therefore, if the goal is to improve resistance to mastitis and decrease the frequency of the disease across the entire productive life, sire evaluations based on progeny records from all parities should be the method of choice. Total Somatic Cell Production
Phenotypic and rg between se and LT were above .95, indicating that both see traits are influenced by nearly the same genetic and environmental factors. Estimates of rp between LT and ML were around .15 in first parity and .05 in later parities. Estimates of rg between LT and ML were around .50 in first parity and .15 in later parities. These values are much higher than estimates of se with ML due to a partwhole relationship between LT and ML. Because of its part-whole relationship with milk yield, LT is less desirable than se for selection. Results for LT regarding estimation of h 2 , rg between parities, and studies of GxE were sut>: stantially the same as those for se. CONCLUSIONS
Heritability estimates of se and LT did not vary considerably across levels of HAvse, 10urnal of Dairy Science Vol. 73,
No.9. 1990
except at highest HAvse and, consequently, high mastitis incidence. In these herds, which represented approximately one quarter of all data, variance among sires and heritability consistently declined, but differences from the remaining herds were small compared with SE of the estimates. Therefore, sire evaluation based on daughter performance across all herds will sufficiently predict response to selection in the general population. Sire effects for both measures of milk see and lactation milk yield were consistent across herds with different average milk see. Genetic correlations between genotypes of the same sires evaluated in different herd levels of sec were around unity, showing that reranking of sires across environments could be attributed to the random error associated with their estimated transmitting ability. The proportion of total phenotypic variance accounted for by SxH variance was generally low but might have been subject to a small negative bias. Sire by herd irlteraction reflects the similarity between daughters of a sire in the same herd, and represents both GxE and covariances between the records of half-sisters herdmates. Failure to account for this interaction will cause an overestimation of the accuracy of sire evaluation, even though reranking of sires may not be affected considerably, as shown in the present study. The impact of this bias becomes more severe in proofs of sires whose daughters are located in one or very few herds. In data used in this study, on the average only about 10% of sires had daughters with records in less than 5 herds. However, these were somewhat selected data sets because of the specific edits applied The proportion of these sires in the entire sire population may determirle the necessity of incorporating c2 effects in a national sire evaluation scheme. There was evidence of change irl variation of see from first to later parities. This was indicated by changes irl the rg of both sec traits with milk yield Also rg between first and later parities was less than unity for see. Somatic cell count in young and mature ages may be two different but correlated traits. Genetic correlations between second and third parity were around unity, meaning that the same genetic factors influence see in these two parities. Results were similar for milk yield Sire evaluation and selection on first parity records would avoid bias due to culling and decrease the generation interval. Such practice,
GENOTYPE-ENVIRONMENT INTERACTION
however, may not necessarily result in the most efficient reduction of see in later parities. Also, records from later parities may be better indicators of resistance to mastitis, because of more frequent mastitis occurrence, than first parity records. Sire evaluation for see should be based on progeny records in all parities. This approach will also increase the accuracy of sire evaluations. ACKNOWLEDGMENTS
The authors wish to acknowledge the Wisconsin DHIA for providing data for this research. The assistance of E. D. Hallman in assembling the data is gratefully acknowledged. REFERENCES I Ali, A.K.A., and G. E. Shook. 1980. An optimum transformation for somatic cell concentration in milk. J. Dairy Sci. 63:487. 2 Calo, L. L., R. E. McDowell, L. D. Van Vleck, and P. D. Miller. 1973. Genetic aspects of beef production among Holstein-Friesians pedigree selected for milk production. J. Anim. Sci. 37:676. 3 Cassell, B. G., B. T. McDaniel, and H. D. Norman. 1983. Impact of culling on modified contemporary comparisons sire evaluations. J. Dairy Sci. 66:1359. 4 Coffey, E. M., W. E. Vinson, and R. E. Pearson. 1986. Potential of somatic cell concentration in milk as a sire selection criterion to reduce mastitis in dairy cattle. J. Dairy Sci. 69:2163. 5 Dabdoub, SAM., and G. E. Shook. 1984. Phenotypic relationships among milk yield, somatic cell count, and clinical mastitis. J. Dairy Sci. 67(Suppl. 1):163. (Abstr.) 6 DaneU, B. 1982. Interaction between genotype and environment in sire evaluation for milk production. Acta Agric. Scand. 32:33. 7 Emanuelson, U., B. Danell, and J. Philipsson. 1988. Genetic parameters for clinical mastitis, somatic cell counts, and milk production estimated by multiple-trait restricted maximum likelihood. J. Dairy Sci. 71:467. 8 Grootenhuis, G. 1981. Mastitis prevention by selection of sires. Vet. Rec. 108:258. 9 Heuven, H.C.M., H. Bovenhuis, and R. D. Politiek. 1988. Inheritance of monthly somatic cell count and its genetic relationship with milk yield. Livest Prod. Sci. 18:115. 10 Hickman, C. G., A. L. Lee, and K. Gravir. 1969. Genotype x season x method interaction in evaluating dairy sires from progeny records. Can. J. Anim. Sci. 49: 151. 11 Kennedy, B. W., M. S. Setbar, J. E. Moxley, and B. R. Downey. 1982. Heritability of somatic cell count and its relationship with milk yield and composition in Holsteins. J. Dairy Sci. 65:843. 12 Lofgren, D. L., B. G. Casell, H. D. Norman, and B. T. McDaniel. 1983. Effects of culling on sire evaluations by
2573
mixed models. J. Dairy Sci. 66:2418. 13 Lush, J. L. 1950. Inheritance of susceptibility to mastitis. J. Dairy Sci. 33: 121. 14 Maijala, K., and M. Hanna. 1974. Reliable phenotypic and genetic parameters in dairy cattle. Proc. 1st World Coogr. Genet. Appl. livest. Prod. 1:541. 15 Meyer, K. 1985. Maximum likelihood estimation of variance components for a multivariate mixed model with equal design matrices. Biometrics 41:153. 16 Meyer, K. 1986. Restricted maximum likelihood estimation of variance components-in practice. Proc. 3rd World Coogr. Genet. Appl. livest. Prod. 12:454. 17 Meyer, K. 1987. Estimates ofvariances due to sire x herd interactions and environmental covariances between paternal half-sibs for first lactation dairy production. Livest. Prod. Sci. 17:95. 18 Miller, R. H., U. Emanuelson, E. Persson, L. Brolund, J. Philipsson, and H. Funke. 1983. Relationships of milk somatic cell counts to daily milk yield and composition. Acta Agric. Scand. 33:209. 19 Monardes, H. G., and J. F. Hayes. 1985. Genetic and phenotypic relationships between lactation cell counts and milk yield and composition of Holstein cows. J. Dairy Sci. 68:1250. 20 Monardes, H. G., and J. F. Hayes. 1985. Genetic and phenotypic statistics of lactation cell counts in different lactations of Holstein cows. J. Dairy Sci. 68:1449. 21 Monardes, H. G., J. F. Hayes, and J. E. Moxley. 1984. Heritability of lactation cell count measures and their relationships with milk yield and composition in Ayrshire cows. J. Dairy Sci. 67:2429. 22 Shook, G. E. 1986. Genetic aspects of mastitis. Proc. 25th Anon. Mtg. Natl. Mastitis Counc. Arlington, VA. 23 Shook, G. E., F. Ruvuna, and AX.A. Ali. 1982. Genetic parameters of lactation average of somatic cell concentration in milk. Proc. 2nd World Congr. Genet. Appl. livest. Prod. 8:142. 24 Smith, S. P., andH. U. Graser. 1986. Estimating variance components in a class of models by restricted maximum likelihood. J. Dairy Sci. 69:1156. 25 Strandberg, E., and G. E. Shook. 1989. Genetic and economic responses to breeding programs that consider mastitis. J. Dairy Sci. 72:2136. 26 Syrstad, O. 1976. Interaction between genotype and nutrition in dairy production - a review. World. Rev. Anim. Prod. 12:33. 27 Tong, AX.W., B. W. Kennedy, and J. E. Moxley. 1977. Sire by herd interactions for milk yield and composition traits. Can. J. Anim. Sci. 57:383. 28 VanRaden, P. M., and A. E. Freeman. 1985. Rapid method to obtain bounds on accuracies and prediction error variances in mixed models. J. Dairy Sci. 68:2123. 29 Vecht, U., G. E. Shook, R. D. Politiek, G. Grootenhuis, W. J. Koops, and D. G. Groothuis. 1985. Effect of bull selection on somatic cell count in fIrSt lactation on cell counts and pathogens in later lactations. J. Dairy Sci. 68: 2995. 30 Wiggans, G. R., and G. E. Shook. 1987. A lactation measure of somatic cell count. J. Dairy Sci. 70:2666.
Journal of Dairy Science Vol. 73,
No.9, 1990
traits. Between second and third parity genetic correlation estimates were around unity for all traits. Records from all parities should be used for sire evaluation. (Key words: genotype environment interaction, genetic correlation, somatic cell count)
Lactation measures of somatic cell concentration and total sec production were developed. Data were separated into three parity groups. Within parity, five data sets were created: four subsets by herd-year average sec, and one with all records. Records on lactation sec, total sec production, and 305- 3.02. 3Second parity Ql: ~ 2.37; Q2: 2.38 to 2.81; Q3: 2.82 to 3.28; Q4: > 3.28. ~d parity QI: ~ 2.64; Q2: 2.65 to 3.14; Q3: 3.15 to 3.64; Q4: > 3.64. Journal of Dairy Science Vol. 73,
No.9, 1990
2570
BANOS AND SHOOK
TABLE 10. Product-moment (PM) correlation coefficients between sire effects for 305-d milk (ML) estimated in different levels of herd avenge somatic cell concentrations (HAVSC), 95% confidence intervals (CI), approximate limits of expectations (EXP) of PM, and genetic correlations (fg) given by the ratio PM:EXP. ML
First pari Q1/Q2 1'y Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4 Secondjarlty Q1/Q2 Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4 Third parity Q1/Q24 Q1/Q3 Q1/Q4 Q2/Q3 Q2/Q4 Q3/Q4
PM
CI
.55 .57 .54 .60 .54 .55
.49 .52 .49 .55 .49 .49
-
.59 .62 .59 .65 .59 .59
.56 .54 .53 .54 .54 .52
-
.59 .57 .57 .58 .58 .56
.93 1.00 .95 1.03 .93 .98
-
.98 1.06 1.02 1.11 1.00 1.06
.53
.48 .53 .54 .44
.47 .39 .42 .47 .48 .37
-
.58 .52 .54 .58 .59 .50
.44 .41 .43 .43 .44 .42
-
.49 .47 .49 .48 .50 .48
1.08 .98 .98 1.10 1.08 .92
-
1.20 1.12 1.12 1.23 .122 1.05
.50 .43 .45 .47 .46 .39
.43 .36 .38 .40 .39 .32
-
.56 .50 .51 .54 .53 .46
.37 .39 .37 .38 .37 .40
-
.46 .47 .46 .46 .47 .46
1.09 .91 .98 1.02 1.02 .85
-
1.35 1.10 1.22 1.24 1.24 .98
.46
EXP
f
g
IThe HAVSC stratification from low (Ql) to high (Q4). 2p irst parity Ql: S 2.30; Q2: 2.31 to 2.64; Q3: 2.65 to 3.02; Q4: > 3.02. 3Second parity Ql: S 2.37; Q2: 2.38 to 2.81; Q3: 2.82 to 328; Q4: > 3.28.
~d parity Ql: S 2.64; Q2: 2.65 to 3.14; Q3: 3.15 to 3.64; Q4: > 3.64.
though there were differences in c2 estimates by level of HAYSC, they showed no trend and they were nonsignificant due to large standard errors associated with them. For ML, c 2 estimates over all records within parity were a little higher than for SC. Within parity by level of HAYSC, they ranged from 1.31 and 7.33% and were in agreement with estimates of others (17, 26, 27). Product-moment correlations between levels of HAYSC for sire effects under Model [1], and approximate limits of expected correlations, are in Tables 9 and 10 for se and ML. Confidence intervals for the observed correlations using Fisher's log transformation are included. Estimates of genetic correlation (rg) for expression of the same genotype in two environments were obtained by the ratio of observed to expected correlations and are also presented in Tables 9 and 10. Lower limits of rg between SCC performance in different environments were above .80, indicating little GxE. For SC, lowest values of rg were between the Journal of Dairy Science Vol. 73,
No.9. 1990
lowest and highest HAYSC subset (Ql and 04) in parities 1 and 2 (Table 9). High estimates of r g for ML (Table 10) demonstrate the similarity of sire effects for ML between herds with different average sec. These estimates of r¥ are sensitive to the effective number of daugnters per sire. In the present study, the average effective number of daughters in each subset ranged from 22.3 to 24.5 in frrst, from 15.9 to 17.8 in second, and from 10.7 to 12.7 in third parity. Sampling error due to the small average effective number of daughters may have caused some estimates of r g to be larger than unity, especially in second and third parities. Phenotypic and Genetic Correlation Between Traits
Estimates of phenotypic (rp ) and genetic correlations (r~) between SC and ML by level of HAYSC WIthin parity are in Table 11. Estimates of rp were always negative, more so in second and third parity than in frrst. This was an expected result, because several reports have
2571
GENOTYPE-ENVIRONMENT INTERACTION
TABLE 11. Phenotypic and genetic correlations between somatic cell concentration (SC) and 305-d milk (ML), by level of herd average somatic cell concentration (HAVSC) within parity. Phenotypic correlation
HAVScl First parity Ql ~ 2.30 Q2 2.31 to 2.64 Q3 2.65 to 3.02 Q4 > 3.02 Total Second parity Ql ~ 2.37 Q2 2.38 to 2.81 Q3 2.82 to 3.28 Q4 > 3.28 Total Third parity Ql ~ 2.64 Q2 2.65 to 3.14 Q3 3.15 to 3.64 Q4 > 3.64 Total
-.04 -.04 -.05 -.08 -.05
Genetic correlation .20 .31 .31 .24
(.10)2 (.09) (.10) (.11)
.24 (.06)
-.06 (.14)
-.14 -.16 -.17 -.17 -.16
-.17 -.11 .12 -.17
(.13) (.14) (.14) (.07)
-.16 -.15 -.16 -.19 -.16
-.29 -.12 -.08 -.17 -.12
(.16) (.17) (.17) (.20)
different genetic factors may influence milk and see in first and later parities. Also culling in frrst parity based on milk yield, mastitis, or both may influence the correlation in later parities, Culling practices would remove low milk producers or potentially high milk producers with mastitis infection and high see. eonsequently, high milk producers with low see would be favored to have second and later parities. Within parity, there were not substantial differences between rg estimates obtained in different HAvse subsets. Genetic Correlation Among Parities
(.09)
I The HAVSC stratification from low (Ql) to high (Q4). 2Approximate standard errors in parentheses.
already shown a similar decline in milk production with increasing see (9, 18). Estimates of rg were positive in first and negative in later parities. Positive r g reflect an antagonistic relationship between milk yield and see, meaning that genetically high milk producers have a tendency toward higher see and greater susceptibility to mastitis. Similar results in first parity have been reported (7, 9, 11). Negative estimates of r between ML and se in later parities have k n observed (9, 19, 21). A possible explanation for this change is that
Product-moment correlations between sire effects for se and ML in different parities and their 95% confidence intervals are in Table 12, Approximate limits of expected values of the correlation coefficients and of estimates of rg between parities are also given. Genetic correlations of frrst with later parities were moderately high for se ranging from .71 to .81, whereas between second and third parity were around unity, eomparative results from the literature for see measures are quite contradictory. Shook et al. (23) estimated rg between measures of lactation see in pairs of the first five parities by simultaneously obtaining estimates of variance and covariance components. They reported rg between adjacent parities ranging between .44 and .77 and averaging .55. Monardes and Hayes (20), however, estimated r g between measures of lactation see in pairs of the first three parities between .90 and .97. For ML, rg estimates of first with later parities were between .77 and .86. Between second and
TABLE 12. Product-moment (PM) correlation coefficients between sire effects for somatic cell concentrations (SC) and 305-d milk (ML) estimated in different parities, 95% confidence intervals (CI), approximate limits of expectations (EXP) of PM, and genetic correlations (rg) given by the ratio PM:EXP.
SC Pl/P21 PIIP3
P2/P3
PM
a
EXP
.39 .39 .52
.33 - .45 .32 - .45 .47 - .57
.50 - .55 .48 - .54 .44 - .50
.78 .72 - .81 1.04 - 1.18
.51 .46 .56
.46 - .56 .40 - .51 .51 - .60
.59 - .62 .55 - .60 .51 - .56
.82 - .86 .77 - .84 1.00 - 1.10
.71 -
ML
PI/P2 PIIP3
nIP3 1PI
= First parity;
P2
= second
parity; P3
= third
parity. Journal of Dairy Science Vol. 73.
No.9. 1990
2572
BANOS AND SHOOK
third parity, correlations were around unity. Similarly, Maijala and Hanna (14), in a review of the literature, reported genetic correlations for milk yield between first and later parities of .80 to .85 and between second and third of .91 to 1.00. They concluded that milk yield may be a somewhat different trait in later parities than in first. A possible explanation for the less than perfect rg of first with later parities observed in the present study is culling based on first parity records, on both milk yield and mastitis. It has been shown that culling reduces correlations between sire evaluations for milk on first and second parity records, under mixed models (12), and modified contemporary comparisons (3). Another explanation is that different sets of genes may influence a trait in first and later parities. H sec in first and later parities are two correlated but different traits, sire evaluation and selection based only on first parity records may not be the most effective scheme in reducing see and mastitis in later parities, although such a strategy would decrease generation interval. Selecting on first parity sec records, however, to reduce overall see is expected to be as effective as selecting on first parity milk yield to increase overall milk yield. Therefore, if the goal is to improve resistance to mastitis and decrease the frequency of the disease across the entire productive life, sire evaluations based on progeny records from all parities should be the method of choice. Total Somatic Cell Production
Phenotypic and rg between se and LT were above .95, indicating that both see traits are influenced by nearly the same genetic and environmental factors. Estimates of rp between LT and ML were around .15 in first parity and .05 in later parities. Estimates of rg between LT and ML were around .50 in first parity and .15 in later parities. These values are much higher than estimates of se with ML due to a partwhole relationship between LT and ML. Because of its part-whole relationship with milk yield, LT is less desirable than se for selection. Results for LT regarding estimation of h 2 , rg between parities, and studies of GxE were sut>: stantially the same as those for se. CONCLUSIONS
Heritability estimates of se and LT did not vary considerably across levels of HAvse, 10urnal of Dairy Science Vol. 73,
No.9. 1990
except at highest HAvse and, consequently, high mastitis incidence. In these herds, which represented approximately one quarter of all data, variance among sires and heritability consistently declined, but differences from the remaining herds were small compared with SE of the estimates. Therefore, sire evaluation based on daughter performance across all herds will sufficiently predict response to selection in the general population. Sire effects for both measures of milk see and lactation milk yield were consistent across herds with different average milk see. Genetic correlations between genotypes of the same sires evaluated in different herd levels of sec were around unity, showing that reranking of sires across environments could be attributed to the random error associated with their estimated transmitting ability. The proportion of total phenotypic variance accounted for by SxH variance was generally low but might have been subject to a small negative bias. Sire by herd irlteraction reflects the similarity between daughters of a sire in the same herd, and represents both GxE and covariances between the records of half-sisters herdmates. Failure to account for this interaction will cause an overestimation of the accuracy of sire evaluation, even though reranking of sires may not be affected considerably, as shown in the present study. The impact of this bias becomes more severe in proofs of sires whose daughters are located in one or very few herds. In data used in this study, on the average only about 10% of sires had daughters with records in less than 5 herds. However, these were somewhat selected data sets because of the specific edits applied The proportion of these sires in the entire sire population may determirle the necessity of incorporating c2 effects in a national sire evaluation scheme. There was evidence of change irl variation of see from first to later parities. This was indicated by changes irl the rg of both sec traits with milk yield Also rg between first and later parities was less than unity for see. Somatic cell count in young and mature ages may be two different but correlated traits. Genetic correlations between second and third parity were around unity, meaning that the same genetic factors influence see in these two parities. Results were similar for milk yield Sire evaluation and selection on first parity records would avoid bias due to culling and decrease the generation interval. Such practice,
GENOTYPE-ENVIRONMENT INTERACTION
however, may not necessarily result in the most efficient reduction of see in later parities. Also, records from later parities may be better indicators of resistance to mastitis, because of more frequent mastitis occurrence, than first parity records. Sire evaluation for see should be based on progeny records in all parities. This approach will also increase the accuracy of sire evaluations. ACKNOWLEDGMENTS
The authors wish to acknowledge the Wisconsin DHIA for providing data for this research. The assistance of E. D. Hallman in assembling the data is gratefully acknowledged. REFERENCES I Ali, A.K.A., and G. E. Shook. 1980. An optimum transformation for somatic cell concentration in milk. J. Dairy Sci. 63:487. 2 Calo, L. L., R. E. McDowell, L. D. Van Vleck, and P. D. Miller. 1973. Genetic aspects of beef production among Holstein-Friesians pedigree selected for milk production. J. Anim. Sci. 37:676. 3 Cassell, B. G., B. T. McDaniel, and H. D. Norman. 1983. Impact of culling on modified contemporary comparisons sire evaluations. J. Dairy Sci. 66:1359. 4 Coffey, E. M., W. E. Vinson, and R. E. Pearson. 1986. Potential of somatic cell concentration in milk as a sire selection criterion to reduce mastitis in dairy cattle. J. Dairy Sci. 69:2163. 5 Dabdoub, SAM., and G. E. Shook. 1984. Phenotypic relationships among milk yield, somatic cell count, and clinical mastitis. J. Dairy Sci. 67(Suppl. 1):163. (Abstr.) 6 DaneU, B. 1982. Interaction between genotype and environment in sire evaluation for milk production. Acta Agric. Scand. 32:33. 7 Emanuelson, U., B. Danell, and J. Philipsson. 1988. Genetic parameters for clinical mastitis, somatic cell counts, and milk production estimated by multiple-trait restricted maximum likelihood. J. Dairy Sci. 71:467. 8 Grootenhuis, G. 1981. Mastitis prevention by selection of sires. Vet. Rec. 108:258. 9 Heuven, H.C.M., H. Bovenhuis, and R. D. Politiek. 1988. Inheritance of monthly somatic cell count and its genetic relationship with milk yield. Livest Prod. Sci. 18:115. 10 Hickman, C. G., A. L. Lee, and K. Gravir. 1969. Genotype x season x method interaction in evaluating dairy sires from progeny records. Can. J. Anim. Sci. 49: 151. 11 Kennedy, B. W., M. S. Setbar, J. E. Moxley, and B. R. Downey. 1982. Heritability of somatic cell count and its relationship with milk yield and composition in Holsteins. J. Dairy Sci. 65:843. 12 Lofgren, D. L., B. G. Casell, H. D. Norman, and B. T. McDaniel. 1983. Effects of culling on sire evaluations by
2573
mixed models. J. Dairy Sci. 66:2418. 13 Lush, J. L. 1950. Inheritance of susceptibility to mastitis. J. Dairy Sci. 33: 121. 14 Maijala, K., and M. Hanna. 1974. Reliable phenotypic and genetic parameters in dairy cattle. Proc. 1st World Coogr. Genet. Appl. livest. Prod. 1:541. 15 Meyer, K. 1985. Maximum likelihood estimation of variance components for a multivariate mixed model with equal design matrices. Biometrics 41:153. 16 Meyer, K. 1986. Restricted maximum likelihood estimation of variance components-in practice. Proc. 3rd World Coogr. Genet. Appl. livest. Prod. 12:454. 17 Meyer, K. 1987. Estimates ofvariances due to sire x herd interactions and environmental covariances between paternal half-sibs for first lactation dairy production. Livest. Prod. Sci. 17:95. 18 Miller, R. H., U. Emanuelson, E. Persson, L. Brolund, J. Philipsson, and H. Funke. 1983. Relationships of milk somatic cell counts to daily milk yield and composition. Acta Agric. Scand. 33:209. 19 Monardes, H. G., and J. F. Hayes. 1985. Genetic and phenotypic relationships between lactation cell counts and milk yield and composition of Holstein cows. J. Dairy Sci. 68:1250. 20 Monardes, H. G., and J. F. Hayes. 1985. Genetic and phenotypic statistics of lactation cell counts in different lactations of Holstein cows. J. Dairy Sci. 68:1449. 21 Monardes, H. G., J. F. Hayes, and J. E. Moxley. 1984. Heritability of lactation cell count measures and their relationships with milk yield and composition in Ayrshire cows. J. Dairy Sci. 67:2429. 22 Shook, G. E. 1986. Genetic aspects of mastitis. Proc. 25th Anon. Mtg. Natl. Mastitis Counc. Arlington, VA. 23 Shook, G. E., F. Ruvuna, and AX.A. Ali. 1982. Genetic parameters of lactation average of somatic cell concentration in milk. Proc. 2nd World Congr. Genet. Appl. livest. Prod. 8:142. 24 Smith, S. P., andH. U. Graser. 1986. Estimating variance components in a class of models by restricted maximum likelihood. J. Dairy Sci. 69:1156. 25 Strandberg, E., and G. E. Shook. 1989. Genetic and economic responses to breeding programs that consider mastitis. J. Dairy Sci. 72:2136. 26 Syrstad, O. 1976. Interaction between genotype and nutrition in dairy production - a review. World. Rev. Anim. Prod. 12:33. 27 Tong, AX.W., B. W. Kennedy, and J. E. Moxley. 1977. Sire by herd interactions for milk yield and composition traits. Can. J. Anim. Sci. 57:383. 28 VanRaden, P. M., and A. E. Freeman. 1985. Rapid method to obtain bounds on accuracies and prediction error variances in mixed models. J. Dairy Sci. 68:2123. 29 Vecht, U., G. E. Shook, R. D. Politiek, G. Grootenhuis, W. J. Koops, and D. G. Groothuis. 1985. Effect of bull selection on somatic cell count in fIrSt lactation on cell counts and pathogens in later lactations. J. Dairy Sci. 68: 2995. 30 Wiggans, G. R., and G. E. Shook. 1987. A lactation measure of somatic cell count. J. Dairy Sci. 70:2666.
Journal of Dairy Science Vol. 73,
No.9, 1990
