Direct Response in Yield and Correlated Response in Components Accompanying Selection for Milk Yield in Jerseys R. R. BONCZEK,1 D. O. RICHARDSON,2 E. D. MOORE,1 R. H. MILLER,3 J. R. OWEN,1 H. H. DOWLEN,4 B. R. BELL, 2 and W. K. LANGHOLFF5 Agricultural Research 5ervlce. USDA and University of TaMessee Lewisburg 37091 ABSTRACT
test for group differences. Groups differed for all traits. Selection was superior to control in breeding value for milk and fat (828 and 31 kg) and for production of milk. fat, and 4% FCM (1066, 42, and 1061 kg). Control was superior to selection in breeding value and production fat test (.15 and .12%). Group differences existed within generation class for all yield traits but not for fat percentage. Realized response closely matched or exceeded expected response as estimated from pedigree information. (Key words: selection, milk yield, direct response)
In 1967, the Jersey herd at the Dairy Experiment Station, Lewisburg, TN was divided into two groups on the basis of ancestry, type, and breeding value for milk as part of a project to detennine effects of single trait selection for high milk yield on yield and correlated traits. Control group was mated randomly to 20 unproven young sires selected randomly from those available from breeding studs in 1967. Selection group was mated to sires selected solely on the basis of their high transmitting ability for milk. Selection sires were selected at intervals and used for 4 yr. By the end of the project (1984), lactation information was available on 672 daughters (520 selection and 152 control) of 37 bulls (17 selection and 20 control). Differences in breeding values for milk, fat, and fat test as calculated from the YfA reported in the July 1989 USDA genetic evaluations and differences in first lactation mature equivalent production of milk. fat, fat test, and 4% FCM were examined. Linear mixed models were used for all analyses and contained the fixed effects group, generation within group, and year. Sires were random, nested within group, and used to
Abbreviation key: BV = breeding value, EBV milk or selection genetic group, ME = DHIA-adjusted lactation record, PDM = PO milk. R % = repeatability, RBV = realized BV, YS = young sire or control genetic group.
= expected BV, HM = high
INTRODUCTION
Received January 31, 1991. Accepted April 24, 1991. lARS-USDA, Dairy Experiment Station. 2Department of Animal Science, University of Tennessee, Knoxville 37901-1071. 3ARS-USDA, Milk Secretion and Mastitis Laboratory, Beltsville, MD 20705. "Dairy Experiment Station, University of Tennessee. SARS-USDA, Delta States Research Center, Stoneville, MS 38776. 1991 J Dairy Sci 74:3209-3222
Comparisons of expected and realized responses to selection for economically important traits in dairy cattle is important in determining whether recommendations derived from selection theory are directly applicable to the dairy industry. Of the many traits that affect economic efficiency of dairy cattle, yields of milk and its components are the most important. Stott and Delorenzo (21) reported that milk yield was the single most important determinant of profitability in Jerseys. Similarly, Andrus and McGilliard (2) reported that milk yield was the most important factor in determining profit per year of herd life in dairy cattle, and Pearson (13) determined that it is highly profitable for cows to produce at the highest level possible.
3209
3210
BONCZEK ET AL.
Rapid improvement in genetic potential and actual yield of milk in dairy cattle has been achieved from selection (I 5). However, realized response to selection tends to deviate from expected when selection is on sire PD milk (PDM), and level and direction of deviation tend to vary with data sample source. Reports using field data (15, 16) indicate lower actual response to selection on sire PDM than data from experimental herds (8, 9, 10, 12, 14). In all but the lowest producing herd classes in their study, Powell and Norman (15) determined that response to sire selection within herd was greater than expected for all breeds and improved with increasing average milk yield. To examine realized responses to long-term selection for high milk yield in dairy cattle, the NC-2 and S-49 Regional Projects established several selection programs in the 1960s and 1970s. Reports from these long-term selection projects have concluded that selection for milk yield results in improved yield of milk in Holsteins (3, 4, 5, 6, 8, 9, 10, 11, 12, 14) and Jerseys (24) that exceeds (3, 5, 9, 14, 24) or closely matches (12) that predicted from the PDM of sires used. However, they also conclude that a decrease in percentage of milk components (3, 5, 12, 14, 23) results, even though actual yields of milk components increase (3, 5, 9, 10, 11, 12, 14, 23). The objective of this study is to examine results of long-term selection for milk yield in Jerseys. Specific objectives are 1) to evaluate the direct response of milk yield to single trait selection for milk yield and 2) to determine the correlated response of milk components to single trait selection for milk yield. Correlated responses in some other traits studied in this project have been reported (18, 19, 20). MATERIALS AND METHODS Project Protocol
Group Formation. In 1967, the Jersey herd at the Dairy Experiment Station, Lewisburg, TN was divided randomly into two breeding groups. Factors balanced between cow groups during assignment were sire, age, type, and breeding value (BV) for milk. Approximately one-third (n = 80) of the breeding herd was assigned to a group to be bred to young sires Joumal of Dairy Science Yo:. 74, No.9, 1991
(YS), and the remainder (n = 184) was assigned to a group to be bred to sires selected solely for high milk yield transmitting ability (HM).
Younger animals were assigned to either HM or YS on the basis of their ancestry; i.e., not all young stock were assigned randomly. All daughters of Bruce Westfield Cherry Noble (546137), Marlu Milestone Milkboy (584458), and Brigham Confident Infred (549809) were assigned to HM as foundations, and all daughters of Tenn Jeweler FOE Lad (595125) and Tenn TOlOno GDO Lad (599997), unproven clean-up bulls, were assigned to YS as foundations. After initial assignment, a ratio of 3 HM animals to 1 YS animal was maintained within the herd. Management. Both groups were managed similarly except for one voluntary culling criterion and the mating plan. Animals from both groups were housed, treated, and fed as a single unit. Housing was in free-stall barns, and milking was in a double-two walk-through parlor. Animals were placed on pasture when possible and received supplemental concentrates in the milking parlor according to production level. Involuntary culling criteria were the same in both groups. Reproductive criteria allowed for culling of producing animals not pregnant by 305 d postpartum or after six services. Young stock were culled as nonbreeders if they had not conceived by 21 mo of age. Animals aborting or reacting to blood test for brucellosis were culled. Animals were culled for production if they declined 40% or more in weekly average milk yield due to illness and did not recover in a reasonable period of time, or if they consistently produced abnormal milk. Functional criteria allowed culling of any animal suffering an obvious injury or having an abnormality that seriously impaired its ability to function. Voluntary culling was practiced within both groups. Excess YS animals were culled randomly if group size became too large, and in later years an effort was made to cull earlier generation YS animals preferentially. Lower producing HM animals were culled voluntarily if group size became too large. Voluntary culling of young stock was minimal in both groups because all females were to be given every opportunity to produce a first
DIRECf RESPONSE TO SINGLE TRAIT SELECTION
lactation record to at least 90 d. Preliminary analyses of survivability data indicate that total losses prior to first calving, although high, did not differ signifIcantly between groups. Some HM females were sold prior to calving for promotional and breeding purposes. Mating Plan and Sire Selection. The mating plan for both groups precluded any close matings (inbreeding >6.25%). Animals in both groups were bred artifIcially at least three times before being exposed to natural service. Originally, progeny of natural service sires were to be included in the project as YS animals regardless of dam's group; however, preliminary analyses of direct response data indicated a positive genetic trend over generations in the YS mean milk BV if daughters of these sires were included in the analyses. Therefore, daughters of these sires and their offspring were removed from the data set to stabilize the YS average milk BV. Except to avoid close matings, all YS matings were random and used semen from 20 bulls selected randomly from those available in sampling programs of the various AI organizations in 1967. An effort was made to equalize use within years, and each sire was available for use in all years until its semen supply was exhausted. Information on YS sires including PTA from the July 1989 bull evaluations is in Table 1. Except to avoid close matings, all HM matings were random. Sires used in HM were selected at intervals and used for a limited period. Requirements for selection were a minimum of 30 AI daughters in at least 20 herds and a repeatability (R %) of at least 60%. Lower R % sires (40 to 60%) were considered for selection if the probability of their being superior to a sire meeting the prior criteria was 75%. Eligible sires were ranked on PDM, and the sire with the highest PDM was chosen if the fertility of his semen was acceptable (average services per conception, 3 or fewer) and if semen price per unit was not more than $5 more than the eligible group's average price per unit. Fertility and price requirements were instituted to ensure that HM sire selection would be similar to that of commercial dairies with registered cattle. Approximately four HM sires were used in every year. Each HM sire was used for about 4 yr with nearly equal numbers of services
3211
planned for each sire. After 4 yr of service and exhaustion of semen supply, use of the HM sire was discontinued; however, the sire was eligible for reselection and use for another 4 yr. Information on HM sires including PTA from the July 1989 bull evaluations is in Table
2. For both genetic groups, sire use over generations was severely unbalanced (as expected); however, because of overlapping generations within the herd, all but three sires (codes 101, 103, and 206) had producing offspring in two or more generations. Only 7 of the 37 sires failed to have producing offspring in three or more generations (codes 101, 102, 103, 104, 105, 206, and 216). Data
Data used to examine direct and correlated responses in yield traits were BV and first lactation production information and included females born between January 1, 1968 and December 31, 1981. Data collection started in January 1968 and stopped after the last born YS female completed her f'ITst lactation in 1984. Dispersal of YS began shortly thereafter. Number born in each year group is in Table 3. Genetic merit data consisted of each animal's realized (RBV) and expected BV (EBV) for milk, fat, and fat percentage as calculated from the July 1989 USDA PTA. Protein information was not included because it was not collected from the herd during the project. The RBV was calculated as twice the animal's PTA. The EBV were determined by summing the PTA of each animal's parents. AU genetic merit estimates were to the 1985 birth year base; hence, most estimates were negative. To be included in the genetic merit data set, an animal was required to have at least one official lactation record. Therefore, data were not available on some animals due to culling or death before or soon after f'ITst calving (n = 378). Also, 6 animals' estimates were not available for unknown reasons. The edited genetic merit data included 672 females (520 lIM and 152 YS). Observations for an additional 78 YS females were excluded because they were sired by or were offspring of a female sired by a herd bred young sire. As noted earlier, preliminary analyses indicated Journal of Dairy Science Vol. 74, No.9, 1991
...
~
0
....NN
[ 0
..... 0 ~. Vol
~. n
TABLE 1. Sires used in the young sire group.
Brigham Confident Infred Sybil Royal Predictor Sybil Carletta's FLOalist Nobleman's Lotus Designer Observer Signal Revival Observer O1ocolate Soldier Milestones Generator Secret Baronet S5 Quicksilver of Pal1neva Vaucluse Sleeping Swville Millcboy Happy Hill Abe Supreme Virginian Glen Meadows 1eweled Mark Briarcliff Bold Torono Starn General Favorite Saint Briarcliffs Soldier Boy
Regislnilion Project code number 549809 573431 577474 583958 578382 596832 602658 6027S4 593883
101 102 103 104 105 106 107
lOS
604694
109 110
613165 615053 606981 616022 619994 623158 620738
112 113 114 115 116 117
III
1101y 1989 evaluation. All values in kilograms.
~.
2Rcliability for milk, fat, and fat percentage.
n
3Rcliability for protein and protein precentage.
Number of daughters Milk
10 56
36 66 9 41 109 52 64 65 41 69 45 47 35 8 12
Predicted
transmiltin8
Fat
Fat %
Rel2
-210 -221 --82 -104 -50 +116 +27 -191 -217
-14 -5 +2 -5
93 97 97
+3
-.06 +.10 +.11 +.01 -.10 -.05
-13
-.24
99
-1 +5
-224
~
+.14 +.27 +.07
-105 +69 -168 +144 +319 +194
-1 -4 -10
+.cn
-10
-.27 -.21 -.21 -.09
99 99 99 99 99 98 99 99 99 99
+406
--8
+2 -4 +14
-.12
-.04
99 97
99
t'
abilityl Protein -9 -8 -2 -3 +1 +4 -7 -5 +1 -7 -2 +2 -7 -1 +4
+0 +10
Years
~
Protein %
Relp3
of use
~
-.03 +.01 +.01 +.01 +.05 .00 -.14 +.04 +.17 +.02 +.02 -.02 -.01 -.11 -.13 -.12 -.08
67 83 88 98 86 98
1967 1967-1971 1967-1970 1967-1974 1969-1970 197(}"1974 197(}..1978 1971-1975 1971-1976 1972-1976 1975--1978 1976-1980 1976-1980 1978-1980 1978-1980 1980 1980
~
99 99 99 99 98 98 97 98
99 99 99
~
Z
en til
a tn
~
I;;
~
tn
~g Z
~ -J .~
~ ~
...
...
~ ~
\,;J
....IV \,;J
3214
BONCZEK ET AL.
TABLE 3. Number of females entering project by year of binh. Group Year
1968 1969 1970 1971 1972
1973 1974 1975 1976 1977 1978 1979 1980 1981 Total
High milk
Young sire
29
21
46
15 20 18
57 50 53 58 61 68 52 64
56 57 62 52
765
26 25
17 25 23 26 19
observation was used to investigate genetic merit; the general linear models procedure of SAS was used (17). The model fit to all genetic merit traits was Yijklm
= J.1
+ ~ + G(L)j(i) + S(L)k(i) + T I + ~jkIm [1]
where Yijklm
17 20 19 291
that herd bred young sires were markedly superior to the original 20 YS sires in transmitting ability for milk. Therefore, their daughters gave a positive trend over generations to the average genetic merit of YS that was undesirable in the control group. Further standardization of genetic merit data included setting all generation ~5 animals to generation 5 due to paucity of higher generation number animals. The number providing data by generation and group is in Table 4. Another analysis was of genetic merit data consisting of the RBV of foundation animals calculated from PIA. These analyses were performed to determine whether the herd's initial randomization was balanced. Data on 303 foundation animals (214 HM and 89 YS) were available. Production data were DHIA-adjusted first lactation values (ME) for milk, fat, fat percentage, and 4% FCM. Information on 623 females (479 lIM and 144 YS) was available. To be included in production data, the first lactation had to meet each of the factors mentioned with genetic merit data and be at least 125 d in length. Only first lactations were used to minimize culling bias. Analyses
Genetic Merit Data. A linear mixed model with observations weighted by reliability of Journal of Dairy Science Vol. 74, No.9, 1991
J.1
~ G(L)j(i) S(L~(i)
Tl ~jIdm
= the
breeding value of female m from generation j and sire k of group i and born in project year 1; = the overall mean; = fixed effect of group i; = fixed effect of generation of selection j nested in group i; = random effect of sire k nested within group i; fixed effect of project year of = birth I; and = random error common to all observations.
The interaction, generation by sire within group, was included in the initial model but pooled with the error tenn after it was found to be not signifIcant (P > .25) (1). Type ill sums of squares from the general linear models procedure were used in all tests of signifIcance (17). Random error was used to test the fixed effects of year and generation nested within group and the random effect of sires nested within group. The effect of group was tested against sires nested within group. Weighted analyses were used for genetic merit to account for the heterogeneity of variance in PIA. Reliability was determined to be the most appropriate weighting variable because it most closely reflected the differing amounts of information incorporated in each estimate of genetic merit (22). Results of analyses without weighting did not vary markedly from weighted analyses. A similar model [2] including only the fixed effect of group and the random effects of sire within group was used to analyze foundation genetic merit. Methodology of testing was identical with model [1].
DIRECT RESPONSE TO SINGLE TRAIT SELECTION
Production Data. For production data, the following mixed linear model was used: Yijklm = 11 + Li + G(L)j(i) + S(L)k(i) + T, + eijklm [3] where Yijklm
=
11 G(L)j(i)
= = =
S(L)k(i)
=
T1
=
t1jklm
=
l.j
ME production of female m from generation j and sire k of group i calving in year-season l', the overall mean; fixed effect of group i; fixed effect of generation of selection j nested in group i; random effect of sire k nested within group i; fixed effect of year-season of calving 1; and random error common to all observations.
All testing used the random error term except for group effect, which was tested against sires within group. The generation by sire within group interaction was included in the initial model but was pooled with the error term after testing determined that it was not significant (P> .25) (1). For these analyses, least squares means were calculated for each generation within group as follows (1): Yij ..
= 11
+ l.j + G(L)j(i) + S.. + T.
where
y Ij.. .. = least squares mean ME production level for generation j of group i; S = average of effect of sire in group; and T = average effect of year-season of calving.
Expected Versus Realized Response. Two approaches were used to compare expected and realized selection responses in milk yield and correlated yield traits. In the first approach, differences of EBV between groups for each generation were compared with the differences between generation within group least squares means from the genetic merit and production
3215
analyses. For these comparisons. the standard errors of differences were generated by computing the square root of the sum of the variances of the estimates, and simple t tests were performed. In the second approach, EBV were regressed on RBV. Analyses were performed both within and across groups. For analyses within group, the initial model contained generation as a fixed effect, EBV as a covariate, and their interaction. As with model [1], regression analyses were weighted by reliability of BV estimate. Comparisons between groups of coefficients and intercepts were made using simple t tests. For regressions over groups, group was included in the model as a fixed effect in addition to generation, and generation was nested within group. All interactions between group, generation, and the covariate, EBY, were examined. RESULTS
Genetic Merit
A summary of significant effects for realized BV for milk, fat, and fat percentage is in Table 5. Group and sire within group differed significantly for each dependent variable. Generation within group differed significantly for BY milk and BV fat but not for BY fat percentage. Year born did not signifIcantly affect any dependent variable. Group least squares means for each dependent variable are in Table 6. As expected from the selection practiced, HM was superior to YS in kilograms of product produced; however, the opposite was true for BV fat percentage. Of more interest are the generation within group least squares means for BV milk and BV fat in Table 7. For YS, a slightly negative trend existed for both traits from generations 1 to 4 with a slight nonsignificant recovery in the last generation group. The HM had a consistent positive trend for both traits with greater improvement in earlier versus later generations. For BV milk, differences between generation least squares means within YS approached significance for generation 1 versus 2 (P < .10) and for generation 1 versus 3 (P < .10). Within HM, the generation 1 least squares mean for BY milk differed significantly from all later estimates (P < .01), but estimates did not differ }omnal of Dairy Science Vol. 74, No.9, 1991
3216
BONCZEK ET AL.
TABLE 4. Number of observations by generation and genetic group. Group Generation
High milk
1
172
2 3 4
155
~5
Total
113
55 25 520
Young sire
45 41 30 24 12 152
significantly for later generations. Between groups, BV milk least squares means for generations differed significantly for each generation class (P < .01); HM had the higher estimate in each case. However, improvement of HM advantage was significant only between generations 1 and 2. For BV fat, differences between generation least squares means within YS approached significance for generation I versus 3 (P < .10) and were significant for generation 1 versus 4 (P < .05) but not elsewhere. Within HM, the generation I least squares means for BV fat differed significantly from all later estimates (P < .01), and the level for generation 4 exceeded significantly that of generation 2 (P < .01). As with BV milk, least squares means for generations differed significantly for groups within each generation class (P < .01); HM had the higher estimate in each case. However, significant improvement in HM advantage over YS existed only between generations 1 and 2. Realized BV of the foundation population did not differ significantly between groups for any trait. Least squares means for foundation generation BV milk, BV fat, and BV fat percentage are in Table 8. These estimates should not be compared directly with estimates from later generations because they are from separate analyses.
tion within group was significant for ME fat and approached significance for ME milk and ME FCM. Group least squares means for each dependent variable are in Table 10. As with genetic merit data, HM was superior to YS for both milk and fat produced but not for fat percentage. Additionally, ME FCM was greater in HM than in YS. Generation within group least squares means for ME milk., ME fat, and ME FCM are in Table 11. For all three traits, HM least squares means were highest in later generations. However, generation 3 estimates had a nonsignificant decrease from generation 2 for each trait. The YS decreased between generations 1 and 3 for all traits but increased thereafter. For ME milk., ME fat, and ME FCM, YS and HM generation least squares means did not differ significantly for generations within group. Significant differences for groups within generation existed, and the patterns of these differences were similar for all three traits. Least squares means of HM were significantly larger than those of YS in generations 2 (P < .01) and 3 (P < .10 for ME milk and P < .05 for ME fat and ME FCM) but not other generations. Change in advantage of HM over YS between generations was not significant in any case. ExpeCted Versus Realized Response to selectIon
Least Squares Means. A comparison of group differences in EBV milk., least squares means ME milk, and least squares means RBV milk by generation and over all generations is in Table 12. Over all generations and for all but generation 4, the realized response, as measured by the difference in ME milk least squares means, exceeded that predicted by the difference in group EBV milk. However, realized response did not differ from expected response as measured by difference in ME milk.
First Lactation Production
A summary of significant sources of variation in ME production of milk., fat, fat percentage, and 4% FCM is in Table 9. Effects of group, sire within group, and year-season of calving were significant for all traits. GeneraJournal of Daily Science Vol. 74, No.9, 1991
Differences between EBV milk and RBV milk for generations 1, 2, and ~ approached significance (P < .10). For generations 1 and 2, realized response as estimated by RBV milk was greater than EBV milk. For generations ~, EBV milk was greater than RBV milk, and EBV milk was greater than RBV milk for
3217
DIRECf RESPONSE TO SINGLE 1RAIT SELECTION TABLE 5. Significance level of independent variables for genetic merit (breeding value). Independent variable Trait
Group
Sire (group)
Generation (group)
Year
Milk.
*** *** ***
*** *** ***
• •••
NS l NS NS
Fat Fat %
NS
INS (P > .10).
*p < .05.
***p < .001.
generations 3 and 4, but these differences did not approach significance. Over all generations, the realized response in BV did not differ significantly from that expected. A comparison of group differences in EBV fat, least squares means ME fat, and least squares means RBV fat is in Table 12. In all cases, realized response as estimated by ME fat or RBV fat did not differ significantly from expected response. Regression. A summary of regressions of EBV on RBV is in Table 13. Within HM, regression coefficients were greater than or equaled 1.0 for all traits. Coefficients for BV milk and BV fat were significantly greater (P < .01) than 1.0. Intercepts for HM regressions were significantly greater (P < .01) than 0 for all regressions except BV fat percentage. Regression coefficients of YS were significantly greater than 1.0 for BV milk (P < .01) and BV fat (P < .01). The regression coefficient for BV fat percentage did not differ significantly from 1.0. Intercepts from YS regressions were significantly greater than 0 for all regressions but BV fat percentage. Using data from both groups, regression coefficients and intercepts were not signifi-
cantly different between groups. For BV milk and BV fat, regression coefficients significantly exceeded 1.0 (P < .01), and the intercepts significantly exceeded 0 (P < .01). For BV fat percentage, the regression coefficient from combined data did not differ from 1.0, and the intercept did not differ from O. DISCUSSION
Direct Response In Milk Yield Selection for milk production using HM sires was effective. However, direct response to selection as measured by difference between selection (HM) and control (YS) genetic merit for milk yield (BV milk) or first lactation milk yield (ME milk) was greater in earlier generations than later. In later generations, the advantage of HM increased at a progressively slower rate (BV milk) or actually decreased (ME milk). Others (3, 5, 6, 8, 9, 10, 11, 12, 14) reported positive direct response to selection for milk yield in Holsteins, and Wilk and McDaniel (24) reported positive direct response in Jerseys. They also reported variation in rate of gain or loss of advantage over generations.
TABLE 6. Least squares means and standard errors by group for breeding value. Genetic group Trait
High milk.
X
SE
X
SE
X
SE
Milk, kg
-147 -9
53.0 2.2 .04
-976
82.3
-39
3.4
+829 +30
97.9 4.0
Fat, kg Fat %
-.02
Young sire
.13
.07
-.15
.08
lHigh milk. minus young sire. Journal of Dairy Science Vol. 74, No.9, 1991
3218
BONCZEK ET AL.
TABLE 7. Least squares means and standard errors by generation within group for breeding values for milk and fat. Genetic group
Milk, kg Generation 1 2 3 4 ~5
Fal, kg Generation 1 2 3 4 ~
Change in difference
Difference!
Young sire
High milk
SE
X
SE
-984 -1001 -1021 -984
40.8 44.9 52.3 58.8 84.3
+616 +826 +870 +924 +906
49.8 51.6 59.8 71.2 103.1
+210 +46 +56 -18
-35 -40 -41 -42 -39
1.8 1.9 2.3 2.6 3.7
+19 +30 +34 +38 +33
2.2 2.2 2.6 3.1 4.5
+11 +4 +4 -5
X
SE
X
-273 -158 -131 -97 -78
28.5 25.5 29.0 40.1 59.3
-16 -16 -7 -4 -6
1.2 1.1 1.3 1.7 2.6
~89
1High milk minus young sire.
Realized response in milk yield closely matched and in most cases exceeded expected response in milk production as determined by the two methods of comparison. Deviation of HM from YS for ME milk was not significantly different from the deviation of HM from YS for EBV milk. However, response as measured by within generation differences between HM and YS for RBV milk was significantly greater than expected in earlier generations but significantly less than expected in the last generation group. Realized response closely matched expected response over all generations. Regression of EBV milk on RBV milk using data from both lines resulted in coefficients that differed significantly from 1.0 (1.1), and the intercept was significantly greater than (31 kg). Hence, realized response in BV milk exceeded expected response over most of the range of data. The relationship became unfa-
°
vorable when EBV fell below -437 kg. Expected versus realized response agreed closely with other reports from similar selection projects using Holsteins (3, 9, 12, 14) or Jerseys (25). Our results were also similar to those of Powell and Norman (15), who, using field data, reported within herd regressions of first lactation production on sire PD averaging 1.20 over all breeds and 1.33 for Jerseys. Correlated Response In Milk Components
Similar to other reports (3, 5, 9, 10, 12, 14, 23), correlated response in fat yield to selection for high milk. yield was positive, and response in percentage composition (fat percentage) was negative. The greatest correlated responses occurred in earlier generations as with direct response in milk yield.
TABLE 8. Least squares means and standard errors by group for breeding value of foundation females. Genetic group Trait
High milk
Milk, kg
-682 -31
X Fal, kg Fat %
Young sire SE
32.4 1.4 .03.02
! High milk minus young sire. Journal of Dairy Science Vol. 74, No.9. 1991
Difference!
X
SE
X
-690 -29 .07
37.7 1.6 .03
+8 -2
SE
49.7 2.1 -.04 .07
3219
DIRECf RESPONSE TO SINGLE TRAIT SELECTION TABLE 9. Significance level of independent variables for DHIA-adjUSled IIlSI lactation production. Independent variable Trait
Group
Milk Fat Fat % 4% FCM
••• ••• • •••
Sire (group)
••
• •••
Generation (group)
Year-season
t
••• •• ••• ••
•
•
lNS (P > .10). tP < .10. • p < .05. •• p < .01.
".p
< .001.
Comparison of expected and realized responses for fat gave results similar to those involving mille Realized correlated response as estimated by either genetic merit or first lactation ME fat were close to the expected. Similarly, regressions of expected on realized response in genetic merit for BV fat indicated a greater realized response than expected. The Young Sire Group
Results from this project agree with Legates and Myers (10) and Meland et aI. (12), who concluded that random bred controls are acceptable as controls in long-term breeding studies in dairy cattle. For all traits examined, no significant change occurred in YS performance level after the second generation. Similarly, regressions of expected on realized genetic merit were not different between groups for any trait. Estimates from regressions indicated that, over the range of data available here, the genetic relationships between parents and daughters for production traits were consistent between YS and HM.
The close agreement between expected and realized response for all traits is of note because a substantial portion of the realized response to selection resulted from declines within YS. For all genetic merit traits except fat percentage, a significant decline in genetic merit occurred between the YS foundation generation and the first YS generation. This result was probably due to the low genetic merit of YS sires relative to the YS foundation cows. Unfortunately, this project began in 1967, and, as Cassell et aI. (7) reported, selection within the Jersey breed as estimated by trends in sire transmitting ability gained markedly in effectiveness after 1968; trends before this date were flat or negative. Therefore, the population of young sires from which the 20 young sires used here were selected was probably of much lower genetic merit than it may have been only 2 yr later. CONCLUSIONS
Improving Jersey milk yield by selecting HM sires from those available from AI organ-
TABLE 10. Least squares means and standard errors by group for DHIA-adjusted first lactation production. Genetic group Trait Milk, kg Fal, kg Fat, % 4% PCM, kg
High milk
Young sire
X
X
SE
6594 239.2 306 10.3 4.7 .04 7234 243.9
SE
5528 431.0 264 18.5 4.8 .07 6173 439.4
x
SE
+1066 +42 -.1 +1061
492.9 21.2 .07 502.6
lHigh milk minus young sire. Journal of Dairy Science Vol. 74, No.9. 1991
3220
BONCZEK. ET AL.
TABLE 11. Least squares means and standard errors by generation within group for mature equivalent 1ml lactation production. Genetic group High milk Milk, kg Generation 1 2 3 4 ~
Fal, kg Generation 1 2 3 4 ~
4% FCM. kg Generation 1 2 3 4 ~
Change in difference
Difference 1
Young sire
X
SE
X
SE
X
SE
6426 6614 6350 6550 7031
187.9 171.9 201.9 254.1 373.1
5685 5419 5389 5555 5591
512.6 366.0 466.9 688.3 881.9
+741 +1195 +961 +995 +1440
546.0 404.4 508.7 733.7 957.6
+454 -234 +34 +445
294 308 301 307 322
7.5 6.9 8.1 10.2 15.0
278 255 251 260 277
20.6 14.7 18.8 27.7 35.5
+16 +53 +50 +47 +45
22.0 16.3 20.5 29.5 38.6
+37 -3 -3 -2
6985 7272 7048 7221 7644
180.9 165.4 194.4 244.6 359.2
413.5 5992 352.3 5914 449.5 6123 662.6 6390 848.4
+539525.6 +1280 389.2 +1134 489.7 +1098 706.3 +1254 921.3
6446
+741 -146
-36 +156
lHigh milk minus young sire.
TABLE 12. Mean group difference (high milk Illinus young sire) in expected breeding value (EBV) versus least squares means group difference in mature equivalent ("mt lactation production (ME) and realized breeding value (RBV). EBV
Milk, kg Generation 1 2 3 4 ~
Overall Fal, kg Generation 1 2 3 4 ~
Overall
ME
RBV
X
SE
X
SE
X
SE
511 718 930 1033 1139 754
34.7 39.1 49.8 61.6 75.2 22.3
741 1195 961 995 1440 1066
546.0 404.4 508.7 733.7 957.6 492.9
616 826 870 924 829
49.8 51.6 59.8 71.2 103.1 97.9
18 28 33 37 34 27
1.8 2.1 2.3 2.6 4.5 1.1
16 53 50 47 45 42
22.0 16.3 20.5 29.5 38.6 212
19 30 34 38 33 30
2.2 2.2 2.6 3.1 4.5 4.0
Journal of Dairy Science Vol. 74, No.9, 1991
906
3221
DIRECT RESPONSE TO SINGLE TRAIT SELECTION
TABLE 13. Intercepts (a), slopes (b), and their standard errors by group and overall from regression of expected on realized breeding values. Milk, kg
Intercept
Fat percentage
Fat, kg Slope
Intercept
Slope
Slope
Intercept
Group
a
SE
b
SE
a
SE
b
SE
a
SE
b
SE
High milk Young sire Combined
30 118 31
11.0 48.4 9.8
1.1 1.2 1.1
.03 .05 .02
2 3 2
.6 1.6 .5
1.1 1.1
.04 .04 .02
.002 .001 .001
.004
1.0 1.0 1.0
.02 .05 .02
izations was effective. However, improvement in fat lagged behind that of milk, and fat percentage decreased as yield increased. For all traits, realized response to selection was similar to that expected; realized response exceeded expected in most cases. ACKNOWLEDGMENTS
The authors would like to thank the staff and crew at the Dairy Experiment Station, Lewisburg, TN for their cooperation with data collection over the term of this project. REFERENCES 1 Anderson, V. L., and R. A. McLean. 1974. Design of experiments: a realistic approach. Marcell Dekker, New York, NY. 2 Andrus, D. F., and L. D. McGilliard. 1975. Selection of dairy cattle for overall excellence. J. Dairy Sci. 58: 1876. 3 Bars, S. 1981. Direct response to selection for milk yield and correlated responses in traits associated with beef characteristics and milk composition. M.S. Thesis, Univ. Minnesora, St. Paul. 4 Bertrand, J. A., P. J. Berger, A. E. Freeman, and D. H. Kelley. 1985. Profitability in daughters of high versus average Holstein sires selected for milk yield of daughters. J. Dairy Sci. 68:2287. 5 BoettCher, P. J., L. B. Hansen, C. W. Young, and H. Chester-Jones. 1989. Milk and fat yields of static controls versus daughters of highest progeny tested bulls for milk from 1964 to 1984. J. Dairy Sci. 72(Suppl. l):68.(Abstr.) 6 Brundage, A. L. 1984. Six generations of single-trait selection for milk production. J. Dairy Sci. 67(Suppl. 1):188.(Abslr.) 7 Cassell, B. G., K. L. Lee, G. R. Kroll, and H. D. Norman. 1986. Trends in estimated transmitting ability from sire evaluations based on different lacrations. J. Dairy Sci. 69:1613. 8 Foster, W. W. 1990. Response in milk yield of Holstein cows sired by bulls selected for milk yield or type score. J. Dairy Sci. 73(Suppl. l):139.(Abstr.)
1.1
.009 .003
9 Freeman, A. E., P. M. VanRaden, and D. H. Kelley. 1983. An experiment selecting for high versus average predicted difference milk in Holsteins. J. Dairy Sci. 66(Suppl. 1):115.(Abslr.) to Legates, 1. E., and R. M. Myers. 1988. Measuring genetic change in a dairy herd using a control population. 1. Dairy Sci. 71:1025. II McAllister, A. J., 1. A. Vesely, T. R. Batra, A. 1. Lee, C. Y. Lin, G. L. Roy, 1. M. Wauthy, K. A. Winter, and L. A. McClel1and. 1990. Genetic changes in protein, milk, and fat yields as a response to selection for protein yield in a closed population of Holsteins. J. Dairy Sci. 73:1593. 12 Meland, O. M., R. E. Pearson. 1. M. White, and W. E. Vinson. 1982. Response to selection for milk yield in Holsteins. J. Dairy Sci. 65:2131. 13 Pearson, R. E. 1971. The effect of age distribution and female culling on the profitability of the dairy herd. Ph.D. Diss., Iowa State Univ., Ames. 14 Pearson, R. E., R. H. Miller, J. W. Smith, L. A. Pulton, M. R. Rothschild, D. S. Balaine, and E. M. Coffey. 1981. Single and multiple trait sire selection, first lactation milk yield and composition, conformation, feed intake, efficiency, and net income. 1. Dairy Sci. 64:77. 15 Powell, R. L., and H. D. Norman. 1984. Response within herd to sire selection. 1. Dairy Sci. 67:2021. 16 Powell, R. L., H. D. Norman, and F. N. Dickinson. 1977. Trends in breeding value and production. J. Dairy Sci. 60:1316. 17 SASlStat~ Guide for Personal Computers, Version 6. 1985. SAS Inst., Inc., Cary, NC. 18 Short, T. H., B. R. Bell, D. O. Richardson, H. H. Dowlen, E. D. Moore, and J. R. Owen. 1990. Correlated responses of health care cost to selection for milk yield in Jerseys. 1. Dairy Sci. 73:2547. 19 Smothm, C. D., B. R. Bell, D. O. Richardson, B. F. Hollon, E. D. Moore, and J. R. Owen. 1986. Correlated response in feed efficiency accompanying selection for milk yield in Jerseys. 1. Dairy Sci. 69:2370. 20 Smothm, C. D., B. R. Bell, D. O. Richardson, E. D. Moore, and J. R. Owen. 1988. Correlated response in classification scores accompanying selection for milk yield in Jerseys. J. Dairy Sci. 71:3446. 21 Stott, A. W., and M A. Delorenzo. 1988. Factors influencing profitability of 1ersey and Holstein lactations. J. Dairy Sci. 71:2753. 22 Wiggans, G. R., and P. M. VanRaden. 1989. USDA-
Journal of Dairy Science Vol. 74, No.9, 1991
3222
BONCZEK ET AL.
DHIA animal model genetic evaluations. Nail. Coop.
Dm Program Handbook Fact Sheet H-2, Washington, DC.
23 Wilk, J. C., H. T. Malon, and B. T. McDaniel. 1979. Correlated response in milk components to selection foe milk yield in Randleigb Jerseys. J. Dairy Sci. 62(Suppl. 1): 169.(Abstr.)
Journal of Dairy Science Vol. 74, No.9, 1991
24 Wilk, J. C., and B. T. McDaniel. 1979. Direct response to selection for milk yield in Randleigb Jerseys. J. Dairy Sci 62(Suppl. 1):169.(Abstr.) 2S Wilk., J. C., B. T. McDaniel, and H. T. Malon. 1979. Regressions of milk yield and components on predicted differences in Randleigb Jerseys. J. Dairy Sci 62(Suppl. 1):170.(Abstr.)
test for group differences. Groups differed for all traits. Selection was superior to control in breeding value for milk and fat (828 and 31 kg) and for production of milk. fat, and 4% FCM (1066, 42, and 1061 kg). Control was superior to selection in breeding value and production fat test (.15 and .12%). Group differences existed within generation class for all yield traits but not for fat percentage. Realized response closely matched or exceeded expected response as estimated from pedigree information. (Key words: selection, milk yield, direct response)
In 1967, the Jersey herd at the Dairy Experiment Station, Lewisburg, TN was divided into two groups on the basis of ancestry, type, and breeding value for milk as part of a project to detennine effects of single trait selection for high milk yield on yield and correlated traits. Control group was mated randomly to 20 unproven young sires selected randomly from those available from breeding studs in 1967. Selection group was mated to sires selected solely on the basis of their high transmitting ability for milk. Selection sires were selected at intervals and used for 4 yr. By the end of the project (1984), lactation information was available on 672 daughters (520 selection and 152 control) of 37 bulls (17 selection and 20 control). Differences in breeding values for milk, fat, and fat test as calculated from the YfA reported in the July 1989 USDA genetic evaluations and differences in first lactation mature equivalent production of milk. fat, fat test, and 4% FCM were examined. Linear mixed models were used for all analyses and contained the fixed effects group, generation within group, and year. Sires were random, nested within group, and used to
Abbreviation key: BV = breeding value, EBV milk or selection genetic group, ME = DHIA-adjusted lactation record, PDM = PO milk. R % = repeatability, RBV = realized BV, YS = young sire or control genetic group.
= expected BV, HM = high
INTRODUCTION
Received January 31, 1991. Accepted April 24, 1991. lARS-USDA, Dairy Experiment Station. 2Department of Animal Science, University of Tennessee, Knoxville 37901-1071. 3ARS-USDA, Milk Secretion and Mastitis Laboratory, Beltsville, MD 20705. "Dairy Experiment Station, University of Tennessee. SARS-USDA, Delta States Research Center, Stoneville, MS 38776. 1991 J Dairy Sci 74:3209-3222
Comparisons of expected and realized responses to selection for economically important traits in dairy cattle is important in determining whether recommendations derived from selection theory are directly applicable to the dairy industry. Of the many traits that affect economic efficiency of dairy cattle, yields of milk and its components are the most important. Stott and Delorenzo (21) reported that milk yield was the single most important determinant of profitability in Jerseys. Similarly, Andrus and McGilliard (2) reported that milk yield was the most important factor in determining profit per year of herd life in dairy cattle, and Pearson (13) determined that it is highly profitable for cows to produce at the highest level possible.
3209
3210
BONCZEK ET AL.
Rapid improvement in genetic potential and actual yield of milk in dairy cattle has been achieved from selection (I 5). However, realized response to selection tends to deviate from expected when selection is on sire PD milk (PDM), and level and direction of deviation tend to vary with data sample source. Reports using field data (15, 16) indicate lower actual response to selection on sire PDM than data from experimental herds (8, 9, 10, 12, 14). In all but the lowest producing herd classes in their study, Powell and Norman (15) determined that response to sire selection within herd was greater than expected for all breeds and improved with increasing average milk yield. To examine realized responses to long-term selection for high milk yield in dairy cattle, the NC-2 and S-49 Regional Projects established several selection programs in the 1960s and 1970s. Reports from these long-term selection projects have concluded that selection for milk yield results in improved yield of milk in Holsteins (3, 4, 5, 6, 8, 9, 10, 11, 12, 14) and Jerseys (24) that exceeds (3, 5, 9, 14, 24) or closely matches (12) that predicted from the PDM of sires used. However, they also conclude that a decrease in percentage of milk components (3, 5, 12, 14, 23) results, even though actual yields of milk components increase (3, 5, 9, 10, 11, 12, 14, 23). The objective of this study is to examine results of long-term selection for milk yield in Jerseys. Specific objectives are 1) to evaluate the direct response of milk yield to single trait selection for milk yield and 2) to determine the correlated response of milk components to single trait selection for milk yield. Correlated responses in some other traits studied in this project have been reported (18, 19, 20). MATERIALS AND METHODS Project Protocol
Group Formation. In 1967, the Jersey herd at the Dairy Experiment Station, Lewisburg, TN was divided randomly into two breeding groups. Factors balanced between cow groups during assignment were sire, age, type, and breeding value (BV) for milk. Approximately one-third (n = 80) of the breeding herd was assigned to a group to be bred to young sires Joumal of Dairy Science Yo:. 74, No.9, 1991
(YS), and the remainder (n = 184) was assigned to a group to be bred to sires selected solely for high milk yield transmitting ability (HM).
Younger animals were assigned to either HM or YS on the basis of their ancestry; i.e., not all young stock were assigned randomly. All daughters of Bruce Westfield Cherry Noble (546137), Marlu Milestone Milkboy (584458), and Brigham Confident Infred (549809) were assigned to HM as foundations, and all daughters of Tenn Jeweler FOE Lad (595125) and Tenn TOlOno GDO Lad (599997), unproven clean-up bulls, were assigned to YS as foundations. After initial assignment, a ratio of 3 HM animals to 1 YS animal was maintained within the herd. Management. Both groups were managed similarly except for one voluntary culling criterion and the mating plan. Animals from both groups were housed, treated, and fed as a single unit. Housing was in free-stall barns, and milking was in a double-two walk-through parlor. Animals were placed on pasture when possible and received supplemental concentrates in the milking parlor according to production level. Involuntary culling criteria were the same in both groups. Reproductive criteria allowed for culling of producing animals not pregnant by 305 d postpartum or after six services. Young stock were culled as nonbreeders if they had not conceived by 21 mo of age. Animals aborting or reacting to blood test for brucellosis were culled. Animals were culled for production if they declined 40% or more in weekly average milk yield due to illness and did not recover in a reasonable period of time, or if they consistently produced abnormal milk. Functional criteria allowed culling of any animal suffering an obvious injury or having an abnormality that seriously impaired its ability to function. Voluntary culling was practiced within both groups. Excess YS animals were culled randomly if group size became too large, and in later years an effort was made to cull earlier generation YS animals preferentially. Lower producing HM animals were culled voluntarily if group size became too large. Voluntary culling of young stock was minimal in both groups because all females were to be given every opportunity to produce a first
DIRECf RESPONSE TO SINGLE TRAIT SELECTION
lactation record to at least 90 d. Preliminary analyses of survivability data indicate that total losses prior to first calving, although high, did not differ signifIcantly between groups. Some HM females were sold prior to calving for promotional and breeding purposes. Mating Plan and Sire Selection. The mating plan for both groups precluded any close matings (inbreeding >6.25%). Animals in both groups were bred artifIcially at least three times before being exposed to natural service. Originally, progeny of natural service sires were to be included in the project as YS animals regardless of dam's group; however, preliminary analyses of direct response data indicated a positive genetic trend over generations in the YS mean milk BV if daughters of these sires were included in the analyses. Therefore, daughters of these sires and their offspring were removed from the data set to stabilize the YS average milk BV. Except to avoid close matings, all YS matings were random and used semen from 20 bulls selected randomly from those available in sampling programs of the various AI organizations in 1967. An effort was made to equalize use within years, and each sire was available for use in all years until its semen supply was exhausted. Information on YS sires including PTA from the July 1989 bull evaluations is in Table 1. Except to avoid close matings, all HM matings were random. Sires used in HM were selected at intervals and used for a limited period. Requirements for selection were a minimum of 30 AI daughters in at least 20 herds and a repeatability (R %) of at least 60%. Lower R % sires (40 to 60%) were considered for selection if the probability of their being superior to a sire meeting the prior criteria was 75%. Eligible sires were ranked on PDM, and the sire with the highest PDM was chosen if the fertility of his semen was acceptable (average services per conception, 3 or fewer) and if semen price per unit was not more than $5 more than the eligible group's average price per unit. Fertility and price requirements were instituted to ensure that HM sire selection would be similar to that of commercial dairies with registered cattle. Approximately four HM sires were used in every year. Each HM sire was used for about 4 yr with nearly equal numbers of services
3211
planned for each sire. After 4 yr of service and exhaustion of semen supply, use of the HM sire was discontinued; however, the sire was eligible for reselection and use for another 4 yr. Information on HM sires including PTA from the July 1989 bull evaluations is in Table
2. For both genetic groups, sire use over generations was severely unbalanced (as expected); however, because of overlapping generations within the herd, all but three sires (codes 101, 103, and 206) had producing offspring in two or more generations. Only 7 of the 37 sires failed to have producing offspring in three or more generations (codes 101, 102, 103, 104, 105, 206, and 216). Data
Data used to examine direct and correlated responses in yield traits were BV and first lactation production information and included females born between January 1, 1968 and December 31, 1981. Data collection started in January 1968 and stopped after the last born YS female completed her f'ITst lactation in 1984. Dispersal of YS began shortly thereafter. Number born in each year group is in Table 3. Genetic merit data consisted of each animal's realized (RBV) and expected BV (EBV) for milk, fat, and fat percentage as calculated from the July 1989 USDA PTA. Protein information was not included because it was not collected from the herd during the project. The RBV was calculated as twice the animal's PTA. The EBV were determined by summing the PTA of each animal's parents. AU genetic merit estimates were to the 1985 birth year base; hence, most estimates were negative. To be included in the genetic merit data set, an animal was required to have at least one official lactation record. Therefore, data were not available on some animals due to culling or death before or soon after f'ITst calving (n = 378). Also, 6 animals' estimates were not available for unknown reasons. The edited genetic merit data included 672 females (520 lIM and 152 YS). Observations for an additional 78 YS females were excluded because they were sired by or were offspring of a female sired by a herd bred young sire. As noted earlier, preliminary analyses indicated Journal of Dairy Science Vol. 74, No.9, 1991
...
~
0
....NN
[ 0
..... 0 ~. Vol
~. n
TABLE 1. Sires used in the young sire group.
Brigham Confident Infred Sybil Royal Predictor Sybil Carletta's FLOalist Nobleman's Lotus Designer Observer Signal Revival Observer O1ocolate Soldier Milestones Generator Secret Baronet S5 Quicksilver of Pal1neva Vaucluse Sleeping Swville Millcboy Happy Hill Abe Supreme Virginian Glen Meadows 1eweled Mark Briarcliff Bold Torono Starn General Favorite Saint Briarcliffs Soldier Boy
Regislnilion Project code number 549809 573431 577474 583958 578382 596832 602658 6027S4 593883
101 102 103 104 105 106 107
lOS
604694
109 110
613165 615053 606981 616022 619994 623158 620738
112 113 114 115 116 117
III
1101y 1989 evaluation. All values in kilograms.
~.
2Rcliability for milk, fat, and fat percentage.
n
3Rcliability for protein and protein precentage.
Number of daughters Milk
10 56
36 66 9 41 109 52 64 65 41 69 45 47 35 8 12
Predicted
transmiltin8
Fat
Fat %
Rel2
-210 -221 --82 -104 -50 +116 +27 -191 -217
-14 -5 +2 -5
93 97 97
+3
-.06 +.10 +.11 +.01 -.10 -.05
-13
-.24
99
-1 +5
-224
~
+.14 +.27 +.07
-105 +69 -168 +144 +319 +194
-1 -4 -10
+.cn
-10
-.27 -.21 -.21 -.09
99 99 99 99 99 98 99 99 99 99
+406
--8
+2 -4 +14
-.12
-.04
99 97
99
t'
abilityl Protein -9 -8 -2 -3 +1 +4 -7 -5 +1 -7 -2 +2 -7 -1 +4
+0 +10
Years
~
Protein %
Relp3
of use
~
-.03 +.01 +.01 +.01 +.05 .00 -.14 +.04 +.17 +.02 +.02 -.02 -.01 -.11 -.13 -.12 -.08
67 83 88 98 86 98
1967 1967-1971 1967-1970 1967-1974 1969-1970 197(}"1974 197(}..1978 1971-1975 1971-1976 1972-1976 1975--1978 1976-1980 1976-1980 1978-1980 1978-1980 1980 1980
~
99 99 99 99 98 98 97 98
99 99 99
~
Z
en til
a tn
~
I;;
~
tn
~g Z
~ -J .~
~ ~
...
...
~ ~
\,;J
....IV \,;J
3214
BONCZEK ET AL.
TABLE 3. Number of females entering project by year of binh. Group Year
1968 1969 1970 1971 1972
1973 1974 1975 1976 1977 1978 1979 1980 1981 Total
High milk
Young sire
29
21
46
15 20 18
57 50 53 58 61 68 52 64
56 57 62 52
765
26 25
17 25 23 26 19
observation was used to investigate genetic merit; the general linear models procedure of SAS was used (17). The model fit to all genetic merit traits was Yijklm
= J.1
+ ~ + G(L)j(i) + S(L)k(i) + T I + ~jkIm [1]
where Yijklm
17 20 19 291
that herd bred young sires were markedly superior to the original 20 YS sires in transmitting ability for milk. Therefore, their daughters gave a positive trend over generations to the average genetic merit of YS that was undesirable in the control group. Further standardization of genetic merit data included setting all generation ~5 animals to generation 5 due to paucity of higher generation number animals. The number providing data by generation and group is in Table 4. Another analysis was of genetic merit data consisting of the RBV of foundation animals calculated from PIA. These analyses were performed to determine whether the herd's initial randomization was balanced. Data on 303 foundation animals (214 HM and 89 YS) were available. Production data were DHIA-adjusted first lactation values (ME) for milk, fat, fat percentage, and 4% FCM. Information on 623 females (479 lIM and 144 YS) was available. To be included in production data, the first lactation had to meet each of the factors mentioned with genetic merit data and be at least 125 d in length. Only first lactations were used to minimize culling bias. Analyses
Genetic Merit Data. A linear mixed model with observations weighted by reliability of Journal of Dairy Science Vol. 74, No.9, 1991
J.1
~ G(L)j(i) S(L~(i)
Tl ~jIdm
= the
breeding value of female m from generation j and sire k of group i and born in project year 1; = the overall mean; = fixed effect of group i; = fixed effect of generation of selection j nested in group i; = random effect of sire k nested within group i; fixed effect of project year of = birth I; and = random error common to all observations.
The interaction, generation by sire within group, was included in the initial model but pooled with the error tenn after it was found to be not signifIcant (P > .25) (1). Type ill sums of squares from the general linear models procedure were used in all tests of signifIcance (17). Random error was used to test the fixed effects of year and generation nested within group and the random effect of sires nested within group. The effect of group was tested against sires nested within group. Weighted analyses were used for genetic merit to account for the heterogeneity of variance in PIA. Reliability was determined to be the most appropriate weighting variable because it most closely reflected the differing amounts of information incorporated in each estimate of genetic merit (22). Results of analyses without weighting did not vary markedly from weighted analyses. A similar model [2] including only the fixed effect of group and the random effects of sire within group was used to analyze foundation genetic merit. Methodology of testing was identical with model [1].
DIRECT RESPONSE TO SINGLE TRAIT SELECTION
Production Data. For production data, the following mixed linear model was used: Yijklm = 11 + Li + G(L)j(i) + S(L)k(i) + T, + eijklm [3] where Yijklm
=
11 G(L)j(i)
= = =
S(L)k(i)
=
T1
=
t1jklm
=
l.j
ME production of female m from generation j and sire k of group i calving in year-season l', the overall mean; fixed effect of group i; fixed effect of generation of selection j nested in group i; random effect of sire k nested within group i; fixed effect of year-season of calving 1; and random error common to all observations.
All testing used the random error term except for group effect, which was tested against sires within group. The generation by sire within group interaction was included in the initial model but was pooled with the error term after testing determined that it was not significant (P> .25) (1). For these analyses, least squares means were calculated for each generation within group as follows (1): Yij ..
= 11
+ l.j + G(L)j(i) + S.. + T.
where
y Ij.. .. = least squares mean ME production level for generation j of group i; S = average of effect of sire in group; and T = average effect of year-season of calving.
Expected Versus Realized Response. Two approaches were used to compare expected and realized selection responses in milk yield and correlated yield traits. In the first approach, differences of EBV between groups for each generation were compared with the differences between generation within group least squares means from the genetic merit and production
3215
analyses. For these comparisons. the standard errors of differences were generated by computing the square root of the sum of the variances of the estimates, and simple t tests were performed. In the second approach, EBV were regressed on RBV. Analyses were performed both within and across groups. For analyses within group, the initial model contained generation as a fixed effect, EBV as a covariate, and their interaction. As with model [1], regression analyses were weighted by reliability of BV estimate. Comparisons between groups of coefficients and intercepts were made using simple t tests. For regressions over groups, group was included in the model as a fixed effect in addition to generation, and generation was nested within group. All interactions between group, generation, and the covariate, EBY, were examined. RESULTS
Genetic Merit
A summary of significant effects for realized BV for milk, fat, and fat percentage is in Table 5. Group and sire within group differed significantly for each dependent variable. Generation within group differed significantly for BY milk and BV fat but not for BY fat percentage. Year born did not signifIcantly affect any dependent variable. Group least squares means for each dependent variable are in Table 6. As expected from the selection practiced, HM was superior to YS in kilograms of product produced; however, the opposite was true for BV fat percentage. Of more interest are the generation within group least squares means for BV milk and BV fat in Table 7. For YS, a slightly negative trend existed for both traits from generations 1 to 4 with a slight nonsignificant recovery in the last generation group. The HM had a consistent positive trend for both traits with greater improvement in earlier versus later generations. For BV milk, differences between generation least squares means within YS approached significance for generation 1 versus 2 (P < .10) and for generation 1 versus 3 (P < .10). Within HM, the generation 1 least squares mean for BY milk differed significantly from all later estimates (P < .01), but estimates did not differ }omnal of Dairy Science Vol. 74, No.9, 1991
3216
BONCZEK ET AL.
TABLE 4. Number of observations by generation and genetic group. Group Generation
High milk
1
172
2 3 4
155
~5
Total
113
55 25 520
Young sire
45 41 30 24 12 152
significantly for later generations. Between groups, BV milk least squares means for generations differed significantly for each generation class (P < .01); HM had the higher estimate in each case. However, improvement of HM advantage was significant only between generations 1 and 2. For BV fat, differences between generation least squares means within YS approached significance for generation I versus 3 (P < .10) and were significant for generation 1 versus 4 (P < .05) but not elsewhere. Within HM, the generation I least squares means for BV fat differed significantly from all later estimates (P < .01), and the level for generation 4 exceeded significantly that of generation 2 (P < .01). As with BV milk, least squares means for generations differed significantly for groups within each generation class (P < .01); HM had the higher estimate in each case. However, significant improvement in HM advantage over YS existed only between generations 1 and 2. Realized BV of the foundation population did not differ significantly between groups for any trait. Least squares means for foundation generation BV milk, BV fat, and BV fat percentage are in Table 8. These estimates should not be compared directly with estimates from later generations because they are from separate analyses.
tion within group was significant for ME fat and approached significance for ME milk and ME FCM. Group least squares means for each dependent variable are in Table 10. As with genetic merit data, HM was superior to YS for both milk and fat produced but not for fat percentage. Additionally, ME FCM was greater in HM than in YS. Generation within group least squares means for ME milk., ME fat, and ME FCM are in Table 11. For all three traits, HM least squares means were highest in later generations. However, generation 3 estimates had a nonsignificant decrease from generation 2 for each trait. The YS decreased between generations 1 and 3 for all traits but increased thereafter. For ME milk., ME fat, and ME FCM, YS and HM generation least squares means did not differ significantly for generations within group. Significant differences for groups within generation existed, and the patterns of these differences were similar for all three traits. Least squares means of HM were significantly larger than those of YS in generations 2 (P < .01) and 3 (P < .10 for ME milk and P < .05 for ME fat and ME FCM) but not other generations. Change in advantage of HM over YS between generations was not significant in any case. ExpeCted Versus Realized Response to selectIon
Least Squares Means. A comparison of group differences in EBV milk., least squares means ME milk, and least squares means RBV milk by generation and over all generations is in Table 12. Over all generations and for all but generation 4, the realized response, as measured by the difference in ME milk least squares means, exceeded that predicted by the difference in group EBV milk. However, realized response did not differ from expected response as measured by difference in ME milk.
First Lactation Production
A summary of significant sources of variation in ME production of milk., fat, fat percentage, and 4% FCM is in Table 9. Effects of group, sire within group, and year-season of calving were significant for all traits. GeneraJournal of Daily Science Vol. 74, No.9, 1991
Differences between EBV milk and RBV milk for generations 1, 2, and ~ approached significance (P < .10). For generations 1 and 2, realized response as estimated by RBV milk was greater than EBV milk. For generations ~, EBV milk was greater than RBV milk, and EBV milk was greater than RBV milk for
3217
DIRECf RESPONSE TO SINGLE 1RAIT SELECTION TABLE 5. Significance level of independent variables for genetic merit (breeding value). Independent variable Trait
Group
Sire (group)
Generation (group)
Year
Milk.
*** *** ***
*** *** ***
• •••
NS l NS NS
Fat Fat %
NS
INS (P > .10).
*p < .05.
***p < .001.
generations 3 and 4, but these differences did not approach significance. Over all generations, the realized response in BV did not differ significantly from that expected. A comparison of group differences in EBV fat, least squares means ME fat, and least squares means RBV fat is in Table 12. In all cases, realized response as estimated by ME fat or RBV fat did not differ significantly from expected response. Regression. A summary of regressions of EBV on RBV is in Table 13. Within HM, regression coefficients were greater than or equaled 1.0 for all traits. Coefficients for BV milk and BV fat were significantly greater (P < .01) than 1.0. Intercepts for HM regressions were significantly greater (P < .01) than 0 for all regressions except BV fat percentage. Regression coefficients of YS were significantly greater than 1.0 for BV milk (P < .01) and BV fat (P < .01). The regression coefficient for BV fat percentage did not differ significantly from 1.0. Intercepts from YS regressions were significantly greater than 0 for all regressions but BV fat percentage. Using data from both groups, regression coefficients and intercepts were not signifi-
cantly different between groups. For BV milk and BV fat, regression coefficients significantly exceeded 1.0 (P < .01), and the intercepts significantly exceeded 0 (P < .01). For BV fat percentage, the regression coefficient from combined data did not differ from 1.0, and the intercept did not differ from O. DISCUSSION
Direct Response In Milk Yield Selection for milk production using HM sires was effective. However, direct response to selection as measured by difference between selection (HM) and control (YS) genetic merit for milk yield (BV milk) or first lactation milk yield (ME milk) was greater in earlier generations than later. In later generations, the advantage of HM increased at a progressively slower rate (BV milk) or actually decreased (ME milk). Others (3, 5, 6, 8, 9, 10, 11, 12, 14) reported positive direct response to selection for milk yield in Holsteins, and Wilk and McDaniel (24) reported positive direct response in Jerseys. They also reported variation in rate of gain or loss of advantage over generations.
TABLE 6. Least squares means and standard errors by group for breeding value. Genetic group Trait
High milk.
X
SE
X
SE
X
SE
Milk, kg
-147 -9
53.0 2.2 .04
-976
82.3
-39
3.4
+829 +30
97.9 4.0
Fat, kg Fat %
-.02
Young sire
.13
.07
-.15
.08
lHigh milk. minus young sire. Journal of Dairy Science Vol. 74, No.9, 1991
3218
BONCZEK ET AL.
TABLE 7. Least squares means and standard errors by generation within group for breeding values for milk and fat. Genetic group
Milk, kg Generation 1 2 3 4 ~5
Fal, kg Generation 1 2 3 4 ~
Change in difference
Difference!
Young sire
High milk
SE
X
SE
-984 -1001 -1021 -984
40.8 44.9 52.3 58.8 84.3
+616 +826 +870 +924 +906
49.8 51.6 59.8 71.2 103.1
+210 +46 +56 -18
-35 -40 -41 -42 -39
1.8 1.9 2.3 2.6 3.7
+19 +30 +34 +38 +33
2.2 2.2 2.6 3.1 4.5
+11 +4 +4 -5
X
SE
X
-273 -158 -131 -97 -78
28.5 25.5 29.0 40.1 59.3
-16 -16 -7 -4 -6
1.2 1.1 1.3 1.7 2.6
~89
1High milk minus young sire.
Realized response in milk yield closely matched and in most cases exceeded expected response in milk production as determined by the two methods of comparison. Deviation of HM from YS for ME milk was not significantly different from the deviation of HM from YS for EBV milk. However, response as measured by within generation differences between HM and YS for RBV milk was significantly greater than expected in earlier generations but significantly less than expected in the last generation group. Realized response closely matched expected response over all generations. Regression of EBV milk on RBV milk using data from both lines resulted in coefficients that differed significantly from 1.0 (1.1), and the intercept was significantly greater than (31 kg). Hence, realized response in BV milk exceeded expected response over most of the range of data. The relationship became unfa-
°
vorable when EBV fell below -437 kg. Expected versus realized response agreed closely with other reports from similar selection projects using Holsteins (3, 9, 12, 14) or Jerseys (25). Our results were also similar to those of Powell and Norman (15), who, using field data, reported within herd regressions of first lactation production on sire PD averaging 1.20 over all breeds and 1.33 for Jerseys. Correlated Response In Milk Components
Similar to other reports (3, 5, 9, 10, 12, 14, 23), correlated response in fat yield to selection for high milk. yield was positive, and response in percentage composition (fat percentage) was negative. The greatest correlated responses occurred in earlier generations as with direct response in milk yield.
TABLE 8. Least squares means and standard errors by group for breeding value of foundation females. Genetic group Trait
High milk
Milk, kg
-682 -31
X Fal, kg Fat %
Young sire SE
32.4 1.4 .03.02
! High milk minus young sire. Journal of Dairy Science Vol. 74, No.9. 1991
Difference!
X
SE
X
-690 -29 .07
37.7 1.6 .03
+8 -2
SE
49.7 2.1 -.04 .07
3219
DIRECf RESPONSE TO SINGLE TRAIT SELECTION TABLE 9. Significance level of independent variables for DHIA-adjUSled IIlSI lactation production. Independent variable Trait
Group
Milk Fat Fat % 4% FCM
••• ••• • •••
Sire (group)
••
• •••
Generation (group)
Year-season
t
••• •• ••• ••
•
•
lNS (P > .10). tP < .10. • p < .05. •• p < .01.
".p
< .001.
Comparison of expected and realized responses for fat gave results similar to those involving mille Realized correlated response as estimated by either genetic merit or first lactation ME fat were close to the expected. Similarly, regressions of expected on realized response in genetic merit for BV fat indicated a greater realized response than expected. The Young Sire Group
Results from this project agree with Legates and Myers (10) and Meland et aI. (12), who concluded that random bred controls are acceptable as controls in long-term breeding studies in dairy cattle. For all traits examined, no significant change occurred in YS performance level after the second generation. Similarly, regressions of expected on realized genetic merit were not different between groups for any trait. Estimates from regressions indicated that, over the range of data available here, the genetic relationships between parents and daughters for production traits were consistent between YS and HM.
The close agreement between expected and realized response for all traits is of note because a substantial portion of the realized response to selection resulted from declines within YS. For all genetic merit traits except fat percentage, a significant decline in genetic merit occurred between the YS foundation generation and the first YS generation. This result was probably due to the low genetic merit of YS sires relative to the YS foundation cows. Unfortunately, this project began in 1967, and, as Cassell et aI. (7) reported, selection within the Jersey breed as estimated by trends in sire transmitting ability gained markedly in effectiveness after 1968; trends before this date were flat or negative. Therefore, the population of young sires from which the 20 young sires used here were selected was probably of much lower genetic merit than it may have been only 2 yr later. CONCLUSIONS
Improving Jersey milk yield by selecting HM sires from those available from AI organ-
TABLE 10. Least squares means and standard errors by group for DHIA-adjusted first lactation production. Genetic group Trait Milk, kg Fal, kg Fat, % 4% PCM, kg
High milk
Young sire
X
X
SE
6594 239.2 306 10.3 4.7 .04 7234 243.9
SE
5528 431.0 264 18.5 4.8 .07 6173 439.4
x
SE
+1066 +42 -.1 +1061
492.9 21.2 .07 502.6
lHigh milk minus young sire. Journal of Dairy Science Vol. 74, No.9. 1991
3220
BONCZEK. ET AL.
TABLE 11. Least squares means and standard errors by generation within group for mature equivalent 1ml lactation production. Genetic group High milk Milk, kg Generation 1 2 3 4 ~
Fal, kg Generation 1 2 3 4 ~
4% FCM. kg Generation 1 2 3 4 ~
Change in difference
Difference 1
Young sire
X
SE
X
SE
X
SE
6426 6614 6350 6550 7031
187.9 171.9 201.9 254.1 373.1
5685 5419 5389 5555 5591
512.6 366.0 466.9 688.3 881.9
+741 +1195 +961 +995 +1440
546.0 404.4 508.7 733.7 957.6
+454 -234 +34 +445
294 308 301 307 322
7.5 6.9 8.1 10.2 15.0
278 255 251 260 277
20.6 14.7 18.8 27.7 35.5
+16 +53 +50 +47 +45
22.0 16.3 20.5 29.5 38.6
+37 -3 -3 -2
6985 7272 7048 7221 7644
180.9 165.4 194.4 244.6 359.2
413.5 5992 352.3 5914 449.5 6123 662.6 6390 848.4
+539525.6 +1280 389.2 +1134 489.7 +1098 706.3 +1254 921.3
6446
+741 -146
-36 +156
lHigh milk minus young sire.
TABLE 12. Mean group difference (high milk Illinus young sire) in expected breeding value (EBV) versus least squares means group difference in mature equivalent ("mt lactation production (ME) and realized breeding value (RBV). EBV
Milk, kg Generation 1 2 3 4 ~
Overall Fal, kg Generation 1 2 3 4 ~
Overall
ME
RBV
X
SE
X
SE
X
SE
511 718 930 1033 1139 754
34.7 39.1 49.8 61.6 75.2 22.3
741 1195 961 995 1440 1066
546.0 404.4 508.7 733.7 957.6 492.9
616 826 870 924 829
49.8 51.6 59.8 71.2 103.1 97.9
18 28 33 37 34 27
1.8 2.1 2.3 2.6 4.5 1.1
16 53 50 47 45 42
22.0 16.3 20.5 29.5 38.6 212
19 30 34 38 33 30
2.2 2.2 2.6 3.1 4.5 4.0
Journal of Dairy Science Vol. 74, No.9, 1991
906
3221
DIRECT RESPONSE TO SINGLE TRAIT SELECTION
TABLE 13. Intercepts (a), slopes (b), and their standard errors by group and overall from regression of expected on realized breeding values. Milk, kg
Intercept
Fat percentage
Fat, kg Slope
Intercept
Slope
Slope
Intercept
Group
a
SE
b
SE
a
SE
b
SE
a
SE
b
SE
High milk Young sire Combined
30 118 31
11.0 48.4 9.8
1.1 1.2 1.1
.03 .05 .02
2 3 2
.6 1.6 .5
1.1 1.1
.04 .04 .02
.002 .001 .001
.004
1.0 1.0 1.0
.02 .05 .02
izations was effective. However, improvement in fat lagged behind that of milk, and fat percentage decreased as yield increased. For all traits, realized response to selection was similar to that expected; realized response exceeded expected in most cases. ACKNOWLEDGMENTS
The authors would like to thank the staff and crew at the Dairy Experiment Station, Lewisburg, TN for their cooperation with data collection over the term of this project. REFERENCES 1 Anderson, V. L., and R. A. McLean. 1974. Design of experiments: a realistic approach. Marcell Dekker, New York, NY. 2 Andrus, D. F., and L. D. McGilliard. 1975. Selection of dairy cattle for overall excellence. J. Dairy Sci. 58: 1876. 3 Bars, S. 1981. Direct response to selection for milk yield and correlated responses in traits associated with beef characteristics and milk composition. M.S. Thesis, Univ. Minnesora, St. Paul. 4 Bertrand, J. A., P. J. Berger, A. E. Freeman, and D. H. Kelley. 1985. Profitability in daughters of high versus average Holstein sires selected for milk yield of daughters. J. Dairy Sci. 68:2287. 5 BoettCher, P. J., L. B. Hansen, C. W. Young, and H. Chester-Jones. 1989. Milk and fat yields of static controls versus daughters of highest progeny tested bulls for milk from 1964 to 1984. J. Dairy Sci. 72(Suppl. l):68.(Abstr.) 6 Brundage, A. L. 1984. Six generations of single-trait selection for milk production. J. Dairy Sci. 67(Suppl. 1):188.(Abslr.) 7 Cassell, B. G., K. L. Lee, G. R. Kroll, and H. D. Norman. 1986. Trends in estimated transmitting ability from sire evaluations based on different lacrations. J. Dairy Sci. 69:1613. 8 Foster, W. W. 1990. Response in milk yield of Holstein cows sired by bulls selected for milk yield or type score. J. Dairy Sci. 73(Suppl. l):139.(Abstr.)
1.1
.009 .003
9 Freeman, A. E., P. M. VanRaden, and D. H. Kelley. 1983. An experiment selecting for high versus average predicted difference milk in Holsteins. J. Dairy Sci. 66(Suppl. 1):115.(Abslr.) to Legates, 1. E., and R. M. Myers. 1988. Measuring genetic change in a dairy herd using a control population. 1. Dairy Sci. 71:1025. II McAllister, A. J., 1. A. Vesely, T. R. Batra, A. 1. Lee, C. Y. Lin, G. L. Roy, 1. M. Wauthy, K. A. Winter, and L. A. McClel1and. 1990. Genetic changes in protein, milk, and fat yields as a response to selection for protein yield in a closed population of Holsteins. J. Dairy Sci. 73:1593. 12 Meland, O. M., R. E. Pearson. 1. M. White, and W. E. Vinson. 1982. Response to selection for milk yield in Holsteins. J. Dairy Sci. 65:2131. 13 Pearson, R. E. 1971. The effect of age distribution and female culling on the profitability of the dairy herd. Ph.D. Diss., Iowa State Univ., Ames. 14 Pearson, R. E., R. H. Miller, J. W. Smith, L. A. Pulton, M. R. Rothschild, D. S. Balaine, and E. M. Coffey. 1981. Single and multiple trait sire selection, first lactation milk yield and composition, conformation, feed intake, efficiency, and net income. 1. Dairy Sci. 64:77. 15 Powell, R. L., and H. D. Norman. 1984. Response within herd to sire selection. 1. Dairy Sci. 67:2021. 16 Powell, R. L., H. D. Norman, and F. N. Dickinson. 1977. Trends in breeding value and production. J. Dairy Sci. 60:1316. 17 SASlStat~ Guide for Personal Computers, Version 6. 1985. SAS Inst., Inc., Cary, NC. 18 Short, T. H., B. R. Bell, D. O. Richardson, H. H. Dowlen, E. D. Moore, and J. R. Owen. 1990. Correlated responses of health care cost to selection for milk yield in Jerseys. 1. Dairy Sci. 73:2547. 19 Smothm, C. D., B. R. Bell, D. O. Richardson, B. F. Hollon, E. D. Moore, and J. R. Owen. 1986. Correlated response in feed efficiency accompanying selection for milk yield in Jerseys. 1. Dairy Sci. 69:2370. 20 Smothm, C. D., B. R. Bell, D. O. Richardson, E. D. Moore, and J. R. Owen. 1988. Correlated response in classification scores accompanying selection for milk yield in Jerseys. J. Dairy Sci. 71:3446. 21 Stott, A. W., and M A. Delorenzo. 1988. Factors influencing profitability of 1ersey and Holstein lactations. J. Dairy Sci. 71:2753. 22 Wiggans, G. R., and P. M. VanRaden. 1989. USDA-
Journal of Dairy Science Vol. 74, No.9, 1991
3222
BONCZEK ET AL.
DHIA animal model genetic evaluations. Nail. Coop.
Dm Program Handbook Fact Sheet H-2, Washington, DC.
23 Wilk, J. C., H. T. Malon, and B. T. McDaniel. 1979. Correlated response in milk components to selection foe milk yield in Randleigb Jerseys. J. Dairy Sci. 62(Suppl. 1): 169.(Abstr.)
Journal of Dairy Science Vol. 74, No.9, 1991
24 Wilk, J. C., and B. T. McDaniel. 1979. Direct response to selection for milk yield in Randleigb Jerseys. J. Dairy Sci 62(Suppl. 1):169.(Abstr.) 2S Wilk., J. C., B. T. McDaniel, and H. T. Malon. 1979. Regressions of milk yield and components on predicted differences in Randleigb Jerseys. J. Dairy Sci 62(Suppl. 1):170.(Abstr.)
