REGRESSION IN EGG PRODUCTION IN THE DOMESTIC FOWL WHEN SELECTION IS RELAXED1

REGRESSION IN EGG PRODUCTION IN THE DOMESTIC FOWL WHEN SELECTION IS RELAXED1 A. W. NORDSKOG AND FRANCIS G. GIESBRECHT Iowa State University, Ames Received March 18, 1964 THE question of what happens to egg production when selection is relaxed in commercial strains of chickens selected over many generations for high performance is of practical importance because of the economic implications to commercial breeding practice. Breeders would like to know whether the genetic gains they have made over many years of selection can be maintained in flocks with no further selection. The consequences of relaxing selection are also of theoretical interest because they provide some insight on the nature of gene action. If the genes influencing a trait behave in a simple additive and independent manner, then according to the Hardy-Weinberg rule, the expected population mean should not regress. If the gene action is more complex, involving epistasis, then selection would favor gene combinations which would decay when selection is relaxed. This assumes that the decline is simply a consequence of Mendelian segregation and recombination. On the other hand, if the selected genes influence fitness, then the regression would be interpreted as a consequence of genetic homeostasis, i.e., that natural selection does not favor the same gene complexes as artificial selection. Several workers have shown a regression when selection in a population is suspended following several generations of intense selection. MATHER and HARRISON (1949) increased the number of abdominal bristles in Drosophila from 36 to 54 in twenty generations, but after selection was relaxed the bristle number declined to 40 in four generations. CLAYTON, MORRIS and ROBERTSON (1957) working with the same trait in Drosophila as MATHER and HARRISON found that the tendency to return to the original number of bristles after relaxation of selection varied between lines and the direction of selection. This was shown to be true also for long and short wing-length in Drosophila by ROBERTSON and REEVE (1952) who concluded that the degree of return to the original population mean from relaxation after several generations of directional selection depended on the genetic constitution of the original populations. In the case of chickens, LERNER (1958) suspended selection for shank length in a Leghorn line selected for that trait over a period of 14 years. A marked regression in shank length was observed over the five generations that selection was relaxed. MOULTRIE, COTTIER and KING (1956) reported an experiment on relaxed selection for adult viability in White Leghorns. After three generations of relaxed selection, the results indicated increased mortality. SHOFFNER and GRANT (1960) were unable to find significant regression for egg production, viability, hatchability, or 8-week-body weight following three generations of relaxed selection in a commercial Journal Paper No. 5.4815 of the Iowa Agricultural and Home Economics Experiment Station, Ames, Iowa. Projects No. 1326 and No. 1448. The former is in cooperation with the North Central Regional Poultry Breeding Project NC-47. Genetics 50: 407416 September, 1964.

408 A. W. NORDSKOG AND F. G. GIESBRECEIT White Leghorn strain initially selected mainly for egg production over some 25 generations. In a more recent study, BOHREN and MCKEAN (1964) analyzed five generations of relaxed selection by least squares technique in a White Leghorn population previously selected for high egg production. No evidence of a decline due to relaxation was found. MATERIALS AND METHODS The cooperation of commercial breeders was solicited for this experiment for two reasons. First, many commercial strains have been intensively selected for egg production and other economically important traits over several generations. Secondly, this procedure reduced the requirements of experimental facilities since only the relaxed selection populations were maintained and not the selected populations. At the same time, this made possible the testing of more than one strain which added greater generality to the experiment. In April or May of each year starting in 1953 and continuing for seven successive years, one case of 360 hatching eggs was obtained from each of three commercial breeders. Two of these were Leghorn breeders (BA and GH) and one was a Rhode Island Red breeder (PA). The eggs from breeders BA and PA were representative samples collected from the current pedigree breeding pens. The GH eggs were from flock matings one generation removed from the pedigree matings. The chicks were handled according to routine procedures used at the Iowa State University farm. Chicks were intermingled and placed in the brooder house to eight weeks of age after which they were moved to a grass range pasture and then placed in permanent housing quarters at approximately 150 days of age. Complete records were kept on fertility, hatchability, egg production and laying house mortality. Unfortunately, the records on age at sexual maturity, egg weight and body weight were incomplete in some years. From 80 to 100 females were placed in a single pen each year and mated to about eight to ten cockerels. From this flock mating, one standard case of 360 hatching eggs was obtained to produce the chicks for the following generation. In the second year, a second sample of hatching eggs was obtained from the original commercial breeder. These two lots were then hatched, the chicks were raised to maturity, and performance was recorded as in the previous year. In the third year, chicks were obtained from these two lots together with a third lot from the original breeder. This general procedurci was followed over a seven-year period, with a new lot of hatching eggs obtained from the breeder each year. The general plan of the experiment is indicated in Table 1. Because total facilities were fixed, i.e., a maximum of 12 test pens were available in any one year, certain lines were discontinued, but this is not thought to seriously influence the analysis and interpretation of the data. For purposes of this discussion, all of the birds of one strain from an original importation TABLE 1 Pattern of observations Line number and year of origin 1 1 3 4 5 6 7 Test year 1953 1954 1955 1956 1957 1958 1959 1. 1954 Yl, 2. 1955 y2, y22 3. 1956 '31 '32 y,, 4. 1957 y41 y42 5. 1958 YS, _- * 6. 1959 -_ Ye1 ye, 7. 1960 y, -_ 1 yi3 =a y44 * Dashes represent llnes discontmued because of space liniitatlorls -- y53 y54 ys, '65 '66 y75 -- y77

REGRESSION WITH RELAXED SELECTION 4'09 together with their subsequent progeny over succeeding generations are called a line. Thus, each column vector of Table 1 represents a line. Statistical Procedures: The analysis requires the assumption that the initial populations were plateaued. Since individual records were not taken, the analysis was made on the 22 line-test year means indicated in Table 1 for each trait. The linear model is, ~ i= j a + ai + 6i-j+l + eij where i = 1, 2,..., 7, and i = i, ifl,..., 7. yii is the observed mean of the birds in the jth line of the ith year, is the general population mean, ai is the ith year effect, 8, is the relaxed selection effect in the kth year, and eij is a random error effect assumed to be normally distributed with zero mean and variance U*. A second model was also used in which the assumptions concerning the initial populations being plateaued were side-stepped. From this model (HINKELMANN 1963), quadratic effects were estimated, i.e. it is possible to discover information concerning acceleration of deceleration when selection is relaxed. On the other hand, the HINKELMANN model does not permit the estimation of simple linear contrasts among the parameters. More particularly, we cannot estimate the linear regression due to relaxed selection. RESULTS The mean values for each trait are presented in Tables 2 and 3. The analyses of variance of the means are presented in Table 4. In the case of egg production, year effects are statistically significant for each of the three strains. The regression due to relaxed selection is statistically significant for the BA and PA strains. For the other characters, with the exception of the laying house mortality year effects for the GH and PA strains, none of the mean squares are statistically significant. Since the GH and BA lines are both Leghorn lines, the results can be pooled into one analysis. If this is done, the pooled mean squares for the regression due to relaxed selection for egg production and percent fertile eggs are 2.76 and 2.58, respectively, and are statistically significant at the 5 percent level. The estimates of the relaxed selection effects on egg production are presented in Table 5. These estimates indicate a marked decline associated with the relaxation of selection in all three lines. The decline for the Leghorn strains seems to be curvilinear, while that for the Rhode Island Red strain seems to be linear. The average decline in rate of egg production per generation of relaxed selection is - 1.6 percent, - 2.8 percent and - 3.5 percent for the GH, BA and TA strains, respectively. Combining overall, the average decline is - 2.6 percent. An incomplete set of data was obtained for body weight, egg weight and age at maturity. Only the mean values for the 1960 test year are presented in Table 6. The results indicate some regression in egg weight in all three strains but not in body weight. Age at 50 percent egg production (a measure of sexual maturity) shows no change with the exception of the PA line originating in 1953. This does not appear to be a relaxed selection effect in view of the consistent results otherwise. Possibly, the PA breeder may have emphasized early maturity in his 1953 or 1954 pedigree selections. Hinkelmann model: Most of the results from the analysis of the HINKELMANN model were negative and are not presented here. The relaxed selection effect for

~ ~~ 41 0 A. W. NORDSKOG AND F. G. GIESBRECHT TABLE 2 Egg production and laying house mortalily for the three strains Percent egg production \-ear of origin Percent laying house mortality Year of origin -. Test vrar 1953 1954 1955 1976 1957 1958 1927 1953 1N4 1955 1956 1957 1958 1959 GH-Leghorn 1954 62 31 1955 50 59 34 23 1956 54 49 54 12 24 18 1957 54 52 61 58 17 14 16 12 1958 59 60 61 68 11 9 5 6 1959 60 70 69 71. 14 19 23 9 1960 54 68 60 60 8 11 8 9 BA-Leghorn 1954 62 8 1955 56 60 16 6 1956 51 57 61 17 24 8 1957 60 53 56 56 15 21 16 15 1958 66 61 58 64 16 13 18 14 1959 68 69 68 68 28 8 25 10 1960 66 77 64 76 10 4 4 16 PA-Rhode Island Red 1954 46 38 1955 44 50 33 15 1956 41 54 50 29 33 22 1957 56 61 61 65 23 17 32 26 1958 59 62 65 65 25 20 14 16 1959 42 54 65 56 40 44 35 30 1960 55 65 70 75 6 8 4 9 egg production was statistically significant only for the BA strain. The interpretation is that the decline in egg production due to relaxed selection in the BA strain was nonlinear. DISCUSSION The interpretation of the results of this experiment on relaxed selection hinges on two important questions: (1 ) The possible contribution of inbreeding depression resulting from the limited flock sizes, and (2) whether egg production has plateaued in the three commercial strains sampled. The question of inbreeding will be considered first. Undoubtedly some of the estimated decline in egg production in the relaxed selection lines is due to inbreeding in the populations sampled. Thus, part of the overall decline of 2.6 percent obtained from the linear model should be accounted for by inbreeding depression. The amount of inbreeding per generation of relaxed selection may be estimated from the formula, F = l/m + 1/8F, where M and F are the number of sires and dams, respectively, per flock mating. For a mating of eight males and 80 females, which is near the minimum number used in the

REGRESSION WITH RELAXED SELECTION 41 1 TABLE 3 Fertile eggs and fertile eggs which hatched, for the three strains Percent of fertile eggs Year of origin Percent of fertile eggs which hatched Year of origin ~~ Test year 1953 1954 1955 1956 1957 1958 1959 1953 1954 1955 1956 1957 1958 1959 GH-Leghorn 1954 92 83 1955 95 93 83 84 1956 95 95 96 49 71 81 1957 95 93 91 87 84 84 71 79 1958 94 90 92 94 85 84 72 80 1959 96 95 91 95 86 84 94 82 1960 95 75 88 85 89 78 80 90 BA-Leghorn 1954 88 71 1955 94 87 86 71 1956 95 95 84 85 84 82 1957 84 92 93 94 79 74 74 79 1958 73 91 92 84 80 75 76 80 1959 96 91 96 93 96 96 94 64 1960 96 94 97 89 85 70 86 73 PA-Rhode Island Red 1954 83 79 1955 93 78 83 80 1956 90 92 96 76 80 55 1957 81 88 84 88 78 80 72 78 1958 76 79 90 84 78 81 74 79 1959 87 88 91 89 93 88 90 89 1960 88 93 89 84 78 89 86 83 TABLE 4 Analyses of variance Mean square due to: Percent Degrees of Egg, Lying house eggs Hatch of Strain Source freedom production mortality fertile fertile eggs GH Years 6 94.0* 140.5** 37.3 93.0 Relaxed selection 6 30.4 25.7 27.4 21.4 Error 9 16.6 14.8 11.8 91.3 BA Years 6 86.l** 24.7 35.2 32.5 Relaxed selection 6 33.0* 58.6 48.9 79.5 Error 9 6.4 30.9 17.8 69.6 PA Years 6 224.6** 206.4* 40.1 117.5 Relaxed selection 6 91.4** 33.1 13.2 27.0 Error 9 12.4 37.6 27.7 30.3 * Significant at the 5 percent level. ** Significant at the 1 percent level

412 A. W. NORDSKOG AND F. G. GIESBRECHT TABLE 5 Estimates of relaxed selection effects on percent egg production A n A n A A ~ Strain s7-s1 67-82 87T-63 87-84 87-85 8*-8, GH -10.1-7.0 4.2 4.6-9.6 +2.1-1.6 BA -7.7-4.5-1.0-8.3-10.3-6.5-2.8 PA -19.3-19.8-14.4-12.3-11.2 -.8-3.5 grand average -2.6 * Ayerage decline in percent egg production per generation. TABLE 6 Body weight, egg weight, and age at 50 percent egg production for the relaxed selection lines tested in 1960 Trait Adult body weight (lb) Year selection wa5 relaxed Strain 1953 1955 1957 1959 GH 4.2 BA 4.3 PA 5.6 4.0 3.9 4.0 4.4 4.4 4.3 5.4 5.6 5.6 March egg weight (gm) GH 57 BA 54 PA 53 56 57 60 54 54 57 56 57 60 Age at 50 percent egg production (days) GH 180 BA 180 PA 201 180 181 179 180 179 180 180 181 179 present study, F comes out to be 1.7 percent. STEPHENSON, WYATT and NORDSKOG (1953) reported that percent egg production declined.43 *.04 for each one percent increase in inbreeding. If this is applied as an inbreeding correction to the present data, we have, 2.6-1.7(.43) = 1.9 percent which then represents the decline in egg production owing to relaxed selection from genetic effects other than that due to inbreeding. Therefore, a rough approximation is that about one quarter of the egg production decline from relaxed selection in the present experiment can be accounted for by inbreeding depression. The second important question is whether egg production in the commercial strains sampled has plateaued. Because commercial breeding has almost completely changed over the past 15 to 20 years from pure strain breeding based on family selection and progeny testing to some form of hybridization, it appears that the rate of genetic improvement from intra-population selection has significantly slowed down or has plateaued to justify the switch by commercial breeders to the more complex and costly procedures or inter-population crossing commonly used today. In an early experiment at the Maine Agricultural Experiment Station reported by PEARL (1915), no apparent improvement in egg production could be

REGRESSION WITH RELAXED SELECTION 41 3 demonstrated from individual selection on first-year egg records over a ten-year period. A comprehensive case study reported by DICKERSON (1955) failed to show any increase in egg production in a commercial Leghorn strain selected over a 22-year period. DICKERSON concluded that this strain had reached a fluctuating plateau even though the amount of genetic variance had not declined and the selection differential applied had been large for the separate components of performance. DICKERSON attributed the refractory response to selection to interenvironmental slippage caused by unfavorable environmental trends or to genotype-environment interactions. On the other hand, the results of a long-time selection experiment at the University of California (LERNER 1958) did not indicate that a plateau had been reached in a White Leghorn strain selected on an intra-population basis. Continuous selection was made for egg production, viability and other traits over a 24- year period. LERNER S results showed substantial gains by selection over the 24- year period and with little if any indication of a population plateau. A selection experiment, covering a 17-year period, reported by HUTT and COLE (1955) also showed substantial improvement in egg production and viability. The above California and Cornel1 University selection experiments suffer from the fact that no controls were included to evaluate yearly trends which may have been wrongly interpreted as genetic improvement. GOWE, JOHNSON, DOWNS, GIBSON, MOUNTAIN, STRAIN and TINNEY (1959) used a randombred control strain to estimate genetic changes in two Leghorn lines selected for high total egg numbers (i.e., hen-housed egg production)--one, an old strain with a past history of selection, and the other a new strain started from a strain cross foundation. After the first test generation of selection, the old population appeared to have plateaued. The new population, starting at a lower level, showed an increase but did not exceed the level of the old strain. However, in neither of these selected populations did rate of egg production appear to have increased from selection. (Hen-housed egg production is an average value influenced by age at sexual maturity, adult mortality and rate of production. GOWE, et al. (1959) gave no information on changes in rate. This was estimated indirectly from their data by using an approximation for hen-day egg production given by NORDSKOG and HILL (1958), and then using a linear correction for maturity.) DICKERSON (1962) used a control repeat mating scheme to evaluate time trends over a four-year period. Again, he was unable to demonstrate any genetic increase in egg production by selecting on an index with egg production as a principal component. NORDSKOG and FESTING (1962) presented preliminary results of four generations of selection solely for rate of egg production. Two strains representing two breeds were used. Egg production in the White Leghorn strain appeared to be essentially plateaued, whereas in the second, a Fayoumi strain from Eygpt with little or no previous selection history for egg production, a small increase in rate of egg production might have occurred. Since publication of this preliminary

414 A. W. NORDSKOG AND F. G. GIESBRECHT study, three additional generations have been obtained. The results (to be published separately) show that rate of egg production has markedly increased in both the Leghorn and Fayoumi lines over the seven years but that, when the yearly means are corrected by using control populations, no improvement can be demonstrated. Each selection line has its own control: for the Leghorn line, the Cornell Randombred Leghorn population (KING 1958) was used and, for the Fayoumi line, a population maintained at Iowa State University as a closed-flock randombred mating, since 1949 was used. The evidence shows that the increase in egg production is due to environmental circumstances associated with time trends. For this reason, the purported genetic improvement in the Leghorn flocks reported by LERNER (1958) and HUTT and COLE (1955) may be only a reflection of time trends. By the same token, there is reason to believe that the strains used in this experiment were probably plateaued or nearly plateaued in egg production. We conclude from the total evidence considered that, egg production, even though it is a component of reproductive fitness, may decline when selection for it is relaxed. Accordingly, the results analyzed by the linear model seem valid. At the same time, it is to be expected that the rates of decline in egg production will vary from one population to another. Thus, we find statistical evidence for decline in only two of the three populations sampled. It is not wholly unexpected, therefore, that the recent study reported by BOHREN and MCKEAN (1964) yielded negative results. All of this further substantiates the conclusions reached by CLAYTON, et al. (1957) and ROBERTSON and REEVE (1952 with Drosophila, that the consequences of relaxed selection depends on the genetic constitution of the original populations. In the typical selection experiment, including the one reported here, it is not possible to separate the statistical regression effects of Mendelian segregation (A) from the consequences of genetic homeostasis (B). GRIFFING (1960) examined a theoretical two-locus epistatic case and concluded that the response to selection and the decline when selection is relaxed mimics the response which would occur if natural selection were operating antagonistically to artificial selection. That is, A mimics B. DICKERSON ( 1962) presented evidence from two different chicken experiments which seems to favor the concept that Mendelian segregation and recombination account for the decay of favorable epistatic effects of genes between generations within a strain. In the first experiment, he found that total egg production was about 4 percent less in three-way crosses compared with two-way crosses involving all combinations of the same pure strain females. The reduction in total egg production was accounted for primarily by a lower rate of production but also by higher mortality and later sexual maturity. In the second experiment, based on a repeat mating scheme (GOODWIN, DICKERSON and LAMOREUX 1960), the progeny of selected parents laid more eggs than the progeny from unselected parents, but this advantage was only transitory, suggesting that selection favored epistatis combinations of genes which in succeeding generations became lost by Mendelian segregation and recombination. DICKERSON (1962) stated that since natural selection would give positive em-

REGRESSION WITH RELAXED SELECTION 41 5 phasis to viability and egg production, it seems an unlikely explanation for the loss in performance for these same traits when deliberate selection is avoided, i.e., that genetic homeostasis is not the primary cause of regression in egg production when selection is relaxed. If we apply FISHER S fundamental theorem, then the additive genetic variance for fitness is zero when fitness is at a maximum. However, this does not require that the additive genetic variance for the separate components of fitness must also be zero. Therefore, natural selection may not necessarily favor maximum egg production and when artificial selection forces egg production above the optimum for fitness, egg production would be expected to decline when selection is relaxed. Conclusions: (1) Rate of egg production may decline in some populations of chickens by as much as 1 to 3 percent per generation when selection is relaxed. (2) Part of the decline in experimental populations of fixed size is expected to be accounted for by breeding. In the populations of the size used in the present study, inbreeding depression is estimated to account for one quarter of the decline due to relaxed selection. (3) The rate of decline is expected to differ among populations, depending on previous selection treatment and the genetic constitution of the populations before relaxation. The authors wish to thank PROFESSOR 0. KEMPTHORNE for his interest and valuable suggestions concerning the statistical aspects of this study. SUMMARY A seven-year experiment on the consequences of relaxing selection in three commercial strains of chickens is reported. Two were White Leghorn strains, and one was a Rhode Island Red. The experiment involved a total of 4080 birds. The typical mating consisted of 8 to 10 males and 80 to 100 females. The amount of inbreeding was estimated to be 1.7 percent per generation. If the assumption is made that these three populations have plateaued for egg production, then the results show a decline in rate of egg production amounting to 1.6 percent and 2.8 percent for the two Leghorn strains and 3.5 percent for the Rhode Island Red. About one fourth of this decline is attributed to inbreeding effects. Evidence also indicated that the rate of decline in egg production was not linear. No significant change was found for adult viability, fertility or hatchability in these populations when selection was relaxed. LITERATURE CITED BOHREN, B. B., and H. E. MCKEAN, 1964 Relaxed selection in a closed flock of White Leghorns. Genetics 49 : 279-284. CLAYTON, G. A., J. A. MORRIS, and ALAN ROBERTSON, 1957 An experimental check on quantitative genetic theory. I. Short-term responses to selection. J. Genet. 55: 131-151. DICKERSON, G. E., 1955 Genetic slippage in response to selection for multiple objectives. Cold Spring Harbor Symp. Quant. Biol. 20: 213-224. - 1962 Experimental evaluation of selection theory in poultry. Proc. 12th World s Poultry Congr. (Symposia), Sydney. 17-25.

416 A. W. NORDSKOG AND F. G. GIESBRECHT GOODWIN, K., G. E. DICKERSON, and W. F. LAMOREUX, 1960 An experimental design for separating genetic and environmental changes in animal populations under selection. pp. 117-138. Biometrical Genetics. Edited by 0. Kempthome. Pergamon Press, New York. GOWE, R. S., A. S. JOHNSON, J. H. DOWNS, R. GIBSON, W. F. MOUNTAIN, J. H. STRAIN, and B. F. TINNEY, 1959 Environment and poultry breeding problems. 4. The value of a random-bred control strain in a selection study. Poultry Sci. 38: 443-462. GRIFFING, B., 1960 Theoretical consequences of truncation selection based on the individual phenotype. Aust. J. Biol. Sci. 13: 309-343. HINKELMANN, K., 1963 metrics 19: 105-117. HUTT, F. B., and R. K. COLE, 1955 27 1-283. A commonly occurring incomplete multiple classification model. Bio- Multiple shifts for testing cockerels. Poultry Sci. 34: KING, S. C., 1958 A program to evaluate breeding systems for chickens. Proc. 11th World s Poultry Congr., Mexico City (English Summary) p. 743. LERNER, I. M., 1958 MATHER, K., and B. J. HARRISON, 1949 131-1 62. The Genetic Basis of Selection. Wiley, New York. Chap. 4, and Chap. 7. The manifold effects of selection. Heredity 3: 1-52, MOULTRIE, F., G. J. COTTIER, and D. F. KING, 1956 The effects of relaxed selection on performance of a strain of disease-resistant White Leghorns. Poultry Sci. 35: 1345-1348. NORDSKOG, A. W., and M. FESTING, 1962 Selection and correlated responses in the fowl. Proc. 12th World s Poultry Congr. (Section papers) Sydney, 25-29. NORDSKOG, A. W., and J. F. HILL, 1958 Correlation between egg production and adult viability in hybrid flocks. Poultry Sci. 37: 1265-1273. PEARL, R., 1915 Seventeen year s selection of a character showing sex-linked Mendelian inheritance. Am. Naturalist 49 : 595-608. ROBERTSON, F. W., and E. REEVE, 1952 Studies in quantitative inheritance. I. The effects of selection of wing and thorax length in Drosophila melanogaster. J. Genet. 50: 414-448. SHOFFNER, R. N., and R. E. GRANT, 1960 Relaxed selection in a strain of White Leghorns. Poultry Sci. 39: 63-66. STEPHENSON, A. B., A. J. WYATT, and A. W. NORDSKOG, 1953 Influence of inbreeding on egg production in the domestic fowl. Poultry Sci. 32: 510-516.