Breeding plans for sheep : past and possible progress.

Abstract

Breeding Plans for Sheep - Past and Possible Progress By HELEN NEWTON TURNER* SUMMARY THIS paper outlines the general structure of the Merino industry in Australia, and reviews its influence on plans for genetic improvement. Various mating plans are discussed, and the conclusion reached that mass selection on measured phenotype has a wide general application. Further information is required, however, as to the extent to which environment may alter ranking on performance. I. INTRODUCTION Various breeding plans are available to animal breeders, and the choice of a plan is influenced by many factors, including the type of animal and the particular aspect of its performance in which the breeder is interested. This paper is designed to review briefly the factors which govern the choice of a plan in relation to sheep-breeding for wool production, particularly under Australian conditions. There will be discussion of existing knowledge concerning theoretical and practical points which have a bearing on the decision, as well as an indication of need for further research. In defining a breeding problem with any animal, the first step is to obtain a general description of the breeding structure of the industry, together with its method of husbandry and its vital statistics. Next comes the problem of defining 'performance' - that is, establishing the criteria for assessing the type of production with which the breeding plan is to be concerned. The breeding plan itself will then cover the two processes of selection and mating, which are sometimes, though not always, distinct. Some aspects of the two processes can be made clearer if they are discussed separately. II. DESCRIPTION OF THE AUSTRALIAN WOOL-GROWING INDUSTRY (a) Breeding structure For present purposes, discussion will be confined to the Merino industry, which contributes the major part of Australian wool production. The Merino population is in two sections - commercial flocks, whose main product is wool, and registered studs, whose main product is breeding stock. Short and Carter (1955) have analysed the flock records of the registered studs, as published annually in the 'Australian Stud Merino Flock Register'. They have outlined a structure commencing with. 24 closed 'parent' studs, mating some 164,000 ewes, from which stock migrate to two other types - 'daughter' studs, dependent on one 'parent' only, and 'general ' studs, dependent on more than one 'parent'. In 1950 there were 258 daughter studs, mating 220,000 ewes, and 685 general studs, mating 497,000 ewes. Annual ram replacements for the non-registered flocks are supplied partly from the non-registered flocks themselves, but are also drawn from the registered studs; Short and Carter give estimates of the proportion from the latter, which vary from 50 per cent. in some years to as high as 90 per cent. in others. Of this, the parent and daughter studs contribute some 25 per cent., and the general studs the remainder. In the existing records there appears to be no way of determining the proportion of non-registered flocks which draw their rams consistently from the same source. The size of a flock has a direct bearing on the type of breeding plan which might be considered, and on the amount of genetic progress which might be made. Short and Carter do not present a complete distribution of registered studs according to flock size, but they quote the mean size in 1950 as 6,842 ewes for parent, 852 ewes for daughter and 726 ewes for general studs, taken over the whole of Australia. Tables of the distribution of all sheep according to flock size can be found in most year books published by the State Government Statisticians. Registered and non-registered sheep are included, *Section of Animal-breeding, Division of Animal Health and Production, C.S.I.R.O., McMaster Animal Health Laboratory, Sydney. 100 and there is no separation by breeds. Figures for New South Wales (1952/53) and Queensland (1952), the States with the greatest number of Merinos, indicate that 80 per cent. of sheep in New South Wales, and 97 per cent. in Queensland are run on holdings with at least 1,000 sheep per holding, while 59 per cent. in New South Wales, and 92 per cent. in Queensland are on holdings of 2,000 sheep or more. In New South Wales 13 per cent., and in Queensland 36 per cent., of sheep are on holdings running 10,000 or more. (b) Husbandry The general husbandry practice might be called 'extensive'. There is no shepherding of the type common in some countries. Most sheep-lands are now fenced, and the sheep are grazed in the open, in many cases being handled only a few times a year. Stocking rates vary considerably with location, running as low as one sheep to 20 acres or more. Mating times also vary with location, the rams being put with the ewes for periods of approximately six weeks, in some areas in the spring, in .others in the autumn. Mass mating is largely practised, though there is pedigree work in some, but not all, studs. Shearing is usually done annually, at times which vary with location. There may be occasional exceptions, due to circumstances. (c) Vital Statistics The average lamb-marking percentage for the industry (i.e., per cent. lambs marked to ewes mated) is of the order of 60, but the variation is very great, both from season to season in the same locality and from locality to locality. Ewes are seldom mated before they are two-tooths, and in some cases, notably the New England district of New South Wales, they may not be mated until they are four-tooths. Figures for annual losses among ewes of different ages are difficult to obtain, and vary considerably with season and locality. In some areas they may be well under 5 per cent. for ewes less than eight years of age. The breeding life of the ewe also varies; flock ewes are cast-for-age at five or six years in parts of South Australia, while in some top studs ewes may be kept as long as they are capable of raising a lamb. In general, however, the casting-age could be said to be six to eight years. The age-distribution of rams also varies considerably, but in general, rams would not be mated at a younger age than eighteen months. The generation-interval, all-important for predictions of genetic progress, is of the order of four to five years. For present purposes, it has been taken as four years. The vital statistics of the industry which are relevant to breeding problems have been summarised and discussed by various authors (Granger, 1944; Turner, 1951 a; C.S.I.R.O., 1952; Moule, 1952). III. DEFINITION OF 'PERFORMANCE' The Merino breeder is usually concerned with producing the maximum quantity of wool of a specified quality. To maintain flocks and allow for some selection, lamb marking percentages must also be kept as high as possible. 'Quantity' is easily defined in measurable terms. For ewes, greasy fleece weight is sufficiently accurate, but for the final selection of top stud rams, it is recommended that a sample of wool should be scoured to get an estimate of clean wool weight. 'Quality' is not readily defined in measurable terms. At present wool is sold in Australia after being classed by visual appraisal. According to an analysis by Young and Dunlop (1956), the visually-appraised characteristic which has the greatest influence on price' per pound is 'spinning count' or 'quality number', a classification supposedly on fineness, but in fact based mainly on number of crimps per inch. At the International Wool Textile Conference held in Australia in 1955, workers in textile research were asked by Ibiological research workers (Turner, 1956 a) for information as to the importance in manufacture of various characteristics of raw wool. The information was required to guide breeders in their selection programmes. It was felt that an analysis of factors influencing price was only a temporary solution, and that more fundamental knowledge was 101 necessary of the manufacturer 's requirements, in measurable terms. Clear answers to the questions were not forthcoming, although some research is in The desirability of uniformity of fibre diameter and length was progress. stressed, but without any specific guidance as to tolerance limits. The information obtained is to be published as a special conference bulletin. In the absence of specific information concerning measured characteristics, the present system of visual appraisal of quality must be used for the time being, with some check on measurable factors such as length and diameter. A system for definition of performance, partly by visual appraisal, partly by measurement, with fleece weight as a major criterion, was agreed on at a Fleece Measurement Conference held in May, 1954. This conference was called by C.S.I.R.O. and was attended by all workers in the field from various Australian Commonwealth, State, and University organisations. The proceedings (C.S.I.R.O., 1954) are for limited circulation, but the findings have been summarised (C.S.I.R.O., 1955). The question has also been discussed in detail by Morley (1955b). The literature relevant to the use cf measurement as an aid to selection has recently been reviewed by Turner (1956b). Recommendations with an emphasis placed on fleece weight apply to Australian conditions, where wool is the major product from most Merino enterprises. Research is shortly to be undertaken by C.S.I.R.O. to investigate the possibility of combining better mutton qualities with wool in the Merino. If the main products of the industry change, then the definition of performance will have to be modified. The same comment applies to the possible results of research on efficiency of wool production, which may lead to criteria for selecting animals which produce the most wool for the least feed consumption. One progress defining for more point which must be stressed, however, is that, for maximum genetic in any character, the number of characters being considered when performance should be kept to a minimum. The problem of selection than one character is discussed later. With the present definition of performance (maximum amount per head of wool of a specified quality, plus maintenance of as high a lamb-marking percentage as possible), a major part of the performance, namely, wool production, can be measured phenotypically on the sire. This situation is in direct contrast to that which obtains with dairy cattle or poultry, and has a considerable influence on the choice of a breeding plan. IV. SELECTION Selection may be required at the following stages of a breeding plan:(a) Selection of the strain of sheep to be used in a particular environment, without crossing. (b) Selection of breeds, strains or lines for crossing. (c) Selection of individual sheep within a flock, (i) For use in the same flock - single characters. (ii) For use in the same flock - more than one character. (iii) For use in different flocks. (a) Selection of strain Certain Merino strains have become quite clearly defined, and in choosing one a breeder has to consider not only the type of wool he wants to grow, but also the possibility that one strain or another might be better adapted to his particular environment. In 1948, the C.S.I.R.O. initiated a comparison of five strains (one fine-, three medium-, and one strong-wool) in three widely different environments. Dr. Dunlop, who is in charge of the trial, has made progress analyses of his results (C.S.I.R.O., 1956). The first point of note is that the strains retained distinctive characteristics in all environments. Secondly, Dunlop found strain x environment interactions which were statistically significant but small in magnitude, and for the most part of no great practical importance. It is of interest to record a problem of possible adaptation of strains which is being investigated in Rajasthan (India), by the Director of Sheep and Wool Improvement, Mr. N. L. Narayan, and his officers. These workers have defined three separate 'strains' within the so-called 'Bikaneri' breed of carpet-wool sheep, which constitute the main sheep population of Rajasthan. Small flocks of these three 'strains' are being maintained and observed at the Sheep Experiment Station at Jaipur, together with different 'strains' of other breeds, eight in all. Some interest centres in investigating whether or not these eight types have been developed in response to the demands of different environments. 102 Rainfall in the sheep-growing areas varies from 5 in. to 40 in., and the vegetation is correspondingly diverse. * (b) Selection for crossing This problem is not easily separable from its associated mating plan, and will be discussed in the next section. (c) Selection of individuals Sometimes selection is more than one. Many of simplify the discussion to complications which arise within a flock based on only one character, but more often it is on the problems in the two cases are similar, and it will deal with single characters first, dealing later with the when several characters are under consideration. Animals being selected from within one flock may be required for use in the same flock or elsewhere. The problems arising in these two cases are quite different, and will be dealt with separately. (i) For use in the same flock - Single characters. The main aim in choosing animals is accurate selection of the genotype. This, however, must always be assessed on the phenotype, either of the individual itself or of a group of its relatives. Accuracy of assessment of the phenotype must, therefore, be considered, together with the extent to which the phenotype is an indication of the genotype. Accuracy of assessment of phenotype: Four main points arise in relation to the assessment of phenotype - the possibility of assessment on both sexes, the actual technique to ,be used, the number of repeated assessments which are required, and the age (or ages) at which they should be made. As already noted, Merino breeders are fortunate in that wool, the major item in the performance record, is produced by the ram and can be assessed directly on him. Throughout most of the industry the present method of assessment for both ewes and rams is visual appraisal for general merit. Progeny-testing is carried out in a few instances, but the progeny are usually assessed by visual appraisal. Observations by Riches and of visual appraisal #by weighing selection differential* for lifetime tial for money return per head. similar increases in the selection Turner (1955) have shown that replacement of individual fleeces would at least treble the fleece weight, and double the selection differenMorley (1954b) and Moule (1954) have found differential for fleece weight. At present, there is no rapid substitute for visual appraisal of wool quality; this question has already been discussed in greater detail under 'Definition of Performance'. One way of increasing the accuracy of an observation is, of course, to repeat it, and to determine the value of repeated measurements on any characteristic its 'repeatability'? must be known. Morley (1951b) has found a repeatability of 0.7 for greasy fleece weight - a figure which indicates that selection on one annual fleece weight is sufficient. Riches and Turner (1955) found no advantage in waiting for a second full fleece before selection. They also found selection at seventeen to eighteen months for ewes preferable to selection at six to eight months. The questions of the earliest possible age of selection for rams and the influence of age on measured characteristics of the ram 's fleece are under investigation in C.S.I.R.O. experiments. Some American workers, including Pohle (1942) have tackled the problem of selection at weaning, but as the animals in their experiments were shorn for the first time at twelve months, their results do not contain information on the predictive value of fleece weight at weaning. Russian workers (Sannikov, 1939) have discussed the possibility of assessing animals at weaning, but with no mention of fleece weight. Phenotype as an indication of genotype: This question is tied up with the type of mating plan, but in view of the decision reached later in this paper *The amount by which the mean of the selected animals exceeds that of the unselected group from which they came. 'f'Repeatability' measures the extent to% which an animal' superiority at selection s is maintained throughout its life. A repeatability of 0.7 means that an animal with a fleece weight at selection one pound above the mean would have, on the average, a superiority of 0.7 lb. in each succeeding year. 103 that mass selection should be advocated at present in Australian sheep-breeding, it is appropriate to introduce at this point a comparison of selection on phenotype with selection by assessment of relatives. Lush (1947) carried out a theoretical investigation of the relative efficiency of various ways of combining an individual' performance with that of its relatives, ending with a linear score s combining the two. Morley (1951 a and 1952) was the first to attack the problem for the Australian Merino, and concluded that progeny-testing was not worth while when selecting a sire from among a group of rams reared together. In a paper (unpublished, 1954) read before the Australian Genetics Society at Canberra, Turner adapted the formulae developed by Lush (1947) to investigate various ways of using half-sib testing for Merinos, and again concluded that selection on measured phenotype of the individual was the most profitable form when choosing a sire from a group reared together. The crux of the problem is the 'heritability'* level. If the heritability level is high, selection on measured phenotype is the most profitable form of individual selection; if low, then information on relatives will be of value. The classification 'high' or 'low' cannot, of course, be absolute. In general, if the heritability level is over 0.3, selection on phenotype will supersede methods involving the examination of groups of relatives. For fleece weight and any measured fleece characteristics so far studied in the Merino, the heritability level has been well over 0.3 (Morley, 1951b; Turner, 1954; C.S.I.R.O., 1955). One advantage of this conclusion is that in a large number of cases pedigrees need not be kept. If a breeder wished to be able to use an outstanding sire extensively, some pedigreeing would, of course, be required so that inbreeding could be estimated and, if necessary, controlled. But for mass selection on phenotype, no pedigrees need be kept. This is an advantage under Australian conditions, in view of the small amount of pedigree work in existence. It should be stressed that the conclusion about selection on phenotype applies only to animals reared in the same environment. For rams of the same age, a direct comparison can be made. For rams of different ages, their deviations from the grand mean of their own drop must be compared. The problem of corndaring rams grown in different environments is discussed in the section entitled 'Selection of individuals - For use in different flocks'. (ii) For use in the same flock - More than one character: The problem of how best to use several characters in selection was first attacked by Smith (1936) and later by Hazel (1943). Their solution was to combine the characters into a single score or 'selection index'. Hazel and Lush (1942) compared three methods, and found the linear score to be most efficient. The other two methods are 'independent culling levels' (where a breeder fixes a certain level for each characteristic and culls all animals which fall below in any characteristic), and 'tandem selection' (where the breeder concentrates on raising one characteristic at a time, ignoring the others meanwhile). Although selection indexes are theoretically most efficient and, therefore, of value in research, there are difficulties about their use in commercial practice. Their calculation depends on the estimation of genetic variances and covariances. If these were relatively constant from flock to flock, a standard set of weighting coefficients could be computed and widely used. In fact, they i;ve not been found to be constant, nor is there any reason why theyE&;u:$l . . Each flock must have its own set of coefficients calculated. eventually, large properties in Australia employ their own geneticists, as is done in some animal industries in U.S.A., there remains the problem that estimates of genetic correlations have a high error until large quantities of data have been collected. This objection should not deter research workers, but it is important in commercial application. The system of 'independent culling levels' appears to offer greater scope for wide commercial use, provided the number of characters is kept small. The system recommended by the Fleece Measurement Conference (C.S.I.R.O., 1954 and 1955) involves a preliminary selection for 'off-type' fleeces, which amounts to fixing a culling level for quality. This culling rate should be as low as possible. Final selection is then on fleece weight. As shown by Hazel and Lush (1942) the loss in efficiency for independent culling levels, compared with the selection index, is not great when the proportion saved is small, as it is with rams. With 1 per cent. of animals saved, the ratio of genetic gain, using a selection index, to genetic gain, using independent culling levels, is 1 .lO for two *'Heritability' measures the proportion of the parents' superiority which will, on the average, appear in their offspring. 104 characteristics or 1.25 for five. These figures apply only to uncorrelated characters; the equations for more complex cases remain unsolved. In addition to wool production, maintenance of a high lamb-marking percentage was included in the definition of performance. At present there is no way of determining a ewe's potential as a breeder except by mating her and observing the results. The possible exception to this statement is Terrill 's observation (1949) that face-cover is negatively correlated with fertility in his Rambouillet flock, an observation which has been repeated in N.Z. on the Romney (Rae, A. L., and Clark, E. A., personal communications, 1955) and in Australia on the Merino (Fail and Dun, 1956). No high degree of correlation has been found, however, in C.S.I.R.O.` Merino breeding flock at Gilruth. s Plains, Cunnamulla, which has a much lower degree of face-cover than Terrill' s flock. Observations are being made by C.S.I.R.O. on the repeatability of lambbearing and rearing, and the possibility of using the early years of a ewe 's record as an indication of her breeding performance. Further investigations are being made of the heritability of the two aspects of fertility. The evidence so far (Barrett, 1956) indicates that no great improvement in lamb-raising can be expected from culling on a ewe' early records. If any culling is to be done, s however, the recommendation is again that the method of independent culling levels should be used - namely, that ewes be selected as two-tooths on wool production, with a culling margin left for subsequent selection on reproductive performance, if required. -This recommendation arises partly from the overwhelming importance of wool in the flock economy. If mutton were to increase in importance and to become of equal or greater importance than wool, as it is in U.S.A., it would become necessary to reconsider the recommendations for selection. (iii) For use in different flocks - One or more characters: Selection of individual animals (e.g., rams) from outside a flock involves two additional problems besides those discussed in the preceding paragraph. Comparison of phenotypes grown in different environments: If the introduced rams are to be compared with home-grown rams, or with rams from a different source, direct comparison of the phenotypes will give no indication of the genotypes. Comparison of rams from two sources at a show or a sale-ring, for example, is of no value as an indication of their worth for genetic improvement. If the mean fleece weights for unclassed rams of the same age on their home properties were known for a period of years, so as to give an accurate base for each property, then the rams could be compared in terms of deviations from the appropriate base. Under existing conditions this information is not likely to be available. There is no certainty at present that it would be conclusive, and further information is required on gene x environment interactions. Performance-testing under standard conditions has been tried for other livestock overseas notably dairy cattle and pigs in Denmark. The actual running of the standard conditions for the dairy cattle is presenting some problems, and A. Robertson (personal communication, 1955) has found greater differences between progeny groups under standard conditions than in the field. Performancetesting for rams could be organised much more simply, as the rams themselves, and not their progeny, could be kept under standard conditions for a given period of time. Before performance-testing on a centralised basis can be advocated, however, answers are required to some of the gene x environment questions raised in the next section. Under existing conditions, progeny-testing appears at first sight to be one way of comparing rams grown in different environments, but there seems to be no experimental work nor theoretical consideration of the comparative efficiency of various methods. Gene x environment interactions: Dunlop' observations on the strain trial, s previously discussed, indicated a lack of strain x environment interactions of any practical importance. However, the question of the possible existence of gene x environment interactions for individual animals remains unanswered. It is one of considerable importance for the whole of the Australian Merino industry, in view of the large concentration of the major studs in the better environments. Will a ram which ranks high among his fellows in one area also give superior progeny to theirs when used in another area? If Dunlop 's results with the strains held for individuals, the answer would be 'yes', as the strains ranked in the same order in all environments. But there were significant variations in the actual differences between strains, and the great magnitude of these differences might have made it difficult for the variations actually to change the 105 l ranking. Differences between individual rams would not be as great as the strain differences, and the possibility remains that gene x environment interactions might be important. Plans are being made to investigate the point, with Merinos, in C.S.I.R.O. I Falconer and Latyzewski (1952) carried out selection experiments with mice to throw light on this point. They selected for large body size under two planes of nutrition (ad lib. and 75 per cent. of normal) and obtained response in both lines, though it was less on the restricted than the full diet. At three points (after 5, 7 and 8 generations of selection) portions of each line were reversed in diet. The mice selected on the poor diet showed growth . rates on the ad lib. ration comparable to those of the mice which had been selected on it. But the mice selected on the high diet fell far below the mice selected on the poor diet when transferred to the latter. To quote the authors ' own summary, 'Thus, improvement of the genotype for rapid growth on a high plane of nutrition carried with it no improvement for growth on a low plane. But improvement of the genotype for growth on a low plane did carry with it a considerable improvement for growth on a high plane'. The authors point out that the physiological responses to selection which have led to these results present 'a complex problem which cannot yet be solved'. They considered that they had probably been selecting, in effect, for different characters in the two lines, although bodyweight was the selection criterion in each case. This point also requires to be investigated with sheep - namely, do the characteristics required for survival and high production in one environment differ from those in another? V. MATING PLANS Before starting a discussion of mating plans, it is necessary to review very briefly some ideas on the cause of genetic variability, as these causes may influence the choice of plan. The total phenotypic variation of a characteristic within a flock mayTkz considered as coming from two sources - genetic and environmental. genetic variability itself may come from three direct sources, with possibly two types of gene x environment interaction as well: 1. Additive genetic variance, which is assumed to depend directly on the number of 'plus' genes for the characteristic which are present. 2. Interaction between genes at the same locus. This may express itself as: Dominance, the heterozygote at any one locus having a value up to, but not greater than, the higher homozygote, gver-Dominance, the heterozygote at one locus then having a value greater than the higher homozygote. 3. Interaction between genes not at the same locus, which is known as 'epistasis'. 4. Unconscious gene x environment interactions of the kind already discussed, when animals within the flock somehow find better environments for themselves, although the whole flock is run together. 5. Conscious gene x environment interactions, when the better animals are picked out by the breeder and given preferential feeding. Gene x environment interactions are not likely to be important in influencing the decision for any breeding plan within a sheep flock. Those mentioned under 4 cannot be estimated, while those under 5 can be avoided. The relative magnitude of the other fractions of the genetic variance will govern the initial decisions regarding a mating plan. If the additive genetic portion (which is the portion expressed in the 'heritability' figure) is large for any characteristic, that characteristic should respond to selection. There have been cases recorded in laboratory experiments, however (e.g., Reeve and Robertson, 1953), where, after long continued selection, the heritability has been still high although response to selection had ceased. In general, however, if there is high heritability, the simple method of mass selection should be profitable - that is, the method of merely selecting the best animals from each generation, and putting them into the breeding flock. With a high heritability, as we have pointed out, selection on phenotype is adequate. If, or when, these simple methods fail, other methods are available for plants and some types of livestock; whether they will be valuable with sheep has not been shown. Some of these methods involve exploitation of hybrid vigour, and the choice of plan involves deciding whether 'dominance' or 'over-dominance' is the most likely contributor to intra-allelic interaction. A considerable controversy on this issue 106 c still rages in some genetic circles, and some interesting fields of research are devoted to developing techniques for separating the non-additive portions of the genetic variance. In the following sections it is proposed to discuss mass selection, to review briefly some of the other breeding plans which are used for cases where mass selection is not effective, and to give reasons for thinking that mass selection is likely to be the optimum mating plan for Australian conditions for some time to come. (a) Mass Selection Two questions arise immediately when mass selection is discussed. Firstly, for how long is any response to mass selection likely to cont