- Email: [email protected]
Genetic improvement of small ruminant local breeds with nucleus and inbreeding control: A simulation study
Small Ruminant Research 120 (2014) 196–203
Contents lists available at ScienceDirect
Small Ruminant Research journal homepage: www.elsevier.com/locate/smallrumres
Genetic improvement of small ruminant local breeds with nucleus and inbreeding control: A simulation study Gustavo Gandini a,∗ , Marcello Del Corvo a,b , Filippo Biscarini c , Alessandra Stella b a b c
Dipartimento DIVET, Università degli Studi di Milano, Via Celoria 10, 20133 Milano, Italy IBBA – CNR, Cascina Codazza, 26800 Lodi, Italy PTP – CeRSA, Cascina Codazza, 26800 Lodi, Italy
a r t i c l e
i n f o
Article history: Received 15 April 2014 Accepted 13 June 2014 Available online 21 June 2014 Keywords: Selection Inbreeding control Breeding programs Small ruminants Local breeds
a b s t r a c t In small ruminant local breeds of Southern Europe genetic selection is often constrained by small population sizes, poor animal identification, inadequate animal performance and pedigree recording, and organizational shortcomings. Under these conditions nucleus breeding schemes can offer practical and cost effective solutions. The paper investigated genetic gain in stochastically simulated dairy small ruminant nuclei of 100, 200 and 400 females, supporting commercial populations from 500 to 5000 females. In the nucleus, a young sire selection scheme was used, with optimum contribution selection on a dairy trait, at an annual inbreeding rate of 0.3%, corresponding to a generation inbreeding rate of 0.001. Sires, both selected and not-select as sires of sires, after 1 year of use in the nucleus were utilized in the commercial population for one, or alternatively, 2 years. Annual genetic gain ranged from a minimum of 0.073 SD with 100 females nucleus supporting a commercial population of 500 females, to a maximum of 0.138 SD with a 400 females nucleus supporting a commercial population of 5000 females. Negligible differences in genetic gain were observed between nuclei and corresponding commercial populations. When sires were used for only 1 year in the commercial population, we observed 7.7 years of genetic lag with the nucleus, that increased to 8.2 years when sires were used for 2 years. Results showed that there are opportunities for selection even in populations of a few hundreds of females. Considering a specific breeds, or a specific farming area, a cost benefit analysis should be carried out to orientate the choice of nucleus size and strategy of use of sires. © 2014 Elsevier B.V. All rights reserved.
1. Introduction Small ruminant local breeds in Southern Europe are mainly selected for dairy traits. Selection is often hampered
∗ Corresponding author. Tel.: +39 0250318046. E-mail addresses: [email protected] (G. Gandini), [email protected] (M. Del Corvo), fi[email protected] (F. Biscarini), [email protected] (A. Stella). http://dx.doi.org/10.1016/j.smallrumres.2014.06.004 0921-4488/© 2014 Elsevier B.V. All rights reserved.
by small population size, incorrect animal identification, inadequate animal performance and pedigree recording, and organizational shortcomings (Serradilla and Ugarte, 2006). The use of two or more sires with natural insemination in a single flock, as often observed, does not allow to unambiguously assign paternity of newborns without the use of genetic markers. The limited use of artificial insemination often results into insufficient connections to allow for across-flock genetic evaluation (Lewis and Simm, 2000). The introduction of exotic, more productive, breeds can result in failures because of their poor adaptability
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
to the harsh conditions of the Mediterranean extensive farming environment (Kalaisakis et al., 1977; Zervas et al., 1975). Alternatively, selection within local breeds has the potential to balance genetic improvement in productive and adaptation traits accounting for the local production system (Kominakis et al., 1997), and can contribute to economic sustainability of local breeds farming (FAO, 2013). Despite the importance of small ruminant farming in Southern Europe, the available information on genetic programs for local breeds farmed under low input and low technology production systems is scarce (Roden, 1995; Kominakis et al., 1997; Smulders et al., 2007). Some information and experience is available from tropical areas, where constraints similar to those found in Mediterranean marginal areas are observed, as reviewed by Kosgey et al. (2006) and Kosgey and Okeyo (2007). In general, in such production systems, genetic improvement can be generated in a small fraction of the population, the nucleus, and then disseminated to the whole population. Within the nucleus, trait and pedigree recording can be carried out at limited cost and organizational effort, and breeding strategies based on sire identification, such as the use of artificial insemination or of a single sire per flock, can be implemented, allowing more reliable breeding values estimation. The nucleus population can be an institutional flock in an experimental or public station, or be constituted by two or more coordinated farmer flocks. With respect to genetic flow, the nucleus can be defined as open or closed. In open nucleus schemes there is an upward flow of animals from the commercial population to the nucleus, while in a closed nucleus there is no such migration. Advantages and disadvantages of the two systems should be considered in specific cases (Roden, 1994). In general, in an open nucleus a larger portion of the population can be used for selection and inbreeding control is expected to be easier; however, pedigree and performance recording is not limited to the nucleus. Closed nuclei require a simpler organization, but inbreeding must be carefully monitored. More generally, selection in local breeds should guarantee the maintenance of within and among-breed genetic diversity. In the last years, different strategies have been proposed to control inbreeding in selected populations (Fernandez et al., 2011), the most efficient being selection with optimal contributions (Meuwissen, 1997; Grundy et al., 2000), which maximizes genetic gain subject to a restriction on inbreeding rate. The aim of the present work is to investigate genetic gain in dairy small ruminant populations with a closed nucleus, where a young sire breeding scheme with inbreeding control and optimal contributions selection is applied.
197
various sizes, supporting populations of up to 5000 females, are stochastically simulated under different schemes of sire use. In the nucleus, 1-year old young sires (YS), the unselected male offspring of sires of sires (SS) and dams of sires (DS), are used for a 1-year mating season on dams of dams (DD). SS are selected among YS and used on DS in the same 1-year mating season. The female offspring of SS × DS matings are kept for replacement. Should these not be sufficient, females born from YS and DD are randomly selected as replacements. YS × DD male progeny are discarded. DS are selected among all reproductive females (FFN: Fertile Females in the Nucleus). Optimum contribution selection (OCS) is applied to selection of SS and DS. After being used within the nucleus, SS and YS are mated to females in the commercial population (FFP: Fertile Females in the commercial Population), either for 1-year mating season (scheme B1) or for two mating seasons (scheme B2). Fig. 1 illustrates the nucleus and the commercial population, and relationships between them. 2.2. The simulation model 2.2.1. Nucleus Age structured nuclei of 100, 200 and 400 FFN were stochastically simulated. The simulation interval was 1 year, corresponding to the parturition interval with one mating season per year, and to the interval between birth and age of reproduction. Table 1 reports the simulated demographic parameters for females and males in the nucleus. A single trait repeatability model was used to estimate animal model BLUP EBVs, including a random genetic effect, random environmental effects and a residual effect. No fixed effects were simulated. Selection was applied for a dairy trait with heritability 0.3 and repeatability 0.5. Inbreeding and genetic level were computed each year from average values of newborn females. The founder population was assumed to be unrelated: true breeding values (TBV) were sampled from a normal distribution – N(0,1) – and genetic change was expressed in standardized genetic units. In the selection of SS and DS, genetic gain was maximized subject to a fixed annual inbreeding rate of 0.3%, by placing a penalty on the average relationship among selected animals. The following objective function (H) was maximized: H = x a − x A x
(1)
A semi-stochastic computer simulation has been developed to analyze a number of breeding scenarios.
where x is the vector of individual candidate parents contributions, i.e. the proportion of matings; a is the vector of EBVs of the candidates; A is the matrix of pedigree relationships among candidates weighted by , the penalty assigned to the average relationships among candidates to control for inbreeding. Candidate contributions (vector x) were multiplied by 1/2 to ensure that contributions from each sex sum to 0.5 (Meuwissen, 1997). Function H returns the genetic level of next generation parents, minus the cost due to increased average relationships. As in Berg et al. (2007), to achieve the target annual inbreeding rate of 0.3%, was empirically determined through trial and error. Eq. (1) was maximized by using an annealing algorithm (Press et al., 1989). Additional details on the simulation model are described in Gandini et al. (2014). Mating among both selected (SS × DS) and unselected animals (YS × DD) was at random, but avoiding closely inbred matings (i.e. between half and full sibs and between parents and their progeny). The number of DS was set to correspond the desired number of male progeny (YS), accounting for sex ratio at birth and random culling. The number of selected SS was obtained through the OCS algorithm, driven by the weight and in order to achieve the target inbreeding rate while maximizing genetic gain. Selection in the nucleus was simulated for 25 years. During the first 5 years of simulation truncation selection was applied in order to build the pedigree used in the following 20 years of OCS. From year 15 onwards, cumulative inbreeding and genetic level both followed a monotonic and linear trend. Eight hundred iterations were used. Average annual genetic gain (G) and inbreeding rate (F) were computed at year 25 subtracting their level at year 15 and dividing by ten:
2.1. The breeding scheme
G =
G25 − G15 10
(2)
F =
F25 − F15 10
(3)
2. Methods
The whole population is divided in two tiers: the closed nucleus where recording and selection are carried out, and the commercial population that receives genetically superior sires from the nucleus. Migration from nucleus to the commercial population is restricted to males. Nuclei of
where G25 , G10 , F25 and F10 are the genetic and inbreeding levels at years 25 and 10, respectively.
198
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
Fig. 1. The simulated breeding scheme: nucleus and commercial population. SS: sires of sires; YS: young sires; DS: dams of sires; DD: dams of dams. B1 and B2 refer to different uses of sires from the nucleus in the commercial population, 1 and 2 years, respectively. 2.2.2. Commercial population Commercial populations from 500 to 5000 FFP were deterministically simulated, with the same age structure of the nucleus. TBVs of the founder population were set equal to zero and genetic change was expressed in genetic standard deviation units. TBVs in age group zero were computed as average of the mean TBV of their fathers -i.e. the sires imported from the nucleus one (B1) or 1 and 2 years (B2) earlier- and mothers (born in the commercial population). In the commercial population, YS and SS performed the same number of matings. Inbreeding rate was only monitored in the nucleus, based on the consideration that sires from the nucleus were homogenously used in the commercial population (low variance of progeny size), yielding similar rates. Simulation in the commercial population began at year 8, when the first males selected with OC methodology in the nucleus were available. As in the nucleus, the annual genetic progress in the commercial population was computed as in Eq. (2).
Table 1 Demographic parameters of the nucleus.
2.3. Simulated scenarios Commercial populations of 500, 1000, 2000, 3000, 4000 and 5000 FFP were studied. Assuming a female to male ratio of 40:1, these populations required each year 13, 25, 50, 75, 100, and 125 sires respectively, under scheme B1. When sires were used for 2 years (scheme B2) the required number from the nucleus was halved and rounded to 7, 13, 25, 38, 50, 63, allowing for higher selection intensity or the use of smaller nuclei. Nuclei of 100, 200, 400 FFN were analyzed, which can produce each year a maximum of 43, 86, and 172 YS, respectively. Given female fertility and male survival constraints (see Table 1) not all schemes, and nucleus/commercial population size combinations, could be simulated. Table 2 summarizes the twenty-six simulated scenarios.
3. Results The target annual rate of inbreeding at 0.3% was met (range: 0.295–0.305) in all simulated nuclei. Standard errors for both F and G were <0.001 across all simulations. In the nucleus, generation interval for DD and DS
Parameter Age at first mating Female fertility Sex ratio at birth Parturition interval No. of matings per year No. of years of use of males Culling (random) females
Culling (random) males
Females: 1; young sires (YS) and sires of sires (SS): 1; One offspring-year, age 2 to 9 1:1 One year Females: 1; males: unlimited YS and SS: 1 Age 0–1: 10%; 1–2: 0%; 2–3: 15%; 3–4: 6%; 4–5: 6%; 5–6:13%; 6–7: 31%; 7–8: 33%; 8–9: 33% Age 0–1: 15%
Table 2 Simulated schemes as a function of the dimension of the nucleus (FFN) and of the base population (FFP). Use of males in the base population for 1 year (scheme B1), or for 2 years (scheme B2). FFN
100 200 400
FFP 500
1000
2000
3000
4000
5000
B1, B2 B1, B2 B1, B2
B1, B2 B1, B2 B1, B2
B2 B1, B2 B1, B2
– B2 B1, B2
– B2 B1, B2
– B2 B1, B2
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
199
Fig. 2. Genetic gain (G) (standard deviation units) in nuclei of 100, 200, and 400 females (FFN), as a function of the corresponding population size, from 500 to 5000 females (FFP), and sire use strategy (B1, B2). Table 3 Genetic gain per year – expressed as genetic standard deviations – in the nucleus under different scenarios.
was 4.6 years and 4.5, respectively. Given that the generation interval for SS and SD was constrained to 2 years, the average generation interval in the nucleus was 3.3 years across all simulations, and the 15 years interval used to estimate genetic gain corresponds to 4.5 generations. Fig. 2 illustrates genetic gain in nuclei of 100, 200 and 400 FFN respectively, as a function of the commercial population size (from 500 to 5000 FFP), and consequently of the number of sires required under strategies B1 and B2. Table 3 reports, within the simulated nucleus/ commercial population combinations (Table 2), genetic gain in the nucleus, ranging from a minimum of 0.073 (100 FFN, 500 FFP, B2) to a maximum of 0.138 (400 FFN, 3000 FFP B1; 400 FFN, 5000 FFP, B2) per year. Table 4 reports the number of SS selected by OCS for given numbers of FFN, FFP, YS, and DS. With a nucleus of 100 FFN and scheme B1, commercial populations of 500 and 1000 FFP required 14 and 26 YS, and on average 8.2 and 9.4 SS were selected, respectively. With scheme B2, YS were 7 (500 FFP), 13 (1000 FFP) and 25 (2000 FFP), and 6.4, 7.7 and 9.4 SS were selected, respectively. With a population of 500 FFP and scheme B1, genetic gain was 0.088, 21% higher than with scheme B2 (0.073). With
FFP
FFN 100
500 1000 2000 3000 4000 5000
200
400
B1
B2
B1
B2
B1
B2
0.088 0.089
0.073 0.086 0.091
0.100 0.111 0.111
0.092 0.097 0.109 0.114 0.111 0.106
0.115 0.121 0.130 0.138 0.132 0.125
0.102 0.113 0.118 0.122 0.130 0.138
FFN: fertile females in the nucleus; FFP: fertile female in the commercial population; B1 and B2 are different timespans (1 and 2 years) of sire use in the commercial population.
a population of 1000 FFP genetic gain was around 0.089 with almost no differences between the two schemes B1 and B2. With a population of 2000 FFP, genetic gain slightly increased to 0.091. Using the nucleus of 200 FFN and scheme B1, the population of 2000 FFP required 50 YS and 11.6 SS were selected. Under scheme B2, by increasing the size of the
Table 4 Number of sires of sires (SS), young sires (YS) and dams of sires (DS), as a function of female population size of the nucleus (FFN), of the population (FFP), and of scheme of sire use (B1, B2). Scheme
FFN
FFP 500 SS
B1
B2
1000 YS
2000
3000
DS
SS
YS
DS
SS
YS
DS
100 200 400
8.2 7.44 6.52
14 14 14
39 39 39
9.4 10.6 9.37
26 26 26
73 73 73
11.6 12.8
50 50
100 200 400
6.4 4.54 4.34
7 7 7
20 20 20
7.7 7.12 6.07
13 13 13
36 36 36
9.4 10.5 9.33
25 25 25
4000
SS
YS
DS
SS
139 139
14.3
74
206
70 70 70
12.1 11.3
37 37
103 103
5000 YS
DS
SS
YS
DS
12.9
100
278
12.3
126
350
11.6 12.8
50 50
139 139
10.7 12.9
63 63
175 175
200
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
Fig. 3. Number of young sires (YS) and of sires of sires (SS) in a nucleus of 400 females (FFN), as a function of commercial population size (FFP), with strategy of sire use B1.
population, the number of YS increased from 7 (500 FFP) to 63 (5000 FFP), and the average number of SS from 6.4 (500 FFP) to 10.7 (5000 FFP). Under scheme B1, genetic gain ranged from 0.099 (500 FFP) to 0.111 (2000 FFP), and in this interval performed better than scheme B2 on average by 7%. With scheme B2, genetic gain from 0.092 (500 FFP) reached a maximum at 0.114 (3000 FFP), then progressively decreased to 0.106 (5000 FFP). With the nucleus of 400 FFN, the number of YS increased respectively from 14 and 7 (500 FFP) to 126 and 63 (5000 FFP), the average number of SS from 6.5 and 4.3 (500 FFP) to 12.3 to 12.9 (5000 FFP), under schemes B1 and B2 respectively. With scheme B1, genetic gain increased progressively from 0.118 (500 FFP) to the maximum of 0.138 (3000 FFP), thereafter decreasing to 0.123 (5000 FFP). With scheme B2, genetic gain increased progressively from 0.110 (500 FFP) to the maximum of 0.136 (5000 FFP). In general, by increasing the number of YS in the nucleus to match the demand of sires in the larger commercial populations, the number of candidates to select SS increased. OCS used the larger number of candidates also by increasing the number of SS selected. Fig. 3 illustrates this aspect in the nucleus of 400 FFN, with scheme B2. With a population of 500 FFP, among the 7 YS candidates, on average 4.3 were selected as SS. The number of SS increased constantly up to the population of 4000 FFP, where among 50 YS candidates 12.8 SS were selected as SS. With larger populations (5000) this increment plateaus, despite the total number of sires
increased to 63. Similar trends were observed in 100 and 200 FFN nuclei. A comparison between genetic gain in 400 FFN nuclei and in the corresponding commercial populations of 500 to 5000 FFP, under both B1 and B2 schemes, is shown in Fig. 4. Small differences were observed and genetic gain was, on average, 0.001 (SD 0.001) higher in the nucleus than in the populations. Similar negligible differences were observed with the nuclei of 100 and 200 FFN. After year 15 of simulation, genetic increase was monotonic and linear (Fig. 5) and we could estimate the genetic lag (Lag), i.e. the difference in mean genetic level between nucleus and commercial population, under both schemes B1 and B2. Fig. 5 shows the typical dynamics of genetic level in the nucleus and in the respective population for 100 FFN nucleus, and 1000 FFP population. In case of use of sires for 1 year (scheme B1), the genetic level of the nucleus was 0.68 SD higher than in the commercial population, corresponding to 7.7 years Lag. When scheme B2 was used, the genetic level of the nucleus was 0.72 SD higher, corresponding to about 8.2 years Lag. 4. Discussion The use of OCS methodology in this study allowed the evaluation of the genetic progress that can be achieved at a predetermined acceptable rate of inbreeding in small populations under closed nucleus selection schemes.
Fig. 4. Genetic gain (G) (standard deviation units) in the nucleus (NU) and in the corresponding commercial population (POP), as a function of the commercial population size (FFP) and strategies of sire use B1 and B2. Nucleus size is 400 females.
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
201
Fig. 5. Genetic level (TBV) (standard deviation units) from years 15 to 25 of simulation in a nucleus of 100 females (NU) and in the corresponding commercial population of 1000 females (POP), with strategy of sire use B1 and B2.
Inbreeding is a major concern in closed nucleus breeding. The annual rate of inbreeding of 0.3% used in this study, considering the average generation interval of 3.3 years, corresponds to a generation inbreeding rate of approximately 1%, as suggested in theoretical studies (Meuwissen and Woolliams, 1994) and in practical conservation strategies (FAO, 2013). Assuming a 120 days milk production with a standard deviation of 16 kg, as reported in the Spanish Manchega sheep (Smulders et al., 2007), the genetic gain observed in these simulations will range from a minimum of 1.2 milk kg, in a population of 500 FFP with the nucleus of 100 FFN and scheme B2, to a maximum of 2.2 milk kg in a population of 3000 FFP with a nucleus of 400 FFN and scheme B1. Smulders et al. (2007), in a literature review of genetic progress observed in twelve European dairy sheep breeds, report a range of genetic gain per year from 0.5 kg to 4.3 kg. The genetic gain in our simulations is within the range observed across the European sheep breeds; however, direct comparisons are not possible because of their much larger population sizes (from 17,000 to 4,700,000 heads) and the different factors involved, including heritabilities, selection schemes and paths evaluated, and in particular inbreeding level. Rather small differences in genetic gain between the nucleus and the commercial population were expected due to of the different use of SS. Sires, both YS and SS, were used homogeneously in the population; conversely, in the nucleus SS contributions (no. of matings) were optimized in order to keep inbreeding at the requested rate and to maximize genetic gain. In our schemes, genetic gain was expected to vary in relation to both the dimension of the nucleus and that of the commercial population, which defines the number of YS in the nucleus. The use of artificial insemination, at least in part of the population, would reduce the number of YS to be produced in the nucleus. Within a given nucleus size, higher numbers of YS correspond to larger numbers of DS selected, and also to larger numbers of candidates for SS selection. With the OCS methodology selection intensity
cannot be computed as animals are selected accounting for both breeding values and genetic relationships. Thus, it is difficult to analyze how the number of young sires affects genetic gain in the selection pathways of DS and SS. Moreover, higher numbers of DS correspond to larger proportions of female replacements born from SS per DS matings. In general, the larger the nucleus, the higher the genetic gain. For example with a population of 1000 FFP, by expanding the nucleus, and using the best performing scheme of use of sires, genetic gain increases from 0.086 in the nucleus of 100 FFN, to 0.111 in the nucleus of 200 FFN and to 0.121 in the nucleus of 400 FFN. Considerable genetic lags were observed with both schemes B1 and B2. Through the paternal line the genetic lag was 3 (B1) or 3.5 (B2) years; this stems from 1-year old new sires being used first 1 year in the nucleus, and only afterwards being transferred to the commercial population. Besides this direct flow, alleles between animals in the nucleus and commercial populations are shared also through females from age class two (born from transferred sires at time t and then mated to new sires at time t + 2) onwards, with a genetic lag from 5 years upwards (contributions through females of age class > 2 yield longer genetic lags). Genetic lags observed in both schemes B1 and B2 could be reduced by decreasing the generation interval of females in the commercial population or, more effectively, by transferring not only males but also females from the nucleus to the commercial population, as discussed by Bichard (1971). Nucleus schemes, open and closed, have been shown to produce higher genetic gain than closed flocks selection (Roden, 1995). A number of investigations have shown that higher rates of genetic gain and a lower variability of selection response are achieved in open nucleus breeding systems rather than with closed nucleus in a variety of schemes (e.g. James, 1977; Sheperd and Kinghorn, 1992), as an open nucleus scheme permits higher selection intensities. A major advantage often reported for choosing open nucleus schemes is the lower inbreeding attained compared to closed nuclei, however these studies did not
202
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
consider the possibility of selection with inbreeding control. In this study, we were interested in selection schemes that can be implemented in populations with no pedigree and milk recording, lack of artificial insemination, difficulties of transfer of animals among flocks that can be associated to health risks or social constraint. Kosgey et al. (2006) following the analysis of nucleus schemes for small ruminant implemented in the tropics, underline that the success of the breeding programs generally is not determined by their inherent structure, but by their compatibility with the farming conditions and the involvement of the farmers. Then simplicity and applicability of the systems should be a major criterion in choosing a breeding scheme. For these reasons, we considered closed nucleus schemes as the only option. In our scheme, selection of sires was based on pedigree information. The use of progeny tested sires will require much larger nuclei, or more complex schemes as two tiers nucleus with a pre-nucleus to provide the capacity for progeny testing, as suggested by Kominakis et al. (1997). OCS allows to precisely control inbreeding rate. In addition, OCS is expected to produce higher genetic gain than truncation selection (TS) at the same inbreeding rate. Gandini et al. (2014) observed in simulated cattle populations, undergoing young bull selection schemes, advantages of OCS versus truncation selection of about 6% and up 20%. Higher advantages of OCS versus TS up to 44% have been observed in simulated populations with high selection intensities by Meuwissen and Sonesson (1998), but the large differences in genetic gain observed in favor of OCS could partially due to the lack of optimization of their TS scheme. Advantages from 13% to 30% has been observed in real large livestock populations, by comparing optimized retrospective selection with the observed values, in one ˜ step of selection with constraints on inbreeding (Avendano et al., 2003; Kearney et al., 2004; König and Simianer, 2006). However, in these analysis OC selection possibly detected rather unbalanced use of ancestors occurred in the population because their past management with no inbreeding control, with larger advantages of OCS versus truncation selection, as suggested by Fernandez et al. (2011). In this study, no comparison were made between OCS and truncation selection. This paper dealt with breeding schemes from the traditional perspective of pedigree and performance recording. During the last decade advancements in genotyping and sequencing technology, however, brought about the so called “genomics revolution”, broadening the set of tools available for animal selection (e.g. Van Eenennaam et al., 2014). This may have implications also for the genetic improvement of small ruminant local breeds. Genomic data could be used to estimate realized instead of expected relationships between animals. The matrix of genomic relationships would be readily available and would replace the additive relationship matrix in Eq. (1), allowing for OCS to be readily used at the settlement of the nucleus. Genetic progress would thus be quicker. Additionally, genomic data could be used to better estimate inbreeding levels. Again, realized instead of expected individual inbreeding could be computed, either directly from the matrix of genomic
relationships or by applying novel approaches based on runs of homozygosity (Purfield et al., 2012). 5. Conclusion The selection scheme analyzed in this work could be of use for the genetic improvement of local small ruminant populations farmed in low input production systems with low technological level, such as those found in the Mediterranean area and also in the tropics. In the commercial population, no pedigree and performance recording is requested, but only the homogeneous use of sires coming from the nucleus. Migration from the nucleus to the population is restricted to males, and assumes no artificial insemination in the population. OCS in the nucleus requires good pedigree and performance recording, however the adoption of a young sire scheme would facilitate selection even at low organizational levels. Where an optimum nucleus size cannot be adopted, a smaller sub-optimal nucleus breeding structure is a convenient option as it allows to start a breeding plan achieving some genetic gain. Whenever the nucleus flocks will not reflect the management conditions of the population farms receiving sires from the nucleus, appropriate considerations to avoid wastage of selection efforts should be done. Considering a specific breed, or a specific farming area, a cost/benefit analysis should be carried out to orientate the choice of the nucleus size. In addition, the male to female ratio and the period of use of sires should be adjusted to the specific breed structure. Acknowledgments This work was supported by the MiPAAF project “SelMol” and by the Accordo Quadro CNR-Regione Lombardia project “Risorse Biologiche e Tecnologie per lo Sviluppo Sostenibile del Sistema Agro-alimentare, WP4”. The research was also supported by the Marie Curie European Reintegration Grant “NEUTRADAPT” and the “Nextgen” project (http://nextgen.epfl.ch/) within the 7th European Community Framework Programme. References ˜ S., Villanueva, B., Woolliams, J.A., 2003. Expected increases in Avendano, genetic merit from using optimized contributions in two livestock populations of beef cattle and sheep. J. Anim. Sci. 81, 2964–2975. Berg, P., Sørensen, M.K., Nielsen, J., 2007. Eva Interface User Manual. http://eva.agrsci.dk/index.html (accessed 18.06.13). Bichard, M., 1971. Dissemination of genetic improvement through a livestock industry. Anim. Prod. 13, 401–411. FAO, 2013. In vivo conservation of animal genetic resources. FAO Animal Production and Health Guidelines. No. 14. FAO, Rome, Italy. Fernandez, J., Meuwissen, T.H.E., Toro, M.A., Maki Tanila, A., 2011. Management of genetic diversity in small farm animal populations. Animal 5, 1684–1698. Gandini, G., Stella, A., Del Corvo, M., Jansen, G.B., 2014. Selection with inbreeding control in simulated young bull schemes for local dairy cattle breeds. J. Dairy Sci. 97, 1790–1798. Grundy, B., Villanueva, B., Woolliams, J.A., 2000. Dynamic selection for maximising response with constrained inbreeding in schemes with overlapping generations. Anim. Sci. 70, 373–382. James, J.W., 1977. Open nucleus breeding systems. Anim. Prod. 24, 287–305.
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203 Kalaisakis, P., Papadimitrou, T., Zervas, N., Boyazoglou, J.G.T., Flamant, J.C., 1977. Comparaison des races ovines Chios et Frisonne avec leur croisements en Grèce continentale II. Production laitière. Genet. Sel. Evol. 9, 181–203. Kearney, F., Wall, E., Villanueva, B., Coffey, M.P., 2004. Inbreeding trends and application of optimized selection in the UK Holstein population. J. Dairy Sci. 87, 3503–3509. König, S., Simianer, H., 2006. Approaches to the management of inbreeding and relationship in the German Holstein dairy cattle population. Livest. Sci. 103, 40–53. Kominakis, A., Nitter, G., Fewson, D., Rogdakis, E., 1997. Evaluation of the efficiency of alternative selection schemes and breeding objectives in dairy sheep of Greece. Anim. Sci. 64, 453–461. Kosgey, I.S., Baker, R.L., Udo, H.M.J., Van Arendonk, J.A.M., 2006. Successes and failures of small ruminant breeding programmes in the tropics: a review. Small Rumin. Res. 61, 13–28. Kosgey, I.S., Okeyo, A.M., 2007. Genetic improvement of small ruminants in low-input, smallholder production systems: technical and infrastructural issues. Small Rumin. Res. 70, 76–88. Lewis, R.M., Simm, G., 2000. Selection strategies in sire referencing schemes in sheep. Livest. Prod. Sci. 67, 129–141. Meuwissen, T.H.E., 1997. Maximizing the response of selection with predefined rate of inbreeding. J. Anim. Sci. 75, 934–940. Meuwissen, T.H.E., Woolliams, J.A., 1994. Effective size of livestock populations to prevent a decline in fitness. Theor. Appl. Genet. 89, 1011–1026. Meuwissen, T.H.E., Sonesson, A., 1998. Maximizing the response of selection with a predefined rate of inbreeding: overlapping generations. J. Anim. Sci. 76, 2575–2583.
203
Press, W.H., Flannery, B.P., Teulosky, S.A., Vetterling, W.T., 1989. Numerical Recipes. Cambridge University Press, Cambridge. Purfield, D.C., Berry, D.P., McParland, S., Bradley, D.G., 2012. Runs of homozygosity and population history in cattle. BMC Genet. 13, 70. Roden, J.A., 1994. Review of the theory of open nucleus breeding systems. Anim. Breeding Abstracts 62, 151–157. Roden, J.A., 1995. A simulation study of open nucleus and closed nucleus breeding systems in a sheep population. Anim. Sci. 60, 117–124. Serradilla, J.M., Ugarte, E., 2006. Emerging genetic programs for small dairy ruminants. In: Proceedings of the 8th World Congress on Genetics Applied to Livestock Production, Belo Horizonte, Minas Gerais, Brazil, 13-18 August 2006, pp. 02–05. Sheperd, R.K., Kinghorn, B.P., 1992. Optimising multi-tier open nucleus breeding systems. Theor. Appl. Genet. 85, 372–378. Smulders, J.P., Serrano, M., Pérez-Guzmán, M.D., Jimenez, M.A., Uribe, H., Jurado, J.J., 2007. Stochastic simulation of Manchega sheep breed selection scheme. Impact of artificial insemination, progeny testing system and nucleus size on genetic progress and inbreeding. Livest. Sci. 106, 218–231. Van Eenennaam, A.L., Weigel, K.A., Young, A.E., Cleveland, M.A., Dekkers, J.C.M., 2014. Applied animal genomics: results from the field. Ann. Rev. Anim. Biosci. 2, 105–139. Zervas, N., Boyazoglou, J.G.T., Kalaisakis, P., Papadimitrou, T., Flamant, J.C., 1975. Comparaison des races ovines Chios et Frisonne avec leur croisements en Grèce continentale I. Viabilité et reproduction. Genet. Sel. Evol. 7, 277–291.
Contents lists available at ScienceDirect
Small Ruminant Research journal homepage: www.elsevier.com/locate/smallrumres
Genetic improvement of small ruminant local breeds with nucleus and inbreeding control: A simulation study Gustavo Gandini a,∗ , Marcello Del Corvo a,b , Filippo Biscarini c , Alessandra Stella b a b c
Dipartimento DIVET, Università degli Studi di Milano, Via Celoria 10, 20133 Milano, Italy IBBA – CNR, Cascina Codazza, 26800 Lodi, Italy PTP – CeRSA, Cascina Codazza, 26800 Lodi, Italy
a r t i c l e
i n f o
Article history: Received 15 April 2014 Accepted 13 June 2014 Available online 21 June 2014 Keywords: Selection Inbreeding control Breeding programs Small ruminants Local breeds
a b s t r a c t In small ruminant local breeds of Southern Europe genetic selection is often constrained by small population sizes, poor animal identification, inadequate animal performance and pedigree recording, and organizational shortcomings. Under these conditions nucleus breeding schemes can offer practical and cost effective solutions. The paper investigated genetic gain in stochastically simulated dairy small ruminant nuclei of 100, 200 and 400 females, supporting commercial populations from 500 to 5000 females. In the nucleus, a young sire selection scheme was used, with optimum contribution selection on a dairy trait, at an annual inbreeding rate of 0.3%, corresponding to a generation inbreeding rate of 0.001. Sires, both selected and not-select as sires of sires, after 1 year of use in the nucleus were utilized in the commercial population for one, or alternatively, 2 years. Annual genetic gain ranged from a minimum of 0.073 SD with 100 females nucleus supporting a commercial population of 500 females, to a maximum of 0.138 SD with a 400 females nucleus supporting a commercial population of 5000 females. Negligible differences in genetic gain were observed between nuclei and corresponding commercial populations. When sires were used for only 1 year in the commercial population, we observed 7.7 years of genetic lag with the nucleus, that increased to 8.2 years when sires were used for 2 years. Results showed that there are opportunities for selection even in populations of a few hundreds of females. Considering a specific breeds, or a specific farming area, a cost benefit analysis should be carried out to orientate the choice of nucleus size and strategy of use of sires. © 2014 Elsevier B.V. All rights reserved.
1. Introduction Small ruminant local breeds in Southern Europe are mainly selected for dairy traits. Selection is often hampered
∗ Corresponding author. Tel.: +39 0250318046. E-mail addresses: [email protected] (G. Gandini), [email protected] (M. Del Corvo), fi[email protected] (F. Biscarini), [email protected] (A. Stella). http://dx.doi.org/10.1016/j.smallrumres.2014.06.004 0921-4488/© 2014 Elsevier B.V. All rights reserved.
by small population size, incorrect animal identification, inadequate animal performance and pedigree recording, and organizational shortcomings (Serradilla and Ugarte, 2006). The use of two or more sires with natural insemination in a single flock, as often observed, does not allow to unambiguously assign paternity of newborns without the use of genetic markers. The limited use of artificial insemination often results into insufficient connections to allow for across-flock genetic evaluation (Lewis and Simm, 2000). The introduction of exotic, more productive, breeds can result in failures because of their poor adaptability
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
to the harsh conditions of the Mediterranean extensive farming environment (Kalaisakis et al., 1977; Zervas et al., 1975). Alternatively, selection within local breeds has the potential to balance genetic improvement in productive and adaptation traits accounting for the local production system (Kominakis et al., 1997), and can contribute to economic sustainability of local breeds farming (FAO, 2013). Despite the importance of small ruminant farming in Southern Europe, the available information on genetic programs for local breeds farmed under low input and low technology production systems is scarce (Roden, 1995; Kominakis et al., 1997; Smulders et al., 2007). Some information and experience is available from tropical areas, where constraints similar to those found in Mediterranean marginal areas are observed, as reviewed by Kosgey et al. (2006) and Kosgey and Okeyo (2007). In general, in such production systems, genetic improvement can be generated in a small fraction of the population, the nucleus, and then disseminated to the whole population. Within the nucleus, trait and pedigree recording can be carried out at limited cost and organizational effort, and breeding strategies based on sire identification, such as the use of artificial insemination or of a single sire per flock, can be implemented, allowing more reliable breeding values estimation. The nucleus population can be an institutional flock in an experimental or public station, or be constituted by two or more coordinated farmer flocks. With respect to genetic flow, the nucleus can be defined as open or closed. In open nucleus schemes there is an upward flow of animals from the commercial population to the nucleus, while in a closed nucleus there is no such migration. Advantages and disadvantages of the two systems should be considered in specific cases (Roden, 1994). In general, in an open nucleus a larger portion of the population can be used for selection and inbreeding control is expected to be easier; however, pedigree and performance recording is not limited to the nucleus. Closed nuclei require a simpler organization, but inbreeding must be carefully monitored. More generally, selection in local breeds should guarantee the maintenance of within and among-breed genetic diversity. In the last years, different strategies have been proposed to control inbreeding in selected populations (Fernandez et al., 2011), the most efficient being selection with optimal contributions (Meuwissen, 1997; Grundy et al., 2000), which maximizes genetic gain subject to a restriction on inbreeding rate. The aim of the present work is to investigate genetic gain in dairy small ruminant populations with a closed nucleus, where a young sire breeding scheme with inbreeding control and optimal contributions selection is applied.
197
various sizes, supporting populations of up to 5000 females, are stochastically simulated under different schemes of sire use. In the nucleus, 1-year old young sires (YS), the unselected male offspring of sires of sires (SS) and dams of sires (DS), are used for a 1-year mating season on dams of dams (DD). SS are selected among YS and used on DS in the same 1-year mating season. The female offspring of SS × DS matings are kept for replacement. Should these not be sufficient, females born from YS and DD are randomly selected as replacements. YS × DD male progeny are discarded. DS are selected among all reproductive females (FFN: Fertile Females in the Nucleus). Optimum contribution selection (OCS) is applied to selection of SS and DS. After being used within the nucleus, SS and YS are mated to females in the commercial population (FFP: Fertile Females in the commercial Population), either for 1-year mating season (scheme B1) or for two mating seasons (scheme B2). Fig. 1 illustrates the nucleus and the commercial population, and relationships between them. 2.2. The simulation model 2.2.1. Nucleus Age structured nuclei of 100, 200 and 400 FFN were stochastically simulated. The simulation interval was 1 year, corresponding to the parturition interval with one mating season per year, and to the interval between birth and age of reproduction. Table 1 reports the simulated demographic parameters for females and males in the nucleus. A single trait repeatability model was used to estimate animal model BLUP EBVs, including a random genetic effect, random environmental effects and a residual effect. No fixed effects were simulated. Selection was applied for a dairy trait with heritability 0.3 and repeatability 0.5. Inbreeding and genetic level were computed each year from average values of newborn females. The founder population was assumed to be unrelated: true breeding values (TBV) were sampled from a normal distribution – N(0,1) – and genetic change was expressed in standardized genetic units. In the selection of SS and DS, genetic gain was maximized subject to a fixed annual inbreeding rate of 0.3%, by placing a penalty on the average relationship among selected animals. The following objective function (H) was maximized: H = x a − x A x
(1)
A semi-stochastic computer simulation has been developed to analyze a number of breeding scenarios.
where x is the vector of individual candidate parents contributions, i.e. the proportion of matings; a is the vector of EBVs of the candidates; A is the matrix of pedigree relationships among candidates weighted by , the penalty assigned to the average relationships among candidates to control for inbreeding. Candidate contributions (vector x) were multiplied by 1/2 to ensure that contributions from each sex sum to 0.5 (Meuwissen, 1997). Function H returns the genetic level of next generation parents, minus the cost due to increased average relationships. As in Berg et al. (2007), to achieve the target annual inbreeding rate of 0.3%, was empirically determined through trial and error. Eq. (1) was maximized by using an annealing algorithm (Press et al., 1989). Additional details on the simulation model are described in Gandini et al. (2014). Mating among both selected (SS × DS) and unselected animals (YS × DD) was at random, but avoiding closely inbred matings (i.e. between half and full sibs and between parents and their progeny). The number of DS was set to correspond the desired number of male progeny (YS), accounting for sex ratio at birth and random culling. The number of selected SS was obtained through the OCS algorithm, driven by the weight and in order to achieve the target inbreeding rate while maximizing genetic gain. Selection in the nucleus was simulated for 25 years. During the first 5 years of simulation truncation selection was applied in order to build the pedigree used in the following 20 years of OCS. From year 15 onwards, cumulative inbreeding and genetic level both followed a monotonic and linear trend. Eight hundred iterations were used. Average annual genetic gain (G) and inbreeding rate (F) were computed at year 25 subtracting their level at year 15 and dividing by ten:
2.1. The breeding scheme
G =
G25 − G15 10
(2)
F =
F25 − F15 10
(3)
2. Methods
The whole population is divided in two tiers: the closed nucleus where recording and selection are carried out, and the commercial population that receives genetically superior sires from the nucleus. Migration from nucleus to the commercial population is restricted to males. Nuclei of
where G25 , G10 , F25 and F10 are the genetic and inbreeding levels at years 25 and 10, respectively.
198
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
Fig. 1. The simulated breeding scheme: nucleus and commercial population. SS: sires of sires; YS: young sires; DS: dams of sires; DD: dams of dams. B1 and B2 refer to different uses of sires from the nucleus in the commercial population, 1 and 2 years, respectively. 2.2.2. Commercial population Commercial populations from 500 to 5000 FFP were deterministically simulated, with the same age structure of the nucleus. TBVs of the founder population were set equal to zero and genetic change was expressed in genetic standard deviation units. TBVs in age group zero were computed as average of the mean TBV of their fathers -i.e. the sires imported from the nucleus one (B1) or 1 and 2 years (B2) earlier- and mothers (born in the commercial population). In the commercial population, YS and SS performed the same number of matings. Inbreeding rate was only monitored in the nucleus, based on the consideration that sires from the nucleus were homogenously used in the commercial population (low variance of progeny size), yielding similar rates. Simulation in the commercial population began at year 8, when the first males selected with OC methodology in the nucleus were available. As in the nucleus, the annual genetic progress in the commercial population was computed as in Eq. (2).
Table 1 Demographic parameters of the nucleus.
2.3. Simulated scenarios Commercial populations of 500, 1000, 2000, 3000, 4000 and 5000 FFP were studied. Assuming a female to male ratio of 40:1, these populations required each year 13, 25, 50, 75, 100, and 125 sires respectively, under scheme B1. When sires were used for 2 years (scheme B2) the required number from the nucleus was halved and rounded to 7, 13, 25, 38, 50, 63, allowing for higher selection intensity or the use of smaller nuclei. Nuclei of 100, 200, 400 FFN were analyzed, which can produce each year a maximum of 43, 86, and 172 YS, respectively. Given female fertility and male survival constraints (see Table 1) not all schemes, and nucleus/commercial population size combinations, could be simulated. Table 2 summarizes the twenty-six simulated scenarios.
3. Results The target annual rate of inbreeding at 0.3% was met (range: 0.295–0.305) in all simulated nuclei. Standard errors for both F and G were <0.001 across all simulations. In the nucleus, generation interval for DD and DS
Parameter Age at first mating Female fertility Sex ratio at birth Parturition interval No. of matings per year No. of years of use of males Culling (random) females
Culling (random) males
Females: 1; young sires (YS) and sires of sires (SS): 1; One offspring-year, age 2 to 9 1:1 One year Females: 1; males: unlimited YS and SS: 1 Age 0–1: 10%; 1–2: 0%; 2–3: 15%; 3–4: 6%; 4–5: 6%; 5–6:13%; 6–7: 31%; 7–8: 33%; 8–9: 33% Age 0–1: 15%
Table 2 Simulated schemes as a function of the dimension of the nucleus (FFN) and of the base population (FFP). Use of males in the base population for 1 year (scheme B1), or for 2 years (scheme B2). FFN
100 200 400
FFP 500
1000
2000
3000
4000
5000
B1, B2 B1, B2 B1, B2
B1, B2 B1, B2 B1, B2
B2 B1, B2 B1, B2
– B2 B1, B2
– B2 B1, B2
– B2 B1, B2
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
199
Fig. 2. Genetic gain (G) (standard deviation units) in nuclei of 100, 200, and 400 females (FFN), as a function of the corresponding population size, from 500 to 5000 females (FFP), and sire use strategy (B1, B2). Table 3 Genetic gain per year – expressed as genetic standard deviations – in the nucleus under different scenarios.
was 4.6 years and 4.5, respectively. Given that the generation interval for SS and SD was constrained to 2 years, the average generation interval in the nucleus was 3.3 years across all simulations, and the 15 years interval used to estimate genetic gain corresponds to 4.5 generations. Fig. 2 illustrates genetic gain in nuclei of 100, 200 and 400 FFN respectively, as a function of the commercial population size (from 500 to 5000 FFP), and consequently of the number of sires required under strategies B1 and B2. Table 3 reports, within the simulated nucleus/ commercial population combinations (Table 2), genetic gain in the nucleus, ranging from a minimum of 0.073 (100 FFN, 500 FFP, B2) to a maximum of 0.138 (400 FFN, 3000 FFP B1; 400 FFN, 5000 FFP, B2) per year. Table 4 reports the number of SS selected by OCS for given numbers of FFN, FFP, YS, and DS. With a nucleus of 100 FFN and scheme B1, commercial populations of 500 and 1000 FFP required 14 and 26 YS, and on average 8.2 and 9.4 SS were selected, respectively. With scheme B2, YS were 7 (500 FFP), 13 (1000 FFP) and 25 (2000 FFP), and 6.4, 7.7 and 9.4 SS were selected, respectively. With a population of 500 FFP and scheme B1, genetic gain was 0.088, 21% higher than with scheme B2 (0.073). With
FFP
FFN 100
500 1000 2000 3000 4000 5000
200
400
B1
B2
B1
B2
B1
B2
0.088 0.089
0.073 0.086 0.091
0.100 0.111 0.111
0.092 0.097 0.109 0.114 0.111 0.106
0.115 0.121 0.130 0.138 0.132 0.125
0.102 0.113 0.118 0.122 0.130 0.138
FFN: fertile females in the nucleus; FFP: fertile female in the commercial population; B1 and B2 are different timespans (1 and 2 years) of sire use in the commercial population.
a population of 1000 FFP genetic gain was around 0.089 with almost no differences between the two schemes B1 and B2. With a population of 2000 FFP, genetic gain slightly increased to 0.091. Using the nucleus of 200 FFN and scheme B1, the population of 2000 FFP required 50 YS and 11.6 SS were selected. Under scheme B2, by increasing the size of the
Table 4 Number of sires of sires (SS), young sires (YS) and dams of sires (DS), as a function of female population size of the nucleus (FFN), of the population (FFP), and of scheme of sire use (B1, B2). Scheme
FFN
FFP 500 SS
B1
B2
1000 YS
2000
3000
DS
SS
YS
DS
SS
YS
DS
100 200 400
8.2 7.44 6.52
14 14 14
39 39 39
9.4 10.6 9.37
26 26 26
73 73 73
11.6 12.8
50 50
100 200 400
6.4 4.54 4.34
7 7 7
20 20 20
7.7 7.12 6.07
13 13 13
36 36 36
9.4 10.5 9.33
25 25 25
4000
SS
YS
DS
SS
139 139
14.3
74
206
70 70 70
12.1 11.3
37 37
103 103
5000 YS
DS
SS
YS
DS
12.9
100
278
12.3
126
350
11.6 12.8
50 50
139 139
10.7 12.9
63 63
175 175
200
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
Fig. 3. Number of young sires (YS) and of sires of sires (SS) in a nucleus of 400 females (FFN), as a function of commercial population size (FFP), with strategy of sire use B1.
population, the number of YS increased from 7 (500 FFP) to 63 (5000 FFP), and the average number of SS from 6.4 (500 FFP) to 10.7 (5000 FFP). Under scheme B1, genetic gain ranged from 0.099 (500 FFP) to 0.111 (2000 FFP), and in this interval performed better than scheme B2 on average by 7%. With scheme B2, genetic gain from 0.092 (500 FFP) reached a maximum at 0.114 (3000 FFP), then progressively decreased to 0.106 (5000 FFP). With the nucleus of 400 FFN, the number of YS increased respectively from 14 and 7 (500 FFP) to 126 and 63 (5000 FFP), the average number of SS from 6.5 and 4.3 (500 FFP) to 12.3 to 12.9 (5000 FFP), under schemes B1 and B2 respectively. With scheme B1, genetic gain increased progressively from 0.118 (500 FFP) to the maximum of 0.138 (3000 FFP), thereafter decreasing to 0.123 (5000 FFP). With scheme B2, genetic gain increased progressively from 0.110 (500 FFP) to the maximum of 0.136 (5000 FFP). In general, by increasing the number of YS in the nucleus to match the demand of sires in the larger commercial populations, the number of candidates to select SS increased. OCS used the larger number of candidates also by increasing the number of SS selected. Fig. 3 illustrates this aspect in the nucleus of 400 FFN, with scheme B2. With a population of 500 FFP, among the 7 YS candidates, on average 4.3 were selected as SS. The number of SS increased constantly up to the population of 4000 FFP, where among 50 YS candidates 12.8 SS were selected as SS. With larger populations (5000) this increment plateaus, despite the total number of sires
increased to 63. Similar trends were observed in 100 and 200 FFN nuclei. A comparison between genetic gain in 400 FFN nuclei and in the corresponding commercial populations of 500 to 5000 FFP, under both B1 and B2 schemes, is shown in Fig. 4. Small differences were observed and genetic gain was, on average, 0.001 (SD 0.001) higher in the nucleus than in the populations. Similar negligible differences were observed with the nuclei of 100 and 200 FFN. After year 15 of simulation, genetic increase was monotonic and linear (Fig. 5) and we could estimate the genetic lag (Lag), i.e. the difference in mean genetic level between nucleus and commercial population, under both schemes B1 and B2. Fig. 5 shows the typical dynamics of genetic level in the nucleus and in the respective population for 100 FFN nucleus, and 1000 FFP population. In case of use of sires for 1 year (scheme B1), the genetic level of the nucleus was 0.68 SD higher than in the commercial population, corresponding to 7.7 years Lag. When scheme B2 was used, the genetic level of the nucleus was 0.72 SD higher, corresponding to about 8.2 years Lag. 4. Discussion The use of OCS methodology in this study allowed the evaluation of the genetic progress that can be achieved at a predetermined acceptable rate of inbreeding in small populations under closed nucleus selection schemes.
Fig. 4. Genetic gain (G) (standard deviation units) in the nucleus (NU) and in the corresponding commercial population (POP), as a function of the commercial population size (FFP) and strategies of sire use B1 and B2. Nucleus size is 400 females.
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
201
Fig. 5. Genetic level (TBV) (standard deviation units) from years 15 to 25 of simulation in a nucleus of 100 females (NU) and in the corresponding commercial population of 1000 females (POP), with strategy of sire use B1 and B2.
Inbreeding is a major concern in closed nucleus breeding. The annual rate of inbreeding of 0.3% used in this study, considering the average generation interval of 3.3 years, corresponds to a generation inbreeding rate of approximately 1%, as suggested in theoretical studies (Meuwissen and Woolliams, 1994) and in practical conservation strategies (FAO, 2013). Assuming a 120 days milk production with a standard deviation of 16 kg, as reported in the Spanish Manchega sheep (Smulders et al., 2007), the genetic gain observed in these simulations will range from a minimum of 1.2 milk kg, in a population of 500 FFP with the nucleus of 100 FFN and scheme B2, to a maximum of 2.2 milk kg in a population of 3000 FFP with a nucleus of 400 FFN and scheme B1. Smulders et al. (2007), in a literature review of genetic progress observed in twelve European dairy sheep breeds, report a range of genetic gain per year from 0.5 kg to 4.3 kg. The genetic gain in our simulations is within the range observed across the European sheep breeds; however, direct comparisons are not possible because of their much larger population sizes (from 17,000 to 4,700,000 heads) and the different factors involved, including heritabilities, selection schemes and paths evaluated, and in particular inbreeding level. Rather small differences in genetic gain between the nucleus and the commercial population were expected due to of the different use of SS. Sires, both YS and SS, were used homogeneously in the population; conversely, in the nucleus SS contributions (no. of matings) were optimized in order to keep inbreeding at the requested rate and to maximize genetic gain. In our schemes, genetic gain was expected to vary in relation to both the dimension of the nucleus and that of the commercial population, which defines the number of YS in the nucleus. The use of artificial insemination, at least in part of the population, would reduce the number of YS to be produced in the nucleus. Within a given nucleus size, higher numbers of YS correspond to larger numbers of DS selected, and also to larger numbers of candidates for SS selection. With the OCS methodology selection intensity
cannot be computed as animals are selected accounting for both breeding values and genetic relationships. Thus, it is difficult to analyze how the number of young sires affects genetic gain in the selection pathways of DS and SS. Moreover, higher numbers of DS correspond to larger proportions of female replacements born from SS per DS matings. In general, the larger the nucleus, the higher the genetic gain. For example with a population of 1000 FFP, by expanding the nucleus, and using the best performing scheme of use of sires, genetic gain increases from 0.086 in the nucleus of 100 FFN, to 0.111 in the nucleus of 200 FFN and to 0.121 in the nucleus of 400 FFN. Considerable genetic lags were observed with both schemes B1 and B2. Through the paternal line the genetic lag was 3 (B1) or 3.5 (B2) years; this stems from 1-year old new sires being used first 1 year in the nucleus, and only afterwards being transferred to the commercial population. Besides this direct flow, alleles between animals in the nucleus and commercial populations are shared also through females from age class two (born from transferred sires at time t and then mated to new sires at time t + 2) onwards, with a genetic lag from 5 years upwards (contributions through females of age class > 2 yield longer genetic lags). Genetic lags observed in both schemes B1 and B2 could be reduced by decreasing the generation interval of females in the commercial population or, more effectively, by transferring not only males but also females from the nucleus to the commercial population, as discussed by Bichard (1971). Nucleus schemes, open and closed, have been shown to produce higher genetic gain than closed flocks selection (Roden, 1995). A number of investigations have shown that higher rates of genetic gain and a lower variability of selection response are achieved in open nucleus breeding systems rather than with closed nucleus in a variety of schemes (e.g. James, 1977; Sheperd and Kinghorn, 1992), as an open nucleus scheme permits higher selection intensities. A major advantage often reported for choosing open nucleus schemes is the lower inbreeding attained compared to closed nuclei, however these studies did not
202
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203
consider the possibility of selection with inbreeding control. In this study, we were interested in selection schemes that can be implemented in populations with no pedigree and milk recording, lack of artificial insemination, difficulties of transfer of animals among flocks that can be associated to health risks or social constraint. Kosgey et al. (2006) following the analysis of nucleus schemes for small ruminant implemented in the tropics, underline that the success of the breeding programs generally is not determined by their inherent structure, but by their compatibility with the farming conditions and the involvement of the farmers. Then simplicity and applicability of the systems should be a major criterion in choosing a breeding scheme. For these reasons, we considered closed nucleus schemes as the only option. In our scheme, selection of sires was based on pedigree information. The use of progeny tested sires will require much larger nuclei, or more complex schemes as two tiers nucleus with a pre-nucleus to provide the capacity for progeny testing, as suggested by Kominakis et al. (1997). OCS allows to precisely control inbreeding rate. In addition, OCS is expected to produce higher genetic gain than truncation selection (TS) at the same inbreeding rate. Gandini et al. (2014) observed in simulated cattle populations, undergoing young bull selection schemes, advantages of OCS versus truncation selection of about 6% and up 20%. Higher advantages of OCS versus TS up to 44% have been observed in simulated populations with high selection intensities by Meuwissen and Sonesson (1998), but the large differences in genetic gain observed in favor of OCS could partially due to the lack of optimization of their TS scheme. Advantages from 13% to 30% has been observed in real large livestock populations, by comparing optimized retrospective selection with the observed values, in one ˜ step of selection with constraints on inbreeding (Avendano et al., 2003; Kearney et al., 2004; König and Simianer, 2006). However, in these analysis OC selection possibly detected rather unbalanced use of ancestors occurred in the population because their past management with no inbreeding control, with larger advantages of OCS versus truncation selection, as suggested by Fernandez et al. (2011). In this study, no comparison were made between OCS and truncation selection. This paper dealt with breeding schemes from the traditional perspective of pedigree and performance recording. During the last decade advancements in genotyping and sequencing technology, however, brought about the so called “genomics revolution”, broadening the set of tools available for animal selection (e.g. Van Eenennaam et al., 2014). This may have implications also for the genetic improvement of small ruminant local breeds. Genomic data could be used to estimate realized instead of expected relationships between animals. The matrix of genomic relationships would be readily available and would replace the additive relationship matrix in Eq. (1), allowing for OCS to be readily used at the settlement of the nucleus. Genetic progress would thus be quicker. Additionally, genomic data could be used to better estimate inbreeding levels. Again, realized instead of expected individual inbreeding could be computed, either directly from the matrix of genomic
relationships or by applying novel approaches based on runs of homozygosity (Purfield et al., 2012). 5. Conclusion The selection scheme analyzed in this work could be of use for the genetic improvement of local small ruminant populations farmed in low input production systems with low technological level, such as those found in the Mediterranean area and also in the tropics. In the commercial population, no pedigree and performance recording is requested, but only the homogeneous use of sires coming from the nucleus. Migration from the nucleus to the population is restricted to males, and assumes no artificial insemination in the population. OCS in the nucleus requires good pedigree and performance recording, however the adoption of a young sire scheme would facilitate selection even at low organizational levels. Where an optimum nucleus size cannot be adopted, a smaller sub-optimal nucleus breeding structure is a convenient option as it allows to start a breeding plan achieving some genetic gain. Whenever the nucleus flocks will not reflect the management conditions of the population farms receiving sires from the nucleus, appropriate considerations to avoid wastage of selection efforts should be done. Considering a specific breed, or a specific farming area, a cost/benefit analysis should be carried out to orientate the choice of the nucleus size. In addition, the male to female ratio and the period of use of sires should be adjusted to the specific breed structure. Acknowledgments This work was supported by the MiPAAF project “SelMol” and by the Accordo Quadro CNR-Regione Lombardia project “Risorse Biologiche e Tecnologie per lo Sviluppo Sostenibile del Sistema Agro-alimentare, WP4”. The research was also supported by the Marie Curie European Reintegration Grant “NEUTRADAPT” and the “Nextgen” project (http://nextgen.epfl.ch/) within the 7th European Community Framework Programme. References ˜ S., Villanueva, B., Woolliams, J.A., 2003. Expected increases in Avendano, genetic merit from using optimized contributions in two livestock populations of beef cattle and sheep. J. Anim. Sci. 81, 2964–2975. Berg, P., Sørensen, M.K., Nielsen, J., 2007. Eva Interface User Manual. http://eva.agrsci.dk/index.html (accessed 18.06.13). Bichard, M., 1971. Dissemination of genetic improvement through a livestock industry. Anim. Prod. 13, 401–411. FAO, 2013. In vivo conservation of animal genetic resources. FAO Animal Production and Health Guidelines. No. 14. FAO, Rome, Italy. Fernandez, J., Meuwissen, T.H.E., Toro, M.A., Maki Tanila, A., 2011. Management of genetic diversity in small farm animal populations. Animal 5, 1684–1698. Gandini, G., Stella, A., Del Corvo, M., Jansen, G.B., 2014. Selection with inbreeding control in simulated young bull schemes for local dairy cattle breeds. J. Dairy Sci. 97, 1790–1798. Grundy, B., Villanueva, B., Woolliams, J.A., 2000. Dynamic selection for maximising response with constrained inbreeding in schemes with overlapping generations. Anim. Sci. 70, 373–382. James, J.W., 1977. Open nucleus breeding systems. Anim. Prod. 24, 287–305.
G. Gandini et al. / Small Ruminant Research 120 (2014) 196–203 Kalaisakis, P., Papadimitrou, T., Zervas, N., Boyazoglou, J.G.T., Flamant, J.C., 1977. Comparaison des races ovines Chios et Frisonne avec leur croisements en Grèce continentale II. Production laitière. Genet. Sel. Evol. 9, 181–203. Kearney, F., Wall, E., Villanueva, B., Coffey, M.P., 2004. Inbreeding trends and application of optimized selection in the UK Holstein population. J. Dairy Sci. 87, 3503–3509. König, S., Simianer, H., 2006. Approaches to the management of inbreeding and relationship in the German Holstein dairy cattle population. Livest. Sci. 103, 40–53. Kominakis, A., Nitter, G., Fewson, D., Rogdakis, E., 1997. Evaluation of the efficiency of alternative selection schemes and breeding objectives in dairy sheep of Greece. Anim. Sci. 64, 453–461. Kosgey, I.S., Baker, R.L., Udo, H.M.J., Van Arendonk, J.A.M., 2006. Successes and failures of small ruminant breeding programmes in the tropics: a review. Small Rumin. Res. 61, 13–28. Kosgey, I.S., Okeyo, A.M., 2007. Genetic improvement of small ruminants in low-input, smallholder production systems: technical and infrastructural issues. Small Rumin. Res. 70, 76–88. Lewis, R.M., Simm, G., 2000. Selection strategies in sire referencing schemes in sheep. Livest. Prod. Sci. 67, 129–141. Meuwissen, T.H.E., 1997. Maximizing the response of selection with predefined rate of inbreeding. J. Anim. Sci. 75, 934–940. Meuwissen, T.H.E., Woolliams, J.A., 1994. Effective size of livestock populations to prevent a decline in fitness. Theor. Appl. Genet. 89, 1011–1026. Meuwissen, T.H.E., Sonesson, A., 1998. Maximizing the response of selection with a predefined rate of inbreeding: overlapping generations. J. Anim. Sci. 76, 2575–2583.
203
Press, W.H., Flannery, B.P., Teulosky, S.A., Vetterling, W.T., 1989. Numerical Recipes. Cambridge University Press, Cambridge. Purfield, D.C., Berry, D.P., McParland, S., Bradley, D.G., 2012. Runs of homozygosity and population history in cattle. BMC Genet. 13, 70. Roden, J.A., 1994. Review of the theory of open nucleus breeding systems. Anim. Breeding Abstracts 62, 151–157. Roden, J.A., 1995. A simulation study of open nucleus and closed nucleus breeding systems in a sheep population. Anim. Sci. 60, 117–124. Serradilla, J.M., Ugarte, E., 2006. Emerging genetic programs for small dairy ruminants. In: Proceedings of the 8th World Congress on Genetics Applied to Livestock Production, Belo Horizonte, Minas Gerais, Brazil, 13-18 August 2006, pp. 02–05. Sheperd, R.K., Kinghorn, B.P., 1992. Optimising multi-tier open nucleus breeding systems. Theor. Appl. Genet. 85, 372–378. Smulders, J.P., Serrano, M., Pérez-Guzmán, M.D., Jimenez, M.A., Uribe, H., Jurado, J.J., 2007. Stochastic simulation of Manchega sheep breed selection scheme. Impact of artificial insemination, progeny testing system and nucleus size on genetic progress and inbreeding. Livest. Sci. 106, 218–231. Van Eenennaam, A.L., Weigel, K.A., Young, A.E., Cleveland, M.A., Dekkers, J.C.M., 2014. Applied animal genomics: results from the field. Ann. Rev. Anim. Biosci. 2, 105–139. Zervas, N., Boyazoglou, J.G.T., Kalaisakis, P., Papadimitrou, T., Flamant, J.C., 1975. Comparaison des races ovines Chios et Frisonne avec leur croisements en Grèce continentale I. Viabilité et reproduction. Genet. Sel. Evol. 7, 277–291.