Estimation of crossbreeding parameters between Large White and Meishan porcine breeds

Summary - A crossbreeding experiment using Large White (LW) and Meishan (MS) pig strains was conducted. Direct, maternal and grand-maternal additive genetic effects along with direct, maternal and paternal heterosis effects were estimated for litter productivity traits: total number born (TNB), number born alive (NBA), number weaned (NW), litter weight at birth (WB) and at 21 days (W21), either adjusted or not for litter size, and survival rate from birth to weaning (SR). Direct, maternal additive and direct heterosis effects were also estimated for sow traits: weight before farrowing (SWF) and at weaning (SWW), weight loss (SWL) and feed consumption (SFC) during lactation. Data from 267 litters farrowed by 117 sows were analysed. Between breeds additive differences in prolificacy are mainly maternal (3.7 f 0.9, 4.2 t 0.8 and 2.8 t 0.8 piglets/litter in favour of MS for TNB, NBA and NW respectively). Maternal effects are also important, but in favour of LW, for adjusted litter weights. However, due to litter size differences, they are non-significant for unadjusted litter weights. Direct and grand-maternal differences are non-significant for all litter traits, except SR where grand-maternal effects are in favour of MS (4.1 f 1.5%). Large additive differences also exist in sow traits: LW dams are heavier 57 ! 8 and 56 t 6 kg for SWF and SWW respectively) and consume

Several other crossbreeding systems can be proposed for taking advantage of the high prolificacy of Chinese breeds (see for instance Sellier and Legault, 1986).
However, the high number of possible systems makes any exhaustive experimental evaluation almost impracticable. In this context, the analytical approach developed by Dickerson (1969Dickerson ( , 1973, based on the knowledge of a limited number of crossbreeding parameters (i.e. direct, maternal and grand-maternal breed effects, direct, maternal and paternal heterosis effects, and the corresponding epistatic recombination loss effects) is a useful tool for predicting and comparing the relative merit of various crossbreeding schemes.
Accordingly, an experiment was designed to estimate crossbreeding parameters relative to the cross between the most promising Chinese breed, the Meishan, and the most widely used French breed, the Large White, for the main traits of economic interest. The purpose of the present article is to evaluate breed additive differences and heterosis effects in reproductive traits.

MATERIALS AND METHODS
A. Animals and experimental design The general &dquo;3-step&dquo; design of the experiment is shown in Fig. 1. The first step is a 2-breed diallel whose main objective is to produce the 4 genetic types of females (MS, LWxMS, MSxLW, LW) and the 3 genetic types of males (MS, F1, LW) used as parents in the second step. Data from this first step have not been analysed because LW founder animals were selected on an index including average daily gain and backfat thickness and selection rates differed according to the sex, so that results would have been biased. The second step is a complete quadrallel; 12-21 boars from the 3 genetic types MS, Fl and LW are mated to the 4 above-mentioned genetic types of females (22-45 sows per group), leading to the production of 12 genetic types of litters. Sows are normally kept to produce 3 litters, each one with a different genetic type of boar. In the third step, females from these 12 genetic types are mated to boars from a third breed (Pietrain) and are kept to produce 5 litters.
Breeding animals in the second and third steps were chosen at random within the greatest number of litters after unthrifty animals were culled.
Data analysed in this article originate from the second step of the experiment. The distribution of the 267 litters produced according to sire and dam genetic types is presented in Table I.

B. Herd management
The sow herd has been managed under a batch farrowing system. Each batch included a maximum number of 24 sows. With the exception of some LW gilts showing delayed puberty, young females were bred at the age of 32 weeks, after a synchronisation treatment with a progestagen. In order to avoid any effect of this treatment on prolificacy, matings were not made on the induced oestrus, but on the following natural one. Natural service was used during the first 2 steps, while artificial insemination was employed in the third one. All females that did not conceive at first mating joined the subsequent farrowing batch where they had the opportunity to be mated once more.
Litters were born in individual farrowing crates. When necessary, some piglets could be moved to another crate within the first few hours after farrowing. With very few exceptions, these procedures were practised within each genetic type. Creep feed was provided to piglets at about 5 days of age. Weaning occurred at around 28 days post-farrowing. A 16% crude protein and 3100 kcal DE/kg diet was fed to all sows, ad libitum during lactation and at the rate of 2-2.2 kg for MS, 2.2-2.5 kg for crossbred and 2.5-2.7 kg for LW during gestation. A 3-4 kg forage complement (Beatruts or alfalfa) was also given during gestation. sow feed consumption during a 30-day lactation period (SFC). Consumption was measured daily during this period. Adjustment to 30 days was computed by truncating long lactations and adding the following quantity (Q) for short lactations: Q = N x CL, where N is the number of missing days and CL the consumption on the day before weaning.
sow weight before farrowing (SWF); -sow weight at weaning (SWW); -total weight loss of the sow between farrowing and weaning (SWL), computed as the difference between SWF and SWW; -the ratio of sow feed consumption to number weaned during lactation (SFC/NW); -the ratio of sow feed consumption to litter weight gain during the first 3 weeks of lactation (SFC/LWG). The latter 2 traits were considered for evaluating feed efficiency of the lactating sow.
Following Matheron and Mauleon {1979), the traits which depended on both sire and dam genetic types were regarded as litter traits. The others were considered as dam traits.

D. Statistical analyses
A 2-step procedure has been used to estimate crossbreeding genetic parameters; they have been computed from genetic type effects using a generalized least-squares method (Fimland, 1983).
1. Estimation of genetic type effects. Genetic type effects were obtained from a mixed model analysis (Henderson, 1984) for all traits except survival rate. The assumed model was as follows: where Y ijklmn = an observable random variable; !, = an unknown constant; b i = fixed effect of the i lh farrowing batch (i = 1... 27); p j = fixed effect of the j th parity (j = 1, 2, 3); d k = fixed effect of the k th dam genetic type (k = 1...4); Sl = fixed effect of the l th sire genetic type (I = 1, 2, 3); (ds) kl = interaction effect between dam and sire genetic types; (pd) jk = interaction effect between dam genetic type and parity; T km = random effect of the m th female nested within the k th dam genetic type with mean 0 and variance o!; E2!!i&dquo;,.! = random residual effect associated with the ijklmn ti ' record, with mean 0 and variance ce; Age at measurement and litter size at birth or at weaning were added as covariables to the model to analyse litter weights.
Preliminary analyses indicated that interactions between genetic type and farrowing batch effects, sire genetic type and individual dam effects, genetic type and age at measurement or litter size were small and not significant. Therefore, these interactions were not considered in the final analyses. The SAS Harvey procedure (SAS Institute, 1986) was used. The individual dam effect was treated as random by including the estimated ratios of residual to sow variances. Equations for sows were then absorbed. Sow variances were estimated from the data with a Restricted Maximum Likelihood method (Patterson and Thompson, 1971) using the same model as above. The SAS Varcomp procedure (SAS Institute, 1985) was used for this estimation. This model does not describe the data quite adequately because the relationships between animals are not taken into account. Estimates of fixed effects remain unbiased, but are not actually best linear unbiased estimates.
Survival rates were analysed with a Maximum Likelihood method (Bishop et al., 1975), using the SAS Catmod procedure (SAS Institute, 1985). The assumed model is the same as above, except that the random individual dam effect is ignored.
2. Estimation of crossbreeding parameters. Crossbreeding parameters were obtained by generalized least-squares analyses of litter or dam genetic type effects. The assumed genetic model was as follows: where y is 12x1 1 or 4 x 1 vector of estimates of litter genetic type effects; b is a 11 x 1 or 6 x 1 vector of crossbreeding genetic parameters; b' -(II. go go gm gm gn gn hO hm hP rO) for litter traits' b ' = ( N ' g MS gLW g MS g LW g MS gLW h' h' h P ) for litter traits; b ' = (! g M S gLW g ns + 9n s !Ew + !Ew h°) for dam traits; where a is an unknown constant; g!, g2 , g! are direct, maternal and grandmaternal effects for breed x (x =LW or MS); h°, h'!, h P are direct, maternal and paternal heterosis effects for the MSxLW cross; and r° is the direct epistatic recombination loss effect. K is a 12 x 11 or 4 x 6 matrix relating y to b. An example of a K matrix (for litter traits) is shown in Table II; e is a vector of residual errors; V is a 12 x 12 or 4x4 4 variance-covariance matrix of y.
The generalized least-squares estimate of b is -. -.

A. Analyses of variance
Mean squares (or chi-squares for survival rate) and significance of Fisher statistics (or Wald statistics for survival rate) are given in Tables III and IV. The farrowing batch effect is significant for all traits except litter weight at birth, but examination of batch means suggests that these effects are not related to any seasonal influence.
The parity effect is significant for litter size and unadjusted litter weights, but not for AWB and AW21. This tends to indicate that parity effects on litter weights are due to differences in litter size. This is, however, not entirely true, for parity tends to influence litter size and weight according to different patterns. Prolificacy remains stable during the 2 first parities and increases steadily in the third one (+3.3 and +2.8 piglets/litter at birth; +1.6 piglet/litter at weaning). On the other hand, litter weights increase linearly with parity, owing to much lower piglet weights in the first litters than in the second and third. Parity also affects sow weight before farrowing (27+6 6 and 17 t 8 kg between subsequent parities) and at weaning (18 ± 5 and 10+6 6 kg between subsequent parities), but has no influence on sow weight loss, feed consumption and efficiency during lactation. These trends are similar for the different genetic types of sows, as indicated by the absence of interaction between parity and dam genetic type. The only exception concerns piglet survival rate from birth to weaning, but this interaction has a complicated structure and is difficult to interpret.
None of the traits except survival rate exhibits any additive variation due to sire genetic type. On the other hand, all traits are greatly affected by dam genetic type. F1 sows have the largest litters at birth and at weaning, with little difference between reciprocal crosses. They farrow about 1 piglet more per litter than MS sows (15.3 vs. 14.2 for TNB; 14.7 vs. 13.7 for NBA) and 4 piglets more per litter than LW sows, whose mean litter size reaches 11.4 (TNB) and 10.3 (NBA). Differences are of similar magnitude at weaning: NW is on average 13.4, 12.2 and 9.4 piglets/litter for Fl, MS and LW sows respectively. Litter weight differences follow a somewhat different pattern. Litters from F1 sows are on average much heavier (17.5 kg at birth; 72.0 kg at 21 days) than litters from MS or LW sows (14.0 and 13.0 kg at birth; 50.7 and 50.9 kg at 21 days, respectively). Differences between adjusted litter weights are less important, but litters from crossbred dams remain higher than litters from LW and especially MS sows (+0.9 and +2.8 kg at birth; +3.6 and +16.5 kg at 21 days respectively). A significant interaction between sire and dam genetic types is obtained for all litter size and weight traits except AW21. Least-squares means for litter genetic types are presented in Table V. This interaction is partly due to an inversion of the ranking of sire genetic types in litters farrowed by purebred dams. Crossbred litters are larger and heavier than purebred ones, indicating the presence of some heterosis effects. Low performance of FlxMS litters also greatly contributes to this interaction.
On the other hand, weight, feed consumption and efficiency of sows during lactation mainly depend on their own genotype. Least-squares means for sow genetic types are presented in Table VI. LW and crossbred females have similar weights before farrowing and are much heavier (around 65 kg) than MS sows. Crossbred females are lighter at weaning, due to higher weight loss at farrowing and during lactation than purebred sows which are comparable from this standpoint. Crossbred sows also tend to consume more feed during lactation than LW (+6 kg) and above all MS (+26 kg). But, in spite of their high feed consumption, crossbred females have a much better feed efficiency (expressed as SFC/LWG) during lactation than purebred. On the other hand, MS sows consume less feed/piglet weaned than LW, Fl being intermediate.

B. Crossbreeding parameters
Because of the significant interaction between dam and sire genetic types, crossbreeding parameters have been estimated regarding prolificacy and litter weights as litter traits. On the other hand, sow weights, feed consumption and efficiency have been considered as dam traits.
Crossbreeding parameters for litter traits are given in Table VII. Additive differences between breeds for prolificacy are mainly of maternal origin. These maternal effects are in favour of MS sows and tend to decrease from birth to weaning (respectively 3.7 f 0.9; 4.2 f 0.8 and 2.8 ! 0.8 for TNB, NBA and NW). They are accordingly negative on piglet survival (&mdash;11.8 ±3.2%). Direct and grand-maternal effects are never significant, except for survival rate where grand-maternal effects are in favour of MS (4.1 f 1.5%). Estimates are close to 0 at weaning, but are not negligible at birth. Unfortunately, due to the large sampling errors of the estimates, it is not possible to know whether they reflect real differences. Additive differences between breeds are less important for unadjusted litter weights so that none of the estimated contrasts approaches significance. In general, the MS breed tends to have higher maternal effects and lower direct effects, but both contrasts are quite imprecisely estimated. On the other hand, adjusted litter weights are quite similar to prolificacy, with large maternal effects (but in favour of LW) and non-significant direct and grand-maternal effects.
Crossbreeding parameters for dam traits are given in Table VIII. Sow weights and feed consumption, either expressed on a sow or on a weaned piglet basis, exhibit important additive breed differences, mainly of direct origin. LW sows are much heavier and consume more feed than MS. On the contrary, no additive breed effect appears for weight loss and efficiency of piglet growth during lactation. All traits except feed conversion ratio per piglet present high heterosis effects, with estimates ranging from 10% (SWW) to 35% (SWL) of parental means.

IV. DISCUSSION
First of all, it must be kept in mind that the MS pigs used in this experiment originate from a very limited sample of animals, so that any extrapolation to the MS breed as a whole is unadvisable. Generally, results of MS and crossbred litters are consistent with those previously obtained in France (Legault and Caritez, 1983) and, for MS sows, with results obtained in China (Cheng, 1983;Zhang et al., 1983;Zhang et al., 1986). One exception concerns feed consumption of lactating LW sows, which is much less important than previously reported by Legault and Caritez (1983). In addition, LW purebred matings lead to somewhat lower litter sizes at birth (NBA) and at weaning (NW) than figures usually obtained in France for that breed (e.g. Benoit et al., 1987). This could have led to some overestimation of direct heterosis effects and inversely to some underestimation of direct additive effects on prolificacy and litter weights.
The effect of parity on prolificacy is somewhat different from the usual literature results, which generally indicate a linear increase in litter size between first and third parities. A similar trend (i.e. a lower than expected performance of second parity females) had already been found by Legault and Caritez (1983). However, this effect is not specific for sows derived from Chinese breeds, as several authors have recently reported similar results (see Clark and Leman, 1986). A common explanation is that high first parity litter size would increase sow weight losses during lactation and affect their subsequent litter size (Hillyer, 1979;Clark and Leman, 1986). This could be the case in the present study; parity does not affect total weight loss of sows but, as first parity litter weights are lighter, net weight loss of gilts during lactation is probably higher than weight loss of sows. Otherwise, the increase of litter weight with parity is a well-known result (see for instance Schneider et al., 1982;Buchanan and Johnson (1984) or Gaugler et al. (1984). However, none of these studies investigated the part taken by prolificacy in litter weight variability.
Contrary to parity differences, variation in litter weights between genetic types cannot be entirely explained through litter size. Indeed, adjusted litter weight means indicate an important additive and non-additive variability of individual piglet weight, which will be analysed in the next article of this series.
The analysis of litter and sow weights and of sow feed consumption provides some information on the respective nursing abilities of LW, MS and crossbred females. So, a comparison of the growth of crossbred litters fostered by LW and MS sows shows a significant superiority of LW females over MS. This indicates a better energy supply to piglets and accordingly a better production and/or composition of milk for LW sows. This superiority is likely to come from their greater appetite. Indeed, milk energy originates either from feed consumption of the sow or from the mobilization of the sow's body reserves. The above results seem to indicate that net weight losses of MS and LW sows fostering crossbred litters are comparable. Therefore, the higher milk energy amount provided by LW dams comes from a better energy availability of their body reserves or more likely from a higher feed energy supply related to their greater appetite. This also explains why feed efficiency per unit of piglet growth does not vary among purebreeds. Similar comparisons between LW and crossbred sows indicate a much better milk production and/or composition for F1 females. But, contrary to the former case, the superiority of crossbred dams comes to a large extent from a higher mobilization of their body reserves. The consequence is a better feed efficiency of piglet growth during lactation.
The estimation of genetic parameters has led to some unusual results. The main feature concerns maternal heterosis effects on litter size and weight. Estimates are from 2 to 4 times higher than usual values (from 14 to 36% of the parental means vs. 6-10% for average literature results (Sellier, 1976;Johnson, 1981;Bidanel, 1988). Litter weights also exhibit surprisingly high direct heterosis effects (21% and 17% v.s. 5% and 4% for average literature values on UWB and UW21, respectively). The large differences in litter size partly explain the high values obtained for litter weights, as shown by adjusting data for litter size. However, even so, estimates remain larger than usual values (14% and 11% at birth; 5% and 19% at 21 days for direct and maternal heterosis respectively).
Obtaining significant heterosis for sow weight is not surprising, as nulliparous and primiparous females are still growing actively and growth traits exhibit important non-additive variations. However, estimates are much larger than usual values.
Moreover, heterosis values should reduce with parity, as sows approach their mature size and weight, which are known to be mainly additive. This is not the case here, as estimates do not vary much with parity (35±2, 31t3, 36 + kg respectively before farrowing; 16 t 2, 17 ! 3, 23 ! 4 kg at weaning). A partial explanation could be a possible earlier maturity of MS (and maybe crossbred) females, which seem to reach their adult size earlier than LW (Bidanel, Caritez and Legault, unpublished data).
The third step of the present experiment will provide more detailed information on this problem. Heterosis for sow feed consumption and efficiency results from complex interactions between body size, appetite, milk production and litter weight gain. More precise studies are necessary to elucidate the respective part played by each of these components. Several general hypotheses can be proposed to explain the high heterosis values obtained in the present study: 1) The great genetic distance between LW and MS breeds. Heterosis level is related to between-breed genetic distance (Glodek, 1982;Lefort-Buson, 1986). This distance can be characterized through the comparison of allelic frequency distribution at marker loci in each breed (Glodek, 1982;Brunel, 1985). Unfortunately, the low number of founder animals of the French MS line makes it difficult to check this hypothesis. The only noticeable indication concerns the highly polymorphic swine major histocompatibility complex (SLA): among the 5 haplotypes found in the French MS line, 2 also exist in the French LW breed (Christine Renard, personal communication).
2) The existence of some inbreeding in the MS line. Crossbreeding involving inbred lines generally leads to high heterosis values (Sellier, 1970). Yet, this hypothesis is quite unlikely. As stated above, the experimental design has kept inbreeding at a low level (< 5%) so that its effect should be negligible on the basis of average literature values (Hill and Webb, 1982). On the other hand, the existence of some prior inbreeding could not be verified. However, it should not be very high, as parents of founder animals were not closely related.
3) The existence of a dominant major gene for prolificacy in the MS breed. Due to the complexity and the high coefficient of variation of litter size, testing this hypothesis requires considerable experimental work. The existence of a major gene for embryo survival can theoretically be tested from the data analysed in this study through the analysis of F2 and backcross litters distribution. Unfortunately, our present data set is insufficient to draw conclusions.
The other genetic parameters are more consistent with literature results. The lack of paternal heterosis observed in this study seems to be a general fact, as pointed out by recent reviews of Buchanan (1987) and Bidanel (1988). Pani et al. (1963) first reported significant grand-maternal effects on litter size at weaning. Since then, several other estimates have been reported by Smith and King (1964), Legault et al. (1975), Nelson and Robison (1976) and Johnson et al. (1978). They are all nonsignificant, in agreement with present results, but are generally estimated with very low precision and do not indicate any consistent trend with respect to the influence of the size of the birth litter of a female on its own reproductive performance. Direct and maternal effects were also estimated with low accuracy. However, the estimates confirm the prominent part played by the dam in the determination of litter size.

CONCLUSION
The first estimation of crossbreeding parameters for Large White and Meishan is of great interest for studying strategies of economic use in crossbreeding of the Meishan breed under intensive European management systems. Because of important maternal heterosis effects on prolificacy, the use of discontinuous crossbreeding plans involving crossbred females a priori constitutes the best short-term solution for using the Meishan breed. However, as shown by Legault et al. (1985) and Gueblez et al. (1987), the economic value of such systems also depends on the extent of the deterioration of production performance in Chinese crossbreds. This deterioration can be predicted from the knowledge of appropriate crossbreeding parameters. Estimation of these parameters for growth traits will be presented in the second article of this series.
Moreover, as pointed out by Hill (1971), short-term analysis is not entirely satisfactory for comparing the merit of various crossbreeding plans. Long-term results can differ widely from short-term conclusions, particularly for composite lines or continuous crossbreeding schemes. The value of these latter strategies greatly depends on the proportion of heterosis retained in advanced generations of crossing, i.e. on the amount of the epistatic recombination loss effects. The third step of this experiment will provide data for estimating these parameters.