The pedigree records (Table 1) of 4,529 Chinese Duroc (CD) pigs born from 1998 to 2009, 16,776 Chinese Landrace (CL) pigs born from 1993 to 2009, and 22,600 Chinese Yorkshire (CY) pigs born between 1993 and 2009 were provided by Sichuan Province Livestock Improvement Station (Chengdu, Sichuan, China). For each breed, a reference population was defined as piglets born from year 2008 to 2009 (two yrs) because the average generation interval for each breed was close to 2 yr (Duroc: 1.79 yr; Landrace: 1.87 yr; Yorkshire: 1.78 yr) and, hence, it comprises the last generation of data evaluated in each breed. The population size of Duroc was the smallest in three breeds because it is mainly used as the terminal paternal line in crossbreeding production system of Duroc×Landrace×Yorkshire (DLY), and the market demand for it is less in China. The parameters were calculated in the reference population for each breed (Table 1). Data used in this study came from eight commercial pig farms (DYZC, DZZC, JYZC, LESH, MYZC, NANC, NJPF, and SCBF, which are members of united breeding program of Sichuan province). The official breeding program in Sichuan province was begun in 1995, however, the size of population selected at that time was very small. The year 1999 was used as a first year where a continuous, formal performance testing scheme with significance size (>500) was implemented based on the project “United-Breeding of Foreign Pigs in Sichuan”. The numbers of animals in each birth year cohort are shown in Figure 1.

The completeness and depth of pedigree affects the estimated inbreeding coefficients and relationships among animals directly, and which also affects the estimated generation intervals and effective numbers of founders and ancestors to a certain extent (Stachowicz et al., 2010). Five different measures were used to assess the completeness of the pedigree: mean maximum generations, mean complete generations, mean equivalent generations, the percentage of known ancestors per generation, and pedigree completeness index (Melka and Schenkel, 2010; Stachowicz et al., 2010). The number of generations between an animal and its earliest ancestor was taken as the maximum generations number traced back, and mean maximum generations was the average maximum generations numbers traced back for all individuals in a pedigree. The pedigree completeness index (PCI), described by MacCluer et al. (1983), was calculated for all individuals for 1, 3, 5, and 7 generations back in the pedigree: where, C_sire and C_dam are the indexes for paternal and maternal contribution, respectively, and where, a_i is the percentage of ancestors known in generation i and d is a preset number of generations traced back in pedigree.

The parameters in genetic diversity in the Chinese Duroc, Landrace and Yorkshire breeds were computed using four software packages. The effective number of founders, founder genomes and non-founders were calculated using CFC (Sargolzaei et al., 2006); percentages of known ancestors in each generation, the proportion of gene pool explained by certain number of ancestors, and cumulated marginal contributions of ancestors were obtained using PEDIG (Boichard, 2002). The software package EVA (Berg, 2003) was used to obtain average inbreeding, average coancestry, expected inbreeding under random mating, degree of nonrandom mating, and pedigree completeness indices. The effective population size and generation interval were computed using ENDOG program (Gutiérrez and Goyache, 2005).

In the Chinese Duroc (CD), Landrace (CL) and Yorkshire (CY) pedigrees, the percentages of animals with both parents known were 86.05, 87.40 and 91.49%, respectively, since the beginning of the 1999 for all animals. Table 1 lists some parameters for the whole pedigree and reference population. The maximal ancestral path traced back in the pedigree for three Chinese breeds were 7, 7, and 6, respectively. The mean maximal generations, mean complete generations and mean equivalent generations in CD were the least (2.08, 1.59 and 1.82, respectively) in three Chinese breeds, followed in CL (2.46, 1.85 and 2.11), while the highest in CY (3.06, 2.32 and 2.65, respectively). Almost all ancestors were known in each breed in the first generation. Investigating the depth of pedigree, CY and CD had the highest (23.78%) and lowest (1.42%) percentage of ancestors known in fifth generation. Figure 2 shows the average PCI by year of birth, accounting for 1, 3, 5 and 7 generations for all animals. The PCI of CD and CY was over 90% since 2005 in 1 generation back, while in 7 generations back, it was under 30% for the two breeds. The PCI of CD was over 80% in recent two years when considering 1 generation back.

Effective population size (Ne) and generation interval for three Chinese swine breeds are listed in Table 2. The Ne of CD, CL and CY in the reference population were 76, 117 and 202, respectively. The percentage of inbred animals in the reference population for three Chinese swine breeds were 29.24, 34.98 and 22.92%, respectively. The longest and the shortest generation interval were observed in CL (1.87 yr) and CY (1.78 yr). The generation interval of females was larger than males, and slightly larger than the same breeds in other countries (such as the paper described by Melka and Schenkel, 2010) because the lifespan of boars and sows were longer in China.

Some parameters derived from the analysis of probability of gene origin in the reference population (2008 to 2009) are listed in Table 3. The lowest total number of founders was observed in CD (295), followed by CL (731), and the highest was CY (958). However, oppositely, the fewest effective number of founder genomes was observed in CY (41.85), followed by CL (46.78), and the most was in CD (49.13). The f_e/f ratio in CD was larger than in CY and CL. This might mean that the selection intensities of interesting traits (such as days to 100 kg, backfat to 100 kg, litter size, etc.) in CY and CL were higher than in CD. The f_ge/f_e ratios of CD, CL and CY were 0.56, 0.32 and 0.31, respectively.

About 24 ancestors in CY explained 50% of gene pool in the reference population, while in CL and CD, more ancestors (36 and 30, respectively) were needed to contribute 50% to the current gene pool (Table 3). A total of 261 ancestors explained the whole gene pool of the reference population in CD. More ancestors of CY and CL were needed to explain 100% of the current gene pool (Table 3 and Figure 3). This indicates CY and CL have more genetic diversity than CD breed based on the ancestral contributions.

The percentages of inbreeding animals in the reference population for CD, CL and CY were low (29.24, 34.98 and 22.92, respectively, Table 2). Observing the changing trend of inbreeding coefficients over the last decade (Figure 4a), CY showed the least average inbreeding coefficient in the whole process, followed by CL, while CD had the highest inbreeding coefficient in the same period. The average inbreeding coefficients of three breeds were not over 1.3%, which were smaller than those values (4 to 18%) described by Melka and Schenkel (2010) and Welsh et al. (2010). This implied the inbreeding level of three imported swine breeds in Sichuan province was low. Figure 4b shows the trend of average coancestry between sires and dams, as expected, the trend was similar to the corresponding trend in expected inbreeding for CD.

There has been considerable deviation of the actual inbreeding coefficient from expected inbreeding under random mating in each breed in the past decade (Figure 5). The deviations between the actual inbreeding and expected inbreeding for all years in CD were almost positive, whereas in CL and CY, it was positive before 2004, but mostly negative afterward.

The amount of genetic diversity loss in the three Chinese breeds since 1995, due to different causes, is shown in Figure 6. In total, the amount of genetic diversity loss in CD increased gradually since 1996 and reached the highest (15% loss of genetic diversity) in the 2007, whereas in CL and CY, it was the highest (11 and 12%, respectively) in 1995 to 1996, and decreased gradually after that time. From this loss, average of about 63% (69.09, 57.95 and 60.57 for CD, CL and CY, respectively) over past decade could be attributed to unequal founder contributions. An average of about 37% (30.91, 42.05 and 39.43 for CD, CL and CY, respectively) of the genetic diversity loss was due to random genetic drift accumulated over nonfounder generations, caused mainly by the small effective population size. Figure 6 also shows the change in trend of effective population size (Ne) in three Chinese imported swine breeds. In total, the Ne of three Chinese swine breeds increased gradually since 1995, especially after 2000, the Ne of three swine breeds improved rapidly, and reached the highest in the period of 2001 to 2004. In the most recent two years (2008 to 2009), the Ne of CY increased rapidly, and exceeded 200 in the latest generation.

The overall depths of pedigree in three Chinese swine breeds were small. In Chinese Landrace (CL) and Yorkshire (CY), the PCI with 3 generation back was about 60%, only the PCI with 1 generation back exceeded 90%. The PCI of Chinese Duroc (CD) was even lower than CL and CY considering all generations back cases (Figure 2). Melka and Schenkel (2010) reported an average 4-generaton PCI of 90.4 and 89.6% for Canadian Duroc and Landrace born from 1978 to 2008, respectively. While in the current study, for the similar period, the PCI with 3 generations back was lower (45 and 65% for CD and CL, respectively) than Canadian swine breeds. The mean complete generations of CD and CL reached 1.59 and 1.85 in the whole pedigree, respectively. These values were lower than these reported in Canadian Duroc and Landrace (13 and 17.5, respectively) described by Melka and Schenkel (2010). Of course, the pedigree data in this study only involved records from 1993 to 2009, while the latter had records of about three decades before 2008. In addition, the Chinese swine production mode is introduction - degradation - re-introduction. In this case, the proportion of introduction was very large, moreover the period of re-introduction was short, which caused the genes of ancestors to be transferred for only a few generations.

The effective population size is the most important criterion to indicate the endangered status of breeds (Ruane, 2000). Melka and Schenkel (2010) reported an effective population size of 50 is necessary to withstand effects of inbreeding in a pig population, while a size of 500 is required to sustain the genetic diversity and evolutionary potential of the population for several generations. In this study, the effective population size for CD, CL and CY in the reference population (the present population) were 76, 117 and 202, respectively. These values are enough to resist the effect of inbreeding, but not to supply abundant diversity for genetic improvement in these populations. In addition, compared to the Ne of Duroc, Landrace and Yorkshire in other countries, the Ne of CD, CL and CY are in a reasonable range. For example, Melka and Schenkel (2010) reported the Ne of Canadian Duroc and Landrace was 72 and 125, respectively. Uimari and Tapio (2010) showed the Ne of Finnish Landrace and Yorkshire was 80 and 55, respectively. Welsh et al. (2010) also reported the Ne of American Duroc, Landrace and Yorkshire was 113, 74 and 113, respectively.

Maintaining within-breed genetic diversity for the long-term depends very much on the effort to increase effective population size, and balancing the contribution of ancestors to avoiding mating of much-related animals and reducing genetic drift (Simianer, 2005). In this study, a reference population was defined as piglets born from year 2008 to 2009 because the average generation interval for each breed was close to 2 yr (Table 2). The average generation interval in the reference population for CD and CL were 1.79 and 1.87, respectively (Table 2). These values were larger than Canadian Duroc and Landrace (1.60 and 1.73, respectively) in the same period as described by Melka and Schenkel (2010). This phenomenon is because the local farmers in Sichuan province prolong the lifespan (the longest lifetime of boars and sows exceeded 6 yrs) of boars and sows in production process. A lengthening lifespan of parents will increase the inbreeding level of a population because more full sibs or half sibs will be generated in the lengthened breeding period of parents. However, interestingly, the corresponding inbreeding level was still kept to a very low level (Figure 4a) in these populations, which was mainly due to new unrelated individuals being introduced often from other countries or regions. In the past, the production mode of Chinese imported pigs was introduction-degradation-re-introduction, which resulted in the swine nucleus population in Sichuan province being updated at intervals with foreign pig breeds. Therefore, a more efficient way to increase genetic diversity in Chinese swine populations may be directed towards increasing the effective population size of each breed by introducing genetically unrelated individuals between local farms or from other countries.

The parameters from the probabilities of gene origin are useful tools in measuring genetic diversity within breeds even after only a small number of generations (Boichard et al., 1997). The f_e was one of those parameters that was used to assess whether or not there was a balanced contribution of founders (Lacy, 1989). If all founders were to contribute equally, the effective number of founders is equal to the total number of founders, however, the effective number is usually lower than total number of founders because of unequal contributions of founders due to selection. The comparison between f_e and f demonstrates a loss in genetic diversity due to unequal contributions of founders, which could happen as a consequence of the excessive use of some animals as parents of subsequent generations (Melka and Schenkel, 2010). Moreover, the unequal contributions of founders will directly cause an increase of average inbreeding and coancestry in the reference population. The smaller the f_e/f ratio, the greater the amount of loss of genetic diversity caused by the unequal contribution of founders. The f_ge is another very important parameter for measuring genetic diversity (Lacy, 1995). The f_ge indicates the loss of genetic diversity due to both unequal founder contribution and random genetic drift. The f_ge/f_e ratio measures the impact of genetic drift excluding the effect of founder contribution on genetic diversity, so that lower ratios are associated with a higher impact of genetic drift. Melka and Schenkel (2010) reported the f_ge/f_e ratio of Canadian Duroc, Lacombe, Hampshire and Landrace were 0.07, 0.07, 0.36 and 0.33, respectively. They concluded that the effect of random genetic drift was substantial in Canadian Duroc and Lacombe populations. In this study, CY and CL showed a high f_ge/f_e ratio (0.31 and 0.32, respectively), which was close to the ratios of Canadian Hampshire and Landrace. This implies random genetic drift has smaller effect on genetic diversity loss than unequal founder contribution. Whereas in CD, both f_e/f and f_ge/f_e ratios were high (0.3 and 0.56, respectively), which showed the percentage of diversity loss caused by unequal founder contribution was larger in CD than in CY and CL.

For CY and CL, the actual average inbreeding coefficient tended to be smaller than expected inbreeding coefficient based on random mating since 2004 (Figure 5). In contrast, for CD, the actual inbreeding coefficient was almost higher than corresponding expected inbreeding coefficient since 2000. This implies that, in CY and CL, the mating strategies of local commercial farms in Sichuan province have improved since 2004, and the mating between animals with a high relationship has been avoided through the use of special breeding software (such as Netpig, registration mark in China: 2004SR05466). This action has efficiently controlled the increase of inbreeding. In CD, like CY and CL, the same mating strategy has been in place since 2004, and the corresponding inbreeding was also observed to decrease (Figure 5). However, the individuals in CD with higher than average coancestry were mated more often than the expected under the random mating, which indicates the mating strategy to minimize inbreeding (implemented in Netpig) has not been effective in this breed due to its smaller population size (<500 per year). Some studies indicate that the average coancestry is a good measure of genetic diversity because it accounts for both founder effects and the effects of genetic drift (Lacy, 1995; Meuwissen, 1997; Caballero and Toro, 2000). Considering the relationship between average coancestry and genetic diversity, Lacy (1995) demonstrated that selection with the objective of minimizing coancestry in the next generation would maximize genetic diversity. Therefore, application of minimum coancestry mating will be a good choice in Chinese swine breeds, especially in CD.

Overall, genetic diversity has been lost over the past decade in all Chinese swine breeds in this study, and this trend was obvious in CD. However, interestingly, for CY and CL, the highest amount of genetic diversity loss happened in 1995 to 1997, and then the amount of loss of genetic diversity decreased gradually after that period. At the latest (2009), the amount of genetic diversity loss for CL and CY decreased to 1.04 and 2.76%, respectively. Tracing the history of swine populations in Sichuan province, a closed breeding scheme with small population (<100 per year) and high selection intensity was implemented in 1993 to 1997, which caused serious unequal contributions of founders and random genetic draft (nearly each half, Figure 6), and a considerable amount (up to 12%) of genetic diversity was lost. After that, the amount of loss of genetic diversity decreased rapidly because the population size of commercial farms increased rapidly and more unrelated individuals from other countries (such as Denmark, Canada and American) were introduced termly.

The main cause (about 63%) of loss of genetic diversity in three Chinese swine breeds was shown to be unequal genetic contribution of founders. Random genetic drift and other factor (such as bottleneck) jointly contributed about 37% of the total loss of genetic diversity. However, Melka and Schenkel (2010) reported genetic drift was main factor (>60%) to cause the loss of genetic diversity in Canadian Duroc, Hampshire, Lacombe and Landrace swine breeds and unequal contribution of founders only contributed small fraction (<40%) of the loss of genetic diversity. Several other authors (e.g., Fernández et al., 2005; Melka et al., 2008) had also reported similar results. The excessive use of excellent boars and sows introduced from other countries in Sichuan province was the dominant cause of this difference, which caused very serious unequal contributions between founders. In addition, selection of breeding individuals and non-random mating between individuals selected are also important factors to cause the genetic diversity loss.