The Torbiscal line was obtained in 1963 by blending four ancient Iberian pig strains preserved by private breeders, and since 1944-45 maintained with pedigree and data recording in the experimental herd of "El Dehesón del Encinar" (Oropesa, Toledo, Spain). Two of these strains were reddish and came from the Portuguese herds of the Count of Ervideira (Evora) and Mr. Picao Caldeira (Elvas). The others, both hairless and black coated, came from the herds of the Donoso brothers (Campanario), and Mr. Fabián Lozano (Puebla de la Calzada) in Odriozola (1976) . The last strain (named today Guadyerbas ) and the Torbiscal line are still kept in a conservation programme. The complete pedigree of Torbiscal from 1944 to 2013 was analysed in this study. It consists of 4,077 entries (individual-sire-dam) including 1,411 ancestor reproducers: 312 Ervideira (RE), 321 Caldeira (RC), 237 Campanario (BHC), 343 Puebla (BHP) and 198 admixed ones.

A complete diallel cross was performed among the four quoted strains in order to test their reproductive and productive performance and the possible heterotic effects among them ( García-Casco et al. , 2012 ). Here we analysed the piglet weight at 50 days of age using the available records from 9,052 piglets born in 1,571 litters representing the 16 genetic types resulting from the 4 × 4 possible crosses. The distribution of these piglets per cross and the number of sires and dams of each strain are shown in Table 1.

A founder is defined, for pedigree analysis purpose, as an animal with no relationship to any member of the pedigree except its offspring. The number of discrete equivalent generations traced ( EqG _i ), which is the expected number of generations from the base population if generation proceeded discretely, and the coefficients of inbreeding ( F _i ) and coancestry ( f _jk ) were computed for each ith individual and for their parents j and k tracing the pedigree back to the founder animals ( Woolliams & Mäntysaari, 1995 ; Caballero, 1995 ; Caballero & Toro, 2000 ). These parameters are closely linked to the classical concept of genetic contributions ( James & MacBride, 1958 ), which also sustains other genealogical parameters calculated in this study, such as the effective number of founders ( Nef ) and non-founders ( Nenf ) and the founder genome equivalents ( Nge ), related among them by the expression 1/ Nge = 1/ Nef + 1/ Nenf ( Toro et al. , 2000 ). For some specific calculations, the entire number of traced generations ( t ) was determined for each individual as the rounded number of discrete equivalent generations. The realized effective population size was estimated from the increase coancestry over generations by measuring IBD (identical by descent) probabilities, through the well-known formula: Ne = 1/2Δ IBD , where Δ IBD is the rate of IBD . Δ IBD values were calculated from individual coancestry rates by the expression ( Cervantes et al. , 2011 ) and averaged each generation as (1/2  Δ f _jk ) , being the effective census To take into account the possible bottlenecks, the effective number of ancestors ( fa ) or minimum number of ancestors necessary to explain the genetic diversity of a population was also calculated by each generation according to Boichard et al. (1997) .

In order to examine the relationship between the Torbiscal composite line and their founder strains, the expected genetic contribution of each founder to any Torbiscal reproducer was measured by means of coancestry coefficients between animals and founders ( James, 1972 ). The proportional genetic contribution of each founder to a specific generation is where a _jk is the additive relationship coefficient between founder j and animal k , being N the number of reproducers in the generation. The extension to the calculus of contributions of each one of the founder strains ( S ) is immediate by means of referred to the founders belonging to each one.

Moreover, to inspect whether the alleles contri­buting to inbreeding of Torbiscal animals came from specific founders, the inbreeding coefficient ( F _i ) can be partitioned into components due to each one of the 116 founders animals: , that measures, for any individual, the probability of being homozygous for a gene coming from each founder. These coefficients, calculated by using a modification of the additive matrix method for calculating inbreeding coefficients ( Lacy et al. , 1996 ; Rodri­gañez et al. , 1998 ), have been grouped into four components ascribable to each one of the founder strain , sums of coefficients with respect to founder individuals belonging to each founder population. The sum of these four partial coefficients is equal to the overall inbreeding.

The animal model used in the analysis of weight records can be represented in matrix notation as

A total number of 61 Torbiscal pigs, 75 Iberian pigs of diverse Portuguese and Spanish origins, 52 Duroc from Spain, USA and several European countries, 65 European Wild boars from Spain, Tunisia, Poland and other countries, and 52 pigs of different Asian breeds (Meishan, Jiangquai, Jinhua and Xiang) were genotyped using the Porcine SNP60 BeadChip (Illumina) accor­ding to the manufacturer's recommendations. Data quality control was performed according to the follo­wing filtering criteria: i ) call rate of the sample > 0.96, ii ) single nucleotide polymorphisms (SNPs) with a call rate > 0.99, iii ) GenTrain score > 0.70, iv ) mean of the normalized r-values for the AB genotypes > 0.35. A SNP was removed if: v ) number of inheritance errors > 5, vi ) unknown position on the genome or mapped on chromosomes X or Y, vii ) minor allele frequency MAF = 0. After all these editing steps only 43,693 SNPs were retained in the data set used for checking the possible ad­mix­ture of Iberian with other porcine populations and for assessing their genetic diversity and differentiation.

A Bayesian clustering method in STRUCTURE software ( Pritchard et al. , 2000 ) was employed, using the quoted genotypes, to assign individuals to one of the K-clusters representing ancestral populations, or jointly to two or more populations if their genotypes indicated that they were admixed. The number K was previously determined by the method of Evanno et al. (2005) using a partial SNP datafile (8,738 genotypes) and a range of possible values from two to five. A warm-up of 20,000 iterations followed by 30,000 preserved samples was obtained in all the performed calculations.

We measured the genetic diversity according to Nei (1973) as the heterozygosity expected under the Hardy-Weinberg equilibrium conditions. Caballero & Toro (2002) showed as, in a metapopulation with n breeds, the total genetic diversity or expected heterozygosity ( GD _T = H _T ), may be partitioned into a within breeds component ( GD _WS = H _S ) and another between breeds ( GD _BS = H _T - H _S ):

We computed the contribution of founder and non-founder reproducers to the pairwise coancestry between individuals of each generation t (rounded value of EqG ) of Ervideira , Caldeira , Campanario and Puebla besides of the coefficients of coancestry ( f ) and inbreeding ( F ). The respective averaged coancestry coefficients increased from 0.055, 0.062, 0.054 and 0.047 in the first generation of the respective strain up to 0.199, 0.222, 0.164 and 0.186 in the last generation. The respective contributions to these values of founders and non-founders were 0.032 and 0.167 (RE), 0.061 and 0.161 (RC), 0.050 and 0.114 (BHC), and 0.048 and 0.138 (BHP). Note that the Campanario complete pedigree only consists of five equivalent generations instead of seven in the other strains.

The parameters N _ef , N _enf , and N _ge constitute an alternative to represent the information concerning genetic contributions to coancestry. The change of these parameters over successive generations of each founder strain is shown in Figs. 1a, b, c, d. The first parameter is related to the genetic contributions of founder reproducers to each generation which stabilize with time. Their averaged values along the stable generations were, respectively, 15.6, 8.2, 10.0 and 10.2. In a regular system and with random mating, N _ef equals half the asymptotic effective population size N _e ( Caballero & Toro, 2000 ), providing respective N _e estimates equal to 31.2, 16.4, 20.0 and 20.4. Because of the accomplishment of non-random mattings avoiding coancestry, these N _e values exceed those of effective census calculated from individual coancestry rates ( N _eΔf ), represented in Fig. 1. The N _enf parameter de­pends on the contributions of non-founders, and it reflects the accumulated effects of genetic drift. Therefore it decreased in each strain from 18.8, 22.6, 76.0 and 90.0 to final values 3.0, 3.1, 4.4 and 3.6, respectively. The founder genome equivalents ( N _ge ) summarize both processes, and their values decreased along each pedigree from 9.1 to 2.5 (RE), from 8.0 to 2.3 (RC), from 9.3 to 3.0 (BHC) and from 10.7 to 2.7 (BHP). The effective number of ancestors ( fa ) is another pedigree-based metrics of genetic variability unrelated to the previous ones. Their values slightly changed between the first and last generations from 12.7 to 7.7 (RE), 10.5 to 11.7 (RC), 10.6 to 6.4 (BHC) and 12.3 to 8.4 (BHP), showing the lack of bottlenecks in these strains.

One particular aspect of the productive differences among strains was examined by the genetic analysis of piglet weight al 50 days using the data file described in Table 1. Mean and standard deviation of this trait were 11.31 and 2.54 kg. The main results obtained from the analysis are summarized in Fig. 2 and Table 2. The relationship between the number of piglets weaned per litter, including cross-fostered, and piglet weight at 50 days is represented in Fig. 2, as deviations from litters of one weaned piglet, showing a pattern with an optimum number of weaned piglets from three to seven and negative deviations for the greatest litters. Given the late weaning of litters at 50 days of age, maternal heritability ( h ² _M ) of this trait showed a greater magnitude than individual heritability ( h ² ), with a negative genetic correlation (ρ _G ) between maternal and direct genetic effects. The proportion of phenotypic variance ( c ² ) explained by the common litter environmental effect was also important. Parity effects expressed as deviations from the first parity showed a remarkable value up to the fifth parity. Significant weight differences between males and females were also inferred. Individual and maternal strain effects were inferred as differences respect to Ervideira, although other contrasts have been carried out from the respective marginal posterior distributions. The only relevant strain difference among individual effects was found between Puebla and Caldeira . The other differences were lower, including the zero value their respective 95% highest posterior density intervals (95% HPD). Neither effects different from zero were found among the four maternal strains. However, the available samples of marginal posterior distributions of all the parameters of interest allow calculating additional contrasts, as between the sum of maternal and individual strain effects [( l _i + m _i ) - ( l _j + m _j )]. Measured by this procedure, the mean differences of piglet weight between Puebla and the other strains were 1.25 ( Ervideira ), 1.87 ( Caldeira ) and 1.21 ( Campanario ), with values of the posterior probability to be greater than zero PProb > 0 equal to 0.980, 0.999 and 0.973, respectively. Four out of the six specific heterotic effects on piglet weight showed significant values ranged from 0.31 to 0.86 kg, equivalent to percentages of the mean trait from 2.7 to 7.8%. However, null heterotic effects were detected between pairs Ervideira / Campanario (PM = 0.054, PProb > 0 = 0.583) and Caldeira/Puebla (PM = 0.034, PProb > 0 = 0.552). Inferred reciprocal strain effects were not different from zero.

Crosses between strains for founding the Torbiscal line were carried out from generation seven to nine. The following seven types of crosses were initially implemented: two between two strains [RE × BHP and BHC × BHP], and five between three strains [RE × (RE × BHP), RC × (RE × BHP), BHP × (RE × BHP), RE × (BHC × BHP) and BHP × (BHC × BHP)]. Other intercrosses were performed later to obtain red hairy coated pigs with contributions of the four ancestral strains, which were identified as the first individuals of the new composite line. Only 71 out of the 116 founders actually contributed to the build-up of Torbiscal .

The proportional genetic contribution of each strain to the successive generations of the Torbiscal line is represented in the Fig. 3. Note that this figure starts at generation nine which is the first one where some born pigs were nominated Torbiscal . The values of the expected contributions of Ervideira, Caldeira, Campanario and Puebla at this generation were 0.22, 0.26, 0.24 and 0.28, respectively, and the effective number of ancestors ( f a) equal to 20.28 ( Boichard et al. , 1997 ). The respective contributions to the last generation were 0.27, 0.23, 0.15 and 0.35, being the value of f a = 14.34.

The changes of average inbreeding coefficient over successive generations of Torbiscal line and compo­nents inbreeding ascribable to founder strains are ex­po­sed in the Fig. 4. The mean Torbiscal inbreeding by generation increased from F ₉ =0.041 to F ₂₈ =0.180 while the four partial components showed an analogous increase from F ₉ =0.007, 0.011, 0.004 and 0.019 to F ₂₈ =0.049, 0.043, 0.023 and 0.066.

The application of Evanno et al. (2005) method to a preliminary clustering analysis of 305 individuals representing five pig populations, determined K = 2 as the number of ancestral groups on the superior hierarchical level of the data structure. This number coincides with the two well-known main focus of pig domestication that took place ten thousand years ago in East Europe and Asia from different wild ancestors diverged 1 M years ago ( Larson et al. , 2007 ). For this reason, clusters were named as European and Asian. The final results of the application of the STRUCTURE algorithm using the whole number of SNP genotypes are presented in Fig. 5. Torbiscal , other Iberian pigs and European Wild pigs were ascribed to a common ancestral group (European) with inferred mean proportions 1.000, 0.979 and 0.945, respectively. A similar proportion of the genome of Asian pigs (0.985) was ascribed to the other ancestral group (Asian), while the Duroc pigs showed remarkable admixing of both ancestors, with respective mean proportions of 0.671 and 0.329.

The mean expected heterozygosities (and 95% CI) of Torbiscal , other Iberian, European Wild pigs, Duroc and Asian pigs were 0.194 (0.193/0.196), 0.248 (0.246/0.250), 0.270 (0.268/0.272), 0.319 (0.317/0.320) and 0.218 (0.216/0.219), respectively. A more detailed partition of the genetic diversity for each i population is stated in Table 3. The values of individual and population self-coancestry ( s _i , f _ii ), average distance between individuals ( D _ii ), inbreeding ( F _i ), and related parameters ( Fabuel et al. , 2004 ), as the Hardy-Weinberg deviations ( α _i ) and proportion of diversity between individuals ( G _i ) point out clear differences in the partition of genetic diversity among the Torbiscal closed line and the other pig populations.

The total heterozygosity of the whole set of five populations is GD _T = H _T = 0.364, with components within breeds GD _WS = H _S = 0.253 and between breeds GD _BS = 0.111. The first one also has two components, the genetic diversity within individuals GD _WI = 0.110 and between individuals GD _BI = 0.143. The GD _BS component is equivalent to the Nei's minimum distance between populations ( D ). The amount of differentiation is F _ST = 0.304, the maximum possible being F _ST(max) = 0.611, so that the standardized value is in this case F ' _ST = 0.476. The mean and 95% CI of pairwise values of F ' _ST among the five populations are reported in Table 4. According to these values, the Torbiscal line shows a small differentiation from the groups of other Iberian pigs and European wild boars, being three and six times greater its divergence from admixed Duroc and Asian pigs, respectively.

A more detailed picture of the divergence between the genomes of Torbiscal and the analysed pool of Iberian pigs of other origins is presented in the Manhattan plot of Fig. 6. The numbers on Y-axis correspond to d _i values between Torbiscal and other Iberian pigs averaged for sliding overlapping windows of ten successive SNPs. Differentiated 10 SNP windows are identified as exceeding the 99.5th percentile their whole distribution, represented as an horizontal line. Only four chromosomes (SSC 9, 10, 11 and 16) do not show significant differentiation. A total of 33 divergent regions are identified in the remaining 14 chromosomes, being SSC6 the chromosome with a greater number (seven) of differentiated regions with a total length of 4.3 Mb. The maximum differentiation is found on SSC3 (95.45-96.50 Mb) and SSC7 (101.96-102.54 Mb). The examination of gene content using the more recent assembly of pig genome ( Sscrofa 11.1) allowed the detection of 94 protein-coding genes inside 33 divergent regions. Seven of these genes code proteins related to muscle growth, eight genes to lipid metabolism and only one ( JAG1 ) involved in hair differentiation. Their symbols and chromosome positions are detailed in Table 5.