Long-Term Impact of Conventional and Optimal Contribution Conservation Methods on Genetic Diversity and Genetic Gain in Chinese Indigenous Pig Breeds

China has rich and vast genetic resources of indigenous pig breeds. Currently, great attention is paid to either crossbreeding or conservation of these indigenous pig breeds, and insufficient attention is paid to the combination of conservation and breeding along with their long-term effects on genetic diversity. The genetic diversity of livestock is essential to increase productivity and respond to future challenges such as climate change. The genetic stability and product consistency of these indigenous pig breeds should be focused on and further improved. Therefore, the objective of this study is to compare the long-term effects of using conventional conservation and optimal contribution selection methods on genetic gain and genetic diversity. A total of 11 different methods including conventional conservation and optimal contribution selection methods were investigated using stochastic simulations with a population size of 600 animals in each generation. Each scenario was run for 20 generations and 100 replicates. The long-term effects of using these methods were evaluated in terms of rate of genetic gain, rate of true inbreeding based on genome-wide identity-by-descient (IBD) markers and various genetic diversity metrices such as expected heterozygosity (He). The results indicated that the rates of true inbreeding in these conventional conservation methods were maintained at around 0.01. The optimal contribution selection methods based either on the pedigree (POCS) or genome (GOCS) information showed more genetic gain than conventional methods, and POCS achieved the largest gentic gain. Furthermore, the effect of using GOCS methods on most of the genetic diversity metrics was slightly better than the conventional conservation methods when the the rate of true inbreeding was the same, but this also required more sires used in OCS methods. According to the rate of true inbreeding, there was no significant difference among these conventional methods.

The genetic diversity of livestock is essential to improve productivity and respond to 2 challenges including food security and climate change mitigation in the future [1]. 3 However, due to agricultural innovation since the beginning of the 19th century and 4 subsequent intensification of production, many local varieties can not adapt to the 5 resulting changes. Pigs are one of the most common domestic animals, and more than 6 one-third of indigenous pig breeds in the world are in China. These indigenous pig 7 breeds generally have high fertility, good meat quality and high tolerance to harsh 8 environmental conditions [2]. However, pig inductries currently pay more attention to 9 crossbreeding for indigenous pig breeds, and insufficient attention to the combination 10 of conservation and genetic improvement. To improve the production and economic 11 value of breeding stocks, indigenous pig breeds usually cross with the foreign breeds 12 which have high production performance. Thus, gene flow usually only occurs from 13 breeds with superior economic characteristics to indigenous pig breeds [3]. 14 Furthermore, due to the epidemic of diseases such as African swine fever, a lot of 15 precious indigenous breeds are on the verge of extinction. The herds of indigenous pig 16 breeds have reduced greatly during this period. To protect these indigenous pig breeds 17 in China, the Chinese government has established national-level breeding farms for 18 most indigenous pig breeds, and a large number of breeding funds are used to protect 19 these unique indigenous pig breeds every year [4]. 20 In these national-level conservation farms, our goal is to maintain each breed's genetic 21 materials and control the rate of inbreeding as much as possible. Inbreeding is an 22 important reason for the loss of genetic variation, and the rate of inbreeding mainly 23 depends on the effective population size [5]. In order to reduce the impact of inbreeding 24 on the loss of population genetic variation, the increase in the inbreeding coefficient of 25 each generation in the conservation population is recommended to be controlled at 1-26 4%, and the conservation population needs to have at least 12-25 sires and 100-250 27 dams [6]. These guides are commonly used in conservation farms. Simultaneously, to 28 keep the genetic diversity and control rate of inbreeding, conservation farms attempts 29 to keep the same number of offspring for each family. However, current conservation 30 methods do not combine conservation of genetic resources and genetic improvement of 31 production performance. In conservation populations, it is also very important to 32 properly select the dominant traits of each local pig breed, which will help to further 33 consolidate the advantages of the breed and maintain the uniqueness of each breed. 34 From a long-term sustainable perspective, how to combine conservation and selective 35 breeding in the conservation field is a crucial issue. Great attention should be paid to 36 genetic improvement of important economic traits while maintaining the overall genetic 37 diversity of locsal breeds to meet pig industry's sustainable development and even other 38 livestock industries. Therefore, we need to re-examine our current conservation strategy. 39 The new conservation strategy should take into account at least two principles at the 40 same time. One of the principles is to keep genetic diversity as high as possible, and the 41 other is to obtain genetic progress of some essential economic traits as large as possible. 42 Optimal contribution selection (OCS) is an effective selection method that balances 43 inbreeding and genetic gain [7,8]. It maximises rates of predicted genetic gain while 44 controlling inbreeding at given rates by optimizing the genetic contribution of each 45 selection candidate to the next generation [9][10][11]. Optimal contribution selection can be 46 based either on pedigree information (POCS) or genome information (GOCS). Several 47 previous studies have investigated the impact of OCS on rate of inbreeding and long-48 term genetic gain based on simulated data [12,13] and real data [9,14]. Gourdine et al. 49 [13] claimed that the genetic gain with optimal contribution selection could be similar 50 to truncation selection, but the inbreeding was lower. Sánchez-Molano et al. [15] 51 showed that genome-based optimal contribution strategies could effectively control 52 inbreeding even when selected traits, adaptive traits and production traits, are 53 negatively correlated using simulated data. Henryon et al. [16] reported that the optimal 54 contribution selection based on pedigree information for controlling inbreeding could 55 achieve more genetic gain than that based on genome information due to less restriction 56 on the change of QTL allele frequencies. Nowadays, there are many ways to measure 57 changes in genetic variation and its diversity. We can calculate the inbreeding 58 coefficient if the pedigree information is known. However, in actual situations, the 59 registration of pedigree is often incomplete and inaccurate, limiting the usage of this 60 method. With the use of molecular markers, more and more genetic diversity indicators 61 are used to assess the degree of diversity of a particular population [17][18][19][20]. 62 The objective of this study is to compare the long-term effects on genetic diversity and 63 genetic gain of using conventional conservation and OCS methods. We acheive this by 64 using a stochastic simulation study approach, where 11 different conservation methods 65 for a small pig population were compared. The results of this simulation study are 66 expected to provide guidance to breeders and government departments on formulating 67 better conservation programs. 68

69
Experimental design 70 We used stochastic simulation to estimate the long-term genetic gain and genetic 71 diversity using different conservation methods. For OCS methods, either pedigree (POCS) or two genomic (GOCS) relationship 89 matrices were used to constrain the rate of inbreeding. We also simulated truncation 90 selection and random selection as reference methods. In total, there were 11 selection 91 methods studied. Each selection method was run for 20 discrete generations, and the 92 animals were selected based on a single trait controlled by 360 quantitative trait loci 93 (QTL). The heritability of the trait was set to 0.2. Furthermore, 36,000 markers were 94 simulated to carry out GOCS. For the methods other than OCS, 12 sires were selected, 95 and each sire was mated to 10 dams in each generation. For OCS methods, the males 96 were allocated 0, 1, 2 … or 120 matings by the program, and females were allocated a 97 single mating in each generation. Each dam produced five offspring with an equal sex 98 ratio. The animals were phenotyped before selection. 99

100
POCS allocated matings of selection candidates in generations = 1 … 20 according 101 to EBV and pedigree relationships between all the involved animals. It was done by 102 maximizing, , with respect to [21] : Where is an dimentional vector of genetic contributions, where is the number 105 of selected candidates, is vector of EBV, and is a × matrix for selected 106 candidates which is a submatrix from the full additive genetic relationship matrix for 107 all animals in the pegigree. In this study, pedigree of the selected candidates was traced 108 back to the base population [22]. Elements of were constrained to 0 ≤ ≤ 109 0.5 ( = 1 … ) and the sum of contributions were 0.5 for each sex.The component 110 is the expected breeding value, and the component is the expected average 111 relationship of the proposed offspring. The penalty, , is applied to the expected 112 average relationship of the next generation, which was constant across generations. 113 GOCS was performed by replacing with a × genomic relationship matrix ( ) 114 which was calculated with genotypes for all markers of all the selected candidates using 115 the method described by Yang et al. [23]. A range of (1, 5, 10 …100) was applied 116 to examine the pattern of genetic gain with different inbreeding rates. 117 We used also an additional method to build a matrix for GOCS, where the markers 118 with a distance from a random QTL less than 1cM were excluded from the (we call 119 the corresponding GOCS the GOCS-1cM). To differentiate this GOCS method from 120 the conventional GOCS method, we called the classical GOCS method the GOCS-0cM. 121 We used EVA [24] to perform POCS and GOCS. 122

123
Simulations of each conservation method were carried out in three stages: 1) a single 124 founder population was generated as the basis for the subsequent stages, 2) a unique 125 base population was sampled from the last generation of founder population, and 3) a 126 selected population was generated based on the base population. Stage 2 and 3 were 127 run for 100 times to produce 100 replicates. To simplify the simulation, instead of direct 128 calculation of EBV, the EBV was approximated by the breeding values of a genetically 129 correlated pseudo-trait [25]. The genetic correlation was set to 0.6, mimicking a 130 genomic selection with an accuracy of 0.6 [26]. 131

Founder population and genetic architecture: Generations -2000 to -1 132
The linkage disequilibrium of QTLs and the markers was generated by simulating a 133 founder population with QMSim [27] using a Fisher-Wright inheritance model. The 134 population had an effective population size of 200 animals (100 males and 100 females) 135 and 2,000 discrete generations. The simulated genome consisted of eighteen 1 Morgan 136 long chromosomes, on which 10,000 loci were equidistantly distributed, resulting in 137 180,000 loci in total across the genome. The recurrent mutation was allowed at a rate 138 of 2.5 × 10 and recombination per chromosome was sampled from a Poisson 139 distribution with a mean of 1. 140 At the last generation of the founder population (generation -1), among all segregating 141 loci, every second locus with a minor allele frequency (MAF)>0.05 were used as 142 potential markers. In total, we selected 36,000 markers from these potential marker loci. 143 In total, 360 QTLs were selected from the remaining segregating loci with MAF > 0.01. 144 The QTL allelic effects were assumed to follow a gamma distribution with a shape 145 parameter of 1.48, which was derived from distributing QTL effects in pig breeds [28]. 146

Base population: Generation 0 147
In generation 0, 200 animals were generated by random mating of 100 males and 100 148 females in generation -1. From these 200 animals, 12 males and 120 females were 149 randomly selected as base animals to produce 600 offspring with an equal sex ratio. 150

Selected population: Generations 1 to 20 151
In each of generations 1 to 20, 120 matings were allocated to sires and dams, and each 152 dam was allocated a single mating to produce five offspring with an equal sex ratio. 153 The offspring in each generation inherited alleles of markers and QTLs from their 154 parents, following Mendel's laws of heredity allowing for recombinations following a 155 Poisson diostribution with a mean of 1.

Tracing identity-by-descent 168
To compute the rate of true inbreeding, 2,000 identical-by-descent (IBD) loci were 169 equidistantly placed on each chromosome of animals in the base populations. Unique 170 alleles were assigned to these IBD loci in the base population to trace each base 171 animal's contribution to their descendants [30]. A descendant was IBD at an IBD locus 172 when it inherited two copies of a unique allele. These IBD loci were not used for 173 prediction or selection. 174

Rate of genetic gain and inbreeding 176
The rates of genetic gain and the rates of true inbreeding are presented as means (± ) 177 of the 100 replicates.

201
The rate of genetic gain and rate of true inbreeding 202 We presented long-term response frontiers by plotting the rate of genetic gain 203 against the rate of inbreeding with all possible solutions by applying different penalties 204 for POCS and GOCS. As shown in Fig.1, the rate of true inbreeding for all the 205 conventional methods was around 0.01, except for the truncation selection scenario, 206 consistent with basic conservation theory. Most importantly, inbreeding increment of 207 each generation in the random scenario was also around 0.01, which indicates that the 208 rate of true inbreeding could also be controlled around 0.01 as long as we maintain an 209 approperiate population size (such as12 males and 120 females in this study) and 210 guarantee complete random mating. Among these conventional conservation methods, 211 when the inbreeding increment was 0.01, the scenarios of Sirehalf-Damtrunc, Sirehalf-212 Damhalf, and Sirehalf-Damfull obtained higher genetic gains. However, the genetic 213 gain was much smaller when dam was randomly selected in three various form (i.e., 214 Sirehalf-DamfullRandom, Sirehalf-DamhalfRandom and Sirehalf-DamRandom). As 215 expected, no genetic gain was obtained when both sire and dam were randomly selected. 216 Compared with the six convervational methods, truncation selection on both sire and 217 dam increased genetic gain by 7.5% (vs. Sirehalf-Damtrunc) to 67.5% (vs. Sirehalf-218 DamfullRandam), but it tripled the rate of inbreeding. 219 All the OCS methods based on the genome and pedigree information realized more 220 genetic gain than the conventional conservation methods when the inbreeding rate was 221 almost around 0.01 ( Fig. 1 and Table 1, p=10). Interestinly, there was no significant 222 difference in both the rate of true inbreeding and genetic gain between these two GOCS 223 methods (GOCS-0cM and GOCS-1cM). However, POCS could achieve more genetic 224 gain than GOCS when the inbreeding rate was the same. POCS with penalty p=10 225 obtained the rate of genetic gain as high as the Truncation scenario, but the rate of 226 inbreeding was only one third. The rate of inbreeding of two GOCS methods and POCS 227 were similar to that of the Sirehalf-Damtrunc scenario when the penalty p =7 ,which 228 used the same number of sires ( Fig. 6 and Additional file 3), but they would be similar 229 to that of the four methods (Sirehalf-Damfull, Sirehalf-DamfullRandom, Sirehalf-230 Damhalf and Sirehalf-DamhalfRandom) when the penalty p=10. 231

232
In terms of He and Ho, as shown in Fig. 2a and 2b, 3a and 3b, GOCS methods were 233 better than the conventional conservation method when the penalty p was increased to 234 10. As the number of generations increases, the declined slope of the He and Ho was 235 smaller in GOCS methods compared to the conventional conservation methods, which 236 indicated that GOCS had better effect than that of the conventional conservation 237 methods. However, POCS was not superior to the conventional conservation methods 238 and was only better than the truncation scenario, in terms of He and Ho. Furthermore, 239 there was no significant difference in He and Ho among the conventional conservation 240 methods. 241 Fig. 2c, 3c and Table 2 showed that there were more effective alleles in two GOCS 242 scenarios when the weight p was 10. Regardless of p=7 or p=10, the POCS led to low 243 Ae (see Fig. 3c, Table 2 and Additional file 2), only higher than the truncation scenario. 244 For M01 and M05 (Fig. 2d, 3d and Additional file 1), several conventional conservation 245 methods, such as Sirehalf-Damfull, Sirehalf-DamfullRandom, Sirehalf-Damhalf, and 246 Sirehalf-DamhalfRandom, were better than those of OCS methods. 247 Changes of additive genetic variances across generations are presented in Fig. 4. 248 From Fig. 4, the additive genetic variance in optimal contribution selection methods 249 had the fastest decline, compared with other scenarios except for the method of 250 truncation selection on both sire and dam. The variance in the POCS method was lower 251 than that in GOCS methods. The four conventional scenarios with different types of 252 random selection had the highest additive genetic variance, and the order was Random, 253 Sirehalf-DamfullRandom, Sirehalf-DamhalfRandom, and Sirehalf-DamRandom. The 254 trends of additive variance and inbreeding were generally inversely consistent. 255

256
As for the number of ancestors for different methods (Fig. 5), the pattern of the 257 number of ancestors in the Sirehalf-Damfull and Sirehalf-DamfullRandom methods 258 were different from the other methods. These two methods remained the same number 259 of ancestors in the first generations. The trends began to decline after the fifth 260 generation, indicating that some ancestors failed to make contributions as selection 261 proceeds due to selection and genetic drift. For other methods, the number of ancestors 262 declined rapidly in the previous generations, and then gruadually fallened out. 263 Therefore, keeping the same number of offspring from each sire and dam family will 264 have the best effect in the first few generations. In addition, Sirehalf-Damfull and 265 Sirehalf-DamfullRandom methods also retained the largest number of ancestors in the 266 last few generations. The second largest number of ancestors was observed in the OCS 267 scenarios, including GOCS-0cM, and POCS. 268 The number of sires was around 12,which is the same as in conservational methods, 269 when the weight p was 7 in two GOCS scenarios ( Fig. 6 and

279
There are huge indigenous pig breed resources in China, and these indigenous pig 280 breeds have formed relatively unique characteristics under long-term environmental 281 and artificial pressure. Nowadays, the focus of conventional conservation methods is 282 only to control rate of inbreeding, and not much attention is paid to the selection of 283 favorable traits for each breed in the existing conservation field. It will be helpful to 284 further improve the advantage and uniqueness of each breed if we can combine the 285 maintenance of genetic diversity and the selection of favorable traits. In this study, we 286 studied conventional conservation and optimal contribution scenarios to conserve 287 indigenous pig breeds with small population sizes using simulation studies. We 288 explored the genetic diversity changes and genetic gain of these conservation scenarios 289 during 20 generations. The founding is helpful in guiding the current conservation 290 programs. 291 To utilise indigenous pig breeds for pig production, genetic improvements for some 292 important economic traits in conservative pig populations is necessary. In the current 293 study the genetic gain obtained by the optimal contribution selection methods show a 294 trend of increasing with increasing weight p when p was small, and then decreases as 295 the weight p increases. This may be because the increase in selection intensity with 296 small p accelerates the reduction of genetic variation within the population, thereby 297 reducing the further improvement of genetic gain. This implies that selection without a 298 restraint on inbreeding will lead to the selection limit [34]. Long-term high-intensity 299 selection will reduce the population's genetic variation, and the reduction of genetic 300 variation will counteract the increase in genetic gain. In addition, from the changes in 301 genetic diversity of different scenarios, we could see that the penalty on genetic 302 relationship should at least 10 if we want to apply GOCS methods for maintaining a 303 higher heterozygosity than the conventional conservation scenarios. 304 As expected, the truncation method caused largest rate of inbreeding (Fig. 1, Table  305 1), and the trend became very significant as the number of generations increased. It 306 indicates that we should not use this method to conserve indigenous pig breeds when 307 the population size is small among conservation farms, which is different from selecting 308 and breeding in the breeds for commercial production. Using conventional conservation 309 methods, the trends of rate of inbreeding were amost the same in the senarios with 310 different methods of selection on dam, except for trucation selection. This indicates that 311 selection of dam within full-sib or half-sib family or random selection of dam from the 312 whole population could all be used in actual conservation operation. genetic diversity of main domestic animals and indicated that the ratio of polymorphic 320 loci and the average expected heterozygosity were the primary parameters to measure 321 genetic diversity. Qian et al. [37] reported that the degree of expected heterozygosity 322 was more effivtive than the ratio of polymorphic loci in accuracy of measuring genetic 323 diversity. The variation of the number of polymorphic loci is relatively small, and the 324 sensitivity to genetic diversity is relatively low [37]. In addition, the number of effective 325 alleles could more effectively measure the change of genetic diversity in one population 326 [38]. Therefore, in this study, we used multiple indicators to measure the impact of 327 different conservation methods on genetic diversity changes after a number of 328 generations to make our results more comprehensive and objective to a certain extent. 329 These indicators could complement each other. For each indicator of genetic diversity, 330 the results of GOCS-0cM and GOCS-1cM were similar. As the results shown in Fig. 4, 331 the additive genetic variance in OCS methods had the fastest decline, which indicates 332 the OCS methods results in larger increases in the frequencies of favorable alleles at 333 QTL, compared with the other methods. Moreover, POCS is larger than GOCS, which 334 is also consistent with the previous study [39]. The results of the number of 335 polymorphic loci such as M01 and M05 (Fig. 3d, Additional file 1) also illustrate this 336 point. When the penalty is 7, the optimal contribution selection methods (POCS and 337 GOCS) were not better than several conventional conservation methods such as 338 Sirehalf  DamhalfRandom. This may be also due to the selection of QTLs that affect the traits, 340 and the directional selection decreases polymorphic loci ratio. 341 From OCS methods, we can see that the genetic gain obtained by the optimal 342 contribution selection method based on the pedigree relationship is higher than that 343 obtained based on genomic relationship when the rate of inbreeding in each generation 344 is controlled at about 0.01. OCS automatically determines the number of male animals 345 required to control inbreeding when we use OCS methods. Through comparison (Fig. 346 6 and Additional file 3), it is found that POCS method requires more sires than GOCS 347 when the rate of inbreeding is controlled in the same level. In actual pig production, 348 POCS method is often easier to put into practice. POCS method is based on pedigree 349 information, and it only requires that the pedigree of each animal is registed in the 350 conservation farms. Unlike POCS, GOCS method is based on genomic information, 351 which requires individuals' genotype data. animals because of an overuse of elite males. Therefore, the priority is to minimize 364 inbreeding in these animal populations. In addition, the focus would be changed to 365 increase conservation values in endangered breeds that get allowance for better 366 conservation. This could be realized by increasing their genetic distance or recovering 367 the native genetic background among these breeds. These goals conflict with each other 368 to a certain extent. In order to maximize genetic gain, people would prefer to choose 369 the animals with the highest breeding values for economic traits, which will increase 370 rate of inbreeding, and may lead to inbreeding depression and new bottlenecks. 371 Generally, commercial breeds often have the highest breeding values for economic 372 traits, which would further lead to the loss of the genetic diversity of native breeds. 373 It is important to protect and conserve the indigenious pig breeds, especially when 374 their population size has dropped sharply. However, it is not sensilbe and conducive if 375 we only focus on protecting but not improving favorable traits. The current study shows 376 that the optimal contribution selection method based on genomic information can 377 maintain a high genetic diversity while improving the traits we want to improve, which 378 is in line with our current needs. 379

380
In conclusion, our study showed conventional conservation scenarios resulted in the 381 rate of inbreeding for each generation was at around 0.01. Different methods to select 382 sow has small impact on inbreeding when we use conventional conservation methods. 383  The trends of genetic diversity metrics for different methods across 20 generations Figure 3 The boxplots of genetic diversity metrics for different methods in 20th generations Figure 4 The trends of additive variance (varAdd) for different methods across 20 generations Figure 5 The number of ancestor trends for different methods across 20 generations Figure 6 The number of sires used in OCS methods with penalty (1 to10) for 20 generations (a, GOCS-0Cm; b, GOCS-1cM; c, POCS)