60 Acta argiculturae Slovenica, Supplement 5, 60–65, Ljubljana 2016 24th Int. Symp. “Animal Science Days”, Ptuj, Slovenia, Sept. 21st−23rd, 2016. COBISS: 1.08 Agris category code: L10 ESTIMATION OF INBREEDING IN SLOVENIAN BROWN-SWISS POPULATION Jana OBŠTETER 1, Betka LOGAR 2 Estimation of inbreeding in Slovenian Brown-Swiss population 1 Agricultural Institute of Slovenia, Hacquetova ulica 17, 1000 Ljubljana, Slovenia, e-mail: jana.obsteter@kis.si 2 Same address as 1, e-mail: betka.logar@kis.si ABSTRACT Breeding programs require control of the level (F) and rate (ΔF) of inbreeding in order to avoid inbreeding depres- sion. Increasing availability of genomic information has enabled a more accurate estimation of F and ΔF. This study aimed to investigate classical (Fped) and genomic inbreeding coefficients (FROH) in 214 genotyped Slovenian Brown- Swiss animals. Fped was obtained from pedigree analysis using PEDIG and FROH was estimated based on runs of ho- mozygosity (ROHs) analysis using PLINK. The results show that year-averaged FROH exceeds year-averaged Fped for all the studied years. ΔFROH>1MB (0.00918) was ~3 times higher than ΔFped (0.00334) and was close to the suggested limit that still allows sustainable population management. While detected ROHs reveal more ancient as well as recent inbreeding, the majority reflects inbreeding dating back ~12 to ~3 generations. When stratified by ROH lengths, FROH reveals some differences in most highly inbred chromosomes according to shorter and longer ROHs suggesting some changes in selection goals during breed’s history. Key words: cattle, breeds, Brown-Swiss, inbreeding, pedigree, genotypes, runs of homozygosity, Slovenia 1 INTRODUCTION Inbreeding is defined as the probability of the two alleles at a locus in an individual being identical by de- scend (IBD). Inbreeding coefficient is computed in respect to the base population which is assumed to be non-inbred and have inbreeding of zero (Falconer and Mackay, 1996). When analysing population’s pedigree this condition is usually difficult to satisfy since the pedi- gree records are not always complete and include a lim- ited definite number of generation. Especially with cattle, it is erroneous to assume that the first generation in the pedigree is unrelated since genetic material of the favour- able bulls is widely and abundantly distributed (van der Werf, 1999). Consequently, the results could be spuri- ous and inbreeding coefficient underestimated. Selection programs require a control of the level and the rate of inbreeding in the population, particularly in small popu- lations, in order to maintain genetic diversity and prevent inbreeding depression that could lead to the reduction in mean fitness and production (Falconer and Mackay, 1996). The increased use of genotyping arrays has allowed the development of new and more accurate methods to estimate inbreeding. Identification of long stretches of homozygous genotypes, i.e. runs of homozygosity (ROHs), enables detection of genome-wide autozygo- sity and IBD regions (McQuillan et al., 2008). There- fore, longer homozygous tracts might provide evidence for genomic regions undergoing selection. It has also been denoted that genomic selection requires genomic control of inbreeding since they are both based on the same level (Sonesson et al., 2012). Genomic selection for Brown-Swiss in Slovenia has started in 2009 by reason of participation in Interbull project InterGenomic (Santus, 2011). Since then 191 Slovenian bulls were added to the international reference population (Rigler et al., 2016). The first genomic breeding values were predicted in 2013 Acta agriculturae Slovenica, Supplement 5 – 2016 61 ESTIMATION OF INBREEDING IN SLOVENIAN BROWN-SWISS POPULATION (Potočnik et al., 2016). Now, all new breeding bulls are selected on the basis of genomic breeding values. 2 MATERIAL AND METHODS A total of 214 Slovenian Brown-Swiss animals, 115 males and 99 females, born between 2003 and 2016, were genotyped on six different genotyping chips (Table 1). Genotyped animals served as a reference population for the construction of the pedigree. Final pedigree included 2653 animals, 880 males and 1773 females. Pedigree re- cords were obtained from database in Cattle Information System (Logar et al., 2005) which caters for most infor- mation requirements in the cattle breeding scheme in Slovenia. The quality of the pedigree was assessed with complete generation equivalent computed as a sum of (1/2)n terms over all known individual’s ancestors, where n is the number of generations separating the individu- als from the ancestor (Maignel et al., 1996). Classical in- breeding coefficient (F) was computed based on pedigree information as defined by Meuwissen and Luo (1992) us- ing PEDIG (Boichard, 2002) software: 2 1 i ii jj j A L D = =∑ where Aii = i th diagonal element of the pedigree relation- ship matrix, L = lower triangel of A matrix, D = diagonal matrix. Generation intervals and pedigree completeness were determined using PopRep software (Groeneveld et al., 2009). Individuals genotyped on GeneSeek chips (n = 86) (Table 1) were imputed onto 50K Illumina chip utilis- ing FIMPUTE (Stachowicz et al., 2011) via ZANARDI (Nicolazzi and Marras, 2015) software. SNPs exclusive to each of the GeneSeek chips and sex chromosome SNPs were excluded prior to the imputation. The accuracy of imputation was assessed with 10x cross validation and allelic concordance levels were reported (Table 1). No genotype quality control was applied prior to SNP impu- tation since this was shown to be the best strategy (Ro- shyara et al., 2014). ROH analysis was performed on the 214 imputed genotypes. Genotype quality control was applied prior to the analysis with the following parameters: call rate per SNP > 90 %, MAF > 0.01 and deviation from Hardy- Weinberg equilibrium p > 0.0001, resulting in 214 geno- typed animals and 42,302 remaining SNPs. The analysis was carried out using PLINK 1.9 (Purcell et al., 2007) with adjusted parameters for the sliding window of 20 SNPs, allowing one heterozygous and two missing SNPs within the window, minimum SNP density one SNP every 120 kB and setting the minimum number of SNPs in a segment to be called a ROH to 15 and minimum length to 1 MB (adopted from Purfield et al., 2012 and Ferenčakovič et al., 2013a). Individual FROH was comput- ed as described in ∑∑= AUTOROHROH LLF where ΣLROH is the cumulative length of individual’s ROH and ΣLAUTO is the total length of the genome SNP cover- age (i.e. 2.51 GB). Pedigree (ΔFped) and genomic (ΔFROH) rate of inbreeding was computed by regressing the natu- ral logarithm of Fped and FROH onto the year of birth: eYFFY ++−=− β)1ln()1ln( 0 LeF )1( β−=∆ where FY is year-averaged F | FROH, β = ln(1 – ΔFY) and L is the average generation interval. Effective population size (Ne) was subsequently estimated as Ne = 1/(2 * ΔF). The identified ROHs were classified into the fol- lowing length classes as proposed by other studies (Ferenčaković et al., 2013a): [1–2], (2–4], (4–8], (8–16] and > 16  Mb. FROH was computed at five different cut- Genotyping chip Number of SNPs Number of animals Accuracy of imputation* onto Illumina 50Kv02 Illumina 50Kv02 54,609 128 - GGP v02 19,720 6 94.0 % GGP v03 26,151 43 96.9 % GGP v04 30,105 22 95.6 % GGP HD 76,883 4 98.7 % GGP HDv02 138,892 11 97.0 % Average = 96.4 % GGP = GeneSeek Genomic Profiler, v = version. *Accuracy of imputation is reported as allelic concordance. Table 1: The number of animals genotyped Acta agriculturae Slovenica, Supplement 5 – 201662 J. OBŠTETER and B. LOGAR offs for ROH length: >1 MB, >2 MB, >4 MB, >8 MB and >16 MB. FROH>1MB and FROH>8MB were used for comparison with other studies, since they represent FROH consisting of all identified ROHs and ROHs that are most likely to reflect true identity by descent, respectively. 3 RESULTS AND DISCUSSION The constructed pedigree consisted of animals born between 1952 and 2016 with an average generation in- terval of 7.3 years. The mean of complete generation equivalents was 3.41 for the whole and 5.97 for the refer- ence population. The pedigree completeness was above 90  % for up to four generation and dropped to 79.3 % for six generation pedigree depth. FROH was computed for animals born between 2003 and 2016 hence this period was used for Fped and FROH comparison. There were 666 animals in the pedigree with unknown date of birth and were therefore used only for inbreeding computation but not for rate of inbreeding and effective population size calculation, since their Fped could not be included in the calculation of the year averages. Figure 1 shows the distribution of the Fped (Fig. 1a) and FROH (Fig. 1b) for the reference animals. The mean values for inbreeding coefficients in this study were 0.0142 for Fped, 0.103 for FROH>1MB, 0.102 for FROH>2MB, 0.0898 for FROH>8MB and 0.0607 for FROH>16MB (Fig. 1c). Therefore the highest genomic inbreeding coefficient was observed at ROH lengths >1  MB. This was expected since FROH>1MB and was computed based on all identified ROHs and therefore captures ancient as well as recent inbreeding. All year-averaged FROH exceeded year-averaged Fped for all of the considered years (Fig. 1c), FROH>1MB exceeded Fped ~5 times. Both, year-averaged Fped and all FROH were low- est in 2003 (Fped = 0.00245, FROH>1MB = 0.0680). Converse- ly, Fped was highest in 2005 (0.0367) and all FROH in 2010 (FROH>1MB = 0.130). Other studies investigating inbreed- ing in Brown-Swiss reported values of FROH>1MB = 0.156, FROH>8MB = 0.074 and Fped = 0.048, which is slightly higher than in our study. The highest correlation of Fped and FROH in this study was observed for FROH>16MB (0.464) and it was significantly different from 0 (1.533e-11). Ferenčakovič et al. (2013a) observed the highest correlation of Fped with FROH>1MB for Brown-Swiss (0.660). Studies investigating other cattle breeds reported similar or even weaker cor- relation and larger discrepancies between Fped and FROH. Study from Hillestat et al. (2015) observed ~5 times larg- er FROH values compared to Fped for Norwegian Red Cat- tle and Gurgul et al. (2016) reported as large as 10 times larger FROH values for Holstein. Studies investigating oth- er cattle breeds reported values of FROH>1MB and FROH>8MB ~0.090 and ~0.020 for Simmental, 0.088 and 0.035 for Fleckvieh, ~0.080 and ~0.0250 for Pinzgau, 0.048 and 0.0140 for Nellore cattle. This illustrates that inbreeding levels differ between breeds and even within the same breed depending on the local population’s demography. Estimates of ΔF from pedigree and genotype analy- sis (ROH > 1MB) were 0.00334 (Ne = 149.5) and 0.00918 (Ne = 54.4). It has been pointed out that managing in- breeding rate is more important than managing inbreed- ing level in a population. While ΔF of 0.025 is consid- ered to be high risk for a population (Ne = 20), ΔF of 0.01 (Ne = 50) was suggested to be sufficient for sustain- able management of a population. However, sometimes lower rate are desirable (Woolliams et al., 1998). In this study both ΔFped and ΔFROH are within the acceptable limits of ΔF. However, although ΔFped according to pedi- Figure 1: a) Distribution of Fped in the reference population; b) Distribution of FROH>1MB in the reference population; c) Year-averaged Fped and FROH in the reference population. Fped = Meuwissen inbreeding coefficient, FROH = genomic inbreeding coefficient. Acta agriculturae Slovenica, Supplement 5 – 2016 63 ESTIMATION OF INBREEDING IN SLOVENIAN BROWN-SWISS POPULATION gree records does not imply a special concern should be dedicated to the management of inbreeding in Slovenian Brown-Swiss population, ΔFROH is close to the proposed rate and illustrates the need for an accurate control of the inbreeding in the breeding scheme. Results show that retrieved pedigree records are not adequate to ac- curately estimate the level and rate of inbreeding in the population. Fped can capture only the inbreeding since the beginning of the pedigree records. The pedigree records for Slovenian Brown-Swiss population date back to 1950, therefore animals from this generation are assumed to be unrelated and non-inbred, an assumption which is most likely violated. FROH does not depend on pedigree infor- mation and is therefore not limited with the period and accuracy of record keeping. A total of 7,470 ROHs larger than 1 MB were de- tected in the analysis (Fig. 2a). Lengths of homozygous segments follow exponential distribution with a mean of ½ g Morgan, where g in the number of generation since the common ancestor (Howrigan et al., 2011). There- fore ROHs >1  MB date back ~50 generation, >2  MB ~25 generations, >4  MB ~12.5 generations >8  MB ~6 generations, and >16 MB ~3 generations. Shorter ROHs reflecting ancient inbreeding are more difficult to detect with pedigree analysis and require higher chip density or sequence information for detection (Zhang et al., 2015). The distribution of identified ROH lengths suggests that the local population of Brown-Swiss experienced ancient as well as some recent inbreeding The majority of identi- fied ROHs in this study fall into 2–4 MB class (29.0 %) dating back ~25–12.5 generations and 4–8  MB class (39.3 %) dating back ~12.5–6 generations. More ancient inbreeding, i.e. dating back 12.5 generations (~75 years), could be due to a bottle neck in Slovenian Brown popu- lation’s history caused by the second world war and in- breeding afterwards. Figure 2b, 2c and 2d illustrate the percentage of ge- nome covered in ROH of different lengths, i.e. 4–8 MB, 8–16 MB and >16 MB. Stratifying by ROHs lengths and chromosomal location revealed some differences in highest inbred chromosomes according to ROHs of dif- ferent length classes. Since ROHs reveal selection signa- tures and further more, ROHs of different lengths direct to a different point in population’s history, this could provide insight into the history of selection decision and goals (Kim et al., 2013). Bovine chromosome (BTA) 6 was among most highly inbred chromosomes according to all classes of ROHs (Fig. 2b, 2c, 2d). This could be ex- plained with BTA6 having been associated with milk and mastitis traits and bearing most of these QTLs among all chromosomes (Ogorevc et al., 2009). This is in con- cordance with other studies observing a ROH hotspot on BTA6 (Ferenčaković et al., 2013a). Chromosomes shown as most highly inbred according to longer ROHs (Fig. 2b, 2c), indicating recent inbreeding (< 6 generations back) have been associated with dairy traits, i.e. BTA5 has been associated with milk production in two breeds (Raven et al., 2014) and BTA12 has been associated with fertil- ity in different cattle breeds (Olsen et al., 2011; Minoz- zi et al., 2013). However, more detailed investigation should be conducted in order to determine the cause of inbreeding of the specific chromosomes. Further on, a Figure 2: a) Distribution of identified ROH lengths in MB; b) Percentage of chromosomes in ROHs 4–8 MB; c) Percentage of chromo- somes in ROHs 8–16 MB; d) Percentage of chromosomes in ROHs >16 MB. ROH = runs of homozygosity. Acta agriculturae Slovenica, Supplement 5 – 201664 J. OBŠTETER and B. LOGAR comparison study between breeds should be performed once genotypes of other Slovenian breeds become avail- able. 4 CONCLUSION To conclude, control of inbreeding level is crucial in order to prevent inbreeding depression, especially in populations undergoing selection. This preliminary study shows that inbreeding computed based on pedi- gree records might not be adequate to capture the true inbreeding of the population and might not enable effi- cient control of the inbreeding. This becomes even a big- ger concern with the introduction of genomic selection since it risks a higher rate of inbreeding comparing to classical selection due to a shortened generation inter- val (Boichard et al., 2015). The development of genomic technologies enabled a more accurate estimation of in- breeding levels by using genotypic data to detect ROHs. The latter could provide a more powerful method to es- timate true inbreeding levels since the pedigree estima- tion relies on the pedigree depth and captures only the inbreeding since the beginning of pedigree recordings. This has been shown also in this study where FROH ex- ceeded Fped for all the studies years. Although the major- ity of detected ROHs in this study are >4 MB reflecting inbreeding dating back less than 12.5 generation (~75 years) ago and pedigree records date back ~65 years, the latter are not complete. Furthermore, while estimated ΔF did not raise any special concern, estimated ΔFROH was close to the suggested upper limit for ΔF implying special attention needs to be paid to the control of inbreeding in the Slovenian Brown-Swiss breeding program. Addition- ally, it has been stressed that genomic selection requires genomic control of inbreeding (Sonesson et al., 2012). However, it has to be taken into mind that it has been shown that 50K chip density can overestimate FROH>1MB due to false positives (Ferenčakovič et al., 2013b). Addi- tion of more animal genotype on higher chip densities and expansion of animals’ birth year will improve the de- tection of short ROHs reflecting more ancient inbreeding and enable a more accurate estimation of the inbreeding rate in the population. 4 REFERENCES Boichard, D. (2002). Pedig: a fortran package for pedigree analysis suited for large populations. In: Proceedings of the 7th World Congress on Genetics Applied to Livestock Pro- duction, Montpellier, France, 19–23 Aug. 2002, CD-ROM communication č. 28–13. Boichard, D., Ducrocq, V., Fritz, S. (2015). Sustainable dairy cattle selection in the genomic era. Journal of Animal Breed- ing and Genetics, 132(2), 135–143. Falconer, D. S., Mackay, T. F. C. (1996). Introduction to Quanti- tative Genetics. Harlow, UK: Longman. Ferenčaković, M., Hamzić, E., Gredler, B., Solberg, T. R., Klem- etsdal, G., Curik, I., Sölkner, J. (2013a). Estimates of autozy- gosity derived from runs of homozygosity: empirical evi- dence from selected cattle populations. Journal of Animal Breeding and Genetics, 130(4), 286–293. Ferenčaković, M., Sölkner, J., Curik, I. (2013b). Estimating autozygosity from high-throughput information: effects of SNP density and genotyping errors. Genetics, Selection, Evolution : GSE, 45(1), 42. Groeneveld, E., Westhuizen, B. v D., Maiwashe, A., Voordewind, F., Ferraz, J. B. S. (2009). POPREP: a generic report for population management. Genetics and Molecular Research: GMR, 8(3), 1158–1178. Gurgul, A., Szmatoła, T., Topolski, P., Jasielczuk, I., Żukowski, K., Bugno-Poniewierska, M. (2016). The use of runs of ho- mozygosity for estimation of recent inbreeding in Holstein cattle. Journal of Applied Genetics, 1–4. Howrigan, D. P., Simonson, M. A., Keller, M. C. (2011). Detect- ing autozygosity through runs of homozygosity: A com- parison of three autozygosity detection algorithms. BMC Genomics, 12, 460. Kim, E.-S., Cole, J. B., Huson, H., Wiggans, G. R., Van Tassell, C. P., Crooker, B. A., Liu, G. et al. (2013). Effect of Artificial Selection on Runs of Homozygosity in U.S. Holstein Cattle. PLoS ONE, 8(11). Retrieved from http://doi.org/10.1371/ journal.pone.0080813 Logar, B., Podgoršek, P., Jeretina, J., Ivanoviè, B., Perpar, T. (2005). Online-available milk-recording data for efficient support of farm management. V: Knowledge transfer in cattle husbandry : new management practices, attitudes and adaptation. Wageningen, Wageningen Academic Pub- lisher, 227–230. Maignel, L., Boichard, D., Verrier, E. (1996). Genetic variability of French dairy breeds estimated from pedigree informa- tion. Interbull Bulletin, 14, 49. McQuillan, R., Leutenegger, A. L., Abdel-Rahman, R., Franklin, C. S., Pericic, M., Barac-Lauc, L., Smolej-Narancic, N., et al. (2008). Runs of homozygosity in European populations. American Journal of Human Genetics, 83(3), 359–372. Meuwissen, T., Luo, Z. (1992). Computing inbreeding coeffi- cients in large populations. Genetics Selection Evolution, 24, 305. Minozzi, G., Nicolazzi, E. L., Stella, A., Biffani, S., Negrini, R., Lazzari, B., Ajmone-Marsan, P., Williams, J.L. (2013). Ge- nome Wide Analysis of Fertility and Production Traits in Italian Holstein Cattle. PLoS ONE, 8, 11. Retrieved from 10.1371/journal.pone.0080219 Nicolazzi, E. L., Marras, G. (2015). Zanardi: an open-source pipeline for genomic analysis using SNP array data. Fon- dazione Parco Tecnologico Padano, Via Einstein, Loc. Cas- cina Codazza (26900) Lodi, Italy. Ogorevc, J., Kunej, T., Razpet, A., Dovc, P. (2009). Database of cattle candidate genes and genetic markers for milk pro- duction and mastitis. Animal Genetics, 40(6), 832–851. Olsen, H. G., Hayes, B. J., Kent, M. P., Nome, T., Svendsen, M., Acta agriculturae Slovenica, Supplement 5 – 2016 65 ESTIMATION OF INBREEDING IN SLOVENIAN BROWN-SWISS POPULATION Larsgard, A.G., Lien, S. (2011). Genome-wide association mapping in Norwegian Red cattle identifies quantitative trait loci for fertility and milk production on BTA12. Ani- mal Genetics, 42(5), 466–474. Potočnik, K., Jenko, J., Gorjanc, G. (2016). Ali imamo možnosti za lastno genomsko selekcijo v slovenski populaciji rjavega goveda? Rjavo govedo, 11, 7–10. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M. A. R., Bender, D., Maller J., et al. (2007). PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genet- ics, 81(3), 559–575. Purfield, D. C., Berry, D. P., McParland, S., Bradley, D. G. (2012). Runs of homozygosity and population history in cattle. BMC Genetics, 13, 70. Raven, L. A., Cocks, B. G., Hayes, B. J. (2014). Multibreed ge- nome wide association can improve precision of mapping causative variants underlying milk production in dairy cat- tle. BMC Genomics, 15, 62. Rigler, M., Sadar, M., Potočnik, K. (2016). Izvajanje genomske selekcije pri rjavi pasmi v Sloveniji. Rjavo govedo, 11, 12–13. Roshyara, N. R., Kirsten, H., Horn, K., Ahnert, P., Scholz, M. (2014). Impact of pre-imputation SNP-filtering on geno- type imputation results. BMC Genetics, 15, 88. Santus, E. (2011). Intergenomics: business rules and transition into services. Interbull Bulletin, 43, 2. Sonesson, A. K., Woolliams, J. A., & Meuwissen, T. H. E. (2012). Genomic selection requires genomic control of inbreeding. Genetics, Selection, Evolution: GSE, 44, 27. Stachowicz, K., Sargolzaei, M., Miglior, F., Schenkel, F. S. (2011). Rates of inbreeding and genetic diversity in Cana- dian Holstein and Jersey cattle. Journal of Dairy Science, 94(10), 5160–5175. Van der Werf J. (1999). Inbreeding and the effects of increased prolificacy. V: Animal Breeding: Use of New Technologies. Sydney, Post Graduate Foundation in Veterinarian Science of the University of Sydney: 188–198. Woolliams, J. A., Gwaze, D. P., Meuwissen, T. H., Planchenault, D., Renard, J. P., Thibier, M., Wagner, H. (1998). Second- ary guidelines for the development of national farm animal genetic resources management plans–management of small populations at risk. Rome: Food and Agriculture Organiza- tion of the United Nations. Zhang, Q., Calus, M. P., Guldbrandtsen, B., Lund, M. S., Saha- na, G. (2015). Estimation of inbreeding using pedigree, 50k SNP chip genotypes and full sequence data in three cattle breeds. BMC Genetics, 16, 88. Retrieved from http://doi. org/10.1186/s12863-015-0227-7