Papachristou et al. Genet Sel Evol (2020) 52:43 https://doi.org/10.1186/s12711-020-00560-8 Genetics Selection Evolution

RESEARCH ARTICLE Open Access Genomic diversity and population structure of the indigenous Greek and Cypriot populations Dimitris Papachristou1, Panagiota Koutsouli1, George P. Laliotis1, Elisabeth Kunz2, Maulik Upadhyay2, Doris Seichter3, Ingolf Russ3, Bunevski Gjoko4, Nikolaos Kostaras5, Iosif Bizelis1 and Ivica Medugorac2*

Abstract Background: The indigenous cattle populations from Greece and Cyprus have decreased to small numbers and are currently at risk of extinction due to socio-economic reasons, geographic isolation and crossbreeding with commer- cial breeds. This study represents the frst comprehensive genome-wide analysis of 10 indigenous cattle populations from continental Greece and the Greek islands, and one from Cyprus, and compares them with 104 international breeds using more than 46,000 single nucleotide polymorphisms (SNPs). Results: We estimated several parameters of genetic diversity (e.g. heterozygosity and allelic diversity) that indicated a severe loss of genetic diversity for the island populations compared to the mainland populations, which is mainly due to the declining size of their population in recent years and subsequent inbreeding. This high inbreeding status also resulted in higher genetic diferentiation within the Greek and Cyprus cattle group compared to the remaining geographical breed groups. Supervised and unsupervised cluster analyses revealed that the phylogenetic patterns in the indigenous Greek breeds were consistent with their geographical origin and historical information regarding crosses with breeds of Anatolian or Balkan origin. Cyprus cattle showed a relatively high indicine ancestry. Greek island populations are placed close to the root of the tree as defned by Gir and the outgroup Yak, whereas the mainland breeds share a common historical origin with Buša. Unsupervised clustering and D-statistics analyses provided strong support for Bos indicus introgression in almost all the investigated local cattle breeds along the route from Anatolia up to the southern foothills of the Alps, as well as in most cattle breeds along the Apennine peninsula to the southern foothills of the Alps. Conclusions: All investigated Cyprus and Greek breeds present complex mosaic genomes as a result of historical and recent admixture events between neighbor and well-separated breeds. While the contribution of some mainland breeds to the genetic diversity pool seems important, some island and fragmented mainland breeds sufer from a severe decline of population size and loss of alleles due to genetic drift. Conservation programs that are a compro- mise between what is feasible and what is desirable should focus not only on the still highly diverse mainland breeds but also promote and explore the conservation possibilities for island breeds.

Background Cattle is a species that is present worldwide and well adapted to diverse environments, and has evolved into *Correspondence: [email protected]‑muenchen.de 2 Population Genomics Group, Department of Veterinary Sciences, LMU the most important species for production purposes Munich, Lena‑Christ‑Str. 48, 82152 Martinsried, Germany since its domestication. Te presence of cattle was inval- Full list of author information is available at the end of the article uable in the evolution of human societies with a great

© The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativeco​ ​ mmons.org/licen​ ses/by/4.0/​ . The Creative Commons Public Domain Dedication waiver (http://creativeco​ mmons​ .org/publi​ cdoma​ in/​ zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Papachristou et al. Genet Sel Evol (2020) 52:43 Page 2 of 23

impact on agricultural development, diet alterations, cattle populations since 1950. After the 1960s, with the cultural heritage and social structure [1, 2]. Molecular implementation of artifcial insemination, many indig- evidence [3] has indicated two domestication events in enous populations were crossbred with highly selected cattle from which the taurine (Bos taurus) and indicine commercial breeds [18, 19]. For example, Brachyceros, (Bos indicus) species arose, sharing the aurochs (Bos which is a Greek autochthonous breed with short horns primigenius) as common ancestor ~ 250,000 years ago resembling the Albanian and Buša cattle, was crossed [4]. During the introduction of domesticated cattle into with Swiss Brown in some regions [20]. In the mid- Europe, the Mediterranean coast played a crucial role dle of the 20th century, eight indigenous cattle breeds [2, 5]. Terefore, Greece and Cyprus, which are closely were reported in Greece. Nowadays, four of these are located to the core region of the domestication of cattle, considered as ofcially extinct (Tinos, Andros, Chios, i.e. in the near East, represented an important crossroad Kerkyra), three as threatened (Brachyceros, Katerini for the dispersion of human groups and their herds from and Sykia) and one (Kea) as a rare breed [21, 22]. Te Anatolia towards Europe [6–8]. aforementioned breeds were mainly bred in mountain- Historically, the Southern Balkan peninsula has been ous regions and/or on islands with poor infrastructure. characterized by the free movement of humans and ani- Te latter combined with geographical and natural bar- mals, especially in the areas near the current borders, riers as well as the absence of artifcial insemination led from almost the Neolithic throughout Antiquity, the to reproductive isolation, fragmentation, and gradual Roman, Byzantine and Ottoman empire times to nearly depletion of the genetic diversity in these breeds [23]. 40 years before the present [6, 9]. Pastoralism, the sea- Genetic diversity studies during the past years have sonal movement of herds and people to exploit the graz- been conducted in a large number of domestic cat- ing lands across the Balkan region, is a common practice tle breeds [13, 24–30]. Improving our knowledge on since centuries [10, 11]. Migration events that enhanced the genetic diversity within and among local breeds is the gene fow among the domesticated cattle populations, considered an issue of crucial importance for enhanc- genetic drift, physical isolation due to geographic barriers ing their efcient use with regard to sustainable animal [12] during the above historic periods, and human weak farming in a harsh and less intensifed environment, selection pressures led to the formation of well-adapted and for implementing further conservation programs local cattle breeds in rather marginal and harsh environ- [31]. Since the indigenous breeds exhibit adaptive abil- ments [6, 13]. ity to their local environment and remarkable longevity, Te farming and breeding practices associated with the genetic pool of unselected local breeds can repre- these local breeds difer substantially from those in high- sent a valuable source of genes [32]. However, with a performing breeds [14]: (i) they are fairly undiferentiated few exceptions [33, 34], little research has been carried and unselected; (ii) pedigree records are incomplete or do out on the genetic diversity and genetic relationships of not exist; (iii) breeding associations either do not exist or indigenous cattle from South East Europe with regard are recently established for conservation purposes only; to Greece and Cyprus. (iv) accordingly, there are no classical breed standards Tus, using the single nucleotide polymorphism but a breed is defned by common origin; (v) systematic (SNP) array technology, our objectives were: (i) to and standardized records of traits do not exist; and (vi) obtain unbiased estimations of the neutral genetic the infrastructure necessary for such recording is rudi- diversity of the Greek cattle populations, which rep- mentary. Terefore, these local indigenous populations of resents the frst comprehensive genome-wide analysis South East European cattle do not meet all the conditions for these populations; (ii) to evaluate within-breeds’/ to be referred as breeds. However, to avoid confusion populations’ diferent sources of genetic variance as between the terms breed, strain and population, we will well as their distinctiveness level; (iii) to predict recent use the term ‘breed’ only with the adjectives ‘indigenous’, admixture patterns of the highly selected and competi- ‘local’ or ‘rare’ or without an adjective. Tese local breeds tive breeds with the non-selected and heterogeneous have been considered to be under constant risk mainly indigenous breeds from Greece and Cyprus; (iv) to pre- since the 1970s and onwards. In fact, in some extreme dict historical admixture patterns in cattle breeds in cases, the current population size of some local breeds Greece and their expansion route towards the southern consists of only a few animals. Tis global trend [15–17] foothills of the Alps; and (v) to build an objective basis refects also, in generic terms, the situation of Greek local for the implementation of conservation programs for breeds. breeder’s associations, and national and international In Greece and Cyprus, economic and social condi- bodies, as a solution towards the uncontrolled mating tions as well as geomorphological and climatic reasons and outcrossing of certain rare breeds under high risk have not allowed the development of high-yield local of extinction. Papachristou et al. Genet Sel Evol (2020) 52:43 Page 3 of 23

Methods (KEA; n = 97), Agathonisi cattle (AGT; n = 6), Crete cat- Sampling tle (CRT; n = 11), Kastelorizo cattle (KAS; n = 4), Nisyros Hair roots or blood were sampled from 285 individuals cattle (NSY; n = 7) and Cyprus cattle (CYP; n = 5). A originating from diferent indigenous cattle populations. detailed description of the local cattle breeds studied Sampling areas are in Fig. 1, which as all subsequent fg- here is provided in Additional fle 2 and Additional fle 1: ures, uses the same color code specifed for each geo- Table S2. Te samples collected in this study were com- graphical group in Additional fle 1: Table S1. All the plemented by whole-genome genotypes for GRB (n = 19) samples were collected by trained personnel according and CYP (n = 9) reported by Flori et al. [34] and SYK cat- to the best veterinary practices. An ethical permission tle (n = 5) reported by Verdugo et al. [35] (see Additional was given by the ethics committee of the Agricultural fle 1: Table S1). In addition, for comparison purposes, University of Athens. Briefy, the following local breeds we included genetic information of 104 international from Greece and Cyprus were sampled in our analy- breeds, based on genetic, historical and geographical sis: (i) from mainland Greece: Greek Brachyceros breed criteria. More precisely the large dataset of the above- (GRB; n = 97), Katerini breed (KTR; n = 20), Prespa cat- selected breeds fell into eight main geographic groups tle (PRG; n = 10), Rodope cattle (ROG; n = 12), Sykia (Minor Asia, South East Europe, East Podolian, Tyrrhe- breed (SYK; n = 16), (ii) from the islands: Kea breed nian (Apennin-Sicily-Sardinia-Corse), Alpine, France,

FINN YAK GIR FJL FINE NDA FIAY

FINW SERC HGL NRC YARO AAN

GLW DXT KRY SHR

HER HF LKF GNS BBB JSY FGV BPN NOR UKP MAN VOG BBV DFV MON PAR CHR OBV MWF TGV PIN LIM BAQ ABO PUST PRDO OVAR BPUS SIC TAR REND MARO GAS AUB PMT REGG HRI SAL BURL HRP BAR CANA RDBI CABA MODE SYG PONT HRB SHB GARF RMG BHB RMB ALEN CALV MCH MNB SRB CORS MPIS CHI LKB DGB RHS MNRQ MARE ATSY MALL SBRU SKB MKB ATSR AGER MAB DBB ROG SARD PRE SYK ATER NGAN MOSA PRG TRG MARI KTR PODO GRB ATBC CINI RSIC AGT KEA MOSI NSY CYP CRT KAS

Fig. 1 Origin of the breeds used in the analyzed dataset. Special square marks represent the infuence of East-Podolian (grey), Alpine (green) and North-West (olive green) groups Papachristou et al. Genet Sel Evol (2020) 52:43 Page 4 of 23

Iberian and North West European breeds) plus one out- Illumina BovineSNP50 BeadChip array was used fol- group that included Gir (GIR), Yak (YAK), and N’Dama lowing standard procedures (http://www.illum​ina.com). (NDA). We included all the above breeds in our study Furthermore, quality controls were applied to obtain a because the geographical origin of some of them have high-quality dataset: (i) only the animals with a call rate an obvious proximity to Greece and Cyprus or because higher than 0.95 were included; (ii) SNPs that mapped to they have more or less afected the genetic pool of local unknown or sex chromosomes were removed; (iii) SNPs breeds from Greece and Cyprus through long-term that were genotyped for less than 90% of the samples crossbreeding events during the past [36] and possibly were removed; (iv) SNPs with a MAF lower than 0.025 till present. For example, the long-term crossing of the and that deviated from Hardy–Weinberg equilibrium indigenous short horned (GRB) cattle and some breeds within breed (P ≤0.01) were removed. Te dataset, at this from the Alpine group (e.g. OBV, BBV and TGV) led to stage, consisted of 46,678 autosomal SNPs genotyped in the formation of the KEA breed. However, the GRB and 3457 individuals. the short horned Buša cattle sampled in the neighbor- ing Balkan countries (South East European group) prob- Haplotyping and unifed additive relationships (UAR) ably share the same origin, which is why we sampled 18 Te Beagle software package (v 5.0) was used for impu- additional Buša cattle individuals from North Macedonia tation of missing genotypes [38] and haplotype phasing (see Additional fle 1: Table S1). Te KTR and SYK breeds [39]. To improve the efciency of phasing and impu- are hypothesized to share ancestry with the East Podo- tation, we considered genotyping data of all available lian steppe geographic cattle group. Te AGT and NSY bovine animals in the in-house database, which includes breeds are hypothesized to share ancestry with some a large number of pairs and trios from other projects. strains of Podolian steppe and Anatolian origin, whereas Genome-wide relationships between individuals were CYP and KAS might share Anatolian and zebu origins. In estimated using a unifed additive relationship (UAR) addition to the Bos taurus breeds that come from Europe matrix [40] implemented in the R package snpReady [41] and Minor Asia, in our phylogenetic analyses, we used as and applied to 46,678 SNP genotypes of 3457 animals. outgroups YAK from Mongolia, NDA from West Africa Diversity, phylogeny and population structure analy- representing African Bos taurus and GIR cattle originat- ses require samples of representative and least related ing from India but bred in Brazil, representing Bos indi- animals in each breed. To retain the most representative cus cattle. Tese 115 breeds and 10 geographical breed animals, we excluded outlier individuals (i.e. erroneously groups used in our analyses are described in Additional sampled and/or admixed animals) by multivariate outlier fle 1: Table S1. We believe that the best way to display analysis [42], and then decreased the level of relation- the grouping of breeds is by geographical origin (color ship by successive exclusion of highly related animals. code) without taking previous genetic knowledge into Both the multivariate outlier analysis and the reduction account, and to improve visualization, we added a sym- of familial structures within breeds rely on the genome- bol for some breeds to indicate possible or known genetic wide additive genetic relationships stored in the UAR similarities with other breeds or groups. More specif- matrix. Finally, the dataset used in subsequent diversity cally, these symbols combine the color of the group from and phylogenetic analyses included the 2858 most rep- which they originate geographically with the color of resentative and unrelated animals. Te starting and opti- the group with which they have some genetic similar- mized sample sizes for each breed are listed in columns N ity. Tus, the breeds BURL (from the Tyrrhenian group), and Nd (see Additional fle 1: Table S1). and PUST and PIN (from the Alpine group) display an additional symbol with a color that indicates a known Haplotype diversity infuence from the North West group, the breeds KEA To design the Illumina Bovine SNP50 BeadChip (Illu- (from Greece), and CABA, AGER, and SBRU (from the mina), only fve taurine breeds and one indicine breed Tyrrhenian group) display an additional color that indi- were considered [43]. Terefore, since this array could cates infuence from the Alpine group, and some local contain biased sets of pre-ascertained SNPs, we adopted breeds from Italy (RMG, MCH, CALV, CHI, MARE, and a 4-SNP-block approach as described previously to PODO) display an additional color that indicates possible reduce this ascertainment bias [13, 14]. Specifcally, podolian infuence [37]. 4-SNP blocks (haplotypes) that spanned less than 150 kb and had an inter-marker distance shorter than than 50 kb DNA isolation and SNPs were defned, leading to a compromise between the max- DNA isolation was performed using a commercial kit imum number of SNPs and the minimum recombina- (QIAamp DNA MiniKit, QIAGEN) according to the tion probability within the block [14]. In total, 5756 SNP manufacturer’s instructions. For SNP genotyping, the blocks were taken into account as multi-allelic markers Papachristou et al. Genet Sel Evol (2020) 52:43 Page 5 of 23

and their haplotypes as alleles in the subsequent unbiased Phylogeny and population structure allelic diversity and heterozygosity analyses. Hereafter, We applied four phylogenetic and population struc- SNP blocks are also referred to as multi-allelic markers. ture analyses to infer relationships between animals and breeds. Two of these analyses rely on bi-allelic SNP geno- types and two on multi-allelic SNP-block genotypes. In Genetic diversity addition, two of these analyses represent supervised clus- Te following population genetic parameters were esti- tering and two represent unsupervised clustering. mated: total number of alleles (nA), mean number of alleles per block (mA), allelic richness (AR), observed Supervised phylogeny of 115 cattle breeds (HO) and expected heterozygosity (HE), number of pri- To elucidate the phylogenetic relationships among the vate alleles (npA; alleles observed only in one popula- studied populations, we implemented maximum likeli- tion), frequency of private alleles (fpA) and number of hood (ML) methods using the TreeMix program [49]. common alleles (ncA; observed in all subpopulations) as In this approach, we used the dataset of SNPs genotyped described previously [13, 44, 45]. In addition, we used in 115 breeds (Table 1), and set YAK as outgroup to root the number of semi-private alleles (nspA), defned as the the tree. Te second supervised approach used the allele alleles observed in two populations only, to further assess frequencies of 5756 haplotype blocks to estimate the the allelic diversity and to consider any genetic and geo- Nei’s unbiased DA-distances [50]. Ten, the ­DA-distance graphical closeness of the included breed pairs. Inbreed- matrix of 115 breeds was used to reconstruct the neigh- ing coefcients of each animal i (Fi) were estimated from bor-net with the SplitsTree4 software and the neighbor- the diagonal elements of the UAR matrix as Fi= UAR​ joining tree with the FigTree 1.4 software (http://tree.bio. (i,i) − 1. To improve the presentation and discussion of ed.ac.uk/softw​are/fgtr​ee/) [51]. We set YAK as outgroup the summary statistics related to diversity, we standard- to root the neighbor-joining tree. ized and then plotted these statistics onto a map with a tessellated projection using the R-script available with Unsupervised population structure analyses the package Tess (http://membr​es-timc.imag.fr/Olivi​ er.Franc​ois/TESS_Plot.html). Similar to the above analysis, frst we examined the To assess the subpopulation-diferentiation, we used population structure based on SNP genotypes. For this purpose, we conducted clustering with the program the DEST estimator, which is analogous to the classical Admixture 1.23 [52] under the assumption that the num- GST for multi-allelic loci but is unbiased and more suit- able when the level of gene diversity is high [46]. In this ber of clusters is equal to K, with K ranging from 1 to approach, we used the dataset of the genotypes for 5756 115, i.e. the number of breeds plus 1. Since the admixture multi-allelic SNP blocks in 115 breeds. analysis does not need an outgroup, we used the data- set without YAK. To assess the quality of clustering and thus infer the most likely K, we performed 10 cross-val- Past efective population size based on linkage idations [53] and estimated the cross-validation error for disequilibrium each K. To illustrate the results of the admixture analy- We applied a linkage disequilibrium (LD) based method ses, we used the Python package Pong [54]. Te second as implemented in the SneP tool [47] to estimate the his- unsupervised approach used the proportion of genome- torical and recent efective population size (Ne) for all wide shared SNP-block alleles (PS) among all pairs of the cattle breeds with sample size larger than 8. Te esti- 2858 animals. Te PS matrix was then converted into an mation was performed on the SNPs with minimum and allele sharing distance matrix ­(DPS) as described by Bow- maximum distances equal to 20,000 and 10,000,000 bp, cock et al. [55]. Multidimensional scaling (MDS) [56] was respectively, and by applying a recombination rate cor- used to project the multidimensional D­ PS distance matrix rection [48] and a sample size correction. Te most onto a two-dimensional (2D) plane for the 115 breeds. For a better illustration and interpretation of the MDS recent efective population is represented by Ne5, (i.e. fve generations ago), the efective population size in pre- results, for each breed, we calculated the mean MDS- industrial times (i.e. 50 generations or 250 years ago) is coordinates of all the individuals, which correspond to the center of each breed symbol (circle), and then we represented by Ne50, and in times close to domestication estimated the standard deviation (SD) around that center. (10,000 years ago) by Ne2000. To improve the presenta- tion and discussion of the efective population size across Specifcally, for each breed, we calculated the spatial dis- tance of each individual to the group center by applying time and space, we standardized and then plotted Ne5, the Pythagorean theorem, assuming that the hypotenuse Ne50 and Ne2000 onto a map with a tessellated projection using the R-package Tess as described above. is the distance between the center of the breed symbol Papachristou et al. Genet Sel Evol (2020) 52:43 Page 6 of 23 2000 3966 2737 4338 2952 2753 3441 3905 Ne 5040 – 1753 3932 4099 3933 3825 3664 3398 3298 4274 – 3748 3709 3312 3679 – 4170 – 2791 2950 3099 3429 50 53 304 123 524 349 125 119 192 308 Ne 831 – 425 396 420 345 275 228 257 337 – 712 406 205 337 – 618 – 160 109 114 210 5 64 40 Ne 5 8 48 15 26 54 67 40 18 38 96 51 46 58 40 34 31 34 40 97 54 25 41 85 28 15 15 29 – Ν e – – – E(SNPs) 0.169 0.308 0.308 0.029 0.230 0.317 0.271 0.218 0.332 H 0.330 0.166 0.249 0.337 0.337 0.335 0.335 0.329 0.322 0.324 0.313 0.298 0.342 0.333 0.325 0.335 0.334 0.224 0.284 0.293 0.270 0.316 o(SNPs) 0.169 0.327 0.292 0.030 0.228 0.321 0.271 0.244 0.334 H 0.322 0.193 0.268 0.326 0.329 0.343 0.343 0.331 0.335 0.334 0.310 0.302 0.339 0.340 0.328 0.336 0.314 0.270 0.295 0.319 0.321 0.300 AR 4.18 3.54 2.63 1.29 3.15 4.17 4.10 3.41 2.74 3.80 4.22 4.25 2.02 2.97 AR 4.27 4.27 4.12 4.28 4.13 3.86 3.87 4.12 4.08 4.20 4.10 4.00 4.16 4.18 2.99 3.31 3.58 3.36 3.91 0.117 0.178 F 0.647 1.056 0.334 0.106 0.261 0.329 0.058 F 0.112 0.071 0.457 0.242 0.175 0.076 0.063 0.028 0.027 0.062 0.053 0.060 0.142 0.161 0.042 0.034 0.070 0.045 0.108 0.261 0.150 0.096 0.094 0.149 0.041 0.406 0.032 0.031 0.115 0.206 0.034 fpA 0.015 0.069 0.376 0.188 0.045 0.025 0.028 0.025 0.032 0.039 0.052 0.038 0.034 0.076 0.013 0.024 0.048 0.032 0.021 0.141 0.026 0.073 0.100 0.050 82.6 234.8 nspA 8 70 55 23 65 16 74 92 41 41 64 48 81 19 23 31 29 71 139 185 158 104 362 156 nspA 120 127 104 185 180 105 208 58.6 191.7 npA 78 79 32 17 70 28 11 11 73 84 30 34 43 54 85 33 65 23 16 13 20 67 188 147 294 106 105 npA 152 101 148 142 0.009 0.022 0.031 0.008 0.021 0.010 0.056 0.050 0.049 def − 0.049 − 1.350 − 0.011 − 0.004 − 0.178 − 0.009 − 0.064 − 0.387 H − 0.102 − 0.025 − 0.025 − 0.004 − 0.041 − 0.031 − 0.019 − 0.020 − 0.013 − 0.001 − 0.261 − 0.038 − 0.094 − 0.202 E 0.464 0.574 0.099 0.725 0.746 0.627 0.504 0.734 0.744 H 0.719 0.695 0.373 0.551 0.742 0.733 0.738 0.724 0.700 0.703 0.745 0.683 0.742 0.685 0.729 0.707 0.730 0.520 0.622 0.643 0.602 0.698 O 0.486 0.569 0.232 0.733 0.730 0.629 0.593 0.741 0.721 H 0.713 0.740 0.518 0.608 0.726 0.752 0.757 0.727 0.729 0.725 0.738 0.696 0.701 0.651 0.743 0.717 0.731 0.656 0.646 0.704 0.723 0.664 mA 7.65 5.89 4.19 5.58 1.60 7.26 9.03 4.75 3.01 7.17 7.88 mA 7.10 2.20 3.41 7.71 7.67 7.46 6.91 6.31 6.55 8.87 5.20 5.46 8.53 7.71 6.18 7.25 2.99 6.45 5.76 4.63 4.50 6.43 nA 44,046 37,879 9193 24,111 32,136 nA 41,775 51,955 27,328 17,302 41,278 45,364 40,859 12,688 19,634 44,355 44,126 42,940 39,748 36,322 37,724 51,066 29,941 49,079 44,381 35,556 41,711 31,406 17,185 33,145 26,666 25,912 37,037 37,105 6 7 7 8 4 9 9 24 27 Nd 26 17 37 12 17 22 17 11 20 29 17 17 21 25 43 41 27 14 19 27 17 19 Parameters of genetic diversity in the 115 examined cattle breeds with 46,678 SNPs breeds cattle in the 115 examined of genetic diversity Parameters GIR NDA YAK ATER ATBC CYP ​ AGT RHS MKB ATSR CRT ​ NSY SRB PRE RMB SHB DGB DBB MAB ATSY GRB LKB SKB MNB TRG ​ KAS KEA PRG ROG SYK KTR 1 Table Breed Outgroups Minor Asia Greece and Cyprus Greece South-east Europe Papachristou et al. Genet Sel Evol (2020) 52:43 Page 7 of 23 2000 2952 3756 3882 3546 3632 3816 3633 3226 3572 2966 3367 2928 2203 2841 3311 3245 2833 3916 3414 2959 3014 2912 3409 3181 3335 3161 Ne 2466 3111 3132 3503 3284 50 86 161 434 434 300 342 349 533 173 535 222 294 166 146 282 191 121 583 322 149 267 250 330 233 321 154 Ne 104 239 258 293 294 5 36 51 44 Ne 5 29 56 60 51 45 53 66 21 67 32 53 24 15 29 45 34 22 68 40 49 26 38 48 44 42 23 17 35 38 45 36 Ν e E(SNPs) 0.298 0.332 0.331 0.311 0.329 0.320 0.333 0.317 0.330 0.320 0.312 0.310 0.246 0.280 0.314 0.300 0.283 0.341 0.327 0.322 0.289 0.309 0.325 0.308 0.324 0.291 H 0.274 0.315 0.319 0.322 0.327 o(SNPs) 0.313 0.325 0.325 0.312 0.337 0.325 0.326 0.319 0.331 0.327 0.313 0.320 0.286 0.288 0.320 0.307 0.294 0.340 0.335 0.323 0.306 0.314 0.335 0.314 0.330 0.311 H 0.287 0.324 0.329 0.333 0.339 AR 3.61 4.10 3.79 3.50 4.13 4.06 3.75 4.02 3.91 4.08 3.95 4.06 3.76 3.74 3.81 2.82 3.24 3.86 3.69 3.21 4.26 3.88 3.38 3.81 3.64 3.87 3.63 3.88 3.47 3.20 3.75 3.81 3.83 4.12 AR 0.103 0.051 0.087 F 0.074 0.075 0.112 0.041 0.068 0.059 0.080 0.109 0.050 0.072 0.105 0.093 0.173 0.172 0.077 0.127 0.156 0.074 0.129 0.088 0.038 0.092 0.052 0.027 0.077 0.113 F 0.170 0.069 0.052 0.049 0.029 0.023 0.025 0.021 0.024 0.037 0.021 0.050 0.053 0.022 0.023 0.028 0.042 0.152 0.057 0.040 0.032 0.077 0.039 0.092 0.038 0.035 0.053 0.032 0.020 0.047 0.064 fpA 0.103 0.045 0.027 0.046 0.033 43.4 55.7 106.5 nspA 8 95 77 46 41 39 90 18 38 15 47 42 19 50 16 58 31 17 29 23 31 19 44 18 31 22 56 119 153 147 171 nspA 22.6 82.2 45.1 npA 2 7 4 3 75 44 26 21 22 76 15 44 21 39 13 27 25 16 15 30 11 25 11 21 27 21 52 120 158 106 171 npA 0.055 0.022 0.016 0.022 0.002 0.001 def − 0.002 − 0.028 − 0.017 − 0.012 − 0.025 − 0.006 − 0.037 − 0.185 − 0.036 − 0.021 − 0.014 − 0.044 − − 0.051 − 0.007 − 0.063 − 0.017 − 0.035 − 0.024 − 0.021 − 0.023 − 0.070 − 0.026 − 0.031 − 0.036 H − 0.026 E 0.731 0.727 0.688 0.719 0.707 0.726 0.689 0.724 0.688 0.690 0.677 0.539 0.617 0.696 0.667 0.618 0.599 0.656 0.704 0.639 0.673 0.706 0.671 0.703 0.745 0.696 0.642 H 0.687 0.693 0.699 0.719 O 0.715 0.715 0.690 0.739 0.719 0.710 0.697 0.722 0.705 0.695 0.702 0.639 0.640 0.711 0.677 0.645 0.631 0.689 0.709 0.679 0.685 0.731 0.687 0.717 0.745 0.713 0.687 H 0.705 0.715 0.724 0.737 5.94 7.44 6.43 mA 7.62 7.51 6.56 7.03 6.87 7.77 5.60 7.77 6.02 6.51 5.64 3.76 5.02 6.52 6.06 4.60 4.48 5.42 7.00 5.16 5.81 6.87 5.90 6.71 8.36 5.38 6.28 mA 6.12 6.26 6.61 6.94 34,188 42,800 37,025 nA 43,858 43,242 37,742 40,468 39,565 44,722 32,211 44,708 34,676 37,470 32,476 21,614 28,886 37,548 34,890 26,502 25,804 40,285 29,711 31,176 33,445 39,534 33,939 38,621 nA 30,993 48,122 35,203 36,019 38,053 39,962 36,125 25 30 29 24 28 30 10 30 22 34 12 15 24 21 18 23 13 28 24 21 23 31 24 24 Nd 18 28 23 22 26 16 24 (continued) PODO CINI MOSI RSIC MOSA SARD SBRU CORS AGER MARE CHI MPIS CALV MCH RMG GARF PONT HRI HRP UKP PRDO OVAR REND BPUS BHB HRB MODE CABA REGG PMT BURL Tyrrhenian East Podolian Alpine 1 Table Breed Papachristou et al. Genet Sel Evol (2020) 52:43 Page 8 of 23 2000 3033 2802 3071 3147 2738 3104 3228 2615 2942 2840 2686 3274 3340 3242 3291 2801 3208 2456 3075 3562 3408 3187 3128 3390 2797 3123 3148 3128 3005 3003 Ne 2868 50 81 200 149 317 265 199 136 342 127 321 349 148 563 482 262 426 168 221 197 355 309 344 252 333 233 445 311 245 240 209 128 Ne 5 28 67 56 Ne 5 20 26 49 45 31 65 19 28 58 51 27 88 34 69 31 28 19 32 49 50 67 45 66 65 98 64 33 37 36 29 106 Ν e E(SNPs) 0.303 0.319 0.314 0.309 0.302 0.314 0.294 0.265 0.305 0.314 0.325 0.284 0.335 0.322 0.322 0.279 0.320 0.236 0.306 0.328 0.326 0.318 0.311 0.313 0.296 0.316 0.316 0.317 0.306 0.298 H 0.277 o(SNPs) 0.305 0.336 0.320 0.313 0.301 0.318 0.297 0.277 0.305 0.317 0.326 0.299 0.335 0.333 0.325 0.292 0.301 0.247 0.321 0.336 0.332 0.322 0.327 0.319 0.301 0.320 0.325 0.331 0.325 0.318 0.292 H AR 3.32 3.73 3.67 3.83 3.75 3.80 3.73 3.68 3.67 3.66 2,99 3.62 3.79 3.85 3.30 3.97 3.93 3.83 3.24 3.96 2.65 3.57 3.97 3.84 3.72 3.58 3.68 3.39 3.69 3.65 3.80 3.60 3.43 3.12 AR 0.169 0.087 0.070 F 0.047 0.129 0.108 0.071 0.145 0.166 0.200 0.117 0.079 0.066 0.147 0.079 0.061 0.044 0.065 0.158 0.143 0.276 0.075 0.035 0.066 0.071 0.058 0.074 0.100 0.075 0.067 0.045 0.054 0.083 0.166 F 0.055 0.063 0.037 0.028 0.051 0.080 0.174 0.043 0.028 0.015 0.116 0.022 0.018 0.035 0.021 0.151 0.053 0.211 0.059 0.032 0.024 0.023 0.039 0.028 0.010 0.012 0.032 0.039 0.040 0.065 0.173 fpA 43.2 58.9 30.0 nspA 46 35 65 65 54 86 34 40 49 75 42 47 63 53 73 36 67 26 17 39 39 45 25 26 15 41 36 22 18 15 27 nspA 31.3 48.9 20.5 npA 28 43 40 29 49 98 25 26 32 65 22 67 52 53 16 18 56 14 16 28 27 36 21 17 10 26 15 12 21 11 28 npA 0.006 0.001 0.000 0.059 def − 0.054 − 0.004 − 0.014 − 0.018 − 0.048 − 0.008 − 0.003 − 0.055 − 0.001 − 0.033 − 0.009 − 0.051 − 0.018 − 0.067 − 0.052 − 0.025 − 0.020 − 0.014 − 0.053 − 0.020 − 0.020 − 0.013 − 0.031 − 0.048 − 0.061 − 0.068 − 0.054 H E 0.689 0.667 0.677 0.663 0.687 0.655 0.578 0.667 0.685 0.705 0.622 0.719 0.701 0.700 0.614 0.680 0.702 0.521 0.667 0.714 0.701 0.690 0.675 0.682 0.643 0.684 0.682 0.688 0.666 0.647 0.605 H O 0.726 0.670 0.687 0.659 0.699 0.654 0.606 0.668 0.691 0.708 0.656 0.720 0.724 0.706 0.645 0.692 0.660 0.556 0.701 0.732 0.714 0.700 0.710 0.695 0.656 0.692 0.704 0.721 0.706 0.691 H 0.637 5.07 6.67 6.38 mA 5.80 5.33 6.09 5.57 5.04 4.55 6.32 6.51 7.45 6.48 5.21 7.55 6.41 7.00 5.48 6.16 6.49 3.56 5.49 6.94 6.71 6.76 6.23 6.42 6.13 6.96 6.51 6.04 5.80 5.51 4.62 mA 29,205 38,387 36,709 nA 33,384 30,662 26,609 35,039 32,044 29,018 26,163 36,384 37,489 42,891 30,016 43,444 36,898 40,280 37,288 31,527 35,429 20,486 nA 31,615 39,920 38,594 38,939 35,855 36,925 35,287 40,079 37,494 34,749 33,385 31,738 37,339 15 11 30 19 14 10 29 26 22 48 22 39 17 33 22 30 14 30 Nd 24 26 29 50 46 35 50 50 50 18 22 28 37 (continued) BPN SYG MNRQ MARO BAR ALEN R DBI SAL AUB LIM MARI CHR PAR BAQ GAS CANA NGAN MALL PUST SIC PIN TGV MWF OBV BBV DFV FGV VOG ABO MON ​ TAR North-west Europe Iberian France 1 Table Breed Papachristou et al. Genet Sel Evol (2020) 52:43 Page 9 of 23 2000 3072 2941 3110 2810 2877 2606 2641 2286 2131 2698 2841 2744 2617 2524 3016 2839 2946 2942 3608 3372 3140 Ne 3265 , E 50 246 234 244 329 187 279 159 223 227 128 131 185 275 249 181 268 239 242 281 321 316 190 Ne 5 57 Ne 5 39 43 44 67 34 72 29 74 79 27 22 31 82 83 57 57 47 42 65 41 56 27 Ν e E(SNPs) 0.315 0.307 0.312 0.325 0.307 0.329 0.286 0.286 0.332 0.274 0.300 0.312 0.304 0.322 0.282 0.325 0.320 0.310 0.318 0.324 0.316 0.305 H o(SNPs) 0.324 0.322 0.320 0.328 0.316 0.334 0.292 0.287 0.320 0.267 0.317 0.292 0.296 0.317 0.281 0.333 0.325 0.319 0.323 0.317 0.317 0.327 H , observed and expected within subpopulation heterozygosity AR 3.58 E (SNP) , average observed H heterozygosity within subpopulation; , average 3.78 3.55 3.64 3.72 3.54 3.65 3.36 3.23 3.57 3.12 3.46 3.64 3.45 3.53 3.18 3.70 3.68 3.72 3.61 3.99 3.77 3.64 AR O and H (SNP) 0.142 F 0.080 0.091 0.123 0.103 0.129 0.127 0.175 0.185 0.229 0.300 0.144 0.202 0.186 0.188 0.231 0.101 0.113 0.093 0.111 0.097 0.095 0.069 F pA, average f pA, average subpopulations; only in two , number and mean of alleles present

nspA 0.068 0.055 0.032 0.028 0.057 0.011 0.039 0.020 0.028 0.039 0.083 0.065 0.047 0.026 0.045 0.033 0.069 0.083 0.018 0.048 0.040 0.067 fpA , efective population number for ffty back number for population thousand generations and two , efective 2000 47.0 Ne nspA and ­ 50 69 61 22 46 25 29 65 28 64 15 67 26 36 42 47 39 35 74 46 81 42 100 nspA Ne , number and mean value of allelic richness; , number and mean value Ho

34.9 npA AR 4 58 38 23 40 10 14 89 20 55 57 32 24 30 33 20 32 67 19 78 59 27 npA 0.034 0.016 0.062 0.025 0.014 0.026 def − 0.029 − 0.052 − 0.029 − 0.010 − 0.030 − 0.016 − 0.016 − 0.007 − 0.053 − 0.001 − 0.023 − 0.018 − 0.030 − 0.014 − 0.004 − 0.073 H , number and mean number of private alleles; n spA and , number and mean of private

E npA 0.686 0.664 0.664 0.690 0.658 0.685 0.623 0.619 0.676 0.575 0.644 0.667 0.652 0.668 0.605 0.686 0.678 0.680 0.677 0.711 0.693 0.670 H O 0.707 0.698 0.683 0.698 0.678 0.696 0.634 0.624 0.653 0.566 0.678 0.625 0.635 0.659 0.606 0.702 0.690 0.700 0.686 0.692 0.696 0.719 H 6.06 mA , total and mean number of observed, total mA , nA/5756 (number of blocks); H alleles within subpopulation;

nA 6.27 5.91 6.07 6.84 5.53 6.58 5.02 5.47 6.37 4.45 4.81 5.50 5.97 6.21 5.00 6.40 6.15 6.16 6.44 6.77 6.77 5.59 mA , inbreeding coefcient per breed and per group; AR and and per group; per breed coefcient , inbreeding

F 34,908 nA , efective population number for fve generations back per breed and per group; ­ and per group; back per breed generations fve number for population , efective

5 nA 34,001 34,940 39,395 31,859 37,888 28,903 31,507 36,659 25,626 27,681 31,660 34,358 35,756 28,798 36,853 35,396 35,464 37,043 38,958 38,966 32,166 36,065 Ne and Nd 30 20 45 22 50 16 49 41 13 14 16 40 48 27 34 24 22 42 20 35 18 20 5 (continued) Ne NOR MAN BBB LKF HF GNS JSY HER SHR KRY DXT GLW AAN HGL NRC SERC FJL FIAY FINE FINW FINN YARO 1 Table Nd , number of genotyped animals; nA and Breed frequency of private alleles, F and alleles, frequency of private by SNPs; ­ SNPs; by average expected heterozygosity within subpopulation; npA and expected within subpopulation; heterozygosity average Papachristou et al. Genet Sel Evol (2020) 52:43 Page 10 of 23

and the position per individual. Ten, we estimated the higher diversity in the central Europe group than in the SD of these spatial distances. Finally, the SD was used as Minor Asian group, is the observed heterozygosity esti- the radius around the breed center symbol, as a proxy mator based on bi-allelic SNPs (HO[SNP]) (Table 1 and for spatial dispersion of animals of each breed. For visu- Fig. 2). Intermediate values were found for the breeds alization purposes, plot dimensions were proportionally in the Greek and Cyprus group for mean number of adjusted in R, considering a 1-inch ( 0.254 cm) length as observed alleles ( nA 37,879), mean number of pri- = npA = nspA the longest radius. vate ( = 58.6) and semi-private alleles ( = 82.6). We calculated the mean DPS within breed to show Te Greek and Cyprus group showed the highest mean F the level of breed diferentiation. Mean DPS values were inbreeding coefcient ( 0.178) and relatively low = AR standardized and then plotted as a tessellated projec- values for: mean allelic richness ( = 3.54), mean HO tion onto a map, using the R-package Tess as described observed ( = 0.656) and mean expected heterozygo- HE above. sity ( = 0.641). Within the Greek and Cyprus cattle breeds, CRT exhibited the lowest values for all genetic D‑statistics analysis diversity parameters and, at the same time, the highest To investigate the historical admixture between taurine frequency of private alleles (fpA = 0.376) as well as the and indicine cattle, the D-statistics [57] was computed highest average inbreeding coefcient (F = 0.457). For by using the qpDstats tool of the AdmixTool software GRB, most of the diversity estimates had the highest val- package [58]. In brief, for a set of three populations P1, ues but the frequency of private alleles (fpA = 0.021) and P2 and P3, and an outgroup O that fts to this phylogeny: the inbreeding coefcient (F = 0.108) were low. Gener- (((P1,P2),P3).O), the numbers of shared alleles between ally, all the analyzed island populations except the KEA P1 and P3 (BABA) and, P2 and P3 (ABBA) are calculated breed, which was sampled on the island of Kea and on by assuming that allele “A” represents the ancestral allele mainland, had very high levels of inbreeding and very low and allele “B” the derived allele. Te signifcant excess of diversity parameters (Table 1). GRB and the South East either “ABBA” or “BABA” indicates admixture between Europe Buša cattle shared similar values for almost all populations P2 and P3, or P1 and P3, respectively. We diversity parameters. Te tessellated projection of diver- used YAK as outgroup O and Bos indicus (GIR) as P3. sity statistics provided strong support for a high allelic We selected Highland cattle that originate from the most diversity in breeds from Anatolia and part of South East northwestern part of Europe (Scotland) as P2 and, thus as Europe. Based on Additional fle 3: Figure S1, a south- a Bos taurus breed with the lowest, if any, admixture level east to northwest gradient of genetic diversity could be with Bos indicus. All other cattle breeds were tested as inferred, but it is interrupted by the genetic diversity P1. To present the gradient of Bos indicus genes in Euro- parameters of the Greek island breeds. However, if the pean taurine cattle, we standardized and then plotted the Greek island breeds are excluded, this possible southeast D-values onto a map with a tessellated projection using to northwest gradient of genetic diversity remains con- the R-package Tess as described above. Te D-values sistent (see Additional fle 3: Figure S1). To illustrate a with Z > |3| were considered as signifcant and indicated possible ascertainment bias of the SNP chip data, we pre- on the map. sent the tessellated projection of the observed heterozy- gosity estimations based on multi-allelic blocks (HO) and Results bi-allelic SNPs (HO[SNP]) side-by-side in Fig. 2. Tis shows Genetic diversity that HO[SNP] suggests a high level of genetic diversity in In total, 590 common alleles were detected among the some Alpine and North West European breeds, whereas 115 breeds studied, which represents only 0.7% of the HO highlights breeds from South East Europe and Anato- total number (80,720) of alleles. All estimators of genetic lia as having the highest level of diversity. Te ascertain- diversity for the breeds studied here were highly dif- ment bias of SNP chip data was further highlighted by ferentiated among the predefned geographical groups the scatterplot of HO[SNP] versus HO in Fig. 2. Both HO[SNP] (Table 1). Te geographic group of Minor Asia displays and HO are estimators of the true diversity. Terefore, the highest average value for almost all the estimators of the diversity of the breeds placed above the overall trend allelic diversity used, i.e. for: total number of alleles (nA), line (e.g. North West Europe) is overestimated by HO[SNP] number of alleles per haplotype block (mA), number of and the diversity of the breeds placed below this line (e.g. private (npA), number of semi-private alleles (nspA) Minor Asia) is underestimated by HO[SNP]. and allelic richness (AR) (Table 1). Te heterozygosity estimates based on multi-allelic SNP blocks and on bi- Efective population size Ne allelic SNPs are highest for the South East European Buša Te weighted mean of the current efective popula- breeds. Te only diversity parameter, which indicates a tion size (Ne5) is relatively small and ranges from 28 in Papachristou et al. Genet Sel Evol (2020) 52:43 Page 11 of 23 70 Ho 0.76 0.71 0.66

65 0.61 0.57 0.52 60 55 Latitude 50 45 40 35 −100 10 20 30 Longitude

70 Ho(SNP) 0.34 0.31 0.28

65 0.25 0.22 0.19 60 55 Latitude 50 45 40 35 −100 10 20 30 Longitude

Fig. 2 Tessellated projection and value distribution plot of observed heterozygosity estimated based on multi-allelic SNP-blocks (HO) and bi-allelic SNPs (HO[SNP]) Papachristou et al. Genet Sel Evol (2020) 52:43 Page 12 of 23

0.35

0.30

Ho (SNP)

0.25

Gir&N'Dama Minor Asia Greece&Cyprus SouthEastEurope East Podolian 0.20 Tyrrhenian Alpine France Iberian NorthWestEurope Linear (Minor Asia) Linear (North West Europe) Linear (All) 0.15 0.45 0.55 Ho 0.65 0.75 Fig. 2 continued

the Iberian to 67 in the French geographic group. Te Ne is inversely correlated to the extent of LD, our results Greek and Cyprus group showed a small average value suggest that the level of LD is high in the fragmented (Ne5 = 40). Only GRB showed a larger Ne5 (Ne5 = 85) breeds under extinction pressure, which is most probably than the weighted average of the group (Table 1). Going caused by uncontrolled inbreeding. back in the past, the efective population size increased faster in the Buša group than in other European cat- tle groups. Consequently, during the pre-industrial time Genetic diferentiation (Ne50 250 years ago), the efective population size was Pairwise population diferentiation values as estimated clearly larger for the South East Europe Buša (Ne50 = 399) by the unbiased estimator developed for multi-allelic Τ group than for other European breed groups. Only the markers, DEST, are in Additional fle 5: Table S4. he Minor Asian group showed a larger value (Ne50 = 549) DEST values between 112 cattle breeds (YAK, GIR and during pre-industrial times. A comparable trend was also NDA excluded) ranged from 0.001 to 0.462. Te breeds observed for efective population size 10,000 years ago and breed groups with high allelic diversity and high (Ne2000). Te values for the GRB and Buša groups were heterozygosity showed a very low level of diferentia- comparable and difered from those for the other Greek tion. For example, fve breeds of the Minor Asia group cattle breeds. Te 11 indigenous CRT animals sampled showed an average diferentiation to each other of only on the island of Crete showed the highest inbreeding 0.006 and the Buša breeds a value of 0.050. On the level and the smallest efective population size (Ne5, Ne50 other hand, the Greek and Cyprus breeds were highly DEST and Ne2000) in the entire dataset (Table 1) (see Additional diferentiated ( = 0.250), followed by the Ibe- DEST fle 4: Figure S2). Te tessellated projections provide rian group ( = 0.189) and the East Podolian group DEST strong support to the observation that as we go back in ( = 0.180) (see Additional fle 5: Table S4). Similar time the efective population size increases in regions trends were also observed when breeds from one pre- closer to the domestication center and decreases in defned group were compared to all the other breeds, i.e. North West Europe (see Additional fle 4: Figure S2). As the level of diferentiation was lowest for the Buša breeds Papachristou et al. Genet Sel Evol (2020) 52:43 Page 13 of 23

(0.113) and highest for the Greek and Cyprus breeds Greek group, is positioned within the cluster of the Buša (0.224). breeds with short branches. On the opposite, for the Within the Greek and Cyprus group, low pairwise DEST island breeds CRT, AGT, KAS and NSY, long branches values were obtained between two mainland breeds i.e. result from a high inbreeding level. KTR and SYK, which GRB-SYK (DEST = 0.073) and high values were obtained represent Greek podolian cattle, are placed between the between two island breeds i.e. CRT-AGT (DEST = 0.413). TRG and Italian podolic breeds. KEA is the only Greek It is remarkable that the CRT breed showed the highest breed that clusters together with some breeds from the diferentiation level among all the investigated breeds Alpine region, as a result of crossbreeding that occurred DEST ( = 0.391). Again, the GRB and Buša breeds shared between indigenous Greek cattle and some Alpine cat- comparable DEST values. tle breeds. Tis cluster includes also some Italian breeds In addition, we used average allele sharing distance with Brown-Swiss infuence (CABA, AGER and SBRU). ­(DPS) between animals within a breed as a measure of the In addition, we used the allele counts of bi-allelic SNPs breed-level diferentiation. As shown in Additional fle 6: and reconstructed the phylogeny with the TreeMix pro- Figure S3, the highest ­DPS values were observed for most gram (Fig. 4) by using YAK to root the tree. Compared to of the Buša and Anatolian breeds together with the Greek other European cattle breeds, CYP, KAS and AGT were GRB and the Italian PODO breeds, whereas the lowest closer to the root of the tree and thus, closer to GIR, the values were observed for the Greek island breeds. Bos indicus representative. Te aforementioned Greek breeds are gradually followed by the cattle breeds from Genetic distances and phylogenetic neighbor net Minor Asia, Greece, Bulgaria, North Macedonia, Kosovo We estimated Nei’s genetic distance DA based on multi- and Serbia. In agreement with the neighbor-net, KEA is allelic SNP blocks and present the values as a neighbor- the only Greek breed that clusters in the Alpine cluster. net (Fig. 3) and as a neighbor-joining tree routed by YAK The remaining Buša breeds sampled along the Ion- (see Additional fle 7: Figure S4). In Fig. 3 to improve the ian-Adriatic route, i.e. Albania, Montenegro and Dal- visibility of the main part of the tree, we shortened the matia, were placed after two clusters of Italian breeds. branch length for YAK. Among the Greek and Cyprus Our phylogenetic analyses, both with multi-allelic group, the island breeds CYP, AGT and KAS were placed (Fig. 3) and bi-allelic markers (Fig. 4), do not aggre- close to Bos indicus. Tis is also the case for the Anatolian gate the so-called Podolian or gray steppe cattle breeds breeds, which form a cluster with the aforementioned (TRG, KTR, SYK, HRI, HRP, and UKP) in a single breeds. Interestingly, the GRB breed from the mainland separate cluster, instead they are scattered along the

Fig. 3 Neighbor-network based on pairwise Nei’s ­DA genetic distances among 115 breeds. Special square marks represent the infuence of East-Podolian (grey), Alpine (green) and North-West (olive green) groups. Dotted lines indicate the shortened branch length of Yak to improve visibility Papachristou et al. Genet Sel Evol (2020) 52:43 Page 14 of 23

Mongolianyak Gir Cyprus Kastelorizo Agathonisi AnatolianSouth Yellow AnatolianSouth Red AnatolianEastRed AnatolianBlack NDama TurkishGrey Crete GreekRodope Katerini Nisyros Rhodopean Shorthorn GreekBrachyceros Sykia GreekPrespa Macedonian Buša BosnianBuša RedMetochian Buša Croatian Podolian Sharri Buša UkrainianPodolian Serbian Buša Rossa Siciliana Sardinia Modicana Sicily Mallorquina Albanian Prespa Croatian Istrian Lekbian Buša MiddleAlbanian Buša Croatian Buša Dibra Buša Skodra Buša Monte-NegroBuša Dilagjini Buša Corsican Sarda Gascon Blonde'd Aquitaine Limousin Aubrac Salers Migration Vosges Barà Pustertaler weight Tarantaise PezzataRossa DO' ropa Abondance Montbéliarde 0.5 Charolais RacodiBiou Pustertaler Cika Murnau Werdenfelser Tiroler Grauvieh Cabanina Original SardoBruna Braunvieh Piedmontese Parthenaise Modenese Reggiana OttoneseVarzese Kea 0 NegraAndaluza Marismeña Alentejana Barrosa Maronesa CastaNavarra Menorquina Pinzgauer Jersey Guernsey Fjaell Northern Finncattle Eastern Finncattle WesternFinncattle Yaroslavskaya Bretonne BlackPied Normande Holstein 10ls.e. Dutch Belted Galloway Highland Hereford Dexter FinnishAyrshire SwedishRed cattle Norwegian Red Kerry Angus MaineAnjou Shorthorn BlancBleuBelge

0.00 0.05 0.10 0.15 0.20 Driftlparameter

Fig. 4 Maximum likelihood (ML) tree inferred from genome-wide allele frequency data by methods implemented in the TREEMIX program. The ML dendrogram of the relationships between the examined cattle populations was rooted with the Mongolian yak as an outgroup breed Papachristou et al. Genet Sel Evol (2020) 52:43 Page 15 of 23

phylogenetic tree and are positioned closer to their Assessment of population structure using unsupervised geographic neighbor than to the hypothetical steppe heuristic and unsupervised model‑based methods cattle or Podolian group. On the one hand, TRG, KTR, We used multi-allelic SNP blocks to estimate the allele and SYK are positioned between the Anatolian and sharing distance matrix ­(DPS) among 2858 animals and some of the Greek and North Macedonian breeds. projected these by multidimensional scaling (MDS) on a On the other hand, HRI, HRP and UKP do not form two-dimensional (2D) plane. Te MDS projection of 115 an own cluster but are placed among some of the Buša breeds is shown in Fig. 5. Αlong the frst dimension of neighbor breeds (Fig. 4). The Italian-podolian breeds MDS (MDS1), we observe that the Anatolian, Greek and form a separate cluster, which is placed between some Cyprus breeds have an intermediate position between of the Buša and other Italian breeds and the East pod- Bos indicus and the remaining European cattle breeds. olian breeds. Finally, in the phylogenetic tree recon- CYP, AGT and KAS cluster together with the Anatolian structed based on SNP allele counts (TreeMix), for breeds, except TRG. Subsequently, the mainland Greek some samples with a high inbreeding level (e.g. CRT breeds (SYK, KTR, PRG, ROG, and GRB) and the Nisyros and BHB) long branches are observed whereas for less island breed (NSY) cluster in the geographic region cor- differentiated and highly diverse breeds (e.g. TRG and responding to some of the Buša breeds (RHS, MKB, PRE, RMB) short or no branches are found. RMD, SHD, and BHB; South East geographic group), some of the breeds from the East Podolian geographic group (UKP, and HRP) and some Italian breeds (Tyr- rhenian group) of South and Central Italy (SINI, MOSA,

SHR 0. 4

AAN SERC NRC

FIAY BBB MAB

HF

0. 2 HER YAK

GIR FINN BURL FINW FJL MDS2

A TER CYP A TSY ATBC CHI DBB JSY Outgroup RHS DGB AGT TRG SKB CHR ATSR GRB KAS PRG MAB NGAN ROG PRE SYG MinorAsia MARE RMB KTR BHB PIN MARO MKB LKB 0. 0 MARE CRT SYK CINISHB A LEN Greece &Cyprus NSY SRB MOSI MNB BAR

HRI SouthEastEurope HRB GARF KEA East Podolian SIC LIM MPIS BPUS NDA FGV Tyrrhenian GAS VOG BAQ

Alpine TAR OBV BBV .2 France ABO P RDO

−0 MON Iberian DFV NorthWestEurope

−1.5 −1.0 −0.5 0.0 0.5 MDS1

Fig. 5 MDS projection of the estimated allele sharing distance matrix ­(DPS) among 115 breeds using multi-allelic SNP blocks. The position of each breed is represented with a circle in which the centre is the average position of all animals of the breed with a radius equal to the SD (standard deviation) Papachristou et al. Genet Sel Evol (2020) 52:43 Page 16 of 23

MOSI, and RSIC) including all Italian podolic breeds cross-validation error is at K = 78. Additional fle 8: Fig- (PODO, MARE, RMG, CALV, CHI, and MCH). CRT is ure S5 presents the Admixture results with the default isolated from the other two main Greek breed clusters. color assignment performed by the Pong package for KEA is positioned in the geographic region of some K = 2 till 26. All breeds from Minor Asia and most of the Alpine and Italian breeds, showing a closer relationship South East geographic groups, as well as GRB, showed to these breeds than to its own (Greek) geographic clus- a high level of complex admixture even at K = 78. All ter. Te second dimension of MDS (MDS2) separates the highly selected breeds and isolated breeds formed their breeds of North Europe and Alpine geographic area from own cluster. GRB and KTR, and SYK to a lesser degree, the breeds of central, west and southern Europe. shared the same pattern with most of the breeds belong- Te Admixture analysis presents the second unsu- ing to the South East group with evidence of shared pervised clustering method applied to 2832 animals (26 ancestry between these and the Anatolian breeds. Te Mongolian yak excluded) using the genotypes of bi-allelic highly inbred and diferentiated island breeds assigned to SNPs. Te lowest cross-validation error (cv error = 0.536) separate clusters. Te same was also observed for small- was determined at K = 78. Figure 6 presents the clus- sized mainland populations, such as PRG and ROG. At tering at K 4, 10, 24 and 78. Te reasons for choosing = K = 24, a signifcant proportion of the Alpine ancestry K = 78 are that: (i) with a K larger than 4, animals from (OBV and BBV) was estimated for KEA, which is also highly selected breeds start to separate into distinct clus- retained at the most probable true K = 78. ters; (ii) the clusters at K = 10 represent the number of pre-defned geographical breed groups plus 1, i.e. nine D‑statistics analysis groups of European Bos taurus plus NDA and GIR; (iii) Te historical admixture between taurine and indicine the cross-validation error decreases almost linearly from cattle was confrmed for most of the cattle breeds that K = 1 to K = 24; and (iv) the clustering with the lowest originate from Anatolia, Cyprus, Greek and South East

K = 78

K = 24

K = 10

K = 4

a a u s e a n a n c e n e A s i p r l i n i p i n e n ' D a m C y u r o po d o A l F r a I b e r i a u r o p r N t E P y r r h e t E G i M i n o r e & a s T e s r e e c h E E a s t h W G S o u t N o r t Fig. 6 Model-based clustering of the estimated membership fractions of individuals from all examined breeds. Each cluster is indicated by a diferent color and each individual is shown as a vertical bar, at K 4, K 10, K 24 and K 78. The K-value of 78 represents the lowest = = = = cross-validation error (cv 0.5358). The heights of the colored segments are proportional to genotype memberships. The names of the breed = groups are given according to their geographical distributions as shown in Fig. 1. For a full defnition of the breed groups (see Additional fle 1: Table S1). From left to right within each group are presented the following breeds: Minor Asia: ATER, ATBC, ATSR, ATSY, TRG; Greece and Cyprus: CYP, AGT, CRT, NSY, GRB, KAS, KEA, PRG, ROG, SYK, KTR; South East Europe: RHS, MKB, SRB, PRE, RMB, SHB, DGB, DBB, MAB, LKB, SKB, MNB, BHB, HRB; East Podolian: HRI, HRP, UKP; Tyrrhenian: PODO, CINI, MOSI, RSIC, MOSA, SARD, SBRU, CORS, AGER, MARE, CHI, MPIS, CALV, MCH, RMG, GARF, PONT, MODE, CABA, REGG, PMT, BURL; Alpine: PRDO, OVAR, REND, BPUS, PUST, SIC, PIN, TGV, MWF, OBV, BBV, DFV, FGV, VOG, ABO, MON, TAR; France: RDBI, SAL, AUB, LIM, CHR, PAR, BAQ, GAS; Iberian: MNRQ, MALL, NGAN, CANA, MARI, ALEN, BAR, MARO, SYG; North West Europe: BPN, NOR, MAN, BBB, LKF, HF, GNS, JSY, HER, SHR, KRY, DXT, GLW, AAN, HGL, NRC, SERC, FJL, FIAY, FINE, FINW, FINN, YARO Papachristou et al. Genet Sel Evol (2020) 52:43 Page 17 of 23

70 D-Statistic 0.085 0.065 0.046

65 0.027 0.007 -0.012 60 55 Latitude 50 45 40 35

−100 10 20 30 Longitude Fig. 7 Tessellated projection of D-statistics values. Red dots represent a signifcant infuence of Bos indicus (Z > |3|)

Europe and most of the Central and South Tyrrhenian (CYP, AGT, KAS, and NSY) were represented by very breeds. Te KEA, AGER and SBRU breeds, which are small sample numbers (Table 1). However, due to the infuenced by the Alpine group, as well as the MPIS and aforementioned limited population sizes and limitations SARD breeds deviate from the above described general regarding the herds (i.e. highly related, almost feral and geographic trend. In Fig. 7, the values of all these signif- difcult to handle), it was not realistic to consider that cant D-statistics (Z < |3|) are shown with a red spot on a the number of samples can be increased. For the remain- tessellated map to highlight the gradient of the Bos indi- ing Greek breeds, a sufcient number of 10 or more of cus introgression from the Southeast to the Northwest the most representative individuals of the local breeds direction. Te results provide strong evidence that sup- were sampled. Terefore, this study represents the most port the infuence of Bos indicus along the Balkan con- complete and up-to-date dataset available for the indig- tinental route and along the Mediterranean route up to enous Greek cattle and integrated into the most complete North Italy. Minor Asia and European sample collection. Improving our knowledge of the genetic diversity Discussion within and among local breeds is an issue of crucial In this study, we focused on the genome-wide structure importance for implementing further conservation pro- of 11 indigenous cattle populations, 10 from Greece and grams that are necessary for sustainable development one from Cyprus. Te dataset included all the indigenous and future animal farming in changing environments cattle populations that are reared in Greece and Cyprus [34]. Tis could be particularly important for traditional and considered to be under constant risk mainly since unselected breeds that cover a geographical area close to the 1970s and onwards. An essential reason for this situ- the domestication center [32]. ation is the increased use of sires from the so-called cos- In spite of the ascertainment bias of the BovineSNP50 mopolitan breeds mostly from the mainland. Sampling chip data [13, 43], it was highly informative for the ana- included as many individuals as possible from the whole lyzed Greek and Cyprus cattle populations as refected Greek territory (Fig. 1). Some island indigenous breeds by the consistent values obtained for various parameters Papachristou et al. Genet Sel Evol (2020) 52:43 Page 18 of 23

of genetic diversity levels. Tese parameters indicated However, the cattle populations from the Greek islands a more profound loss of genetic diversity as well as a have constantly sufered from declining population sizes higher risk of inbreeding in the Greek island populations during recent decades. In the case of the CRT breed, even than the Greek mainland breeds. Experimental studies 60 years ago, the statistical data for livestock numbers of [59] have demonstrated that a higher level of nucleotide indigenous cattle in Crete identifed only 149 of the 1954 diversity is associated with a stronger selection response (7.6%) recorded animals as indigenous or local, which is under stressful conditions. Furthermore, Vilas et al. [60] a rather small percentage [62]. Tis statistic also docu- reported that the high adaptive potential of a popula- ments a very small number of animals for such a large tion is better indicated by the allelic diversity of neutral island (one cow per 4.23 km2). For the whole territory markers than by expected heterozygosity. Here, we inves- of the Dodecanese islands, on which the AGT, NSY, and tigated cattle breeds, whose geographical origin ranged KAS breeds are raised, the same source identifed 2434 of from the cattle domestication center to the most west- the 6210 recorded animals (39%) as local. Today, the pop- ern and the most northern parts of Europe. Te overall ulation size of all the above-mentioned cattle breeds is diversity estimators suggest higher allelic diversity in very small (see Additional fle 1: Table S3 and Additional breeds from the region of South East Europe up to Ana- fle 2). In addition to small population sizes, genetic drift tolia. However, this general trend is interrupted by some is strengthened by geographical isolation that further highly fragmented and inbred Greek cattle breeds (see hampers animal exchange. Tus, in such cases, inbreed- Additional fle 3: Figure S1), especially by island breeds ing and genetic drift are unavoidable factors that shape such as CRT, AGT, NSY and KAS which are represented the diversity of these populations. More generally, there by samples of the last remaining indigenous animals. In is an opposite trend between the economic signifcance complete accordance with the strong efect of genetic of cattle and of small ruminants when the regions along drift, the corresponding average frequency of private the line from the Aegean islands toward the Greek main- alleles (fpA; Table 1) reached the highest value in the land, Balkan and Middle Europe are compared. Greek island breeds. Comparable high fpA-values were In contrast to the island breeds, the lowest levels of observed only in genetically isolated island breeds such genetic diferentiation (DEST) were observed for the breed as MNRQ and MALL (Table 1). We observed a general groups from Minor Asia and South East Europe and for trend that characterizes the fragmented breeds at risk of some mainland Greek breeds [e.g. GRB, PRG, SYK, and extinction, i.e. that they are highly inbred with a small KTR (see Additional fle 5: Table S4)]. Tis low diferen- number of high-frequency private alleles. Tis is refected tiation is accompanied by a high level of allelic diversity in the high positive correlation between F and fpA (r(F, including a large number of low-frequency private and fpA) = 0.72) and low negative correlation between F and semiprivate alleles. Low diferentiation and high diversity npA (r(F, npA) = − 0.21). are probably the consequences of a low artifcial selection As outlined in the Background section, all indigenous pressure combined with low genetic drift in efectively breeds of South East Europe, Greece, Cyprus, and Anato- large (Table 1) and less isolated populations. lia are either under weak artifcial selection or not under All supervised clustering methods applied in this any kind of coordinated artifcial selection. However, study clearly reveal the level of genetic diferentiation. there are substantial diferences with respect to isolation. Te highly diferentiated breeds show long branches in Te estimated DEST levels (see Additional fle 5: Table S4) the neighbor network (Fig. 3), neighbor joining tree (see suggest a substantial genetic diferentiation of the Greek Additional fle 7: Figure S4) and maximum likelihood tree and Cyprus group compared to the remaining groups. (Fig. 4). Tis is independent of the evolutionary sources Tis high diferentiation is attributed to the genetic drift of the diferentiation, i.e. the artifcially selected and arti- and inbreeding in the highly fragmented and naturally fcially isolated breeds (e.g. breeds from the UK) and isolated island breeds (CRT, AGT, NSY, and KAS) and the naturally isolated and naturally selected breeds (e.g. is confrmed by the low diferentiation between animals Greek island breeds) show very long branches. However, within these breeds (see Additional fle 6: Figure S3). breeds that are under strong random sampling (island However, empirical studies [59] provide evidence that, breeds) show even longer branches especially in the ML rather than inbreeding, genetic drift led to a reduction in analysis. On the contrary, for local breeds with a low dif- diversity in comparable scenarios. Moreover, purging in ferentiation and high diversity level, the three methods stressful environments (i.e. natural selection) can main- reveal short branches or even no branches (Figs. 3 and tain a higher diversity level than expected by inbreeding. 4; see Additional fle 7: Figure S4). Te unsupervised In general, each island of the Aegean represents a dif- clustering methods provide a partly diferent tale. Most ferent entity with a variety of environmental and socioec- of the animals from the artifcially unselected breeds onomic factors that afect the livestock populations [61]. of South East Europe, Greece, Cyprus and Anatolia Papachristou et al. Genet Sel Evol (2020) 52:43 Page 19 of 23

remain unclustered with the admixture approach, even but long haplotypes, which contribute to the high LD. at very high K-values (Fig. 6 and Additional fle 8: Fig- Te industrial revolution, modern breeding practices ure S5). Te CRT breed which has the highest inbreed- and changes in customers’ preference have led to the ing level (0.457) among the 114 breeds investigated here replacement of local cattle breeds with commercial cat- was the frst unselected breed that was clearly clustered tle breeds, which mostly originate from northwestern with the admixture approach at K = 7. Te K-value of Europe. Tis factor, coupled with uncontrolled breed- 78 is associated with the smallest cross-validation error ing in the remaining fragmented populations due to the and should represent the true number of clusters in our absence of modern reproduction and breeding manage- design. Among the aforementioned unselected breeds, ment practices contributed to population shrinkage of we observed clear clustering for the island breeds and most of the previously highly diverse cattle breeds from the most inbred Buša breed BHB only, which are charac- southeastern Europe. terized by high inbreeding. We observed a similar trend Based on Fig. 5 and Additional fle 8: Figure S5, and regarding clustering for the Greek KTR, SYK, ROG and on the results of the previous studies (e.g. [34, 63]), we PRG populations but in diferent proportions, which assume that introgression of indicine ancestry into Ana- refects their isolation status (see Additional fle 2 and tolian and some Mediterranean cattle occurred. To test Fig. 1). Te remaining unselected breeds and/or animals this putative indicine introgression, which could have remained as an unresolved mixture even at the K-value occurred along the migration route from Anatolia, with the lowest cross-validation error. Greece and South East Europe to the southern foothills Te MDS projection places almost all the unselected of the Alps and North West Europe, we calculated the breeds from Minor Asia, Greece and Cyprus, and geo- D-statistics. For all the breeds from these areas (Minor graphic breed groups from South East Europe and East Asia, Greece and Cyprus, South East Europe and East Podolian, in the overlapping space with suggestive trends Podolian) except KEA and also all the Italian Podolian (Fig. 5), i.e. Anatolian and Cyprus breeds and the Greek breeds and some other breeds from South and central island populations KAS and AGT close to the root of Italy, we obtained signifcant D-statistic values (Fig. 7). the tree and to GIR. Te remaining breeds that were Te frst breed from the southern Alps for which no sig- sampled in Greece were placed among the above-men- nifcant Bos indicus infuence was found is SIC (Z < |3|), tioned island breeds, the South East European Buša and with a D-value close to zero. Te D-values clearly some Italian local breeds. Tereby the breeds that are decrease as the spatial distance to the origin of Bos indi- geographic neighbor of each other overlap in the MDS cus increases. Interestingly, three breeds from the East presentation. Only the KEA breed, which is the histori- Podolian group also show a signifcant indicine intro- cal product of a cross between indigenous animals simi- gression, among which UKP that was sampled east of the lar to the modern-day breeds of Brachyceros and some Carpate Mountains. As discussed recently by Verdugo Alpine breeds (OBV, BBV and TGV), is placed between et al. [35], indicine introgression started ~ 4000 years ago the Alpine and Tyrrhenian breeds, which is also con- and may have been stimulated by the onset of a period of frmed by the admixture results (K = 8 to 26; see Addi- increased aridity known as the 4.2-thousand-year abrupt tional fle 8: Figure S5). climate change event. Te increased level of diversity in Estimating the LD-based efective population size the breeds from South East Europe, and in the Anatolian (see Additional fle 4: Figure S2) for diferent time and some Greek cattle breeds could be, also, the result of intervals during the evolution of cattle provides an various demographic events, including the introgression interesting insight into their demographic history. For of Bos indicus alleles. It should be noted that we cannot example, the pattern of Ne at the time of domestica- distinguish the proportion of diversity caused by intro- tion (~ 2000 generations ago) suggests that compared gression from that caused by other evolutionary forces, to western European breeds, modern South East Euro- e.g. low pressure of artifcial selection. We also note that pean cattle breeds had larger founder population sizes some estimators of allelic diversity, such as the number as a result of their proximity to the center of domes- of private and semi-private alleles measures the propor- tication. In fact, up to ~ 50 generations ago, cattle tion of unique alleles not present even in GIR or other breeds from southeastern Europe had larger Ne than neighbor breeds. Te fact that we sampled and geno- those from northwestern Europe. However, the cur- typed some of the close neighboring breeds intrinsically rent Ne indicates that the population size of the South reduced the number of private and semi-private alleles. East European cattle breeds such as Buša and GRB, has However, although GIR and many neighbor breeds from decreased substantially in the last 50 generations or so. South East Europe and Greece were sampled (Fig. 1), the Furthermore, this observation implies that such popu- numbers of private and semi-private alleles remained lations have accumulated a small number of common larger than for breeds from other parts of Europe. Papachristou et al. Genet Sel Evol (2020) 52:43 Page 20 of 23

Conservation of genetic diversity for sustainable develop- should focus not only on the still highly diverse mainland ment must be understood as a global long-term task. Te breeds but also promote and explore the conservation repeatedly confrmed evolutionary trade-of hypothesis [59, possibilities for island breeds. 64] suggests that increased ftness in the environment of selection is accompanied by a decrease in ftness in other Supplementary information environments. Tis trade-of is applied for high-perfor- Supplementary information accompanies this paper at https​://doi. org/10.1186/s1271​1-020-00560​-8. mance breeds that are adapted to benign (temperate) envi- ronments and for breeds that are adapted for production Additional fle 1: Table S1. Sample description, group allocation, RGB in stressful environments. In spite of the above-mentioned (color) code assigned to the pre-defned groups within this paper, breed trade-of, replacement of well-adapted local breeds by high- names, breed code, number of sampled and genotyped individuals (N), performance breeds and/or animals from their crosses is number of genotyped and unrelated individuals used to estimate the diversity parameters (Nd), current breeding purposes as well as sporadic forced by the free-market mechanism that intrinsically or recent past breeding purposes in parenthesis, breed origin, and source causes the extinction of local breeds [65]. Tis replace- of the samples or genotypes used in this study or from previous studies ment process has either already reached the terminal stage [17, 25, 28, 30, 34, 35, 67–70]. Table S2. Phenotypic, productive and repro- ductive traits. Description of phenotype, productive and reproductive or is close to reaching it in many regions [14]. If we con- traits of the 11 Greek and Cypriot analyzed breeds. Table S3. Evolution of sider animal diversity as a globally shared-resource and as bovine population sizes on the islands during the last 60 years. Statistical a prerequisite for sustainable development in the changing information regarding the evolution of cattle population on the Greek environments, then gradual depletion of neutral diversity island based on [62]. present in local breeds under soft selection pressure is a Additional fle 2. Detailed description of the indigenous Greek and Cyp- riot cattle. Compilation of information on Greek indigenous cattle breeds kind of “tragedy of the commons” [65, 66]. Terefore, con- from diverse sources [16, 33, 71–87]. This includes also information from servation breeding programs must be seen as a regulated records in Greek, which were formerly not accessible to the international long-term exploitation of common resources. scientifc community due to language barriers. As already discussed above, most probably, because of Additional fle 3: Figure S1. Tessellated projection. Spatial geographic presentation of the herein estimated diversity parameters (mA, AR, H , an abrupt climate change event, farmers started to cross E HE(SNP), npA, nspA). taurine cattle with Bos indicus [4, 9] during the early Additional fle 4: Figure S2. Tessellated projection. Spatial geographic Bronze Age. Such crosses increased the already high level presentation of the estimated efective population number (Ne5, Ne50, of diversity of local soft-selected breeds. Te subsequent Ne2000). long-term adaptation to local environments shaped the Additional fle 5: Table S4. Genetic diferentiation index. Pairwise values bovine mosaic genomes with an unknown but low pro- for DEST, per analyzed breed. Maximum and minimum values within each of the predefned geographical groups are depicted in bold letters in portion of indicine alleles in modern-day cattle breeds a gray rectangle. Maximum and minimum values considering all 115 from Anatolia up to the southern foothills of the Alps. breeds are in bold letters in green and yellow, respectively. Terefore, conservation of a high diversity level in these Additional fle 6: Figure S3. Tessellated projection. Spatial geographic partly fragmented breeds could ofer valuable genetic presentation of the estimated allele sharing distance matrix (D­ PS) among breeds using multi-allelic SNP-blocks. resources for future human needs and future abrupt or gradual climate-change events. Additional fle 7: Figure S4. Phylogenetic tree. Neighbor-joining tree based on Nei’s genetic distance DA using multi-allelic SNP blocks. Mongo- lian yak (YAK) was used as a root. Dotted lines indicate the reduced length Conclusions of YAK and GIR to improve visibility. Special square marks represent the infuence of East-Podolian (grey), Alpine (green) and North-West (olive Te phylogenetic patterns derived from genome-wide green) group. information were quite consistent with the geographic Additional fle 8: Figure S5. Admixture analysis. Population structure positions and the historical information regarding of the 114 studied breeds/populations presented for each of K inferred ancestral population value, using a genome‐wide set of bi-allelic SNPs. crossbreeding between Greek and Cyprus cattle. All All the analyzed individuals are presented as colored vertical bars, divided investigated Cyprus and Greek breeds present a com- at most K segments and with proportional heights, according to their plex mosaic genome of historical and recent admixture genotype membership. In the depicted plots (K 2 to K 26), a diferent color is presented for each cluster. From left to right= within= each group the events between neighbor and well-separated breeds. presented breeds are as follows: Minor Asia: ATER, ATBC, ATSR, ATSY, TRG; While the contribution of some mainland breeds to the Greece and Cyprus: CYP, AGT, CRT, NSY, GRB, KAS, KEA, PRG, ROG, SYK, KTR; pool of genetic alleles seems important, some island and South East Europe: RHS, MKB, SRB, PRE, RMB, SHB, DGB, DBB, MAB, LKB, SKB, MNB, BHB, HRB; East Podolian: HRI, HRP, UKP; Tyrrhenian: PODO, CINI, MOSI, fragmented mainland cattle strains sufer from a severe RSIC, MOSA, SARD, SBRU, CORS, AGER, MARE, CHI, MPIS, CALV, MCH, RMG, decline in population size and a loss of alleles due to GARF, PONT, MODE, CABA, REGG, PMT, BURL; Alpine: PRDO, OVAR, REND, strong bottlenecks and genetic drift. However, in spite of BPUS, PUST, SIC, PIN, TGV, MWF, OBV, BBV, DFV, FGV, VOG, ABO, MON, TAR; France: RDBI, SAL, AUB, LIM, CHR, PAR, BAQ, GAS; Iberian: MNRQ, MALL, a markedly reduced genetic diversity level in most island NGAN, CANA, MARI, ALEN, BAR, MARO, SYG; North West Europe: BPN, NOR, breeds, these show high fertility and longevity in stressful MAN, BBB, LKF, HF, GNS, JSY, HER, SHR, KRY, DXT, GLW, AAN, HGL, NRC, environments. Conservation programs that are a com- SERC, FJL, FIAY, FINE, FINW, FINN, YARO. promise between what is feasible and what is desirable Papachristou et al. Genet Sel Evol (2020) 52:43 Page 21 of 23

Acknowledgements 2. Howe T. Domestication and breeding of livestock. Horses, mules, asses, Part of the samples was collected and fnanced by the BUSHALIVE project cattle, sheep, goats and swine. In: Campbell GL, editor. The Oxford hand- (FAO, PO 299432). We acknowledge all breeders and breeding organizations book of animals in classical thought and life. Oxford: Oxford University that provided samples. Exchange of researchers from Greece and Germany Press; 2014. for the current study was also funded by the German Academic Exchange 3. Loftus RT, MacHugh DE, Bradley DG, Sharp PM, Cunningham P. Evidence Service (DAAD) under the running project entitled “Diferent forces behind the for two independent domestications of cattle. Proc Natl Acad Sci USA. development of genomic diversity in indigenous cattle strains and breeds in 1994;91:2757–61. Greece and Germany”. Hellenic Foundation for Research and Innovation (HFRI) 4. Pitt O, Sevane N, Nicolazzi EL, MacHugh DE, Park SDE, Colli L, et al. supported this work with a fellowship to DP under the HFRI Ph.D. Fellowship Domestication of cattle: two or three events? Evol Appl. 2018;12:123–36. frame (Fellowship Number: 1301). The authors would also like to acknowledge 5. Sampson A. Τhe mesolithic hunter-gatherers in the Southeastern Medi- members of AMALTHIA (Stamatina Trivizaki, Kostas Papaioannou, Yiannis Kazo- terranean and their contribution in the Neolithisation of the Aegean. glou, Vasilios Lekkas, Achilleas Tsaprailis, Stefano Dellepiane) for their useful Archaeol Cult. 2018;1:11–36. information and their support during the implementation of the project. 6. Ripoll MP. The Knossos fauna and the beginning of the neolithic in the Mediterranean islands. In: Efstratiou N, Karetsou A, Ntinou M, editors. Authors’ contributions The neolithic settlement of Knossos in Crete: New evidence for the DP carried out data analysis, interpreted data and drafted the manuscript. early occupation of Crete and the Aegean islands. Philadelphia: INSTAP PK and GPL assisted in data interpretation and drafting the manuscript. EK Academic Press; 2013. p. 133–69. assisted with feld work, data analysis and critically revised the manuscript. 7. Peters J, Arbuckle BS, Pöllath N. Subsistence and beyond: Animals in MU carried out the D-statistics analyses, interpreted data and revised the neolithic Anatolia. In: Özdoğan M, Başgelen N, Kuniholm P, editors. The manuscript. DS and IR coordinated the genotyping of samples and provided Neolithic in Turkey. 10500-5200 BC: environment, settlement, fora, fauna, data. GB contributed samples and genotypes. NK contributed samples and dating, symbols of belief, with views from North, South, East, and West. information on characteristics of the studied Greek breeds. IB designed the Istanbul: Archaeology and Art Publications; 2014. p. 135–203. study, provided samples and critically revised the manuscript. IM conceived 8. Arbuckle BS, Kansa SW, Kansa E, Orton D, Cakirlar C, Gourichon L, et al. and guided this study, assisted with feld work, contributed analysis tools and Data sharing reveals complexity in the westward spread of domestic genotype data, analyzed data and critically revised the manuscript. All authors animals across neolithic Turkey. PLoS One. 2014;9:e99845. read and approved the fnal manuscript. 9. Lenstra JA, Felius M. Genetic aspects of domestication. In: Garrick DJ, Ruvinsky A, editors. In The genetics of cattle. 2nd ed. Wallingford: CABI; Funding 2014. p. 19–32. Open access funding provided by Projekt DEAL. Part of the samples was col- 10. Chang C. Pastoral transhumance in the Southern Balkans as a social ide- lected and fnanced by the BUSHALIVE project (FAO, PO 299432). Exchange of ology: Ethnoarcheological research in Northern Greece. Am Anthropol. researchers from Greece and Germany for the current study was funded by the 1993;95:687–703. German Academic Exchange Service (DAAD) under the running project entitled 11. Hadjigeorgiou I. Past, present and future of pastoralism in Greece. Pastor “Diferent forces behind the development of genomic diversity in indigenous Res Pol Pract. 2011;1:24. cattle strains and breeds in Greece and Germany” (project code 57418683, IKYDA 12. Hedrick PW. Genetics of populations. 3rd ed. Sudbury: Jones and Bartlett program between the DAAD and the Greek State Scholarship Foundation (I.K.Y.)). Publishers; 2005. 13. Simčič M, Smetko A, Sölkner J, Seichter D, Gorjanc G, Kompan D, et al. Availability of data and materials Recovery of native genetic background in admixed populations using The datasets used and/or analysed during the current study are available from haplotypes, phenotypes, and pedigree information—using Cika cattle as the corresponding author on reasonable request. a case breed. PLoS One. 2015;10:e0123253. 14. Bett RC, Okeyo MA, Malmfors B, Johansson K, Agaba M, Kugonza DR, et al. Ethics approval and consent to participate Cattle breeds: extinction or quasi-extant? Resources. 2013;2:335–57. Ethical permission for hair root and blood sampling was given by the ethics 15. Taberlet P, Valentini A, Rezaei HR, Naderi S, Pompanon F, Negrini R, committee of the Agricultural University of Athens. All samples were collected Ajmone-Marsan P. Are cattle, sheep, and goats endangered species? Mol by trained personnel according to best veterinary practice. Ecol. 2008;17:275–84. 16. Medugorac I, Veit-Kensch CE, Ramljak J, Brka M, Marković B, Stojanović S, Consent for publication et al. Conservation priorities of genetic diversity in domesticated metap- Not applicable. opulations: a study in taurine cattle breeds. Ecol Evol. 2011;1:408–20. 17. Ramljak J, Bunevski G, Bytyqi H, Marković B, Brka M, Ivanković A, et al. Competing interests Conservation of a domestic metapopulation structured into related and The authors declare that they have no competing interests. partly admixed strains. Mol Ecol. 2018;27:1633–50. 18. Zervas N, Boyazoglou JG. La préservation et mise en valeur des popula- Author details tions locales en Méditerranée: état actuel du cheptel et des méthodes 1 Laboratory of Animal Breeding and Husbandry, Department of Animal d’élevage en Grèce. Paris: Société d’EthnoZoologie; 1977. Science, Faculty of Animal Biosciences, Agricultural University of Athens, 75 19. Constantinou A. Ruminant livestock genetic resources in Cyprus. Anim Iera Odos, 11855 Athens, Greece. 2 Population Genomics Group, Depart- Genet Resour Inf. 1985;4:1–8. ment of Veterinary Sciences, LMU Munich, Lena‑Christ‑Str. 48, 82152 Mar- 20. Mason IL. A world dictionary of livestock breeds, types and varieties. tinsried, Germany. 3 Tierzuchtforschung e.V. München, Senator‑Gerauer‑Str. Wallingford: CABI; 1988. 23, 85586 Poing, Germany. 4 Livestock Department, Faculty of Agricultural 21. Bizelis I. Current situation of autochthonous breeds in Greece. 2015. Sciences and Food Institute of Animal Biotechnology, University Ss. Cyril http://www.save-found​ation​.net/image​s/konfe​renze​n/2015/Iosif​_Bizel​ and Methodius, 16‑ta Makedonska Brigada 3, 1000 Skopje, North Macedonia. is_Amalt​heia.pdf. Accessed 15 Mar 2019. 5 AMALTHIA, Network for the Protection of Greek Indigenous Farm Animals, 51 22. Domestic Animal Diversity Information System (DAD-IS). FAO. 2019. Argyrokastrou, 15669 Athens, Greece. http://www.fao.org/dad-is. Accessed 13 Mar 2019. 23. FAO. The second report on the state of the world’s animal genetic Received: 30 July 2019 Accepted: 15 July 2020 resources for food and agriculture. 2015. http://www.fao.org/3/a-i4787​e. Accessed 10 Feb 2018. 24. Rothammer S, Seichter D, Förster M, Medugorac I. A genome-wide scan for signatures of diferential artifcial selection in ten cattle breeds. BMC Genomics. 2013;14:908. References 25. Decker JE, McKay SD, Rolf MM, Kim J, Molina Alcalá A, Sonstegard TS, et al. 1. Reed CA. Animal domestication in the prehistoric near east. Science. Worldwide patterns of ancestry, divergence and admixture in domesti- 1959;130:1629–39. cated cattle. PLoS Genet. 2014;10:e1004254. Papachristou et al. Genet Sel Evol (2020) 52:43 Page 22 of 23

26. Ramljak J, Ivankovic A, Veit-Kensch CE, Förster M, Medugorac I. Analysis 53. Alexander DH, Lange K. Enhancements to the ADMIXTURE algorithm for of genetic and cultural conservation value of three indigenous Croa- individual ancestry estimation. BMC Bioinformatics. 2011;12:246. tian cattle breeds in a local and global context. J Anim Breed Genet. 54. Behr AA, Liu KZ, Liu-Fang G, Nakka P, Ramachandran S. Pong: fast analysis 2011;128:73–84. and visualization of latent clusters in population genetic data. Bioinfor- 27. Simčič M, Lenstra JA, Baumung R, Dovč P, Cepon M, Kompan D. On the matics. 2016;32:2817–23. origin of the Slovenian Cika cattle. J Anim Breed Genet. 2013;130:487–95. 55. Bowcock AM, Ruiz-Linares A, Tomfohrde J, Minch E, Kidd JR, Cavalli-Sforza 28. Gautier M, Laloë D, Moazami-Goudarzi K. Insights into the genetic history LL. High resolution of human evolutionary trees with polymorphic micro- of French cattle from dense SNP data on 47 worldwide breeds. PLoS One. satellites. Nature. 1994;368:455–7. 2010;6:e13038. 56. Kruskal JB, Wish M. Quantitative applications in the social sciences: multi- 29. Kukučková V, Moravčíková N, Ferenčaković M, Simčič M, Mészáros G, dimensional scaling. Thousand Oaks: SAGE Publications, Inc.; 1978. Sölkner J, et al. Genomic characterization of Pinzgau cattle: genetic con- 57. Green RE, Krause J, Briggs AW, Maricic T, Stenzel U, Kircher M, et al. A draft servation and breeding perspectives. Conserv Genet. 2017;18:839–910. Ssequence of the Neandertal genome. Science. 2010;328:710–22. 30. Mastrangelo S, Ciani E, Ajmone Marsan P, Bagnato A, Battaglini L, Bozzi 58. Patterson N, Moorjani P, Luo Y, Mallick M, Rohland N, Zhan Y, et al. Ancient R, et al. Conservation status and historical relatedness of Italian cattle admixture in human history. Genetics. 2012;192:1065–93. breeds. Genet Sel Evol. 2018;50:35. 59. Ørsted M, Hofmann AA, Sverrisdóttir E, Nielsen KL, Kristensen TN. 31. Groeneveld LF, Lenstra JA, Eding H, Toro MA, Scherf B, Pilling D, et al. Genomic variation predicts adaptive evolutionary responses better than Genetic diversity in farm animals—a review. Anim Genet. 2010;41:6–31. population bottleneck history. PLoS Genet. 2019;15:e1008205. 32. Medugorac I, Medugorac A, Russ I, Veit-Kensch CE, Taberlet P, Luntz B, 60. Vilas A, Pérez-Figueroa A, Queasada H, Caballero A. Allelic diversity for et al. Genetic diversity of European cattle breeds highlights the conserva- neutral markers retains a higher adaptive potential for quantitative traits tion value of traditional unselected breeds with high efective population than expected heterozygosity. Mol Ecol. 2015;24:4419–32. size. Mol Ecol. 2009;18:3394–410. 61. Spilanis G, Kizos A. The Atlas of the islands (in Greek). Mytilene: Editions of 33. Cymbron T, Freeman AR, Malheiro MI, Vigne JD, Bradley DG. Microsatel- the University of the Aegean; 2015. lite diversity suggests diferent histories for Mediterranean and Northern 62. Statistique nationale de La Grèce. Résultats du recensement de European cattle populations. Proc Biol Sci. 2005;272:1837–43. l’Agriculture de la Grèce: Année 1951, 1961, 1991, 2001. Athènes: Imprim- 34. Flori L, Moazami-Goudarzi K, Alary V, Araba A, Boujenane I, Boushaba erie Nationale. https​://www.stati​stics​.gr/en/home/. N, et al. A genomic map of climate adaptation in Mediterranean cattle 63. Upadhyay M, Bortoluzzi C, Barbato M, Ajmone-Marsan P, Colli L, Ginja breeds. Mol Ecol. 2019;28:1009–29. C, et al. Deciphering the patterns of genetic admixture and diversity 35. Verdugo MP, Mullin VE, Scheu A, Mattiangeli V, Daly KG, Maisano Delser in southern European cattle using genome-wide SNPs. Evol Appl. P, et al. Ancient cattle genomics, origins, and rapid turnover in the Fertile 2019;12:951–63. Crescent. Science. 2019;365:173–6. 64. Richard E, Lenski RE. An experimental test of evolutionary trade-ofs dur- 36. Papadopoulos D. The cattle of the Kea Island. Thessaloniki: Editions of the ing temperature adaptation. Proc Natl Acad Sci USA. 2007;104:8649–54. Agricultural Bank of Greece (in Greek); 1946. 65. Cebeci AF, Ince H, Mercan MA. The intrinsic fallacy of market mechanism 37. Appuhn Κ. Ecologies of beef: eighteenth-century epizootics and and private property rights in alleviating the tragedy of the commons. the environmental history of early modern Europe. Environ Hist. Appl Econ. 2020. https​://doi.org/10.1080/00036​846.2020.17169​38. 2010;15:268–87. 66. Hardin G. The tragedy of the commons. Science. 1968;162:1243–8. 38. Browning B, Browning S. Genotype imputation 631 with millions of refer- 67. Medugorac I, Graf A, Grohs C, Rothammer S, Zagdsuren Y, Gladyr E, et al. ence samples. Am J Hum Genet. 2016;98:116–26. Whole-genome analysis of introgressive hybridization and characteriza- 39. Browning S, Browning B. Rapid and accurate haplotype phasing and tion of the bovine legacy of Mongolian yaks. Nat Genet. 2017;49:470–5. missing-data inference for whole-genome association studies by use of 68. Decker JE, Pires JC, Conant GC, et al. Resolving the evolution of extant localized haplotype clustering. Am J Hum Genet. 2007;81:1084–97. and extinct ruminants with high-throughput phylogenomics. Proc Natl 40. Yang J, Benyamin B, McEvoy BP, Gordon S, Henders AK, Nyholt DR, et al. Acad Sci USA. 2009;106:18644–9. Common SNPs explain a large proportion of the heritability for human 69. Iso-Touru T, Sahana G, Guldbrandtsen B, Lund MS, Vilkki J. Genome-wide height. Nat Genet. 2010;42:565–9. association analysis of yield traits in Nordic Red Cattle using imputed 41. Granato ISC, Galli G, de Oliveira Couto EG, Bandeira e Souza M, Freitas whole genome sequence variants. BMC Genet. 2016;17:55. Mendonça L, Fritsche-Neto R. SnpReady: a tool to assist breeders in 70. Upadhyay MR, Chen W, Lenstra JA, Goderie CRJ, MacHugh DE, Park SDE, genomic analysis. Mol Breed. 2018;38:102. et al. Genetic origin, admixture and population history of aurochs (Bos 42. Filzmoser P, Garrett RG, Reimann C. Multivariate outlier detection in primigenius) and primitive European cattle. Heredity. 2017;118:169–76. exploration geochemistry. Comput Geosci. 2005;31:579–87. 71. Pitcairn A. The agricultural resources of Cyprus: the efect of natural and 43. Matukumalli LK, Lawley CT, Schnabel RD, Taylor JF, Allan MF, Heaton MP, other factors on development. Cyprus Agric J. 1935;30:6–18. et al. Development and characterization of a high density SNP genotyp- 72. Bevan W. Notes on agriculture in Cyprus and its products. 1919. http:// ing assay for cattle. PLoS One. 2009;4:e5350. www.helle​nicaw​orld.com/Cypru​s/Liter​ature​/Willi​amBev​an/en/Notes​ 44. El Mousadik A, Petit RJ. High level of genetic diferentiation for allelic OnAgr​icult​ureIn​Cypru​s.html#Page_16. Accessed 19 Jul 2019. richness among populations of the argan tree [Argania spinosa (L) Skeels] 73. Cypriot ministry of agriculture, rural development and environment. endemic to Morocco. Theor Appl Genet. 1996;92:832–9. Census of the local cattle breed (in Greek); 2016. http://www.moa.gov. 45. Nei M. Molecular evolutionary genetics. New York: Columbia University cy/moa/da/da.nsf/All/A3F8DE5D6B78AEC5C22581A3003FFE51/$fle/%C Press; 1987. E%95%CE%A0%CE%99%CE%A3%CE%9A%CE%9F%CE%A0%CE%97%CE 46. Jost L. GST and its relatives do not measure diferentiation. Mol Ecol. %A3%CE%97%20%CE%9D%CE%A4%CE%9F%CE%A0%CE%99%CE%91 2008;17:4015–26. %CE%A3%20%CE%A6%CE%A5%CE%9B%CE%97%CE%A3%20%CE%92 47. Barbato M, Orozco-terWengel P, Tapio M, Bruford MW. SNeP: a tool to %CE%9F%CE%9F%CE%95%CE%99%CE%94%CE%A9%CE%9D%202016. estimate trends in recent efective population size trajectories using pdf?OpenElement. Accessed 19 Jul 2019. genome-wide SNP data. Front Genet. 2015;6:109. 74. Hatziolos B. The issue of animal production in Greece (in Greek). Athens: 48. Sved JA, Feldman MW. Correlation and probability methods for one and Greek Publishing Company; 1941. two loci. Theor Popul Biol. 1973;4:129–32. 75. Keller C. Studien über die Haustiere der Mittelmeer-Inseln. Zürich: 49. Pickrell JK, Pritchard JK. Inference of population splits and mixtures from Kommissions-Verlag Von Georg & Cie; 1911. genome-wide allele frequency data. PLoS Genet. 2012;8:e1002967. 76. Papadopoulos D. The cattle of the Kea Island (in Greek). Thessaloniki: Edi- 50. Nei M, Tajima F, Tateno Y. Accuracy of estimated phylogenetic trees from tions of the Agricultural Bank of Greece; 1946. molecular data. J Mol Evol. 1983;19:153–70. 77. Manetti C. L’Anatolia. Firenze: R Bemporad & Figlio; 1922. 51. Huson DH, Bryant D. Application of phylogenetic networks in evolution- 78. Dimitriadis IN. The cattle of West Macedonia (in Greek). Thessaloniki: Edi- ary studies. Mol Biol Evol. 2006;23:254–67. tor Triantafyllou M; 1933. 52. Alexander DH, Novembre J, Lange K. Fast model-based estimation of ancestry in unrelated animals. Genet Res. 2009;19:1655–64. Papachristou et al. Genet Sel Evol (2020) 52:43 Page 23 of 23

79. Karantounias A. Specifc zootechnia, cattle and bufalo breeding (in 84. Kazoglou IE, Xega N, Logotheti Α, Doleson F. Rare bovine breeds in the Greek). Athens: Editions of the Highest Agricultural School of Athens; transboundary Prespa Park (in Greek with summary in English). In: Sidi- 1967. ropoulou A, Mantzanas K, Ispikoudis I, editors. In: Proceedings of the 7th 80. Keller O. Die antike Tierwelt. Leipzig: Wilhem Engelmann; 1909. Panhellenic Rangeland Congress Conference Grassland Science: 14–16 81. Kazoglou Y. Development of pastures’ management and livestock October 2010; Xanthi; 2010. supporting tools for the ecosystems of the Prespa Municipality (Prespa 85. Grünenfelder HP, Trivizaki S. Description of Prespa cattle (Mistrece). St. National Park, Ladopotamos Valley, Krystallopigi Municipality). Deliv- Gallen: SAVE Foundation, Monitoring Institute for Rare Breeds and Seeds erables of the second reporting period January-March 2015 (in Greek), in Europe Report; 2014. under the frame of “Development of research and technological develop- 86. Psaltis G. Our cattle strains (in Greek). Athens: Editor Apatsidis N; 1931. ment innovation projects (AgroETAK)”. KYPE 3324/102 (MIS: 453350). 87. La Hatziolos B. race bovine de Sykia (Chalcidique) (in Greek and French). Hellenic Agricultural Organization “DEMETER”; 2015. Athens: Ministère de l’Agriculture Report; 1935. 82. Tsaprailis A, Kazoglou Y. Extensive breeding of the Greek shorthorn cattle breed: high quality products from the natural pastures to the consumer’s plate (in Greek with summary in English). In Proceedings of the Food Publisher’s Note Expo 2017 Meat and Products: 18–20 March 2017. Athens; 2017. Springer Nature remains neutral with regard to jurisdictional claims in pub- 83. Vezzani V. Rodi e il suo Problema Zootecnico. Milano: Natura; 1929. lished maps and institutional afliations.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission • thorough peer review by experienced researchers in your field • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions