Hdomestication and Phylogeography of Taurine and Zebu Cattle (Bostuum and Bos Indicus)

Home , Bovine genome, Taurine cattle, Zebu

hDomestication and Phylogeography of Taurine and Zebu Cattle (BOStuum and Bos indicus)

David E. MacHugh, * Mark D. Shriver, Ronan T. Loftus, * Patrick Cunningham" and Daniel G. Bradley*

*Department of Genetics, Trinity College, Dublin 2, Ireland and tDepartment of Human Genetics, University of Pittsburgh, Pittsburgh, Pennsylvania 15261 Manuscript received December 9, 1996 Accepted for publication March 26, 1997

ABSTRACT Genetic variation at 20 microsatellite loci was surveyed to determine theevolutionary relationships and molecular biogeography of 20 different cattle populations from Africa, Europe and Asia. Phylogenetic reconstruction and multivariate analysis highlighted a marked distinction between humpless (taurine) and humped (zebu)cattle, providing strong support for a separate origin for domesticated zebu cattle. A molecular clock calculation using bison (Bison sp.) as an outgroup gave an estimated divergence time between the two subspecies of 610,000-850,000 years. Substantial differences in the distribution of alleles at 10 of these loci were observed between zebu and taurine cattle. These markers subsequently proved very useful for investigations of gene flow and admixture in African populations. When these data were considered in conjunction with previous mitochondrial and Y chromosomal studies, a distinctive male-mediated pattern of zebu genetic introgression was revealed. The introgression of zebu-specific alleles in African cattle afforded a high resolution perspective on the hybrid nature of African cattle populations and also suggested that certain West African populations of valuable disease-tolerant taurine cattle are under threat of genetic absorption by migrating zebu herds.

ICROSATELLITES or simple tandemrepeat toral economies throughout the world. Through milk, M (STR) genetic markers have become the main- they provide the bulk of the animal protein consumed stay of genetic linkage mapping efforts. Recently, they by many human societies, andcontribute other im- have also been used to address questions concerning portant commodities including meat, hides, traction the genetic diversity and evolutionary history of various and dung. The biological systematics and evolutionary organisms. In particular, the genetic origins of human relationships of cattle have always been highly conten- populations have been usefully explored using micro- tious. In particular, the relationship between the two satellite polymorphisms ( BOWCOCKet al. 1994; DEKAet main types of cattle, humped zebu (Bos indicus) and al. 1995; GOLDSTEINet al. 1995a). We have also pre- humpless taurine (Bos taums) , has been an active area viously used microsatellite polymorphisms to investigate of research with various hypotheses having been pro- genetic variation amongEuropean breeds of cattle posed to account for the morphological and genetic ( MACHUGHet al. 1994). differences observed between the two subspecies. Microsatelliteanalysis of genetic diversityprovides One school of thought asserts that domesticated cat- two distinct levels of information. In addition to allele tle were firstdeveloped from a single wild ancestor, the frequency differences among populations, it also prc- aurochs (Bos pimigenius primigenius) during the early vides information about the cladistic relationships be- Neolithic phase of the agricultural communities in the tween allelesand groups of alleles on the basis of differ- Near East (circa 8000-9000 BP ) . Bos indicus popula- ences inallelic repeatlength. Novel methods have tions are then thoughtto have been produced at a later therefore been developed to investigate the mutational date through breeding and selection from Bos taums dynamics of microsatellites and their application to cattle ( EPSTEIN1971; EPSTEINand MASON 1984; PAYNE population genetic problems. The stepwise mutation 1991 ) . The alternative viewpoint contends that domes- model (SMM) , in which microsatellite allele length is ticated zebu populations were developed indepen- treated as an incremental quantitative character, has dently by a separate group of early pastoralistsand that been an important componentof these new analytical the candidate domestication centers were the Neolithic methods ( SHRTVERet al. 1993, 1995; GOLDSTEINet al. societies of Baluchistan in present-day Pakistan. The 1995b). southern Asian subspecies of aurochs (Bos primigenius Domesticated cattle are the major component of pas- namadicus) would then be the most likely progenitor of domesticated zebu cattle. This interpretation is sup Corresponding authw: David MacHugh, Genetics Department, Trin- ported by recent archaeological and genetic evidence ity College, Dublin 2, Ireland. E-mail: [email protected] (MEADOW 1993; LOFTUS et al. 1994a; BRADLEYet al.

Genetics 146 1071-1086 (July, 1997) 1072 D. E. MacHugh et al.

1996). Inparticular, through analysis of mitochondrial MATERIALSAND METHODS DNA (mtDNA) sequencevariation in extant cattle pop- Sample collection and DNA extraction: DNA samples were ulations, Loftus et al. (1994a) were able to demonstrate collected from 728 individual cattle representing 20 different that taurine and zebu cattle probably diverged at least populations originating from Africa, Asia and Europe. Sam- ples were collected during a series of field sampling missions 210,000 years ago, well outside the range required for from artificial insemination (AI) stations,agricultural re- human-mediated development of zebu cattle from tau- search institutions and village herds. Every attempt was made rine progenitors. to ensure eachpopulation sample was representative. Astrict, The history and biogeography of cattle populations distributed sampling strategy was employed and herd book andbreeding records were consulted while farmers and in Africa represent acomplex interaction of ecological, breeders were questioned about the origins and familial rela- genetic and anthropological factors. The original indig- tionships of individual animals. The populations, their sample enous cattle of Africaare universally considered to have sizes and other relevant information are detailed in Table 1. been exclusively taurine.These populations are In addition,a small number of samples were taken from three thought to have arisen from migrations of early pasto- related species. These were American bison (Bison bison; n = 2), European bison (Bison bonasus; n = 2)and Banteng ( ) ralists from the Near East EPSTEIN1971; PAYNE 1991. ( Bibos bantmg; n = 2) . DNA was extracted from blood samples Recent archaeological evidence has, however, ques- using standard procedures ( SAMBROOKet al. 1989 ) . Semen tioned this viewpoint, suggesting that the African au- straws were used as a DNA source for some of the European rochs ( Bos primigenius opisthonomus) may have been do- breeds and these were processed according to a protocol described by ANDERSONet al. (1986). mesticated independently somewhere on the African Microsatellite genotype analysis: A battery of 36 randomly continent ( WENDOWand SCHILD1994). chosen microsatellite loci were used from the array of public- Although, zebu cattle are thought to have been first domain markers available in the bovine genome mapping introduced into Africa about 4000 years ago, they only literature. These microsatellite markers were variants of the started to become widespread about 700 AD with the (CA),class and were characterized and developed using standard methods. Eventually, 20 of the 36 markers that amplified Arabic migrations into North and East Africa. At pres- robustly and repeatedly were used for full screens of the total ent, Africa represents a mosaic of cattle morphologies sample panel. These 20 microsatellites are detailed in Table with zebu cattle and intermediate forms, often referred 2 with associated reaction conditions. PCR amplifications of to as "sanga," predominating over most of the conti- individual loci were performed as follows. Reactions were performed using 96-well microtitre plates with 5-10 ng template nent ( PAYNE1970; EPSTEIN 1971). It haspreviously DNA in 11 p1 reaction volumes using 0.5 unit of Taq polymer- been suggested that the genetic signatureof zebu gene ase with reaction buffer comprising 50 mM KCl; 10 mM Tris- flow in Africa mayrepresent anunusual pattern of male- HCI pH 9.0; 1.O-2.5 mM MgC12 (see Table 2) ; 1% Triton X- mediated introgression ( BRADLEYet aZ. 1994; LOETUSet 100; 200 p~ dATP, dGTP, dTTP; 10 p~ dCTP; 0.3PM of each al. 1994a; BRADLEYet al. 1996). Taurine populations primer was added, as was 0.5 pCi [(Y-"~P]dCTP. A 10 pl oil overlay was then added and amplifications were performed predominate only in subtropical regions of West Africa in a Hybaid OmniGene thermalcycler using a 4 min denatur- where an inheritedresistance to trypanosomiasis ( trypa- ation step at94", followed by 35 cycles of 45 sec at 93", 45 sec notolerance) allows them to inhabit areas infested with at 54-63' (see Table 2), 45 sec at 72" and a final extension tsetse flies ( Glossina sp.) . Zebu cattle possess no innate step at 72" for 4 min. Samples were then mixed with 10 PI formamide solution. After heat treatment at93", 3p1 aliquots resistance to trypanosomiasis and have only started to of these mixtures were then loaded on 6% denaturing poly- penetrate these regions with the assistance of veterinary acrylamide sequencing gels. M13mp18 plasmid DNA was used prophylaxis and the destruction of the tsetse habitat to provide sequence ladders for initial allele size calibration. through widespread deforestation ( LHOSTE1991 ) . Samples with appropriate genotypes were then selected to These recent migrations of zebu cattle pose a serious provide unambiguous length standards for each microsatellite. The allelic data from each autoradiogram were entered threat to the genetic integrity of valuable trypanotoler- manually into a computer database. Data for each gel was ant populations of taurine cattle in the southern areas then printed out and double-checkedagainst the original of West and Central Africa. autoradiogram. To clarify the genetic relationships among the major Basic data analysis: Allele frequencies were determined by direct counting.Allele frequency distributions for the20 micro- groups of cattle and to characterize the extent andpat- satellite loci are available via anonymous file-transfer at the tern of zebu genetic introgression in African popula- following ftp address: acer.gen.tcd.ie/pub/cow-microsat/. tions, we have screened 20 different microsatellite loci As discussed below, 10 out of the 20 microsatellites dis- in a large number of cattle representing 20 distinct played alleles that were present at high frequency in Asian zebu breeds, intermediate frequencies in African zebu popu- populations from Africa, Europe and Asia. The patterns lations, low frequencies in African taurine populations and of geneticvariation observed within and amongpopula- that were either absent or at low frequencies in European tions should provide a large body of data to test hypoth- breeds. As an illustration of this phenomenon, Figure 1 shows eses concerning the genetic affinities and origins of the allelic distributions for two of these loci in pooled popula- domesticated cattle, the dynamics of zebu genetic intro- tions samples. These private ( NEEI.1973) or zebu-diagnostic alleles were used to evaluate gene flow and genetic admixture gression in Africa and should also provide a method to in African individuals and populations (see below).Observed assess the genetic integrity of threatened taurine popu- heterozygosity and unbiased estimates of gene diversity (ex- lations in West Africa. pected heterozygosity) with associated standard errors were Microsatellite Variation in Cattle 1073

TABLE 1 Characteristics of the 20 cattle populations surveyed

ClassificationSample originGeographical Breed size Comments

European taurine Aberdeen Angus Scotland 33 - European taurine Hereford England 34 - European taurine Jersey Channel Islands 34 - European taurine Kerry Ireland 40 Endangered“ European taurine Charolais France 36 - European taurine Friesian Netherlands 40 - European taurine Simmental Switzerland 36 - African taurine N’Dama Gambia Trypanotolerant’ 58 African taurine N’Dama Guinea 63 Trypanotolerant’ African taurine N’Dama Guinea BissauTrypanotolerant’ 54 African taurine N’Dama Nigeria Trypanotolerant’ 19 African taurine N’Dama Senegal 48 Trypanotolerant’ African zebu Butana North Sudan 24 - African zebu Kenana South Sudan 38 - African zebu Gobra Senegal 59 - African zebu Maure Mauritania 55 Sampledin Senegal‘ African zebu White Fulani Nigeria 24 - Asian zebu Hariana North India 10 - Asian zebu Sahiwal Pakistan 13 - Asian zebu Tharparker North India 10 - “The Kerry cattle breed has been classified as endangered by the Food and Agriculture Organization of the United Nations (SCHERF1995). ‘N’Dama cattle, in common with many West African taurine populations, display an inherited tolerance or resistance to the effects of tsetse-borne trypanosomiasis. ‘Because of a border dispute between Mauritania and Senegal, the Maure animals were sampled from a dwindling remnant population in Senegal.

computed for all locus/population combinations according Genetic differentiation: RAYMOND and ROUSSET( 1995b) have to NEI ( 1987). proposed a powerful method to assess population differentia- Genetic stratzjication: Three tests for deviations from Hardy- tion based on whether allelic composition is independent of Weinberg Equilibrium (HWE) were employed. The first two population assignment. This statistical test is based again on methods were performed as described by HAMMOND et al. analysis ofcontingency tables using aMarkov chain procedure (1994) and DEKAet al. (1995). In the first method a chi- to derive an unbiased estimate of the exact probability in square test was used to evaluate the overall discordance of being wrong in rejecting the null hypothesis (&) ; i.e., allelic genotype frequencies at eachlocus / population combination. composition is independent of population assignment (no The second methodused was based on a likelihood ratio (L) differentiation). This test was performed for all 190 pairwise test criterion (Gstatistic) that was used to contrast observed interpopulation comparisons on contingency tables con- and expectedgenotype frequencies.The levels ofsignificance taining datafrom each of the 20 microsatellite loci. The for these two tests were determined empirically by shuffling GENEPOP package was employed with the same Markov (permutation) of alleles across individuals with 1000 replica- chain parameters detailed above. tions ( CHAKRABORTYet al. 1991; EDWARDSet al. 1992). The Geneticdistance computations and phylogenetic recon- third test for deviations from HWE was performed using the struction: The Dsu. “stepwise-weighted genetic distance” was GENEPOP package version 1.2 (RAYMOND and ROUSSET computed according to SHRIVERet al. ( 1995), using a modi- 1995a). This method uses an exact test of HWE with the fied version of the DISPAN computer package (DSW, T. OTA, addition that, for loci/population combinations with five or Center for Human Genetics, Boston University). SHRIVERet more alleles, a Markov chain algorithm is used to obtain an al. ( 1995) have shown that theaW measure is an appropriate unbiased estimate of the exact probability of being wrong in measure for microsatellite loci that are presumed to evolve rejecting HWE. In all cases, the Markov chain was set to 50,000 in a stepwise or near-stepwise fashion (SHRIVERet al. 1993; steps with 1000 steps of dememorization. The significance VALDESet al. 1993). Neighbor-joining (N-J) trees ( SAITOU of the HWE probabilities for the three tests across loci and and NEI 1987) were constructed from aWTdistances and the populations was determined using Fisher’s method for com- reliability of the trees obtained were examined by a bootstrap bining probabilities with the following formula as discussed test with 1000 replicate resamplings of loci with replacement by RAYMOND and ROUSSET( 1995b). ( FELSENSTEIN1985 ) . Multivariatestatistical analysis To condense the genetic variation revealed with the panel of 20 microsatellites, princi- x& = -2 (CIn P,), ,=I pal components analysis (PCA) was performed according to the proceduredescribed by CAVALIMFORZA et al. ( 1994) . PCA where r is the number of loci when pooling across breeds or involves a linear transformation of the observed allele fre- the number of breeds when pooling across loci (in either quencies (in geometrical terms a rotation of the coordinate case r = 20), Pi is the Pvalue for the jthlocus or breed.This axes) where the coefficients are chosen so as to maximize the statistic follows a chi-square distribution with 2 X r d.f. variation of the transformed data measured along each new 1074 E. D. MacHugh et al.

TABLE 2 Chromosomal locations primer sequences and experimental parametersfor 20 microsatellite markers

~ ~~ Locus" ChromosomePrimer sequences (5'-3') (deg)Tm SizeMgC12(mM) range(bp)

ETH152 (D5S1) 5 TACTCGTAGGGCAGGCTGCCTG 59 1.o 191-211 GAGACCTCAGGGTTGGTGATCAG ETH225 (D9S1) 9 GATCACCTTGCCACTATTTCCT 57 1.5 140-160 ACATGACAGCCAGCTGCTACT HELl (D15S10) 15 AGTCCATGGGATTGAAAGAGTTGG 55 1.o 101-117 CTTTTATTCAACAGCTATTTAACAAGG HRHl 22 GGCTTCAACTCACTGTAACACATT 55 2.5 180-190 TTCTTCAAGTATCACCTCTGTGGCC ILSTSOOl (D7S13) 7 GGTGCTGTTATCTAGAATTTGG 58 1.o 77-97 GGAGTCATACACAACTGAGC ILSTS005 (D10S25) 10 GGAAGCAATGAAATCTATAGCC 55 2.0 181-193 TGTTCTGTGAGTTTGTAAGC ILSTS014 (D19S10) 19 CTGACTATGGTGATAATCCC 58 2.0 128-134 TCTTTTCCCTTTCCTTCCCC OCAM 25 CCTGACTATAATGTACAGATCCCTC 57 1.5 178-190 GCAGAATGACTAGGAAGGATGGCA RASA 7q24-qter CCCTTCCGCTTTAGTGCAGCCAG 63 1.5 182-196 GGGCCACAGCCCAGGATCGGGAGC TGLA48 (D7S26) 7 AAATGTTTTATCTTGACTACTAAGC 57 1.5 73-79 ACATGACTCTGCCATAGAGCAT BM2113 (D2S26) 2 GCTGCCTTCTACCAAATACCC 55 1.o 123-147 CTTCCTGAGAGAAGCAACACC BoLA DR2B 23 AGGCAGCGCCGAGGTGAGCGA 60 1.5 144-152 TCCAACACTCACCTGGACGTAGC BoLA DRPl 23 ATGGTGCAGCAGCAAGGTGAGCA 55 1.o 118-142 GGGACTCAGTCTCTCTATCTCTTTG BTMICROS Unassigned CTAGAAGATTTAGAAAACTTGCC 54 1.5 141- 187 ATAGCAAGACATATCTCCATTCC ETHl3l (D21S4) 21 GTGGACTATAGACCATAAGGTC 55 1.5 138-168 GCTGTGATGGTCTACGAATGA HBB 15q22-q27 GATATAAAAAAGAAGACCCAGTAG 54 2.0 98-120 TACCTGAGTCATATGTAATATTCC HEL5 (D21S15) 21 GGTAATGGTTTTCAGACGTTAGTG 54 1.5 161-181 GTAGCAGGATCACTTGTTAGGG PRL 23 GGAAAGTGAACATGACTGTCTAG 60 1.5 158-164 GCCCTCTCTTCTACAATGAACAC RBP3 28 TGTATGATCACCTTCTATGCTTC 55 1.5 141-153 GCTTTAGGTAATCATCAGATAGC TGLAll6 (D4S4) 4 GCACAGTAAGAGTGATGGCAGA 57 1.5 79 - 85 TGGAGAAGATTTGGCTGTGTACCCA

" Bovine genome map codes are shown in parentheses (EGGENand FRIES 1995) coordinate axis (PC). Thefirst three principal components averaged to give an estimate of the frequency of zebu-specific ( PCs) , by definition, are themost informative and these were alleles in each of the African populations. The zebu-specific plotted on a three-dimensional scatter diagramfor all 20 pop- allele for one of these loci (a 182-bp OCAM allele) was pres- ulations. ent in the Guinean N'Dama at a very low frequency ( 1.6%) . Analysis of African gene flow and genetic admixture: Two However, its distribution in India andacross the rest of Africa methods were used to estimate Asian zebu introgression in would suggest that it is a genuine zebu-specific allele and it African cattle populations. As mentioned above, 10 of the 20 was considered as such for the purposes of this survey. The microsatellites displayed zebudiagnostic alleles (see Figure presence of this allele in Guinean N'Dama may therefore 1) . The criteria used to determine whether allelean was zebu- represent residual low-level zebu introgression or perhaps al- diagnostic was presence at high frequency in the Asian zebu lelic homoplasy. populations and absence in the N'Dama population sampled Estimation of genetic admixturepoportionsfrom all allelejkequer- in Guinea. Guinean N'Dama cattle are considered to be a cies: A second, more sophisticated approach was used to esti- completely pure African taurine genepool and previous anal- mate direct zebu admixture proportions from the total data ysis of zebu Y-chromosome polymorphisms has indicated that set. This method has been developed by CHAKRABORW( 1975, they have notbeen subject to zebugenetic introgression 1985, 1986) anduses the concept of the gene identity coeffi- (BRADLEYet al. 1994).In total, 18 alleles from the 10 diagnos- cient, theprobability that two genes chosen at random (from tic loci were classified as zebudiagnostic. The frequencies of one or more populations) areidentical in state ( NEI 1972). these alleles or groups of alleles at a particular locus were The underlying rationale to this method is that genetic simi- Microsatellite Variation in Cattle 1075

ETH152 (D5S1) lLSTsoOl (D7S13) 1.o Europemn taurlne 1 0.0

FIGURE1 .-Allele fre- quencydistributions for two zebudiagnostic loci. 1 1.0 The pooledsamples are as r African taurine 0.0 follows: ( 1 ) seven Euro- pean taurine breeds, (2) 0.6 five African N’Dama tau- 0.4 rine populations, (3) five 0.2 African zebu breedsand (4) threeIndian zebu 0.0 breeds. The diagnostic al- 1.o leles for each locus are in- African zebu dicated by a solid black 0.0 color.The diagnostic al- 0.6 leles are 191 / 193 bp for 0.4 the ETH152 microsatel- n lite and 77 bp for the 0.2 ILSTS001 locus. The eight 0.0 1” other zebudiagnostic loci are as follows: ETH225, 1 1.0 i r HEL1, HRHl, ILSTS005, 0.0 ILSTS014, OCAM, RASA 0.6 and TGIA48. 0.4 0.2 0.0 191193 195 197 199 201 203 205 207 209 211 77 03 06 07 00 09 9593 91 96 97 Allele size (bp) Allele size(bp) larity between populations can beexpressed as a simple linear sition of individual animals using the presence or absence of function of admixture proportions. This method requires that zebudiagnostic alleles. This approach may prove particularly data be available from two putative source or that parental valuable for gaining an insight into the genetic erosion of populations represent the original populations that produced the trypanotolerant taurine populationsin West Africa. Seven the dihybrid populations of interest. An example of this is in populations were analyzed: Guinean N’Dama, Guinea Bissau studies of human admixture where presentday West African N’Dama, Senegalese N’Dama, Gambian N’Dama, Senegalese and Caucasian populations areused to represent the original Gobra zebu, Mauritanian Maure zebu and Sudanese Kenana source populations forinvestigations of genetic admixturein zebu. Eachanimal canbe analyzed at eachlocus that possesses African-American populations (WORKMANet al. 1963; CHAK- zebu-specific alleles, scoring a 1 for each chromosome where RAROKIY et al. 1992). In a similar fashion, the Indian zebu a diagnostic allele is present. With 10 zebudiagnostic loci, cattle breeds were pooled and used to represent a putative this gives an individwal zebu index ranging from zero to 20. Asian zebu parentalpopulation and the Guinean N’Dama were used to represent a putative African taurine sourcepopu- lation. The computations were performed using a computer RESULTS program ADMIX (R. CIiAKRARORIY, Genetics Center, Univer- AUelic diversities, heterozygosities, gene diversities, sity of Texas Health Science Center at Houston). This program uses a vector-matrix approach to produce a weighted Hardy-Weinbergequilibrium and geneticdifferentia- least-square solution for each individual admixture proportion: In total, 168 alleles were detected across the 20 tion with an associated standard error ( CHAKRARORTY1985, lociwhen they were screened in the 20 taurine and 1986). It also producescorrelation coefficients forthe zebu cattle populations. This gives a mean number of weighted least-squares solutions thatgive an indicationof the alleles at each locus of 8.4. The total number of detect- validity of the underlying admixture model (i.e., do presentday Indian zebu breeds and the GuineanN’Dama population able alleles increased dramatically to 200 when the serve as adequate surrogates for the original parentalpopula- other bovid samples were screened (American bison, tions?). European bison and banteng). When the allelic data is Hybrid comjlosition of individual animals: The population ge- compared among cattle populations, some interesting netic resolution that can heachieved using microsatellites has patterns emerge (Table 3). The mean number of al- been demonstrated with previous studies where phylogenetic analysis was carried outusing individuals as operational taxo- leles (MNA) observed over a range of loci for different nomic units (OTUs) ( BOWCOCKet al. 1994; ESTOUPet al. populations is considered to be a reasonable indicator 1995). It is therefore temptingto estimate the hybrid compo- of genetic variation with the provisos that the popula- 1076 D. E. MacHugh et al.

TABLE 3 Observed heterozygosities, gene diversitiesand mean number of alleles for 20 populations

GeographicalMNA MNA Breed origin n H” He sample)”(uniform(full sample)“

Aberdeen Angus Scotland 33 0.477 i- 0.019 0.470 2 0.054 3.7 3.5 Hereford England 34 0.465 i- 0.019 0.456 i- 0.064 3.8 3.6 Jersey Channel Islands 34 0.440 t 0.019 0.432 i- 0.059 3.4 3.1 Kerry Ireland 40 0.479 i- 0.018 0.473 5 0.044 3.4 3.1 - British Isles taurine Four breeds -141 0.466 i- 0.009 0.514 2 0.057 -5.2 Charolais France 36 0.525 i- 0.019 0.549 t 0.055 4.5 4.0 Friesian Netherlands 40 0.551 i- 0.018 0.545 i- 0.052 4.8 4.3 Simmental Switzerland 36 0.499 i- 0.019 0.480 i- 0.053 4.4 3.9 - Continental Europe taurine Three breeds -112 0.526 t 0.011 0.552 i- 0.054 -6.0 N’Dama Gambia 58 0.584 i- 0.014 0.583 t 0.049 5.6 4.7 N’Dama Guinea 63 0.513 t 0.014 0.503 i- 0.053 4.3 3.8 N’Dama Guinea Bissau 54 0.538 i- 0.015 0.534 t 0.054 4.7 3.9 N’Dama Nigeria 19 0.529 2 0.026 0.556 i- 0.050 4.8 4.8 N’Dama Senegal 48 0.532 i- 0.016 0.550 i- 0.053 5.4 4.5 - West African taurine Five populations -242 0.540 i- 0.007 0.549 -+- 0.051 -6.4 Butana North Sudan 24 0.652 i- 0.022 0.599 i- 0.036 4.3 4.1 Kenana South Sudan 38 0.643 2 0.017 0.619 i- 0.041 4.9 4.6 Gobra Senegal 59 0.633 i- 0.014 0.634 i- 0.045 5.9 4.8 Maure Mauritania 55 0.631 5 0.015 0.658 i- 0.040 6.2 5.3 White Fulani Nigeria 24 0.610 i- 0.022 0.636 i- 0.044 4.8 4.7 - African zebu Five breeds 200 0.634 5 0.008 0.652 i- 0.039 -7.0 Hariana North India 10 0.550 i- 0.035 0.548 t 0.048 4.2 - Sahiwal Pakistan 13 0.542 i- 0.031 0.569 i- 0.043 4.1 - Tharparker North India 10 0.635 i- 0.034 0.575 t 0.046 3.6 - Asian zebu Three breeds -33 0.573 t 0.019 0.588 i- 0.044 -5.2 4.8 Total 20 breeds 728 0.551 i- 0.004 0.643 i- 0.045 8.4 - Sample sizes (n)are the number of animals typed and pooled samples are underlined. Observed heterozygosities (H”) and gene diversities (HI)are shown with their respective standard errors. “The mean number of alleles (MNA) was calculated using all the animals in each population sampled. ’ Uniform population sizes were used for the second column of MNA values. A random sample of 20 individuals was used for each population except for the three Indian breeds and Nigerian the N’Dama. In thecase of the Indian breeds, a pooled sample of 20 different animals from the three breeds was used. In the case of the Nigerian N’Dama, the total sample of 19 animals was used. tions are at mutationdriftequilibrium and that sample to reflect distinctive population histories rather than a size is more or less the same for each population ( NEI sampling artifact. 1987).When the present datawere analyzed, a positive Among the different biogeographical groupings the correlation was observed between the MNA and the four breeds from the British Islesexhibited the greatest sample size of each population (Pearson product-mo- allelic paucity and the African zebu breeds displayed ment correlation = 0.547). Consequently, to remove the highest allelic diversity. The total pooled sample (n any sample bias, the MNA for a random sample of 20 = 200) for the five African zebu breeds possessed a animals was calculated for each population except the significantly higher MNA than the pooled sample ( n = Nigerian N’Dama and the three Indian zebu popula- 241) from seven European breeds (Wilcoxon’s signed tions. In the case of the Nigerian N’Dama, the original rank test, P = 0.020). Two breeds showed the lowest sample of 19 animals was used, and in the case of the number of detectable alleles with a mean of 3.4 per three Indian zebu breeds, a random pooled sample of locus (3.1 when n = 20). These were the Kerry breed 20 animals was used to represent a combined Indian from southwest Ireland and the Jersey breed from the zebu population. Themarked pattern of allelic diversity island of Jersey in the Channel Islands. This reduction based on geographic origin evident when all animals in genetic variation may be attributable to genetic isola- were typed for each population was maintained when tion, historical population bottlenecks or founder ef- only 19-20 animals were used for each population. fects. Kerry cattle are an endangered indigenous breed Therefore, the marked pattern of allelic diversity based that experienced asevere population bottleneck within on geographical origin shown in Table 3 is more likely the last 20 years ( CURRAN1990). The Jersey cattle M icrosatellite VariationMicrosatellite in Cattle 1077 breed originate from a small island and have existed Senegal and notin the breed’s country of origin (Table in effective genetic isolation because of strict breeding 1) . These animals may have been representatives of a practices for -200years (FRENCHet al. 1966). The number of small insular clusters. Either of these factors population with greatest allelic diversity was the Maure alone would tend to increase the likelihood of devia- breed from Senegal in West Africa with a MNA of 6.2 tions from HWE and taken together, they may account (5.3 when n = 20). This population, in common with for the observed disequilibrium. all African zebu breeds, has been influenced by histori- An exact testfmpopulation differentiation: When the test cal zebu-taurine crossbreeding and the high allelic di- for population differentiation was performed for the versity observed is undoubtedly anartifact of admixture 190 painvise population comparisons, the results were and the consequent input of both taurine and zebu essentially uninformative. In all but onecase, the proba- alleles. bilities were highly significant (P-=S 0.01 ) that the two Obseruedheterozygosities and genediversities: Table 3 populations being compared were differentiated. The shows mean heterozygosities computed across the 20 only two populations that showed a remote genetic af- loci for each breed and biogeographical grouping. The finity using this test were the N’Dama sampled in Nige- mean observed heterozygosities ranged from 0.440 2 ria and the N’Dama collected in Gambia (P= 0.036). 0.019 for the Jersey breed to 0.652 2 0.022 for the It is known that the Nigerian population was originally Butana breed. The overall pattern is similar to the mean derived from Gambian imports brought to Nigeria number of alleles;the populations from the British Isles within the last 20 years. This recent common identity displaying the lowest observed heterozygosities and the may explain the relatively high Pvalue observed for the African zebu breeds displaying the highest. The pooled comparison between these two populations. African zebu sample from five breeds possess a signifi- An evolutionary tree for domesticated cattle con- cantly higher observed heterozygosity than the pooled structed from microsatellite data: A neighbor-joining European sample consisting ofseven breeds (Wil- phylogeny was constructed using a matrix of hWge- coxon’s signed rank test, P = 0.002). The mean esti- netic distances among the 20 cattle populations and matedgene diversities ranged from 0.432 -+ 0.059, the three related species. However, the topology of this again for the Jersey breed, to 0.658 5 0.040 for the tree was heavily distorted by the inclusion of popula- Maure breed. The overall pattern concords with the tions known to be admixtures between zebu and taurine previous observations that,for the most part, cattle populations. Because hybrid populations are not dis- breeds from the British Isles showthe least genetic varia- crete evolutionary lineages, trees with admixed population and crossbred African zebu populations display tions violate the fundamental principles of phylogeny the most genetic variation. Again, the pooled European reconstruction ( FELSENSTEIN 1982; NEI 1987). sample displayed a significantly lower valuefor this mea- Figure 3 shows a neighbor-joining tree subsequently sure of diversity than the pooled African zebu sample constructed from a matrix of distances using only (Wilcoxon’s signed rank test, P = 0.024) . Dsw Perturbations”om Hardy- Weinberg equilibrium: Figure 2 12 non-admixed cattle populations and the three re- shows a diagrammatic summary matrix of the tests per- lated species. All of the African zebu breeds were ex- formed for deviations from HWE. Three tests were per- cluded for thepurposes of this tree. Also excluded were formed for each locus/population combination. In to- three of the N’Dama populations known to be cross- tal, 33 locus / population comparisons gave one or more bred from previous Ychromosome studies and the ad- significant result for each test for HWE. Of these, 15 mixture datapresented in alater section. Figure 3 were instances when all three tests concurred at the clearly illustrates the extent of the divergence between P I0.05 level, seven were instances when two tests the taurine and zebu clades. A 97% bootstrap probabil- concurred at P 5 0.05 level and 11 were cases where ity supports the separation of the Bos taurus and Bos only one test gave a significant result. Out of a total of indicus lineages. The separation between the European 1200 tests, 70 gavea significant result, which is not many and African taurine groupsis supported by a 95% boot- more than the 60 that would be expected by chance. strap value, indicating that these two groups are also However, when the results were pooled across loci and quite distinct evolutionary lineages. The internal struc- breeds using Fisher’s combined probability method, a ture of the European clade is not strongly supported small number of consistent deviations emerged. The by the bootstrap tests. This probably reflects a shallow Aberdeen Angus and the Maure breeds gave a signifi- depth for the radiation of European cattle and histori- cant result at the P 5 0.01 level for all three pooled cal admixture among thebreeds. Three of the microsat- tests. It is difficult to envisage factors that have caused ellites screened areknown to be tightly physicallylinked such significant genetic substructure in the Aberdeen on chromosome 23(BoLA DRZB, BoLA DRPl and Angus breed. However, the Maure breed was collected PRL, EGCENand FRIES 1995). Therefore an additional as two separate samples in different years (24 animals phylogenetic tree was constructed using the other 17 in 1992 and 31 animals in 1993). In addition, these markers plus the most polymorphic of the three linked cattle were sampled from a dwindling population in microsatellites (BoLA DRPl ) . The topology of this new 1078 D. E. MacHugh et al.

B $m -0

c5 -In 111

n I1 y‘i y

KL: POOLED (BREED) 111 r: 111 FIGURE2.--Summa1y matrix of three tests performed for HWE proportions for all 400 locus/population combinations. The results of the shuffled chi-square test are presented in the first cell of each comparison, the shuffled likelihood ratio (L) test results are shown in the second cell of each comparison and the Markov chain exact probability test results are shown in the third cell. A dark gray cell indicates a significant deviation at the P 5 0.05 level and a solid black cell indicates a significant result at the P I0.01 level. Pooled results across loci and across breeds were calculated using Fisher’s method for combining probabilities (see MATERIAIS AND METHODS). tree was identical to that shown in Figure 3 and the the microsatellite allele frequency distributions in 20 bootstrap values were very similar (data not shown). cattle populations. The first PC accounts for 52.4% of Estimation of timm of divergence: It has been shown that the variation and clearly distinguishes the taurine and DSw is approximately linear with respect to time zebu groups. The second PC, which also differentiates ( SHRIVERet al. 1995; L. JIN,unpublished data). Using the taurine and zebu groups, summarizes 14.4%of the the two bison species as an outgroup, it is therefore variation. The third PC describes 7.2% of the variation possible to obtain approximate estimates of the time of and clearly separates the European and African taurine divergence between Bos taurus and Bos indicus and also populations. In the three-dimensional space encom- between the two major types of taurine cattle. The esti- passed by the first three PCs, the African zebu popula- mated time since divergence between Bison and Bos is tions form an almost linear array stretching towards the 1.4 million years based on allozyme data ( HARTL et al. Asian zebu cluster at one end and to the five N’Dama 1988), or at least one million years based on paleonto- populations at the other end. Onclose inspection, the logical evidence (A. W. GENTRY,personal communica- position of the various African populations corresponds tion). Using this information and all the pairwise ge- to the relative proportion of zebu and taurine alleles netic distance values among thebison, taurine andzebu in their gene pools (see next section). This distinctive clades, a divergence time between Bos tuum and Bos pattern provides a clearer representation of the true indicw was extrapolated as between 610,000 to 850,000 evolutionary history of the admixed zebu populations years, depending onwhich estimate for the Bison / Bos than a phylogenetic reconstruction. split was employed. In a similar manner, thedivergence Zebu genetic introgression and admixture in African between the African and European taurine clade was cattlepopulations: Table 4 shows the percentage of estimated as between 180,000 to 250,000 years. zebu-specific alleles at 10 diagnostic loci in five African Principal component analysis:Figure 4 illustrates the zebu breeds, five African taurine populations and the first, second and third principal components (PC) for pooled Asian zebu sample. Also shown are the zebu V ariation in Cattle Microsatellitein Variation 1079

Aberdeen Angus FIGURE %-Neighbor- joining tree summarizing geneticdistances 32 Slmmentai 32 among 12non-admixed 46 L Friesian cattlepopulations and threerelated species. ,LI- Kerry Bootstrap values indicat- * 97 1 Charolais ing the degree of support N’Dama (Guinea Bissau) for each branch point are shown beside the node as N’Dama (Guinea) thepercentage of repli- r Tharparker cates in which the cluser I to the right of the node was recovered. The linear 53- Hariana scale relates thebranch lengthsto units of 4~. European Bison The root of the tree was American Bison placed at the midpoint of the longest branch sepa- ratingthe banteng from D,, units the other groups. 0.00 0.20 0.40 0.60 0.80 1.00 admixture proportions calculated using the gene iden- zebu cattle into East Africa and also the lack of zebu tity method. Itis interesting to note that thedistribution penetration into areaswhere tsetse-borne trypanosomi- of zebu alleles and the zebu admixture proportions de- asis is endemic ( EPSTEIN1971 ) . The correlation coeffi- cline from East to West Africa and then follow a steep cients calculated from the weighted least square solu- north-south gradient in West Africa. The five N’Dama tions for the admixture proportions are all highly sig- populations display varying degrees of zebu influence nificant, indicating that the three North Indian zebu depending on their proximity to the edge of the sub- breeds and the GuineanN’Dama population are appro- tropical Guinean biome, thearea with a high tsetse priate surrogates for the original source populations challenge. This pattern of hybridization and genetic that contributed to the genetic exchange between tau- exchange reflects the original historical influx of Asian rine and zebu cattle in Africa.

FIGURE 4.-Principal componentsdiagram showing the first three PCs from transformed allele frequencies in 20 cattlepopulations. The three-lettercode corresponds to the populations sampled as follows: ANG, Aberdeen Angus; HER, Hereford; JER, Jersey; KER, Kerry; CHA, Cham lais; FRI, Friesian; SIM, Simmental; NGA, N’Dama (Gambia) ; NGU, N’Dama (Guinea); NGB, N’Dama (Guinea Bissau) ; NNI, N’Dama (Nigeria) ; NSE, N’Dama (Senegal) ; BUT, Butana; KEN, Kenana; COB, Cobra; MAU, Maure; FUL, White Fulani; HAR, Hariana; SAH, Sahiwal; THA, Tharparker.

Ta urine populations Zebu populations@ populationsZebu Taurine 1080 D. E. MacHugh et al.

TABLE 4 Zebu-diagnostic alleles and zebu admixture proportionsin African and Asian populations

Breed GeographicalBreed Zebuorigin alleles” (%) Admixtureproportions (%)”

Asian zebu Three breeds 67.0 f 3.6 100‘ Kenana South Sudan 55.8 f 3.5 83.2 -+ 2.6 (0.988) Butana North Sudan 47.3 -+ 4.5 73.6 f 1.7 (0.925) White Fulani Nigeria 43.3 f 4.0 65.4 1- 1.8 (0.953) Maure Mauritania 39.7 f 2.9 61.3 f 1.8 (0.962) Gobra Senegal 36.4 f 2.7 59.8 -+ 1.3 (0.981) N’Dama Nigeria 8.7 f 2.8 14.8 f 1.5 (0.994) N’Dama Gambia 8.6 ? 1.6 18.9 f 1.6 (0.992) N’Dama Senegal 5.1 -+ 1.4 12.4 f 1.4 (0.995) N’Dama Guinea Bissau 1.5 f 0.7 6.9 f 2.0 (0.993) N’Dama Guinea 0.2 t 0.2 0‘ “The mean percentage of zebu-specific alleles across 10 diagnostic loci. Standard errors were computed as the standard error of a binomial distribution with sample size equal to the number of chromosomes tested (10 diagnostic loci, therefore: 20 chromosomes X population size). ’Admixture proportions calculatedfrom all allele frequenciesusing the gene identity method with standard errors calculated accordingto CHAKRA~ORTY(1985). Weighted least-squarescorrelation coefficients are shown in parentheses. ‘Source population.

A high-resolution map of zebu introgression in West 1990’s and these studies were generally broader in African cattle: Figure 5 shows a map that links the ap- scope, encompassing bothEuropean and non-Euro- proximate geographical location of each of 337animals peanpopulations (SUZUKIet al. 1993; BRADLEYet al. sampled in and around theSenegambia region of West 1994, 1996; LOETUSet al. 1994a,b). These reports have Africa to the proportion of zebu alleles present at 10 been based on analysis of mtDNA or Y-chromosome zebudiagnostic microsatellites in each individual ge- haplotypes and the presentstudy is the first comprehen- nome. Also shown for comparative purposes are the 38 sivesurvey of genetic variation at autosomal nuclear Kenana animals sampled in southern Sudan. The north- DNA markers in domesticated cattle. south gradient of zebu introgression running through Microsatellite variation and cattle genetic diver- the semi-desert Sudano-sahelian biome into the sub- sity: Striking patterns were evident in the distribution tropical Guinean region is clearly evident. The N’Dama and pattern of genetic variation observed among the population sampled at the edge of the tsetse zone in 20 cattle populations examined. Gambia have an average of 8.6% alleles that are zebu The seven European cattle breeds and particularly in origin. Only 12.1% ( 7 / 58) of these animals did not those from the British Isles displayed reduced levels of display any zebu alleles. This contrasts with a N’Dama genetic variation as estimated by allelic diversity, ob- population sampled 350 km further south, deepwithin served heterozygosity and genediversity. These observa- the equatorialtsetse zone. These GuineanN’Dama have tions may be a consequence of the recent evolutionary an average of0.2% zebu-specific alleles with 96.8% history of European cattle populations. Genetic isola- ( 61 / 63) of the cattle showing no trace whatsoever of tion and biological manipulation by humans have been zebu influence. Intermediate values were noted for the consistent features of the development of European two populations situated between these two extremes. breeds since the original demarcation of landraces into The Senegalese sample displayed 5.1% zebu-specific al- recognized breeds - 150 yearsago ( FELIUS1995). The leles with 41.7% (20/48) of the sample giving no indi- introduction of scientific breeding and reproductive cation of zebu ancestry. Cattle sampled in Guinea Bis- technology inthe 1950’s has accelerated these pro- sau showed 1.5% zebu-specific alleles and 72.2% (32/ cesses, and in many respects, European cattle popula- 54) of these animals did not betray any zebu influence. tions have a unique genetichistory that may have given rise to the attenuated patterns of genetic variation ob- DISCUSSION served in extant breeds. Half (10) of the microsatellite markers displayed Domesticated cattle have been the subject of numer- markedly different allelic distributions when analyzed ous surveys of genetic variation within and amongpopu- in taurine and zebu cattle populations. These 10 loci lations (for review see BAKERand MANWELL 1991 ) . Pre- showed a single allele or group of alleles that were vious studies have tended to focus on European breeds present athigh frequencies in Asian zebu breeds, inter- of cattle using allozyme or immunoproteintechniques. mediate frequencies in crossbred African zebu popula- The first DNA-based surveys did not appear until the tions, low frequencies in African taurine populations Microsatellite Variation in Cattle 1081

FIGURE 5.-Genetic introgression of zebu-specific alleles in West Afri- can cattle. Individual cattle are represented by circles (taurine animals) or squares (zebu animals).The colorcoded scale represents the proportion of zebu-specific alleles in percentage terms from an original scale of zero to 20. The percent- agebeside each population is the proportion of zebu-specific allelesin the population as a whole (Table 4). With the exception of the Maure, the position of each animal corresponds approximately to the location BanJuY where they were sampled. Although the Maure breed was actuallysam- pled in Senegal, they are recent migrants to this country and wererelo- catedto Mauritania for the purposes of this diagram. For comparative purposes, the graywindow shows the EastAfrican zebuKenana breed that I". was sampledin Sudan. .... 1 The green shaded area "_ denotes the arboreal tsetse-infested zone. and were either absent or present at very low frequen- Africa. Zebu populations sampled in the north of the cies in European taurine populations (Figure 1) . The region may have been influenced by early crossbreed- presence of thesezebu-specific or diagnosticalleles ing due to trade and migration between the Fertile underlines the substantial divergence betweenthe tau- Crescent and the Indian subcontinent (KULKE and rine and zebu genomes and supports the conclusions ROTHERMUND1990). On the other hand, zebu popula- of LOFTUSet al. ( 1994a,b) concerning the separate ori- tions originating further south would have been less gins of domesticated taurine and zebu cattle. exposed to taurine introgression from northern Asia Another aspect of the biological history of Indian and, atany particular zebudiagnostic locus, theyshould zebu cattle revealedby the distribution of these diagnos- display zebu-specific allele frequencies that are closer tic alleles isthe suggestion that the zebu breeds sampled to unity or fixation. Ongoing analysis of microsatellite in northern India mayhave been subject to limited variation in zebu breeds fromsouthern India supports genetic exchange and admixture with taurine cattle. this scenario and suggests that a cline of taurine intro- This hypothesis emerges because taurine-specific alleles gression may run from the north to the south of the equivalent to those present in zebu genomes were not Indian subcontinent (C.GALLIARD, personal communi- detected in this study. If the two groups had followed cation ) . mutuallyexclusive evolutionary pathways since com- Implications for the evolutionaryorigins of domesti- monancestry, symmetrical divergence would be ex- cated cattle: The most widespread view of the origins pected and European taurine should display high fre- of domesticated cattle is that the Bos tuum and Bos quency allelesthat are not present in pure Indian zebu indicus subspecies originate from the same early Neo- genomes. It is likely that these alleles do actually exist, lithicdomestication centers ( EPSTEIN1971; EPSTEIN but that taurine introgression hasintroduced them into and MA~oN1984; PAYNE1991 ) . The pervasive influence northern Indian cattle populations and masked their of this viewpoint is evidenced by the classification of exclusivity in this survey. In effect, the Indian subconti- zebu cattle as Bos tuurus in a standard text on mamma- nent may represent a mirror-image of the situation in lian taxa ( NOWAK1991 ) . 1082 D. E. MacHugh et nl.

Phylogenetic analysisof microsatellite polymor- genetic data provide some support for an autochtho- phisms in domesticated cattle has underlined the sub- nous domestication of Afiican cattle, genetic exchange stantial divergence between zebu and taurine breeds. between populations emerging from the Middle East Previous studies of allozyme and immunoprotein varia- and indigenous African aurochs cannot be discounted. tion have also indicated a large divergence between the Until unambiguous evidence for thedomestication pro- two cattle subspecies (BREND 1972; NAIK 1978; MAN- cess such as faunal shifts or clear size diminution is WELL and BAKER1980; VON GRAMLet al. 1986) . How- discovered, it is unlikely that this issue will be resolved ever, until mtDNA sequence data became available, no throughstandard archaeological methods. However, attempt was made to date the time of the original split. molecular studies, in the form of analysis of ancient Analysis of mtDNA control region sequence variation DNA from faunal remains may provide new insights using bison sequence as an outgroupsuggested that the into this problem. two subspecies diverged a minimum of 210,000 years Male-mediated zebu genetic introgression in African ago ( LOFTUSet al. 1994a; BRADLEYet al. 1996). This cattle populations: The unique history of African do- workmay be criticized on the basis that phylogenies mesticated cattle is reflected in both the morphology derived from a single genetic locus may not reflect or- and genetic composition of extant populations.For the ganismal evolution ( PAMILOand NEI 1988; BALL et al. most part, the original taurine substrate of African cat- 1990; WV 1991) . However, the total array of DNA-based tle has been displaced and absorbed by successive waves information supporting a deep split between the tau- of zebu migration from Asia and Arabia. This historical rine and zebu clades now includes mtDNA sequence or influx is evident in the cline of zebu genomic identity RFLP haplotypes from 130 cattle ( LOFTUSet al. 199413; observed running in aneast-west direction across Africa BRADLEYet al. 1996), Y-chromosome haplotypes in 184 and in anorth-south direction in WestAfrica (Ta- cattle (BRADLEYet al. 1994), and the allele frequency ble 4). variation at 20 independent autosomal microsatellites The genetic composition of African cattle popula- in 728 cattle reported here. A substantial body of ar- tions has been assayed using three different genomic chaeological evidence also exists that provides support systems. Mitochondrial studies of European, African for a separate origin of zebu cattle somewhere on the and Asian cattle have revealed two distinct mtDNA lin- Indiansubcontinent (ALLCHIN 1969; GRICSON1980; eages, one taurine and one zebu. These lineages were MEADOW1984; MEADOW1993). Thealternative hypoth- detected using bothcontrol region sequence differ- esis supporting a recent evolution of zebu cattle from ences and RFLP analysis of the whole mtDNA molecule domesticated taurine ancestral populations now seems ( SUZUKIet al. 1993; LOFTUSet al. 1994a,b; BRADLEYet untenable. One problem remains however, the discrep- al. 1996).In total, mtDNA haplotypes (mitotypes) have ancy between the microsatellite-derived divergence been characterized from 90 Africananimals ( 35 taurine time and thedate estimated from mtDNA sequence and 55 zebu) originating from both West and East Af- evolution (BRADLEYet al. 1996) . The microsatellite esti- rica. Without exception, all of these animals possessed mate (610,000-850,000 years ago) is approximately a taurine mitotype. In contrast to this, analysis of the Y three times larger than the upper bound of the date chromosome in African populations using a bovine Y- derived from mtDNA sequence analysis (275,000) . This specific DNA probe (btDYZ-1) and a Fspecific RAPD disparity may reflect differences in evolutionary dynam- marker (IL065) has shown a markedly different pat- ics or, more plausibly, may result from the potentially tern ( BKADLEYet al. 1994; TEALEet al. 1995). A zebu Y- large errors inherent in these estimates. chromosome haplotype has become widespread Recent mtDNA analysishas tentatively suggested that throughout Africa and has reached high frequencies African and European taurine cattle may have also had in breeds and populations nominally classified as pure different domestic origins (BRADLEYet al. 1996) . Analy- taurine. These asymmetrical distributions of the unipa- sis of microsatellite variation in European and African rentalinherited genetic systems, highlight the male- taurine cattle also provides some support for this inter- mediated natureof zebu gene flow on the African conti- pretation. In particular, the tree constructed from nent. The autosomal genomic composition of African genetic distances indicates that the African and Euro- cattle populations presents an intermediate picture be- pean taurine gene pools diverged between 150,000 and tween the two uniparental extremes and represents the 250,000 yearsago. Resolution of the microsatellite varia- overall penetration of the zebu genome most accu- tion into principal components also emphasizes a clear rately. Figure 6 shows a summary diagram of zebu ge- separation between the African and European taurine netic introgression for the three systems. groups. Although they lack zebu mtDNA, as shown in Table A number of workers have presented archaeological 4 and Figure 6, the two East African zebu populations evidence for the presence of domesticated cattle in that were surveyed for microsatellite variation (Butana North Africa during the early stages of the Neolithic and Kenana) have predominantly zebu autosomal ge- period (CARTER andCLARK 1976; GAUTIER1987; GRIC nomes with admixture proportions of 73.6 ? 1.7% and SON 1991; WENDORFand SCHILD 1994).Although the 83.2+2.6%, respectively. They also display zebu Y chro- Microsatellite Variation in Cattle 1085

FIGURE&-The introgression of three different zebu genomic comptr nents into African cattlr. Sample size is proportional to the arca of each piecircle as indicated on each of thethree dia- grams. The distribution of taurineand zebu mitotypes are shown in A. Zebu admixture proportions are shown in R. The black portion of each circle represents zebu admixture proportions from Table 4 and thearea of each circle is proportional to sample size. The introgression of the zebu Y Gambiao' . ., chromosome is shown in '1 C. Again, the black por- GuineaBissau ' ' ' ' tion of the circles corresponds to the proportion of zebu Ychromosomes in the population. Data for A and C were taken from SLT- .??.&S. @ = 50 animals I Tsetsezone (N'Dama shown by country) ZL'KI Pt fll. ( 1995), BRAD- LEY et nl. (1994), LO~VS ~t nl. (1994b) and TEAI.E et nl. ( 1995). Populations marked with an asterisk (*) were not analyzed in our laboratory. Sample sizes arenot givenin TEAI.Eel nl. (1995) and for the purposes of C an arbitrary size of 20 is as- sumed for the Roran and KenyanSahiwal populations.

@ = 20 Y-chromosomes +?eTsetse zone (N'Dama shown by country) mosomes. The West African zebu breeds show a similar all 28 males screened with the btDYZ-1 Y-specific probe pattern, although with less zebu influence on theirau- yielded zebu haplotypes (Figure 6 and BRADLEYd al. tosomal genomes. The most striking result is that oh 1994). This result dramatically illustrates the asymmet- tained for the Gambian N'Dama. This population is rical nature of zebu gene flow, highlighting the genetic morphologically taurineand was sampled onthe inertia of theAfrican taurine mtDNA gene pool and the fringes of the tsetse zone near the Gambia river. Their rapid dispersal of the zebu Y chromosome. It suggests a autosomal genome is composed of 8.6 t 1.6% zebu mechanistic pattern forzebu genetic introgression that alleles with a zebu admixture proportionof almost 20% probably applies to the whole of Africa. (18.9 t 1.6%).In common with all the other African There are number of factors that could account for populations, they possess no zebu mitotypes. However, the unusual dynamics of zebu gene flow in African cat- 1084 D. E. MacHugh et al. tle. However, it seems clear that the genetic exchange were taken from these mountains and may represent the between zebu bulls and taurine females may have been last remnants of pure N’Dama left in West Africa. It is particularly important. The unusual population smc- therefore imperative that this population is targeted in ture of domesticated cattle with consequent dispropor- future conservation efforts. tionate contributions to future generations from indi- The systematicclassification of cattle genetic re- vidual males may have also contributed to the observed sources in Africa has been actively researched for a long pattern. The Arab peoples who brought zebu cattle to period of time and a broad consensus has yet to be Africa commencing about 700 AD may, for economic reached (MASON 1951; MASON and MAULE 1960; PAYNE or logistical reasons, have broughtdisproportionate 1970; EPSTEIN1971; BAKERand MANWELL 1980; MAULE numbers of males. Zebu cattle would have been prized 1990; GRIGSON1991). Previous commentators have for their ability to withstand arid conditions and bulls beenpreoccupied with traits of dubious systematic in particular may have been particularly valuable. A value such as horn shape, skull morphology and the single zebu bull could have been crossed with a herd location of the hump in zebu/ taurine crossbreds. The of taurine females to produce an arid-resistant hybrid molecular data demonstrate that many of the different FI population that possessed no zebu mitotypes and physical characteristics of African cattle (for example, males fixed for the zebu Y chromosome. Under these thoracic us. cervicethoracic humps) are probably just circumstances, the composition of the African gene manifestations of the variable penetrance of admixed pool may have been transformed very rapidly and nu- zebu and taurine morphological traits. clear genomes could have become progressively more Based on analysis of mtDNA, no African zebu breed zebu without any appreciable input of zebu mitotypes. should be classified as true zebu. The retention of tau- Another factor that may have hastened the spread of rine mitotypes is de facto evidence for some input of the zebu nuclear genome was the periodic rinderpest taurine genes during the developmentof these breeds. epizootics that were introduced from Asiawith zebu It may actually be more appropriate to adopt the term cattle ( EPSTEIN1971 ) . Through longer exposureto the “zeboid,” first suggested by MASON ( 1951 ) . Likewise, disease, zebu cattle are more resistant to the effects of many taurinepopulations should not be considered rinderpest than taurine cattle ( PAYNE1970). Themost purebred based on the infiltration of zebu Y chromo- destructive rinderpest outbreak to happen within the somes and autosomal zebu-specific alleles. Accurate es- last 200 years was that of 1890-91. During this period, timates of admixture in African cattle would therefore over 80% of the livestock of East Africa weredestroyed seem to provide the most rational biological framework and many wild species were also decimated ( EPSTEIN for classification of African cattle populations. A set of 1971 ) . The majority of the cattle left after this disease zebu-diagnostic microsatellites could be used to disen- episode had zebu ancestry and the last vestiges of pure tangle and simplify classification systems. The propor- East African taurine cattle were practically eradicated. tion of zebu admixture could be correlated with pro- During the time since the large-scale influx of zebu duction and fitness-related characteristics such as fertil- cattle about 1300 years ago, there would have been a number of rinderpest outbreaks and it is possible to ity, feed-intake and conversion, growth rate, milk envisage a precipitous decline in taurinepopulation production, water requirements, disease resistance and numbers during each epizootic. mortality. The relative ease with which microsatellites Implications for genetic conservationand breed char- may be screened in a large number of samples makes acterization: The introgression of the zebu genome into this approach particularly attractive. trypanotolerant populations of taurine N’Dama cattle Conclusions: This study demonstrates the resolution shown in Figure 5 illustrates the precarious state of some that can be achieved using microsatellite polymor- of these populations. The sharp cline in zebu-specific phisms for studies of genetic microdifferentiation. The alleles at the tsetse zone transition reflects the disease primary outcome was the affirmation of the large diver- challenge faced by zebu animals inhabiting the equate gence between the Bos taurus and Bos indicus lineages rial region. It also suggests that the introductionof zebu previously reportedfor allozyme and mitochondrial alleles into the southernforested areas must be predomi- data ( MANWELL and BAKER1980; LOFTUSet al. 1994a,b; nantly through admixed taurine animals. The effects of BRADLEYet al. 1996). This provides strong support for deforestation, desertification and the subsequent migra- the dualdomestication theory proposed by LOFTUSet tion of zebu populations are apparent. If tsetse densities al. (1994a). The discordance observed among three decrease because of destruction of the species’ natural different geneticsystems in African cattle highlights the habitat, even populations such as the Guinean N’Dama potential for misinterpretationif mtDNA is used as the would quickly succumb to genetic absorption and intre sole source of data when investigating phylogeographic gression from zebu populations moving into deforested processes. Important sex-related demographic patterns areas. The mountainous Fouta Djallon region in Guinea may be overlooked unless a broad approach, encom- is reputed to be the area where the N’Dama breed origi- passing mtDNA,Ychromosome polymorphisms and au- nate ( FELIUS1995). The Guinean population surveyed tosomal variation is used. M icrosatellite VariationMicrosatellite in Cattle 1085

We acknowledge the assistance of the following people with sample morphisms in world populations. Am. J. Hum. Genet. 56: 461- collection and data analysis: A. M. BADI, D. S. BALAIN, M. DIALLO,H. 474. EDWARDS,A., H. A. HAMMOND,L. JIN, C. T. &KEYand R. CHAKRA- DJATA, A. FAYE, B. FYE, A. GUEYE, A. HASSAN,L. JIN, A. T. LLOYD,C. BORTY, 1992 Genetic variation at five trimeric and tetrameric MEGHEN,M. MURRAY, L. 0. NGERE, NORRIS,T. OTA, RIZGALLA, J. J. tandem repeat loci in four human populationgroups. Genomics V. J. SHANKER,P. M. SHARP, B. SAUVEROCHE,R. S. SOW,M. J. STEAR, 12: 241-253. C. S. TROY,H. WAGNER,T. WILSON, Y. ZHONG.This work was funded EGGEN,A,, and R. FRIES,1995 An integrated cytogenetic and meiotic by the European Commission DGXll Science and Technology in map of the bovine genome. Anim. Genet. 26 215-236. Developing Countries Program. D.E.M. is supported by a Research EPSTEIN,H., 1971 The Origin of theDomestic Animals of Afnca. Afri- Fellowship in Bioarchaeology from the Wellcome Trust. M.D.S.is cans, New York. supported by a Fellowship from the W. M. Keck Foundation for Ad- EPSTEIN,H., and I. L. MASON, 1984 Cattle, pp. 6-27 in Evolution of vanced Training in Computational Biology and a grant from the Domesticated Animals, edited by I. L. MASON.Longman, London. National Institute of Justice. ESTOUP,A.,L. GARNERY,M. SoLIcNAcandJ-M.CORNUET, 1995 Micro- satellite variation in honey bee (Apis mellijkra L.) populations: hierarchical genetic structure and test of the infinite allele and stepwise mutation models. Genetics 140: 679-695. LITERATURECITED FELIUS,M., 1995 CattleBreeds-An Encyclopedia. Misset, Doetinchem, The Netherlands. ALLCHIN,F. R., 1969 Early domestic animals in India and Pakistan, FEISENSTEIN,J., 1982 How can we infer geography and history from pp. 317-322 in The Domestication and Exploitation of Plants and gene frequencies? J. Theor. Biol. 96 9-20. Animals, edited by P. J. UCKOand G. W. DIMBLEBY.Duckworth, FEISENSTEIN,J., 1985 Confidence limits on phylogenies: an ap London. proach using the bootstrap. Evolution 29: 783-791. ANDERSON, L., J. BOHME,L. RASK and P. A. PETERSON, 1986 Geno- FRENCH,M. H., I. JOHANSSON,N. R. JOSHI and E. A. MC~UGHI.IN, mic hybridization of bovine class I1 major histocompatibility 1966 European Breeds of Cattle. Food and Agricultural Organiza- genes: 1. Extensive polymorphism of DQaand DQO genes. tion of the United Nations, Rome. Anim. Genet. 17: 95-112. GAUTIER,A,, 1987 Prehistoric menand cattle in North Africa: a BAKER,C. M.A., and C. MANWELL,1980 Chemical classification of dearth of data and a surfeit of models, pp. 163-187 in Prehistoly cattle. 1. Breed groups. him. Blood. Grps. Biochem. Genet. 11: of Arid North Afnca: Essays in Honor of Fred Wmdorf; edited by A. E. 127-150. CLOSE.Southern Methodist University Press, Dallas. BAKER,C. M. A,, and C. MANWELL,1991 Population genetics, molec- GOLDSTEIN,D. B., A. RUIZLINARES, L. L. CAVAIJJ-SFORZAand M. W. ular markers and gene conservation of bovine breeds, pp. 221- FELDMAN,1995a Genetic absolute dating based on microsatel- 304 in Cattle Genetic Resources, edited by C. G. HICKMAN.Elsevier, lites and the origin of modern humans. Proc. Natl. Acad. Sci. Amsterdam. USA 92: 6723-6727. BALL,R. M., J. E. NEIGELand J. C. AVISE, 1990 Gene genealogies GOLDSTEIN,D. B., A. RUIZLINARES, L. L. CAVAI.I.I-~FORZAM. and W. within the organismal pedigrees of random-mating populations. FELDMAN,199513 An evaluation of genetic distances for use with Evolution 44 360-370. microsatellite loci. Genetics 139: 463-471. BOWCOCK,A. M., A. RUIZ-LINARES,J. TOMFOHRDE, E. MINCH,J. R. GRIGSON,C., 1980 The craniology and relationships of four species KIDD et al., 1994 High resolution of human evolutionary trees of Bos. 5. Bos indicus L. J. Arch. Sci. 7: 3-32. with polymorphic microsatellites. Nature 368: 455-457. GRIGSON,C., 1991 An African origin for African cattle?-some ar- BRADLEY,D. G., D. E. MACHUGH, R. T. LOFTUS,R. S.SOW, C. H. chaeological evidence. Afric. Archaeol. Rev. 9: 119-144. HOSTE et al., 1994 Zebu-taurine variation in Y chromosomal HAMMOND, A,,H. L. JIN, Y. ZHONG, C. T. CASKEY and R. CHAKRA- DNA a sensitive assay for genetic introgression in West African BORTY, 1994 Evaluation of 13 short tandem repeat loci for use trypanotolerant cattle populations. Anim. Genet. 25: 7-12. in personal identification applications. Am. J. Hum. Genet. 55: BRADLLY,D. G., D. E. MACHUGH, P. CUNNINGHAM andR. T. LOFTUS, 175-189. 1996 Mitochondrial diversity andthe originsof African and HARTL,G. B., R. GOLTENBOTH,M. GRILLITSCHand R. WILLING,1988 European cattle. Proc. Natl. Acad. Sci. USA 93 5131-5135. On the biochemical systematics of the Bovini. Biochem. Syst. BREND, M., 1972 Studies on the relationships between cattle breeds Ecol. 16: 575-579. in Africa, Asia and Europe:evidence obtained by studies of blood KULKE,H., and D. ROTHERMUND,1990 A Histmy of India. Routledge, groups and protein polymorphisms. World Rev. Anim. Prod. 8: London. 9-14. LHOSTE,P., 1991 Cattle genetic resources of West Africa, pp. 73- CARTER,P. L., and J. D. CLARK,1976 Adrar Bous and African cattle, 89 in Cattle Genetic Resources, edited by C. G. HICKMAN.Elsevier, pp. 487-493 in Proceedingsof 7th Pan-African Congress on Amsterdam. Prehistory, Addis Ababa, 1971. LOFTUS, R. T., D. E. MACHUGH, D. G. BRADLEY, M. P. SHARPand E. P. CAVALLI~FORZA,L. L., P. MENOZZIand A.PIAZZA, 1994 The Histoly CUNNINGHAM,1994a Evidence for two independent domestica- and Geography of Human Genes. Princeton University Press, tions of cattle. Proc. Natl. Acad. Sci. USA 91: 2757-2761. Princeton, NJ. LOITUS, R. T., D. E. MACHUGH,L. 0. NGERE, D. S. BUN, A. M. BAD[ CHAKRABORTY,R., 1975 Estimation of raceadmixture-a new et al., 1994b Mitochondrial genetic variation in European, Afri- method. Am. J. Phys. Anthropol. 42: 507-511. can and Indian cattle populations. Anim. Genet. 25 265-271. CHAKRABORTY,R., 1985 Gene identity in racial hybrids and estima- MACHUGH,D. E., R. T. LOFTUS,D. G. BRADLEY,P. M. SHARP and E. P. tion ofadmixture rates, pp. 171-180 in Genetic Microdiffereiation CUNNINGHAM,1994 Microsatellite DNA variation within and in Man and Other Animals, edited by J. V. NEELand Y. AHUJA. among European cattle breeds. Proc. R. SOC.Lond. Ser. B 256: Indian Anthropological Assoc., Delhi, India. 25-31. CHAKRABORTY,R., 1986 Geneadmixture in human populations: MANWELL,C., and C. M.A. BAKER,1980 Chemical classification of models and predictions. Yearbook Phys. Anthropology 29 1-43. cattle. 2. Phylogenetic tree and specific status of the Zebu. Anim. CHAKRABORTY,R., M. FORNAGE,R. GUEGUE andE. BOERWINKLE,1991 Blood Grps Biochem. Genet. 11: 151-162. Population genetics of hypervariable loci: analysis of PCR-based MASON, I. L., 1951 The Chsification of West Afncan Livestock. Com- VNTR polymorphism within a population, pp. 127-143 in DNA monwealth Bureau of Animal Breeding, Edinburgh. Finge@inting: Approaches and Applications, edited by T. BURKE,G. MASON, I. L., and J. P. MAULE,1960 TheIndigenous Livestock ofEastem DOU, A. J. JEFFREYS and R. WOLFF.Birkhsuser, Basel. and Southern Afnca. Commonwealth Bureau of Animal Breeding, CHAKRABORTY, R., M. I. KAMBoH, M. NWANKWOand R.E. FERRELL, Edinburgh. 1992 Caucasian genes in American Blacks:new data. Am. J. MAULE, J. P., 1990 The Cattle of the Tropics. University of Edinburgh, Hum. Genet. 50: 145-155. Edinburgh. CURRAN, P. L., 1990 Keny and Dexter Cattle and Other Ancient Irish MEADOW, R. H., 1984 Animal domestication in the Middle East: a Breeds: A Histq. The Royal Dublin Society, Dublin, Ireland. view from theEastern margin,pp. 309-337 in Animals andArchm- DEKA,R., L. JIN, M.D. SHRIVER,L. M. Yu, S. DECROOet al., 1995 ology: 3. Early Herah and TheirFZocks, edited by J. CISITTON-BROCK Population genetics of dinucleotide (dC-dA); (dGdT). poly- and C. GRIGSON.Oxford University Press, Oxford. 1086 D. E. MacHugh et al.

MFADOW,R. H., 1993 Animal domestication in the Middle East: a SCHEW,B. D., 1995 World Watch List for Domestic Animal Diversily. revised view from the Eastern Margin, pp. 295-320 in Harappan Food and AgricultureOrganization of theUnited Nations, Civilizc~tion,edited by G. POSSEHI..Oxford & IBH, New Delhi, Rome. India. SHRIWR,M. D., I,. JIN,R. CHAKRABORTYand E. BOEKWINKI.E,1993 NAIK, S. N., 1978 Origin and domestication of zebu cattle (Bos in& VNTR allele frequency distributions under the Stepwise Muta- ms).J. Hum. Evol. 7: 23-30. tion Model: a computer simulation approach. Genetics 134: 983- NEEI.,J. V., 1973 “Private” genetic variants and thefrequency of 993. mutation among SouthAmerican Indians. Proc. Natl. Acad. Sci. SHKIVEK, M. D., L. JlN, E. BOERM~INKI.E,R. DBKA, R. E. FER REI.^. Pi d., USA 70: 3311-3315. 1995 A novel measlire of genetic distance for highly polymor- NEI, M., 1972 Genetic distance between populations. Am. Nat. 106: phic tandem repeat loci. Mol. Biol. Evol. 12: 914-920. 283-292. SUZIW,R., S.J. KEMP and A.J. TEKE, I993 Polymerase chain reac- NEI, M., 1987 Molecular Evolutionaq Genetirs. Columbia University tion analysis of mitochondrial DNA polymorphism in N’Dama Press, New York. and Zebu cattle. him. Genet. 24: 339-343. NOWAR,R. M., 1991 Walker’s Mammals of the World. The John Hop- Twl.~,A. J.,J. WAMHI~(;A,P. S. GWAKISA,G. STRGWNGER,D. BKAI)I.EY kins University Press, Baltimore. rt cd., 1995 A polymorphism in randomly amplified DNA that Pnwr.o, P., and M. NEI, 1988 Relationships between gene trees and differentiates the Y chromosomes of Hos zndicu.r and Bos tnunc.~ species trees. Mol. Bid. Evol. 5: 568-583. him. Genet. 26: 243-248. VAI.I)ES,A. M., M. SIATK~Nand N. B. FR€:IMEK, 1993 Akle frequen- PA~E,W.J. A., 1970 Cnttk Production zrz ihr Tropics. Longman, Lon- don. cies at microsatellite loci: the Stepwise Model revisited. Genetics PAWE,W. J. A,, 1991 Domestication: a fonvard step in civilization, 133: 737-749. VON GKAMI., R., G. OHMWER, F. PIRcHNER, L. ERHAKD,J. BUCHBEK(;F.K pp. 51-72 in Cattle Gnrtir Resourres, edited by C. G. HICKMAN. rt nl,, 1986 Biochemical polymorphism in Egyptian Baladi cattle Elsevier, Amsterdam. and their relationship with other breeds. Anim. Genet. 17: 61- FL\woru~,, M., and F. ROLXET, 1995a GENEPOP (Version 1.2): 76. population genetics software for exact tests and ecumenicism. J. WENDOW, F., and R. SCHII.l), 1994 Are the early Holocene cattle in Hered. 86: 248-249. the Eastern Sahara domestic or wild!Evol. Anthropol. 3: 118- RAYMOND, M., and F. Rouss~I,1995b An exact test for population 128. differentiation. Evolution 49: 1280-1285. WORKMAN,P. L., B. S. BILMBERGand A.J. CM)I%R, 1963 Selection, SAITOU, N., and M. NE!, 1987 The neighbor-joining method: a new gene migration and polymorphic stability in US White and Negro method for reconstructing phylogenetic trees. Mol. Biol. Evol. populations. Am. J. Hum. Genet. 15: 429-437. 4: 406-425. WL, GI., 1991 Inferences of species phylogeny in relation to segrc- SAMRROOK,,J,, E. F. FRITSCHand T. MANIATIS, 1989 Molecular Clon- gation of ancient polymorphisms. Genetics 127: 429-435. ing: A Labomtor?;Manual. Cold Spring Harbor Lahordtory Press, Cold Spring Harbor, NY. Communicating editor: M. SlAl hlN