Genetic analysis reveals the wild ancestors
of the llama and the alpaca

Miranda Kadwell(1), Matilde Fernandez(2), Helen F. Stanley(1), Ricardo Baldi(3),
Jane C. Wheeler(4,2)*, Raul Rosadio(2) and Michael W. Bruford(5,1)**

(1) Institute of Zoology, Regent’s Park, London NW1 4RY, UK
(2) Facultad de Medicina Veterinaria, Universidad Nacional Mayor de San Marcos, Apartado 41- 0068, Lima 41, Peru
(3) Centro Nacional Patagonico, Consejo Nacional de Investigacions Cienti¢cas yTechnologicas, 9120 Puerto Madryn, Argentina
(4) CONOPA, Los Cerezos 106, Salamanca-Lima 3, Peru
(5) Biodiversity and Ecological Processes Group, Cardi¡ School of Biosciences, Cathays Park, Cardi¡ CF10 3TL, UK
* Author for correspondence for camelids (webmaster@conopa.org).
** Author for correspondence for genetic analysis (brufordmw@cardi¡.ac.uk).

The origins of South America’s domestic alpaca and llama remain controversial due to hybridization, near extirpation during the Spanish conquest and di¤culties in archaeological interpretation. Traditionally, the ancestry of both forms is attributed to the guanaco, while the vicu·a is assumed never to have been domesticated. Recent research has, however, linked the alpaca to the vicu·a, dating domestication to 6000^7000 years before present in the Peruvian Andes. Here, we examine in detail the genetic relationships between the South American camelids in order to determine the origins of the domestic forms, using mitochondrial (mt) and microsatellite DNA. MtDNA analysis places 80% of llama and alpaca sequences in the guanaco lineage, with those possessing vicu·a mtDNA being nearly all alpaca or alpaca^vicu·a hybrids. We also examined four microsatellites in wild known-provenance vicuña and guanaco, including two loci with non-overlapping allele size ranges in the wild species. In contrast to the mtDNA, these markers show high genetic similarity between alpaca and vicu·a, and between llama and
guanaco, although bidirectional hybridization is also revealed. Finally, combined marker analysis on a subset of samples con¢rms the microsatellite interpretation and suggests that the alpaca is descended from the vicu·a, and should be reclassi¢ed as Vicugna pacos. This result has major implications for the future management of wild and domestic camelids in South America.

Keywords: domestication; alpaca; guanaco; llama; vicuña; mitochondrial DNA

1. INTRODUCTION

Four South American camelids are recognized today, two of which are wild species, the guanaco (Lama guanicoe, MÏller 1776) and the vicuña (Vicugna vicugna, Molina 1782), and two of which are domestic forms, the alpaca (Lama pacos L.) and the llama (Lama glama L.), whose evolutionary origins are debated (Wheeler 1995). The guanaco and vicuña diverged from a common ancestor around two million years ago, and are the only representatives of the lamini to survive the Pleistocene period (Stanley et al. 1994; Wheeler 1995).

Archaeozoological evidence from the central Peruvian Andes links alpaca origins to the vicuña at 6000^7000 years before present (Wheeler 1995). Because all potential ancestral forms are
extant, South American camelid domestication represents an unusual and useful opportunity to gain insight into the origin and biodiversity of domesticated animals, an issue which is of increasing interest due to the recognized potential economic bene¢ts of indigenous genetic
resources and the threats that face marginal and extensive agriculture today (Hall & Bradley 1995). In contrast with many other domestic farm animals, there is no written
history associated with the llama and alpaca.

Orally transmitted herding knowledge was largely lost during the Spanish conquest, and breeds disappeared as both the human and the native domestic live stock populations were reduced by 80^90% during the ¢rst 100 years of contact (Wheeler 1995). At present, although llama and
alpaca rearing is a central element of the economy in the high Andes, it is often not pro¢table due to the poor quality of the animals and their fibre. The reconstruction of fine-fibre breeds and the breeding strategies needed are therefore uniquely dependent upon the contributions of
archaeozoology and genetic analysis (Wheeler et al. 1995).

Understanding the domestication of the llama and alpaca using morphological analyses has been hampered by the lack of species-specifc post-cranial skeletal characteristics, and because archaeological remains are often in poor condition (Bahn 1994). Fibre analysis has been
pursued at sites where £eece has been preserved (Dransart 1991a; Reigadas 1993; Wheeler et al. 1995), but the decimation of traditional llama and alpaca breeds at the time of the conquest, and probable subsequent hybridization among the survivors, make inference from phenotypic characters problematic.

Molecular genetic analysis using methods such as mitochondrial DNA (mtDNA) sequencing and microsatellite and Y-chromosome analysis has recently proved illuminating when examining the origins of modern domestic livestock, even in the face of often overwhelming recent changes in domestic populations due to genetic drift, selection and/or hybridization (e.g. Bradley et al. 1996;
MacHugh et al. 1997; Hiendleder et al. 1998; Luikart et al. 2001; Vila et al. 2001). Here, we apply mitochondrial cytochrome b sequence and microsatellite typing to the question of the evolutionary relationships between and domestication of South American camelids throughout
their range in South America today.

2. MATERIAL AND METHODS

(a) Samples
Sample collection sites span as far as possible the geographical range of the two wild species (see ¢gure 1). Our sample comprises: two vicu·a subspecies (V. v. vicugna and V. v. mensalis) from Argentina, Chile and Peru, guanaco (L. g. guanicoe and L. g. cacsiliensis) from Argentina and Peru, llama (a range of morphological types) from Argentina, Bolivia and Peru, alpaca (including `suri’ and `huacaya’ £eece types) from Argentina, Bolivia, Chile and Peru, and 10 known hybrids (seven between llama and alpaca (locally known as wari) and three between alpaca and vicu·a (locally known as pacovicu·a). Samples were taken from only those individuals whose
phenotype conformed to accepted morphological criteria for domestic forms. The localities and genetic data for each individual included in this study can be found at http://www.cardi¡.
ac.uk/biosi/research/biodiversity/sta¡/mb.html.

(b) MtDNA analysis
The phylogenetic afiliations of the llama and alpaca were¢rst confirmed by sequencing a short (158 bp) but highly informative region of the cytochrome b gene of the mitochondrial genome sensu Stanley et al. (1994) (GenBank accession numbers U06425^U06430). In total, 211 South American camelids were analysed from throughout the geographical range (comprising 21 guanaco, 42 vicuña, 54 llama, 84 alpaca and the 10 hybrids described in table 1). One Arabian camel was also analysed.

Briefly, DNA was extracted from blood or skin using standard proteinase-K digestion followed by organic extraction using phenol and phenol^chloroform, and total DNA was precipitated
in 100% ethanol (Stanley et al. 1994; Bruford et al. 1998). DNA samples were stored in TE bu¡er (10mM Tris-HCl, 1mM EDTA, pH 8.0). The cytochrome b primers L14724 and H14900 were used for polymerase chain reaction (PCR), which was carried out as in Stanley et al. (1994). PCR products were purified and DNA sequencing was carried out as previously described (Stanley et al. 1994).

Sequences were aligned manually, and unique haplotypes have been deposited in GenBank under accession numbers AF373809^AF373833. Mitochondrial haplotype divergence and
frequencies were analysed for the guanaco, vicu·a, llama and alpaca samples as previously described (Stanley et al. 1994), and a minimum-spanning network (Kruskal 1956; Bandelt at al. 1999) was generated using the program MINSPNET (Exco¤er 1993;¢gure 2). The distribution patterns of domestic South American camelid haplotypes were then compared with those of the wild South American camelid sample.

(c) Microsatellite analysis
Since the strict maternal inheritance of mtDNA in most mammals restricts its use in studies of hybridization, especially in domestic livestock (e.g. MacHugh et al. 1997), we also applied
nuclear DNA markers. Four microsatellite loci (YWLL 38, YWLL 43,YWLL 46 and LCA 19; Lang et al. 1996; Penedo et al. 1998) were typed for 669-771 individuals, including the 211 individuals for cytochrome b (table 1).

The genetic distances between the four South American camelid taxa were measured in three ways: first, by using 1-(proportion of shared alleles); second, by using Reynold’s distance (Reynolds et al. 1983), a measure commonly used in livestock analysis where genetic drift
has a major impact on allele frequencies; and finally, by using δμ2 (Goldstein et al. 1995), which calculates population subdivision using allele size di¡erences under a stepwise mutation
model. All distances were estimated using the program MICROSAT v. 1.5d (Minch 1999).

Factorial correspondence analysis was then performed on pairwise allele frequency di¡erences using GENETIX v. 4 (Belkhir 1999). Here, the genetic di¡erentation between populations (in
this case vicu·a, guanaco, llama and alpaca) is expressed as factors, which explain the correspondence between samples in a number of dimensions (Benzécri 1973).

Thus, the relationships among populations can be judged by examining how individuals
from each population cluster in two or three dimensions.

(d) Combined analysis
To assess the extent of introgression in llama and alpaca populations, we used both the microsatellite and the mitochondrial data to calculate the admixture proportions mC and mY
(mC: estimator based on allele frequencies, Chakraborty et al. (1992); mY: estimator based on the number of substitutions for mtDNA and squared allele size di¡erence for microsatellite
data, Bertorelle & Exco¤er (1998)) for alpaca and llama compared with the vicu·a and guanaco. This was implemented using the program ADMIX 1__0 (Bertorelle 1998).

Finally, to examine the concordance between mitochondrial and microsatellite data and patterns of introgression in more detail, we returned to the 211 individuals typed for both mtDNA and microsatellites. Genotypes were coded `V’ (vicuña) or `G’ (guanaco) for mtDNA and `V’, `G’ or `H’ (hybrid) for YWLL 46 and LCA 19, depending on their allele sizes with reference to the guanaco and vicuña ranges, and we examined the data for each locus separately and combined.

3. RESULTS

(a) Mt DNA
We found 26 unique haplotypes within the 211 South American camelids analysed from throughout the geographical range. Uncorrected distances within the South American camelids ranged from 0.006 (one substitution) to 0.089 (14 substitutions).

The minimum-spanning network (figure 2) recovers two groups, recapitulating the reciprocally monophyletic clades found previously (Stanley et al. 1994). The first group contains all vicuña
(V), and the other contains all guanaco (G). Both groups are connected by 21 substitutions to the Arabian camel.

The domestic South American camelids are found in both groups, but 81% (120 out of 148, including 61 out of 84 alpaca) are found within the `G’ group. A minority (28 out of 148) comprising alpaca (n=23), pacovicuña (n= 3) and llama (n = 2) are found within the `V’ group.

(b) Microsatellites
A feature of the microsatellite loci analysed here is the large number of private alleles found when comparing vicu·a and guanaco. These private alleles, which range between 33% for YWLL 38 and 100% for LCA 19 and YWLL 46, occupy predominantly di¡erent allele size ranges. Figure 3a(i) and 3b(i) show allele frequency histograms for LCA 19 and YWLL 46 for wild vicuña
and guanaco: the allele sizes do not overlap between the

(a)

(b)

Figure 1. (a)Geographical distribution and photographs of the guanaco (L. g. cacsilensis) (left) and vicu·a (V. v. mensalis) (right). Map taken fromWheeler et al. (1995). Photographs:guanaco, P. Daulesberg; vicu·a, J.C.W. Shaded areas mark the extent of present ranges, solid circles represent Pleistocene and Early Holocene fossil localities. (b)Geographical distribution and photographs of the llama (left) and alpaca (right). Map taken fromWheeler et al. (1995). Photographs: J.C.W. Solid linesmark the distribution prior to 1532, dashed linesmark the present range, solid circles represent archaeological sites containing llama and alpaca remains.

two species. These loci therefore provide potentially powerful tools for the discrimination of ancestral genomes in modern domestic stock. Figure 3a(ii) and 3b(ii) show histograms for the same loci in llama and alpaca, which display similar patterns. However, the patterns of genetic similarity are in contrast to those

Table 1. South American camelid samples analysed for mitochondrial DNA (cytochrome b) and four microsatellite loci.

	cytocromeb mtDNA	YWLL 38	LCA a	YWLL 46 a	YWLL 43
vicuña	42	434	439	438	440
Guanaco	21	112	120	122	115
wari	7	7	7	7	7
pacovicuña	3	3	3	3	3
llama	54	50	60	60	60
alpaca	84	63	138	141	103

a Indicates loci with non-overlapping allele size ranges.

Figure 2. Minimum-spanning network representing the relationships between cytochrome b mitochondrial haplotypes as circular nodes, where the number of substitutions (if greater than one) are represented by multiple dashes or numbers on each connecting line. The relative frequency of each haplotype is represented by the area of the circle. Unfilled circles indicate vicuña haplotypes, and filled circles represent guanaco haplotypes. Wild samples are specifically referred to where present. Phylogenetic analysis (maximum parsimony and neighbour joining based on uncorrected p, JC and K2P sequence distances) recovered an equivalent pattern to the network, and high (490%) bootstrap support was always found for the major split between `V’ and `G’ haplotypes (not shown).

revealed by mtDNA. Visual inspection reveals strong similarities between the allele size distributions of vicuña and alpaca, and between guanaco and llama. For example, for YWLL 46, the 98 bp allele has a frequency of 0.95 in the vicuña sample and 0.75 in the alpaca, while the 104 and 106 bp alleles have a combined frequency of 0.91 in the guanaco and 0.64 in the llama.

Analysis of all four loci reveals that genetic distances between the vicu·a and alpaca and between the guanaco and llama (table 2) are almost always much lower than those between the vicuña and guanaco, the vicu·a and llama or the guanaco and alpaca. Distances between the alpaca and llama are intermediate, with the exception of δμ2, where the distance between the guanaco and llama is slightly larger.

However, a second feature of the histogram in figure 3b is the presence, at low frequencies, of `vicuña’ alleles in the llama sample and of `guanaco’ alleles in the alpaca sample. For example, in LCA 19, the 99 bp and 103 bp `vicu·a’ alleles are present at a combined frequency of 0.063 in the llama sample, and the 87 bp `guanaco’ allele has a frequency of 0.119 in the alpaca. The presence of these alleles in our sample con¢rms bidirectional introgression in both domestic forms.

A striking pattern emerges from the factorial correspondence analysis (¢gure 4), where almost half of the

Figure 3. Allele frequency histograms for a large sample of (i) vicuña and guanaco and (ii) llama and alpaca for (a) LCA 19 and (b) YWLL 46. Distributions were generated using the following: guanaco, n = 104 (LCA 19) and n = 177 (YWLL 46); llama, n = 56 (LCA 19) and n = 58 (YWLL 46); alpaca, n = 80 (LCA 19) and n = 82 (YWLL 46); and vicuña, n = 227 (LCA 19) and
n = 231 (YWLL 46).

Table 2. Pairwise genetic distances between the four South American camelids. (a) Above diagonal: Reynolds distance; below diagonal: δμ2. (b) Below diagonal: 17proportion of shared alleles.

Figure 4. Two-dimensional factorial correspondence plot for allele frequencies at four microsatellite loci in all South American camelids. Almost half of the explained correspondence (15%) is found in factors 1 and 2, represented on the horizontal and vertical axes, respectively.

explained variance (15%) is found in factors 1 and 2, represented on the horizontal and vertical axes, respectively.

It can be clearly seen that the guanaco (black squares) and vicuña (grey, ¢lled diamonds) form two tightly clustered and highly distinct groups. The alpaca (grey, filled triangles) also form a cohesive group and cluster strongly with the vicuña. In contrast, the llamas (open squares) and hybrids form a much more di¡use group. The llamas, although tending to cluster with guanaco on axis 1, are more intermediate with respect to the wild species when compared with the alpaca sample, and are also the most genetically di¡use group on axis 2.

(c) Combined data
The admixture results are non-concordant between the mitochondrial and microsatellite analyses in the alpaca, where the estimated microsatellite proportion of vicuña genome is two to three times higher than the proportion estimated using mtDNA (0.310 ± 0.121 for mtDNA; 0.903 ± 0.108 for mC and 0.823 ± 0.087 for mY for microsatellites). Furthermore, in the llama, although both estimates are relatively low, the microsatellite admixture proportions are an order of magnitude higher (0.0231 ± 0.242 for mtDNA; 0.220 ± 0.088 for mC and 0.389 ± 0.109 for mY for microsatellites). Importantly, it is evident using both mC and mY that the proportion of
vicuña DNA is much lower in llama than in alpaca.

The combined three-locus analysis of the 211 individuals produces striking results (table 3). Out of the 54 llamas, 52 (96%) possessed a `G’ mtDNA haplotype, and 47 (90%) and 33 (61%) possessed a pure `G’ genotype for LCA 19 and YWLL 46, respectively. Out of the 84 alpacas, only 23 (27%) possessed a `V’ mtDNA haplotype, while 63 (75%) and 66 (79%) possessed pure `V’

Table 3. Three-locus genotypes for samples where all three types of data are available. Loci are ordered mitochondrial DNA, LCA 19 and YWLL 46; so, for example, GVH indicates guanaco mitochondrial DNA, vicuña genotype at LCA 19 and a hybrid genotype at YWLL 46 (X signi¢es that the sample could not be typed).

LCA 19 and YWLL 46 genotypes, respectively. Out of the llamas tested in this study, 32 (59%) exhibited a `GGG’ three-locus genotype; only 17 (20%) alpacas exhibited a`VVV’ three-locus genotype. Extensive nuclear introgression is detected in the llamas studied here, with 20 (37%) showing one or more `vicu·a’ alleles at LCA 19 and/or YWLL 46. In contrast, much of the presumed introgression in the alpaca is mitochondrial, with 34 (40%) samples showing a `GVV’ three-locus genotype.

4. DISCUSSION

The data resulting from this study present a complex picture of how modern-day domestic South American camelid genetic diversity has been shaped by domestication, selection, hybridization and genetic drift, with the added uncertain impact of events relating to the Spanish conquest. Such complexity is not without precedence in studies of domestic livestock (e.g. MacHugh et al. 1997), and requires detailed analysis and cautious interpretation.

Taken in isolation, the expanded mtDNA analysis presented here largely supports the findings of Stanley et al. (1994). Wild vicu·a and guanaco mtDNA is reciprocally monophyletic with 5.8-8.9% uncorrected sequence divergence being found between the two lineages, recapitulating the suggestion in Stanley et al. (1994) that these species diverged from a common ancestor two to three million years ago. Furthermore, the ¢nding that nearly all modern llamas possess a `guanaco’ haplotype is also supported by the present dataset (table 3), where all except two llamas from a sample of 54 individuals had guanaco mtDNA. However, our much expanded alpaca dataset reveals a di¡erent pattern from that previously found, with only 27% of individuals possessing vicuña mtDNA (table 3), in contrast to the 50% reported by Stanley et al. (1994). It is worth noting here that our sample comprises 84 individuals (as opposed to 14) sampled from throughout the geographical range.

In isolation, the ¢nding that a large proportion of modernday alpacas possess guanaco mtDNA is in accordance with hypotheses that alpacas, in common with llamas, are descended from the guanaco. However, clearly, as in Stanley et al. (1994), the presence of substantial numbers of alpaca possessing vicuña mtDNA also raises the possibility that the alpaca is of mixed origin or has undergone substantial hybridization during domestication or subsequently.

However, the limitations of mtDNA, which is maternally inherited, in the context of gene flow and
evolution in domestic populations are obvious, since historical and modern-day agricultural practices have often used phenotypically desirable males to sire large numbers of females. Therefore, analysis of biparentally inherited markers (in this case microsatellites) capable of
discriminating between the wild ancestors was desirable.

The microsatellite data provide a contrast to the mtDNA data, and the existence of two loci with nonoverlapping allele size ranges in the wild ancestors allowed us to compare patterns of divergence in relatively large numbers of domestic animals. In contrast to the mtDNA results, visual inspection of allele frequency distributions, genetic-distance analysis using three measures with contrasting models and assumptions, and factorial correspondence analysis all reveal a striking similarity between the alpaca and the vicuña (figures 3a,b, 4 and table 2a,b).

Each genetic-distance estimate is lowest for the alpaca-vicuña comparison, and the factorial correspondence analysis (figure 4) shows that the alpaca and vicuña occupy an almost overlapping region of the plot. These data, therefore, point towards a very close genetic afinity between the alpaca and vicuña, a¢nding in complete contrast to the mtDNA data.

The microsatellite data in part agree with the mtDNA data by supporting a close relationship between the llama and guanaco. Out of the genetic-distance estimates, both Reynold’s and allele sharing distances are second lowest for the guanaco-llama comparison (table 2a,b), and the
Reynold’s distance estimate is almost identical to that for the alpaca-vicuña comparison. However, other data are more equivocal, with the factorial correspondence plot revealing a dispersed pattern for the llama, which could be described as intermediate between the vicuña and
guanaco, and the δμ2 distances being slightly lower for the llama-alpaca comparison than for the llama-guanaco comparison. Although none of the above are indicative of a close relationship between the llama and vicuña, they suggest either that there has been nuclear gene flow between the llama and vicuña (or, more likely, between the llama and alpaca) or that the ancestral stock for the present-day llama was genetically highly diverse and that this is re£ected in present-day microsatellite allelic diversity.

The guanaco and llama have much greater geographical ranges than the vicuña and alpaca,
which may have led to greater historical intraspecific di¡erentiation, re£ected in greater diversity in nuclear and mtDNA (where we found 21 guanaco haplotypes, as opposed to five in the vicuña).

Analysis of greater numbers of microsatellite loci (only four were used here) is highly desirable, and could include some of those described by Sarno et al. (2000), where at least two additional loci suggest non-overlapping allele distributions in the wild species. Archaeozoological evidence is
inconclusive concerning llama domestication, but it is possible that there may have been more than one centre (Wheeler 1995).

The suggestion of substantial mitochondrial introgression in the alpaca and nuclear introgression in the llama is substantiated when admixture is measured for both marker types. The relatively low estimated admixture proportion of vicu·a mtDNA present in the alpaca (0.31) is in contrast with the high proportion estimated for the microsatellites (0.82-0.90). Further, the extremely
low admixture proportion of vicu·a mtDNA in the llama (0.02) contrasts strongly with microsatellite estimates (0.22-0.39), which also suggest substantial nuclear admixture in the llama. These data must, however, be treated with caution, since the assumptions underlying the admixture models used in this analysis are highly likely to be violated in South American camelids, where introgression events are likely to have been recurrent, and may have increased in frequency during and after the Spanish conquest. Furthermore, the drastic reductions in domestic South American camelid populations associated with the conquest are likely to have had signi¢cant and
unpredictable impacts on allele frequencies.

Inspection of the distribution of the three-locus genotypes in modern alpacas and llamas conforms many of the above ¢ndings. Only 27% of alpacas are mitochondrially `vicuña’, although 40% of alpaca possess only vicu·a microsatellite alleles with guanaco mtDNA haplotypes.

Such a pattern suggests that introgression of guanaco (or, more likely, llama) mtDNA some time in the past may have occurred recurrently within alpaca populations, but may have been accompanied more recently by a reversion to line or stock breeding within local alpaca populations.

The lack of written records means that any such inference is speculative. Table 3 suggests that, in contrast, mitochondrial introgression has occurred much less frequently in the llama. However, although table 3 suggests that nuclear introgression is similar in the alpaca and llama, the expanded microsatellite dataset suggests that it has occurred at a higher level in the llama
(possibly two or three times higher than in the alpaca from the admixture analysis), which may partly account for the more dispersed factorial correspondence pattern; this warrants further investigation.

The implications of these data are potentially important for the way in which these genetic resources are managed in the future. In our sample, only 35% of domestic animals have not undergone any detectable hybridization based on mtDNA and the two diagnostic
microsatellites. In particular, there are a very large number of detectable hybrids within the alpaca population (80%), which are accentuated when using mtDNA.

Also, 40% of llama show detectable signs of hybridization, but mitochondrial introgression is virtually absent. During the last 20-25 years large-scale hybridization between llamas and alpacas has been carried out in Peru (Bustinza 1989).

Speci¢cally, male alpacas have been bred to female llamas to increase the population of animals
producing higher priced `alpaca’ ¢bre, and male llamas have been bred to female alpacas to obtain greater fleece weights and, thus, increased income.With sale price traditionally
determined by weight, and no consideration given to fineness, the quality of alpaca fibre has decreased markedly over the past 25 years.

Indigenous Quechua- and Aymara-speaking herders subdivide the hybrids into llamawari or waritu (llama-like) and pacowari or wayki (alpaca-like), respectively, depending upon physical
appearance (Flores Ochoa 1977; Dransart 1991b).

The o¡spring are fertile, tend to be intermediate in size and can be backcrossed to either parental type. Further, recent intensive selection for white £eece in modern alpaca may also have involved bidirectional hybridization. A combination of these practices and our results could explain the
taxonomic confusion surrounding the domestic forms in the recent past, as it is likely that many specimens used in previous taxonomic studies were hybrids. Given the extreme hybridization in present-day alpacas, DNA analysis has been critical in resolving the origin of this domestic form. Since our results suggest the vicu·a as the ancestor of the alpaca, we propose that the classi¢cation of the alpaca should be changed from Lama pacos L. to Vicugna pacos L.
We suggest, therefore, that the degeneration of quality and value in present-day alpacas and llamas has been due to extensive hybridization, probably beginning with the conquest and continuing to the present day.While it was believed that these crosses were between di¡erent forms of a single domestic animal descended from the guanaco, there was little concern about the economic impact of such introgression.

However, given that the alpaca is probably descended from the vicu·a, the negative impact
of such crosses is now evident. The use of DNA analysis to identify and eliminate hybrid animals from the breeding pool is essential, since the antiquity of the ongoing hybridization
process makes it impossible to accurately identify all hybrids on the basis of phenotypic characteristics.

Additionally, the knowledge that the alpaca is descended from the vicuña opens new routes for the improvement of alpaca fibre production, not only through the identification of hybrids and their elimination from pure-bred elite herds, but also via the backcrossing of pure-bred alpacas
to their vicuña ancestor in order, possibly, to improvefibre fineness. While 90% of the alpaca ¢bre produced in Peru today has a diameter of greater than 25 mm and fetches low prices on the world market (US$3-30per kg, 1980-1995), pre-conquest animals produced fibre of 17-22mm in
diameter (Wheeler et al. 1995), similar to cashmere (15- 17 mm; US$60-120 per kg, 1980-1995). It is possible, therefore, that identification of the remaining pure alpacas may aid in recovery of the fine fibre characteristics of pre-conquest animals.

The knowledge that the alpaca is the domestic vicuña also necessitates a re-evaluation of vicuña conservation policy. Although the vicuña has been listed as endangered under the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES Appendix I) since its inception in 1975, all Peruvian vicuñas and large segments of the Chilean, Argentine and Bolivian populations have been reclassi¢ed as threatened (CITES Appendix II), permitting controlled
commercialization of live-shorn fibre.

With unprocessedfibre currently valued at ca. US$405 per kg, vicuña fleece is the most expensive natural fibre in the world, and represents an important potential source of income for the extremely poor rural populations on whose lands the animals live. To date, Peru’s rational use policy has produced an important increase in vicuña numbers, but demands for greater control over the species through construction of fences, intensive rearing and selection are growing.

Judging by the alpaca, such interventions will, in the long run, lead to a deterioration of fibre quality and fi¢neness (which, at 12-14 mm, is the basis of its value), and increased limitation on movement, especially of the non-territorial male bachelor bands, represents a significant new threat to this species (Wheeler et al. 2001).

We acknowledge the following people who generously helped by providing information, samples or permits for this study. In Argentina: M. Knobel, Argentine Guanaco Products, Esquel; S.
Poncet; Eduardo Frank, Universidad de Cordoba; Freddy Sossa, EU Supreme Project, Jujuy; Gustavo Rebu¤, INTA Abra Pampa; Daniel Almeida, INTA Bariloche; Hector Guillermo
Villanueva, Recursos Naturales, Salta; Teresa Raquel Chalabe, Universidad de Salta; G. Moseley, BBSRC^IGER, Aberystwyth, UK. In Chile: Hernan Torres, Conservation International, Washington DC, USA; Eduardo Nu·ez and Rafael Fernández, CONAF, Arica; Calogero Santoro, Universidad de Tarapacá, Arica. In Peru: Alfonso Martinez, Domingo Hoces, Jorge Herrera and Marco Antonio Zuñiga, CONACS, Lima; Alex Montufar and Marco Antonio Escobar, CONACS, Puno; Roberto Bombilla, CONACS, Junín; Carlos Ponce del Prado, Conservación Internacional, Lima; Maximo Gamarra and Santiago Baudilio, SAIS Tupac Amaru, Pachacayo; Rosa Perales, Jose Alva and Néstor Falcón, Facultad de Medicina Veterinaria, Universidad Nacional Mayor de San Marcos, Lima; Leoncio Ruiz Ríos, Congreso Constituyente del Perí; Clive Woodham, Labvetsur, Arequipa; Felipe San Martín, Instituto Veterinario de Investigaciones Tropicales y de Altura, Lima. Nicola
Anthony, Peter Arctander, Mark Beaumont, Jon Bridle, Kate Byrne, Loune© s Chikhi, Gordon Luikart, Benoit Goossens and two anonymous referees provided invaluable comments on the
manuscript, and Georgio Bertorelle gave valuable advice on the admixture analysis. This study was supported by the Institute of Zoology, InstitutoVeterinario de InvestigacionesTropicales y de
Altura, Natural Environment Research Council grant GST/02/828 to H.F.S. and J.C.W. and Darwin Initiative grant 162/06/126 to H.F.S, J.C.W and M.W.B.

REFERENCES

Bahn, P. G. 1994 Archaeozoologyötime for a change. Nature 367, 511-512.

Bandelt, H. J., Forster, P. & RÎhl, A. 1999 Median joining networks for inferring intraspeci¢c phylogenies. Mol. Biol. Evol. 16, 37-48

Belkhir, L. 1999 GENETIX v. 4.0. Belkhir Biosoft, Laboratoire des Genomes et Populations, Université Montpellier II, France. Benzécri, J. P. 1973 L’analyse des donnees:T. 2, I’analyse des correspondances. Paris: Dunod.

Bertorelle, G. 1998 ADMIX 1_0. See www.fe.it/genetica/Giorgio/ giorgio_soft.html#ADMIX.

Bertorelle, G. & Exco¤er, L. 1998 Inferring admixture proportions from molecular data.Mol. Biol. Evol. 15, 1298-1311.

Bradley, D. G., MacHugh, D. E., Cunningham, P., Loftus, R. T. 1996 Mitochondrial diversity and the origins of African and European cattle. Proc. Natl Acad. Sci. USA 93, 5131-5135.

Bruford, M. W., Hanotte, O., Brook¢eld, J. F. Y. & Burke, T. 1998 Single and multilocus DNA ¢ngerprinting. In Molecular genetic analysis of populations: a practical approach, 2nd edn (ed.
A. R. Hoelzel), pp. 287-336.Oxford University Press.

Bustinza,V. 1989 Algunas consecuencias de la agresión cultural en la ganadería Andina. In Crianza de llamas y alpacas en los Andes (ed. PAL/PRATEC), pp. 115^130. Lima, Talleres
Gra¢cos de ART Lautrec, S.R.L.

Chakraborty, R., Kamboh,M. I., Nwankwo,M. & Ferrell, R. E. 1992 Caucasian genes in American blacksönew data. Am. J. Hum. Genet. 50, 145-155.

Dransart, P. 1991a Llamas, herders and the exploitation of raw materials in the Atacama desert. World Archaeol. 22, 304-319.

Dransart, P. Z. 1991b Fibre to fabric: the role of fibre in camelid economies in prehispanic and contemporary Chile. DPhil. thesis, University of Oxford, UK.

Exco¤er, L. 1993 MINSPNET. See http://lgb.unige.ch/software/ win/min-span-net/.

Flores Ochoa, J. A. (ed.) 1977 Pastores de alpacas de los Andes. In Pastores de puna, pp.15-52. Lima: Instituto de Estudios Peruanos. Goldstein, D. B., Linares, A. R., Cavalli-Sforza, L. L. &

Feldman, M. W. 1995 An evaluation of genetic distances for use with microsatellite loci. Genetics 139, 463-471.

Hall, S. J. G. & Bradley, D. G. 1995 Conserving livestock breed diversity.Trends Ecol. Evol. 10, 267-270.

Hiendleder, S., Mainz, K., Plante, Y. & Lewalski, H. 1998 Analysis of mitochondrial DNA indicates that domestic sheep are derived from two ancestral maternal sources: no evidence for contributions from urial and argali sheep. J. Hered. 89, 113-120.

Kruskal, J. B. 1956 On the shortest spanning subtree of the graph and the travelling salesman problem. Proc. Am. Math. Soc. 7, 48-57.

Lang, K. D. M.,Wang,Y. & Plante,Y. 1996 Fifteen polymorphic dinucleotide microsatellites in llamas and alpacas. Anim. Genet. 27, 293.

Luikart, G., Gielly, L., Exco¤er, L., Vigne, J. D., Bouvet, J. & Taberlet, P. 2001Multiple maternal origins and weak phylogeographic structure in domestic goats. Proc. Natl Acad. Sci. USA 98, 5927-5932.

MacHugh, D. E., Shriver, M. D., Loftus, R.T., Cunningham, P.& Bradley, D. G. 1997 Microsatellite DNA variation and the evolution, domestication and phylogeography of taurine and zebu cattle (Bos taurus and Bos indicus). Genetics 146, 1071-1086.

Minch, E. 1999 MICROSAT v. 1.5. See http://hpgl.stanford.edu/projects/microsat/.

Penedo, M. C. T., Casetano, A. R. & Cordova, K. I. 1998 Microsatellite markers for South American camelids. Anim. Genet. 29, 411-412.

Reigadas, M. C. 1993 La punta del ovillo: determinación de domesticación y pastoreo a partir del análisis microscopico de fibras y foliculos pilosos de camélidos. Arqueologica 2, 9-52.

Reynolds, J., Weir, B. S. & Cockerham, C. C. 1983 Estimation of the co-ancestry coeficient-basis for a short-term genetic distance. Genetics 105, 767-779.

Sarno, R. J., David, V. A., Franklin, W. L., O’Brien, S. J. & Johnson, W. E. 2000 Development of microsatellite markers in the guanaco Lama guanicoe: utility for South American camelids. Mol. Ecol. 9, 1922-1924.

Stanley, H. F., Kadwell, M. & Wheeler, J. C. 1994 Molecular evolution of the family Camelidae: a mitochondrial study. Proc. R. Soc. Lond. B 256, 1-6.

Vila, C., Leonard, J. A., Gotherstrom, A., Marklund, S., Sandberg, K., Liden, K.,Wayne, R. K. & Ellegren, H. 2001 Widespread origins of domestic horse lineages. Science 291, 474-477.

Wheeler, J. C. 1995 Evolution and present situation of the South American Camelidae. Biol. J. Linn. Soc. 54, 271-295.

Wheeler, J. C., Russel, A. J. F. & Redden, H. 1995 Llamas and alpacas: pre-conquest breeds and post-conquest hybrids. J. Archaeol. Sci. 22, 833-840.

Wheeler, J. C., Fernandez, M., Rosadio, R., Hoces, D., Kadwell, M. & Bruford, M. W. 2001 Diversidad genetica y Manejo de Poblaciones de Vicuñas en el Peru. RIVEP 1(Suppl.), 170-183.

As this paper exceeds the maximum length normally permitted, the authors have agreed to contribute to production costs.