Ancient DNA of Guinea Pigs (Cavia spp.) Indicates a Probable New Center of Domestication and Pathways of Global Distribution

Guinea pigs (Cavia spp.) have a long association with humans. From as early as 10,000 years ago they were a wild food source. Later, domesticated Cavia porcellus were dispersed well beyond their native range through pre-Columbian exchange networks and, more recently, widely across the globe. Here we present 46 complete mitogenomes of archaeological guinea pigs from sites in Peru, Bolivia, Colombia, the Caribbean, Belgium and the United States to elucidate their evolutionary history, origins and paths of dispersal. Our results indicate an independent centre of domestication of Cavia in the eastern Colombian Highlands. We identify a Peruvian origin for the initial introduction of domesticated guinea pigs (Cavia porcellus) beyond South America into the Caribbean. We also demonstrate that Peru was the probable source of the earliest known guinea pigs transported, as part of the exotic pet trade, to both Europe and the southeastern United States. Finally, we identify a modern reintroduction of guinea pigs to Puerto Rico, where local inhabitants use them for food. This research demonstrates that the natural and cultural history of guinea pigs is more complex than previously known and has implications for other studies regarding regional to global-scale studies of mammal domestication, translocation, and distribution.


Results
DNA preservation, sequence authenticity and contamination. Of 66 archaeological samples processed, 59 produced DNA libraries adequate for mitogenome sequencing. Post sequencing, 42 samples provided over 90% coverage of the mitogenome and were retained for further analysis, giving an overall success rate of 64% (SI Table S2). Samples had an average read depth of 654×, ranging from 4.7× to 3297× (SI Tables S3 and S4). All samples showed damage patterns and fragment lengths characteristic of aDNA with no indication of contamination from lab reagents or other sources (SI Table S3).
Poor DNA preservation and lower success rates were generally seen in samples from coastal Caribbean and Peruvian sites. Four sites, in Jamaica, Nevis and St Lucia in the Caribbean, and Carrizales in Peru, failed to yield any usable data (SI Table S4). This is likely due to the coastal environment not being conducive to good DNA preservation, a result reported in other aDNA studies for the Caribbean, as well as from other coastal, tropical environments [38][39][40][41][42][43][44] . In addition, only one of three samples from Aguazuque I (1885-1815 BC), in the eastern Colombian Highlands, produced a complete mitogenome, likely due to a combination of age and environmental factors. Four of the five samples processed from the older site of Checua II (5940-5670 BC), also in the eastern Colombian Highlands, however, produced complete mitogenomes, as did all six of the samples from El Venado (AD 800-1600) as well as one sample from the Madrid site (200-100 BC). The cultural affiliations and additional information regarding the context of all archaeological samples used in our phylogenetic analyses are provided in Table S5. www.nature.com/scientificreports www.nature.com/scientificreports/ phylogenetic identity of colombian samples. When we initially attempted to map all of our sequences to the C. porcellus reference mitogenome (Genbank ID: NC_000884.1), we found that none of the 12 Colombian samples mapped efficiently and all showed an identical pattern of alignment/coverage (SI Fig. S1). This pattern led us to suspect that they may not be C. porcellus. Thus, a de novo assembly was carried out using one high coverage Colombian sample (MS10677, from Checua II) to produce a consensus Colombian reference sequence, to which all other Colombian samples were subsequently mapped (SI Fig. S2). This produced a much higher mapping success from all Colombian samples.
Given the lack of comparative complete mitogenomes for Cavia species, we compared the Cytochrome B region from our Colombian samples with all Caviidae data available in Genbank (SI Table S6). The resulting Maximum Likelihood consensus tree (Fig. 1B) suggests that the ancient Colombian archaeological remains appear to represent a distinct Cavia species; none of the 12 sequenced Colombian samples fell within the C. porcellus clade.
Relationships identified using complete mitogenomes. Once all Colombian samples were realigned to the consensus Colombian reference sequence, a Median Joining (MJ) network (Fig. 1C) was constructed using all of the complete mitogenomes generated in this study, as well as the C. porcellus reference mitogenome (Genbank: NC_000884.1). The C. porcellus and ancient Colombian samples are separated by 1406 mutations, as seen in the CytB tree (Fig. 1B), suggesting that they are from two distinct species.
Within the Colombian samples, two distinct groups are identified, with the two clades separated by 91 mutations. The separation of these samples, all identified as belonging to the same species based on analysis of the Cyt B region, appear to correlate temporally. The first, labelled Colombia A (Fig. 1C), contains samples from the older contexts, Checua II (5940-5760 BC) and Aguazuque (1885-1815 BC). The second group, Colombia B, consists of samples from the younger contexts, Madrid (200-100 BC) and El Venado (AD 800-1600), except for a single sample from Checua II.
Within the C. porcellus samples, we see three distinct groups: Caribbean and Peruvian samples, the Bolivian samples, and the historic/modern European samples. Our analyses of eight samples from four sites in the Caribbean identify two haplotypes separated by two mutations. One haplotype contains samples from northern Peru (Kuelap), a single sample from coastal Peru (Pachacamac), Puerto Rican samples from both Tibes (Tibes A) and Finca Valencia and a sample from Antigua. The second haplotype was found in samples from Tibes (Tibes B) and Carriacou, as well as samples from southern and coastal Peru (Pachacamac, Lo Demás, Moqi and Torata Alta).
The two historic samples had sequences that clearly place them within C. porcellus. The South Carolina sample shared a haplotype with the ancient Peruvian and Puerto Rican (Kuelap, Tibes A) samples, whereas the Belgian sample clustered with the modern samples (the reference sequence NC_000894.1, and the modern sample from Lord et al. 32 ), and the modern Caribbean sample.

Discussion
Cavia taxonomy and a unique process of domestication in Colombia. Our analyses of the Cyt B data show that all 12 of our ancient Colombian samples fall into a distinct clade that is basal to all other members of the Cavia genus (Fig. 1B, SI Fig. S4). When we compare the analysis of the complete mitogenomes ( Fig. 1C), we see that the number of mutations between the ancient Colombian samples and the C. porcellus samples (n = 1406) is similar to the level of variation reported between Rattus spp., such as R. norvegicus and R. exulans differing by 1577 mutations across the mitogenome 45 . We therefore suggest that the ancient Colombian samples we sequenced fall within the Cavia genus, but do not belong to C. porcellus, or other Cavia species for which there are mitochondrial data available. Interestingly, the ancient Colombian samples are genetically distinct from previously published modern Colombian (C. porcellus) sequences (SI Table S6), although these were sourced from southern Colombian markets, as opposed to the highlands of Bogotá where the ancient samples were located. These data suggest the utilization of two different species of guinea pig in Colombia through time: a native Cavia species, identified through the analyses of the archaeological remains studied here, and a secondary, more recent introduction of C. porcellus to coastal Colombia for the meat market.
All of the ancient Colombian samples in this study were previously morphologically identified as C. porcellus, however, as both C. porcellus and C. anolaimae have been described from the same occupation layers at Checua and Aguazuque, we tentatively suggest that all of the Colombian samples sequenced here represent C. anolaimae. First described by Allen in 1916 46 , the taxonomic position of C. anolaimae has been debated, most recently described as a subspecies (C. aperea anolaimae) or a synonym for the domestic C. porcellus 15 . Morphological and molecular evidence generally demonstrate it is a distinct species [46][47][48] , with the exception of one study 15 . Unfortunately, there is only a single Cytochrome B sequence available for C. "aperea" anolaimae in public databases (GU136758 in SI Fig. S4). The identification of this sample appears to be problematic based on the analyses reported here and by that of Dunnum and Salazar-Bravo 15 , as it falls clearly within the C. aperea guianae clade (SI Fig. S4). We suggest that this sample is either mislabeled or misidentified, meaning there are no reliable Cytochrome B data for C. anolaimae. Obtaining a reliable reference sequence for this species is crucial for a definitive species identification of our Colombian samples, though we expect that they may represent C. anolaimae.
Our mitochondrial genome analyses (Fig. 1C) show the presence of two distinct groups/clades within the ancient Colombian guinea pigs, which are separated temporally by more than 1000 years. Campos and Ruiz-Garcia 48 also found a distinction between two modern populations of C. anolaimae, which the authors attributed to geographical isolation. However, given the antiquity, the use of guinea pigs as food, and their inclusion in burials at some archaeological sites in the highlands of Colombia, we suggest the genetic change over time represented by our data may be indicative of an independent domestication of a native Cavia, tentatively identified as C. anolaimae, from a diverse wild population. It is plausible that sustained interaction with humans www.nature.com/scientificreports www.nature.com/scientificreports/ in Colombia resulted in a process of commensal domestication as defined by Zeder 22 , possibly in association with agriculturalists moving into the highlands 21 , or even in the Late Glacial as suggested by Pinto et al. 17 . This supports previous suggestions of an independent Cavia domestication in Colombia based on morphological analyses 14,[17][18][19][20] , although specific the timing of this process remains unclear.
The significance of the two distinct groups of Colombian guinea pigs warrants further study of this possible process of domestication, independent of that which led to C. porcellus in the Central Andes. Alternatively, our samples may represent either a wild Cavia species or an extinct species that did not contribute genetically to the modern populations, even though this species is present in archaeological sites up until AD 1600 indicating that it was present in the highlands until after European colonization began. Furthermore, we suggest that the modern Colombian guinea pigs (from markets) are not related to the archaeological population from the highlands, but are descended from the historic and modern redistribution of varieties bred intentionally for the meat market. Further sequencing and morphological study of ancient and modern Cavia species from the highlands of Colombia will clarify the origins and evolutionary history of Colombian guinea pigs, and confirm whether this species represents an independent process of domestication of Cavia. the origins of guinea pig translocation to the caribbean Antilles. Previous aDNA studies 30,32 suggested a possible Colombian origin for Caribbean guinea pigs. Our results, as shown in the haplotype network (Fig. 1C), indicate that all Caribbean samples share haplotypes with samples from Peru. Thus, we suggest a probable Peruvian origin of the guinea pigs that people translocated to the Caribbean Antilles. This is significant, as no direct connections to Peru have been established empirically through previous archaeological studies. Although similarities in pottery style and raptoral bird iconography are noted between sites on Puerto Rico, Vieques, and in northwestern South America, including the eastern foothills of the Andes 49,50 , the geographic specificity of links between the Caribbean Antilles and South American Andes have remained largely speculative (e.g. 51 ). Thus, this study significantly extends our knowledge of the Late Ceramic Age Caribbean-South American interaction spheres.
The ancient Caribbean guinea pig samples are most closely related to samples from Kuelap (AD 1100-1535) in the northern highlands of Peru, as well as to samples from the Peruvian coastal sites of Pachacamac (AD 600-1000 and AD 1470-1572) and Lo Demas (AD 1480-1540). Both Kuelap and Pachacamac were large sites/polities in late pre-Columbian Peru. The site of Kuelap was first inhabited as a religious center from AD 500, although it is dominated by monumental architecture of the Chachapoyas culture (AD 900-1470), and later the Inca (AD 1470-1535). The site of Pachacamac (AD 600-1500) was also associated with monumental architecture, though has successive Lima, Wari, Ychsma, Inca and early Colonial occupations 52 . This site likely was a 'pilgrimage' center for those travelling throughout Peru, such as from Lo Demas and beyond, although no Caribbean style artifacts have been recorded there. The identification of haplogroups shared between these sites in Peru and sites in the Caribbean Antilles, suggests that interaction between these groups and their wider community extended more broadly than previously documented.

number of translocations and possible routes to the caribbean Antilles. Previous analyses by
Lord et al. 32 identified two haplotypes in the three complete mitogenomes of ancient guinea pigs from three sites in the Caribbean, which also was supported by the increased sample size reported here. We suggest that the initial introduction likely occurred from a guinea pig population originally sourced from northern Peru that reached Puerto Rico around AD 600 (Fig. 2). Significantly, the Puerto Rican samples from both Tibes A and Finca Valencia share a single haplotype, suggesting an initial migration of a non-genetically diverse population. From this founding population, we suggest that guinea pigs were then dispersed to Antigua, in the northern Lesser Antilles after AD 1000, in line with cultural interaction during this time period [53][54][55] . The second Caribbean haplotype, found at Tibes B (Puerto Rico) and Carriacou, potentially represents a distinct translocation, either directly or indirectly from an ancestral population in coastal Peru, possibly around the site of Pachacamac. This scenario for a coastal origin is also supported by the presence of domesticated guinea pigs in coastal Ecuadorian sites and evidence for long-distance maritime trade of spiny oyster 25,26 . The possibility of multiple introductions of guinea pigs to the Caribbean further supports archaeological and isotopic evidence that the Late Ceramic Age cultures of the Caribbean, called Ostionoid (ca. AD 500-European contact ca. AD 1492), were highly mobile and had large, ongoing interaction networks reaching across the Caribbean to both Central America and northern South America (19,(44)(45)(46)54).
The exact route of the initial or subsequent introduction(s) of guinea pigs through northwestern South America and into the Caribbean Antilles is difficult to determine. Translocation may have been the result of direct or indirect down-the-line trade networks via either, or a combination of: a northwestern coastal route, a Pacific maritime route and an eastern Andean overland route (given the genetic similarity to guinea pigs from the montane site of Kuelap). However, few archaeological specimens of domestic guinea pigs exist in northern South America, which makes potential translocation routes through this area difficult to discern. A single specimen has been identified from the inland site of Turen in northern Venezuela 56 and eight individuals (morphologically identified as C. aperea) are reported from the Palmasola site near the northwest coast of Venezuela 57 but these specimens could not be obtained for this study. To date, no known ancient guinea pig remains have been recovered from the coastal regions of Colombia. We observed poor aDNA preservation from samples recovered from coastal localities even when the archaeological specimens appeared to be well-preserved (SI Section 1.1, Table S4). Thus, it may be that remains of guinea pigs brought along either the western and northern coasts or the northeastern interior of South America just do not survive in the archaeological record of these regions. If taphonomic and preservation factors cannot account for the lack of archaeological guinea pig remains in coastal Colombia and Venezuela, it is possible that the movement of guinea pigs was rapid, thus leaving no archaeological signature. (2020) 10:8901 | https://doi.org/10.1038/s41598-020-65784-6 www.nature.com/scientificreports www.nature.com/scientificreports/ From northern South America, guinea pigs may have been introduced to Puerto Rico via guanín trade with the Isthmo-Colombian region (as described in 50,58 ) or the trade of Guatemalan-sourced jadeitite/jadeite artifacts from Central America to the Caribbean 59 . This could offer an alternative route of introduction of guinea pigs to the Caribbean and explain the second haplotype in Puerto Rico (Tibes B), which was then dispersed to Carriacou in the southern Lesser Antilles. This introduction possibly occurred around AD 900, which correlates with the development of increasingly hierarchical societies in Puerto Rico 60 . Thus, almost simultaneously, guinea pigs could have then been transported down to the Lesser Antilles through sustained inter-island trade and interactions.
It is also possible, however, that only a single introduction of guinea pigs to the Caribbean occurred, via the routes described above, with later dispersal within the Antilles. This hypothesis would support a previous proposition by Wing 29 that the significant quantity of guinea pig bones from Finca Valencia on the north coast of Puerto Rico may indicate that Ceramic Age colonists established a founding population of guinea pigs on Puerto Rico and then translocated the animals to other islands. Additional analyses of DNA recovered from directly dated ancient Caribbean samples may be able to discern between the multiple and single introduction hypotheses and allow for better interpretation of the timing and route(s) of the introductions.

Modern reintroduction to the Caribbean.
To date, no archaeological remains of guinea pigs have been identified or verified from well-excavated and well-analyzed early Spanish colonial-associated Caribbean sites 27,61,62 . Our analysis of the mitogenome of a modern specimen from San Sebastían, Puerto Rico indicates that it is not related to any of the archaeological specimens in the Caribbean. Instead, it grouped within the modern European samples, suggesting a recent reintroduction of guinea pigs to Puerto Rico occurred from Europe or, perhaps more likely, from the markets of southern Colombia or Peru (Fig. 2). Further complete mitogenome analyses of modern samples from South America will allow for the origins of the modern translocation to be elucidated, as currently only Cytochrome B data are available.
Historic translocations and interactions with europe and north America. The historic samples from Mons, Belgium and Charleston, South Carolina appear to have origins in the Central Andes (Fig. 2). The Belgian sample falls within the modern European group; however, this group is related most closely to samples from either Bolivia or Peru. We suggest that during the Spanish colonial period, guinea pigs from the Andean www.nature.com/scientificreports www.nature.com/scientificreports/ region of Bolivia/southern Peru were transported to Europe. It is also possible that guinea pigs came to Europe from northern Chile, although samples from this area have not yet been analysed. In order to clarify the precise origins of European guinea pigs and the timing of their translocation, more analysis of samples from historic era contexts in South America and Europe are needed.
The Charleston, South Carolina guinea pig, dating to AD 1820, represents the earliest evidence of guinea pigs in North America. This sample shares a haplotype with samples from Kuelap in northern Peru, and with specimens from Puerto Rico (e.g. Finca Valencia and the Tibes A specimen). As previously mentioned, there is no archaeological evidence for guinea pigs in Puerto Rico or elsewhere in the Caribbean during the colonial period, thus the Charleston specimen probably originated from Peru. Furthermore, the Charleston guinea pig was discarded in a privy, along with the remains of an Amazonian parrot 63 , suggesting that both animals were acquired as South American 'curiosities' (see SI information on Heyward-Washington House).

conclusions
Here we use mitogenome sequence data obtained from archaeological, historical, and contemporary guinea pigs to identify diachronic human networks of interaction spanning several millennia. Ancient mitogenomes obtained from Cavia remains recovered from archaeological sites in the Colombian Highlands were clearly genetically distinct from all other populations studied, suggesting that all of the ancient Colombian samples we analysed belong to a different Cavia species. We tentatively identify this species as C. anolaimae, based on previous literature describing the biogeography of this species. Unfortunately, no comparative mitogenome data for C. anolaimae exists in any public databases, so we cannot confirm this tentative species identification. Our result, however, supports previous suggestions that an independent guinea pig domestication may have occurred in the Colombian Highlands. Further genetic sampling as well as morphometric analyses of modern and archaeological guinea pig specimens will be necessary to broaden our understanding of the domestication of guinea pigs, and aid in identifying processes and biological outcomes of guinea pig domestication.
Our results also demonstrate that beginning around AD 600, people transported guinea pigs (C. porcellus) beyond the South American mainland, introducing them to the Caribbean Antilles from a Peruvian-derived population, possibly two times, through existing social networks of interaction and trade routes. In the absence of archaeological evidence for guinea pigs along coastal Central America, coastal Colombia, or elsewhere in northern coastal South America, this connection to modern Peru is the first empirical evidence demonstrating interactions between groups in the Caribbean Antilles and this region of western South America. We also identify a Central Andean origin of the earliest historic-era introductions of domestic guinea pigs into Europe and North America. Lastly, we identify a modern reintroduction of guinea pigs to the island of Puerto Rico, from Europe or, perhaps more likely, from the modern market breeds of South America.
This study adds to the growing body of evidence suggesting highly diverse geographical, temporal and cultural histories of mammal domestication and subsequent distribution. Through this analysis of ancient guinea pig mitogenomes, the history of such variability within the context of human social interactions over thousands of years and across three continents provides a critical historical perspective of the genetic diversity of one of the world's oldest and enduring domestic animals.

Materials and Methods
A total of 66 archaeological guinea pig samples from 21 sites were obtained from four archaeological sites in Colombia (n = 15), six sites in Peru (n = 32), two sites in Bolivia (n = 4), seven sites from the Caribbean (n = 13), two historic sites from Europe (n = 1) and North America (n = 1), and one modern market in Puerto Rico (Fig. 1A, Table S1). In addition, the modern European sample analysed in Lord et al. 32 was included in downstream analyses. Specific details of archaeological sites and contexts of the analyzed specimens are presented in the Supplementary Information Section 1.1.
All archaeological samples were processed in a purpose-built ancient DNA facility at the University of Otago 64 . DNA extraction, library preparation, hybridization capture and sequencing methods, and description of bioinformatic processing of sequences are described in the electronic Supplemental Information Section 1.2.