Introduction: the commensal model

In the mid 1990s a program developing a new genetic model for understanding the human settlement of the Pacific was undertaken at the University of Auckland (Matisoo-Smith 1994; Matisoo-Smith et al. 1997, 1998). Given the concerns of indigenous peoples about the use and study of human tissues at the time, coupled with the recognised lack of genetic variation in Polynesian populations (Hertzberg et al. 1989; Hill and Serjeantson 1989), it was concluded that perhaps an alternative method for identifying population origins and tracking prehistoric human migration patterns might be to use a proxy. Instead of studying the genetic relationships of the people themselves, we could trace their migration patterns by studying the things that they carried with them in their colonising canoes. The archaeological evidence strongly suggested that Polynesians and their ancestors transported numerous plants and animals with them and introduced those species to the islands of Remote Oceania (Kirch 2000). If we could identify the genetic relationships and track the origins of those plants and animals around the Pacific, these might indicate the immediate origins and the movement of the people who carried them. Thus we began to develop and test what we now refer to as the commensal model for the human settlement of the Pacific (Matisoo-Smith 1994; Matisoo-Smith et al. 1998).

In this paper we will review the development of the R. exulans commensal model for tracking prehistoric human migrations in the Pacific. We will also discuss the projects that have been generated from some of the methodological problems encountered in the commensal study, specifically the development of species identification methods for both modern and ancient Rattus samples. Finally, we will briefly discuss the impact and possible applications of the results of these projects for island conservation programs.

Rattus exulans and the commensal model

It is generally accepted that the first people to settle the islands of Remote Oceania, or those islands east of the main Solomon Island chain, were those associated with the Lapita culture. These Lapita people transported with them, in their canoes, dogs, pigs, chickens and rats, and introduced them to the pristine island environments they settled. Lapita settlement resulted in the relatively rapid settlement of islands as far east as Samoa and Tonga and it is from these Lapita settlements that the ancestors of the Polynesians originated. Polynesians continued to transport the commensal animals throughout the Polynesian triangle.

The first animal that was used to develop this commensal model for Pacific settlement was the Pacific rat, R. exulans, or kiore as it is known in New Zealand. This rat was chosen for a number of reasons. First, it was the most widely distributed of all of the commensal animals that were transported by Pacific peoples. R. exulans bones are found in early if not the earliest archaeological layers throughout most islands of Polynesia. They are also found in early layers of most Lapita sites in both Near and Remote Oceania. Extant populations are still found on most islands across the Pacific, and since R. exulans are a different species from those rodents introduced by Europeans (Rattus rattus and Rattus norvegicus), they do not interbreed with those later arrivals. Unlike the dogs, pigs and chickens carried by Pacific peoples, that have since interbred with European introduced animals, the R. exulans found on Pacific islands today are the direct descendents of those rats introduced by the early Pacific colonists. In addition, it appears that they have not been transported in historic vessels and therefore their distribution remains directly related to prehistoric human dispersal.

Study of mtDNA variation in Polynesian R. exulans

The first test of the commensal model involved the study of mtDNA variation in extant populations of R. exulans throughout Polynesia (Matisoo-Smith 1994; Matisoo-Smith et al. 1998). Samples were collected from Fiji, Samoa, New Zealand, the Cook Islands, the Society Islands, the Marquesas, the Kermadecs, the Chathams and Hawaii. A total of 94 samples were analysed for variation in 432 base pairs (bp) of the hypervariable control region within the mitochondrial genome. The results of the analyses were remarkably consistent with both archaeological data and oral traditions. A central region encompassing the Societies and the Southern Cook Islands was identified from which the other central East Polynesian populations were derived. Interestingly, the Marquesas did not appear to be part of this central “homeland region”, yet the Hawaiian R. exulans populations were clearly derived from those in both the Marquesas and the central homeland, which is consistent with linguistic and archaeological models for Hawaiian origins. The R. exulans populations in New Zealand also appeared to be the result of multiple introductions, most likely from this central homeland region. The Kermadec Islands seemed to have been a stepping stone location for movements between New Zealand and the central homeland as predicted by Irwin (1992). The introduction of rats to the Chathams was most likely the result of a single or very limited number of voyages from a single location, most likely the South Island of New Zealand (Matisoo-Smith et al. 1998, 1999).

Development of ancient DNA methods

Once it was shown that the commensal model for tracking migrations through Polynesia did work, there were a few more issues that needed to be addressed. For example, while the mtDNA phylogenies suggested that the New Zealand R. exulans were most likely derived from both Cook Island and Society Island populations, we could not identify the timing of those introductions. In addition, since R. exulans are no longer present on the North Island of New Zealand (due to competition with European rodents), we did not know for sure if the remnant R. exulans populations, found primarily on the off-shore islands around New Zealand, were truly representative of those populations initially introduced to the mainland. Similarly, other islands, such as Rapa Nui no longer had extant populations of R. exulans, so our study was limited. Luckily, the 1990s saw the rapid growth of ancient DNA studies, and we were able to develop and apply these methods to archaeological remains of R. exulans (Matisoo-Smith et al. 1997).

R. exulans and chickens were the only two commensal animals introduced to Rapa Nui and R. exulans bones are found in large numbers throughout early sites on the island. The potential impact of this apparently large rat population on the native flora of Rapa Nui, in particular on the Jubaea palm, has recently been discussed by Hunt (2007) who suggests that they may have contributed significantly to the ecological collapse there that has received so much attention (Diamond 2005). Unfortunately, mtDNA analyses of the archaeological exulans bones were unable to provide evidence as to the specific origin of the canoes that first introduced them because all belonged to the most common lineage (known as R9) found throughout central East Polynesia (Barnes et al. 2006).

Analyses of archaeological remains allowed us to further test the reliability of the commensal model in numerous ways. Comparisons between archaeological and extant mtDNA sequences in R. exulans from the Chatham Islands indicated that there was little variation in populations separated in time by 500 years or so, suggesting that there was little in situ evolution taking place over the relatively short periods of time represented in Polynesian prehistory (Matisoo-Smith et al. 1999). This study also allowed us to test Irwin’s (1992) ideas about island accessibility and the implications of accessibility on the number of R. exulans introductions to islands. It is most likely that islands that were accessible would receive more R. exulans introductions and therefore would possess higher levels of mtDNA variation. Similarly, islands that were isolated would receive fewer introductions and thus would have rat populations with lower levels of variation. Our results were consistent with this prediction. The R. exulans from the Chatham Islands, which are located in the roaring 40s and are particularly difficult to reach safely according to Irwin’s voyaging models, showed almost no variation. The Kermadec Islands, on the other hand, which Irwin argued were a stepping-stone island for colonising voyages to New Zealand and post-colonisation voyages between New Zealand and Central East Polynesia, showed a much higher degree of mtDNA variation in their R. exulans populations (Matisoo-Smith et al. 1999). The lack of variation identified in Rapa Nui rats was also indicative of a limited number of introductions followed by relative isolation of the island (Barnes et al. 2006).

Analyses of archaeological rat remains from New Zealand also demonstrated what we assumed was a likely situation—that there are ancient lineages in archaeological samples which are no longer present in extant R. exulans populations in New Zealand. We found, for example, that the rats on the North Island of New Zealand belonged to two major haplogroups, whereas the rats from the South Island looked more like the extant populations in that they belonged to only one of those major haplogroups (Matisoo-Smith et al. 2001; Matisoo-Smith 2002). The possible absence or extinction of ancient lineages in extant populations is a major problem facing many molecular studies that focus exclusively on modern populations to infer past behaviours or relationships.

Identifying more distant origins

The development and application of aDNA methods to R. exulans remains also opened up the opportunity to study the bigger question of the ultimate origins of Polynesian populations and perhaps even the origins of Lapita populations. With archaeological remains, museum samples and additional tissue samples of R. exulans from Near Oceania and Island Southeast Asia we were able to study mtDNA variation across time and space (Matisoo-Smith and Robins 2004). One of the major surprises of this larger study was that we were able to identify three distinct lineages of R. exulans in the Pacific region—identified as Groups I, II and III. The distribution of each of these haplogroups was fairly well defined geographically: Group I rats were found in the region spanning the Philippines, Borneo and Sulawesi. Group II rats were found from the Philippines through to New Guinea and as far east as the Southeast Solomon Islands, and perhaps further into Remote Oceania. Group III rats were found almost exclusively in Remote Oceania—including Vanuatu, New Caledonia, Fiji and throughout both Polynesia and Micronesia. These Group III rats, one might suggest, were most likely dispersed as part of the Lapita expansion into Remote Oceania. We were rather shocked, however, by the fact that there appeared to be no Group III rats anywhere in Near Oceania except from the island of Halmahera in Wallacea. This was particularly surprising given the accepted view linking the Lapita dispersal in Remote Oceania to the earliest Lapita settlements in the Bismarck Archipelago, in Near Oceania.

We decided that the surprising lack of connection between Near and Remote Oceanic lineages of R. exulans could have three possible explanations. First, the distribution could mean that those human populations who introduced Group III rats to Remote Oceania did not pass through Near Oceania, in which case we had to seriously either reconsider our current ideas about Pacific prehistory and the Lapita culture, or we would have to reject the premise that R. exulans were transported by the first colonists into Remote Oceania. An alternative explanation, and one that appeared to be much more likely given the archaeological and linguistic evidence linking the colonisation of Remote Oceania with Near Oceania, was that our sampling of populations within Near Oceania was incomplete and that there were indeed Group III rats in the region. This promulgated our most recent research project which focuses on more precise and directed sampling of R. exulans populations throughout Near Oceania, but particularly on the most likely “Lapita target” islands in the Bismarck Archipelago. In the last 2 years we have collected both extant and archaeological R. exulans remains from many islands including Manus, New Ireland, Lihir, Tabar, New Hanover and some of the small islands in the St. Matthias group. Interestingly, we are now finding some overlap between Type II and Type III R. exulans in Near Oceania which may help us tease apart and better understand various episodes of human migrations and rat introductions in the region.

Species identification

As part of this extended research focus on R. exulans populations in Near Oceania and Island Southeast Asia we encountered another difficulty which opened up a new area of research for our group: the problem of species identification. In Polynesia and throughout most of Remote Oceania, there are a very limited number of rodent species present. In Polynesia, R. exulans was the only rat that was introduced prehistorically. European ships later brought new species, specifically Rattus rattus and Rattus norvegicus and the house mouse (Mus musculus). These four species are fairly easy to distinguish from one another morphologically (Cunningham and Moors 1983, McCormack unpublished data), though not always with 100% reliability, particularly when only working with skeletal remains. In western Micronesia, the Reef/Santa Cruz Islands in the Southeast Solomons, Vanuatu and Fiji, prehistoric voyagers introduced other rat species in addition to R. exulans. Archaeological remains of the Asian rat, Rattus tanezumi predate the appearance of R. exulans in western Micronesia (Wickler 2004), and it appears that Lapita canoes carried a New Guinea native rat, Rattus praetor as far as Vanuatu and Fiji (White et al. 2000). The islands of New Guinea and those of Southeast Asia of course have numerous native rodent species and are in fact the two most speciose regions for the genus Rattus (Musser and Carleton 2005). It has been estimated that there are between 11 and 14 native species of Rattus in New Guinea with another five species introduced to the region by humans (Musser and Carleton 2005; Taylor et al. 1982). It is not surprising therefore that with all of the additional possible species appearing in both the archaeological record and as fresh tissue samples, species identification was becoming a problem for our commensal studies. We therefore set out to see if we could develop an mtDNA database to allow us to reliably identify the rat species we were collecting (Robins et al. 2007).

Our lab obtained DNA sequence from tissues of 118 rats which included 18 named species and 1 sub-species of Rattus from Island Southeast Asia, New Guinea, Australia and several Pacific Islands as shown in Fig. 1 of Robins et al. (2007). We then built a reference phylogeny using DNA sequence from three regions of the mitochondrial genome, d-loop, cytochrome b and cytochrome oxidase I, and developed a rat identification system similar to the DNA Surveillance project designed for marine mammal identification (Ross et al. 2003). The phylogeny from the Bayesian partitioned likelihood analysis (Huelsenbeck and Ronquist 2001) of the concatenated sequences of all three regions of the mitochondrial genome (approximately 2,000 bp in total) resolved several well differentiated clades.

The phylogeny in Fig. 2 of Robins et al. (2007) showed a major divergence that separated the rats of Asia and Island Southeast Asia from those of New Guinea and Australia and also resolved 16 well supported clades. The term “nominal” species was used for the rat identification as determined by the collectors and or museums who provided us with the tissue samples because these identifications were not always consistent with the phylogenetic identifications. Nevertheless, 63% (10/16) of the clades had the same nominal and phylogenetic identification and these were largely compatible with the revised Rattus taxonomy of Musser and Carleton (2005).

The inconsistencies between the nominal and the phylogenetic species assignments were thought to be due to a number of contributing factors: (1) The complex and changing taxonomy makes species assignment difficult and can result in name confusion, e.g., Musser and Carleton (2005) list 49 synonyms for R. exulans and over 80 for R. rattus; (2) The lack of good morphological characters makes many rats difficult to identify even from whole carcases. The clade labelled exulans included all the R. exulans samples and also, almost certainly because of misidentification, one sample each of R. steini and R. verecundus; (3) Some species appeared to be over split, e.g., the clade labelled PNG I contained six nominal species of native New Guinean rats, R. mordax, R. niobe, R. novaeguineae, R. praetor, R. ruber and R. steini all of which have in the past been classified as subspecies of R. ruber (Taylor et al. 1982; Musser and Carleton 2005); (4) Some species appeared to be undersplit e.g., R. tanezumi samples fell into three different clades although the singleton that was in the clade labelled tiomanicus was probably misidentified.

One question that has often confused researchers is the description and taxonomic status of R. tanezumi—specifically whether it is a subspecies of R. rattus or indeed a distinct species. Musser and Carleton (2005) identify seven groups within the Rattus genus. One of these, the Rattus rattus group, contains approximately 21 different species including R. rattus and R. tanezumi. Within the species R. rattus, two subgroups were identified based on chromosome number: the Oceanian or European type that has 2n = 48 and the Asian type with 2n = 42 (Yosida et al. 1974). It is this Asian type that Musser and Carleton (2005) have designated R. tanezumi. This Asian R. tanezumi is indigenous to Southeast Asia, but it is also found in Japan, the Philippines, Island Southeast Asia, New Guinea and several Pacific islands mostly in Near Oceania and Micronesia, though it has been reported in Fiji (IUCN 2007). The European/Oceanian variety, R. rattus, is believed to have originated in India, reaching Europe by the third century AD from where it was taken in sailing vessels along the trading routes, around the world (Innes 1990; Atkinson 1985). When R. rattus is introduced to locations where R. tanezumi is already present, R. rattus is generally only found around the port areas. In the tree presented in Robins et al. (2007) the European/Oceanian variety is most likely represented by the rattus I clade (which includes samples from New Zealand, Samoa, the Society Islands and coastal New Guinea) and the Asian type is represented by the tanezumi clade (made up of samples from Japan, Hong Kong and Indonesia). R. tanezumi is thought to be a complex; and Musser and Carlton suggest a two taxon division is probably appropriate. The two clades labelled tanezumi and diardii, each containing a number of specimens, offer support for this hypothesis (Robins et al. 2007).

It is clear that much more work needs to be done to clarify a number of aspects of Rattus taxonomy, but the sequence database we have built and the results of our phylogenetic analyses demonstrate the value of this approach not only for commensal studies and understanding prehistory, but perhaps more importantly for ecological and conservation studies. The identification of rat introductions to islands is increasingly important as islands are now being cleared of rats and set aside for bird conservation and recovery programmes. Regular monitoring of these islands is necessary in order to identify any possible re-introductions. The ability to quickly identify either trapped rodents or even rodent droppings found on important rat-free islands can help identify the source of that introduction. We have so far been able to assist the New Zealand Department of Conservation and other organisations by using our species identification database to help identify and in some cases source the introduced rodents. The database and tree-based species identification programme ‘What rat is that?’ is now publicly available on-line as part of the DNA surveillance project at http://www.dna-surveillance.auckland.ac.nz.

Conclusion

It has been nearly 15 years since we started working on genetic variation in R. exulans and the value of the commensal approach for tracking human presence on and paths of migrations to Pacific islands has been clearly demonstrated. A most interesting current and future development for better understanding the human settlement of the Pacific is the comparison of mtDNA phylogenies of R. exulans and the other Pacific commensal animals, the dog, pig, and chicken, as well as commensal plants such as taro, sweet potato and bottle gourds. Each of these commensal stories can be telling us about various human activities and voyages, all leading to a better understanding of human mobility in Pacific prehistory. In addition much will be illuminated by fully considering and comparing the commensal genetic data with the rapidly increasing human genetic data from Pacific and Island Southeast Asian populations. We are already recognising through commensal studies that the history of human settlement of the Pacific is much more complex than some early models suggest (Matisoo-Smith 2007). Additional studies will most likely only add to our appreciation of how complex that history was.

Perhaps the surprising additional benefit from the development of the commensal model is the potential use of the large amount of genetic data generated. As many of the papers in this volume attest, the arrival of rats on islands has had a significant negative impact on the native flora and fauna. Understanding not only the timing and sources of those past rodent introductions helps us to evaluate and document the impacts of prehistoric human presence on islands. This in turn may also allow us to model and manage island biodiversity more effectively today and in the future. The ability to quickly, efficiently and reliably identify invasive rodent species and also potentially identify the source population could be valuable for management and conservation agencies in the Pacific and many other island environments.