Ancient DNA from Guam and the peopling of the Pacific

Abstract

Humans reached the Mariana Islands in the western Pacific by ∼3,500 y ago, contemporaneous with or even earlier than the initial peopling of Polynesia. They crossed more than 2,000 km of open ocean to get there, whereas voyages of similar length did not occur anywhere else until more than 2,000 y later. Yet, the settlement of Polynesia has received far more attention than the settlement of the Marianas. There is uncertainty over both the origin of the first colonizers of the Marianas (with different lines of evidence suggesting variously the Philippines, Indonesia, New Guinea, or the Bismarck Archipelago) as well as what, if any, relationship they might have had with the first colonizers of Polynesia. To address these questions, we obtained ancient DNA data from two skeletons from the Ritidian Beach Cave Site in northern Guam, dating to ∼2,200 y ago. Analyses of complete mitochondrial DNA genome sequences and genome-wide SNP data strongly support ancestry from the Philippines, in agreement with some interpretations of the linguistic and archaeological evidence, but in contradiction to results based on computer simulations of sea voyaging. We also find a close link between the ancient Guam skeletons and early Lapita individuals from Vanuatu and Tonga, suggesting that the Marianas and Polynesia were colonized from the same source population, and raising the possibility that the Marianas played a role in the eventual settlement of Polynesia.

Keywords: Micronesia; Polynesia; ancient DNA; human settlement.

Fig. 1.

Map of the western Pacific, showing locations and areas mentioned in the text. The Inset shows the location of the Ritidian Site on Guam. Location names in red have been suggested as potential sources for the settlement of the Mariana Islands. Wallace’s Line divides biogeographic regions and lies at the boundary of the prehistoric continental landmasses of Sunda and Sahul. The dashed blue line indicates the boundary between Near and Remote Oceania: The islands of Near Oceania were colonized beginning 45 to 50 kya and involved relatively short, intervisible water crossings, while the islands of Remote Oceania required substantial water crossings that were not intervisible and that were not achieved until ∼3.5 kya or later. Red dots indicate the locations of the early Lapita samples from Vanuatu and Tonga; the blue arrow indicates the conventional route for the Austronesian expansion to the Bismarck Archipelago, which was then the source of initial voyages to Remote Oceania; the solid red arrow indicates the route for the settlement of the Marianas supported by this study; and the dashed red arrow indicates the potential additional contribution of Mariana Islanders to further settlement of the Pacific, suggested by this study.

Fig. 2.

PCA and ADMIXTURE analyses of the ancient Guam samples merged with modern samples genotyped on the Affymetrix 6.0 platform and with SGDP samples. (A) Plot of the first two PCs. The ancient Guam samples are projected. (B) ADMIXTURE results for K = 6. Population names are color-coded as in the PC plot.

Fig. 3.

PCA and ADMIXTURE analyses of the ancient Guam samples merged with Human Origins Array data for modern and ancient samples. (A) Plot of the first two PCs. Ancient samples are projected. (B) ADMIXTURE results for K = 9. Population names are color-coded as in the PC plot.

Fig. 4.

Outgroup-f3 and -f4 results for the relationships of the ancient Guam samples with other populations. (A) Outgroup-f3 results comparing the ancient Guam samples to other modern and ancient samples, with Mbuti used as the outgroup. Bars indicate 1 SE. Larger values of the f3 statistic indicate more shared drift, and hence a closer relationship with the ancient Guam samples. (B) results for an f4 test of the form f4(test, Kankanaey; New Guinea highlands, Mbuti). f4 values that are significantly different from zero are in red.

Fig. 5.

Tree and graph depictions of the relationships of ancient Guam, early Lapita, and select Asian and Oceanian populations. (A) Maximum-likelihood tree with two migration edges. All residuals (SI Appendix, Fig. S12) are within 3 SE. (B) Consensus graph with nodes present in at least 50% of the topology sets recovered with AdmixtureBayes. (C) Admixture graph obtained with qpGraph for the topology found by AdmixtureBayes with the highest posterior probability (SI Appendix, Fig. S13). The colored arrows in B and C indicate drift (ancestry) shared between the ancient Guam and early Lapita samples.