Testing migration patterns and estimating founding population size in Polynesia by using human mtDNA sequences

Sequences from 31 unrelated New Zealand Maori were identical to Polynesian sequences already known (2, 3) and had the characteristic one copy of the 9-bp repeat in the COII/tRNA^Lys intergenic region. We found sequence 1 (Table 1) 27 times, sequence 11 once, and sequence 15 three times. Combined with the 23 samples from Maori described in ref. 3, there is a total of 54 samples, but these still contain only four distinct sequences (haplotypes) between bases 16,189 to 16,388 (40). On the combined data set, one haplotype (sequence 1) occurs in 47 of the 54 people and sequence 15 five times. The frequencies of all haplotypes reported in different parts of Polynesia, together with some Melanesian samples, are shown in Table 1. In some cases, because not all publications report the same length of sequence, two or more haplotypes are identical for the region 16,189 to 16,388. Sequence 1 was found in 87% of the Maori (Ma) samples, and it occurs in 64% of the samples from eastern Polynesia (EP), in 56% of the western Polynesian (WP) samples, and in 23% of Melanesians (MN) sampled. The apparent heterozygosities [h = (1 − Σx_i²) (41)] for Maori, other eastern Polynesians, and western Polynesians are 0.233, 0.564, and 0.659, respectively, over this region of the mitochondrial genome.

Given the generally low diversity in Polynesians (2, 3), a low number of variants is anticipated in Maori, although the similarity was greater than expected. The lack of variability does not appear to be a sampling artefact. For the present sample, the mother’s and maternal grandmother’s iwi (tribal affiliation) and place of birth were recorded and showed a wide distribution from throughout New Zealand (Table 2) and from all but two of the major tribal areas (38, p. 52). Of these two areas, at least one would be well represented in the sample of ref. 3, which was collected in Auckland but without recorded iwi affiliation. In addition, the two groups of samples were collected in centers 600 km apart and both have a similar preponderance of the same sequence. Thus, we consider the sample to be representative.

Table 1 includes the overall frequency of haplotypes from two data sets for New Zealand Maori (54 sequences), two data sets for other eastern Polynesians (108 sequences), and three data sets for western Polynesia (73 sequences) and Melanesia (57 sequences of 87 total are included in Table 1; others are not directly relevant to this study).

A standard method for illustrating the level of diversity in a population is the pairwise distance distribution (or mismatch distribution) (42). Fig. 1 shows the results for Maori, other eastern Polynesians, western Polynesians, and Melanesians. New Zealand Maori sequences are the most similar to each other, then eastern Polynesian. Melanesian sequences show the highest diversity and are comparable with those from other regions of the world (see below). With one exception, Maori sequences are identical or have just 1 bp different. A single sequence (from ref. 3) lacks the 9-bp deletion and gives a small peak at 10 differences in this region of the D-loop. This bimodal distribution is more marked in the eastern Polynesian data and more so again in western Polynesian data, again resulting from the differences between sequences with and without the 9-bp deletion. The Melanesian sequences are more diverse again and have considerable diversity within groups lacking the 9-bp deletion.

The bimodal distribution could be interpreted as a “ragged” distribution (43) and could be used to favor a stable population size over a long time period. However, in the present work, given that humans have not been in Polynesia for a long time period and given the evidence for a small founding population relative to the present population size, an alternative hypothesis is more reasonable. This would be a fusion of two groups of people, probably in Melanesia (20) between the Austronesian-speaking ancestors of Polynesians (with a 9-bp deletion) and an earlier Melanesian population (without this feature).

Comparison of mitochondrial diversity with other parts of the world is shown in Fig. 2 for five groups: Maori, Polynesian, Asian, African, and the Turkana of East Africa. All major groups have higher diversity in sequences than Polynesians in general and Maori in particular. Overall, the results are consistent with the findings that mitochondrial diversity in humans is highest in East Africa and decreases steadily away from that center. The Turkana of East Africa have the highest sequence diversity of any group yet studied (44) yet have a similar population size to the Maori of New Zealand. Population size alone cannot be an explanation for the level of diversity of sequences of either group. The origin of a people and their migration history must be considered.

Simulation estimates of the numbers of women who originally arrived in New Zealand (the founding population) are shown in Fig. 3. The size of this founding population was varied from 4 (the minimum to give four haplotypes) to 250. For each population size, the number of times the final sample of 54 people would have 1 to 11 different haplotypes is shown. With small founding population sizes, only one or two distinct sequences are expected in the final sample. At higher founding population sizes, five, six, or more different sequences are expected, but even with a founding population of 250, only 5 or 6 of the 11 haplotypes in eastern Polynesia are expected. For a founding population size from ≈40 to 90, it is most likely that four distinct haplotypes will be found, and this is the “best estimate” when considering just the number, not the frequency, of haplotypes. This result has a wide confidence interval, so a more strict criterion then was used.

A tighter lower bound on the numbers is given in Fig. 4 where the frequency of finding four haplotypes with the very unequal distribution of frequencies is shown. This more strict criterion includes simulations where the product of the frequencies of the four haplotypes is ≤235 (47 × 5 × 1 × 1). A distribution of 46,6,1,1 also was allowed. The results (Fig. 4) give a slightly larger estimate of the founding population with the most likely number of women being 70. By using a one-tailed test, we found that the 95% confidence interval is 50 or more women and the 99% confidence interval is 30 or more (Table 3). The additional factor in this calculation (compared with Fig. 3) is that, with larger founding populations, the distributions of the 54 sampled haplotypes are more unequal than for small founding populations (data not shown). Thus, the two factors leading to the result in Fig. 4 are the frequency of finding four haplotypes in the final sample (Fig. 3) and an increased chance of an unequal distribution of haplotypes when the founding population is larger.

A more relaxed criterion also was simulated that allowed any sample with four haplotypes in which one of the rare haplotypes occurs twice and the other just once. These results are similar to Fig. 4 (data not shown) and support a similar size for the number of females in the founding population. It is apparent that the highly unequal distribution of haplotype frequencies is an unusual feature of the data. The main feature of the simulation results is that they are inconsistent with a very small population from just one or a small number of canoes but support most strongly a founding female population size of 50–100 and possibly more.

It is important to have an estimate of the sensitivity of the model to variations in the parameters; those in which small changes in their value lead to significant differences in output require close attention (39). For the sensitivity analysis, simulations were re-run: omitting one sequence from each of the four haplotypes observed in Maori; allowing one additional sequence for each haplotype; and adding a fifth haplotype. Each case was simulated 5,000 times, the most likely number of women was determined for the strict criterion (that includes frequencies of haplotypes), and the 1% and 5% cutoffs were estimated (Table 3). As expected, the most sensitive parameter is the number of rare haplotypes (see rows 4 and 8 in Table 3), but even if the Maori sample, just by chance, had not included one rare haplotype, the results still reject a single, or small number of, canoes. Similarly, if a fifth allele had been observed in Maori, it would increase the expected founding population size to over 100. This point is interesting because, during sequencing, one sample indicated an additional allele, but there was insufficient DNA to verify all positions in the sequence, so it was omitted from the analysis. Again, as expected, changes to the Maori sample (with four haplotypes) had more effect than gaining or losing a haplotype from eastern Polynesia (with 11 haplotypes), as shown by the last two rows in Table 3. Thus, the sensitivity analysis has been especially important in showing that our conclusions about the size of the maternal founding population are robust with respect to the frequencies of haplotypes in the samples.

As expected, only a subset of eastern (and western) Polynesian genetic diversity is represented in New Zealand Maori. Although a minimum of four founding female settlers would be needed for four haplotypes, it is most unlikely that all four haplotypes would survive during the phase of population expansion with sufficient frequency to occur in the final sample of 54. Many more settlers appear to have been necessary to explain the observed skewed distribution of the four haplotypes. By comparing the genetic diversity in eastern Polynesian, and in Maori, we estimate that the female founding population of New Zealand included ≈50–100 women. Because of the skewed distribution in Fig. 4 the upper limit (100) is less clear, but given the difficulty in sailing canoes to New Zealand, the value is reasonable [however, Anderson (27) has briefly considered a large founding population]. Our estimates are quite inconsistent with models (29, 30) that assume that only one canoe, or a very small number of canoes, arrived. In contrast, it is in very good agreement with a common understanding of the Maori oral tradition of 8–10 canoes with a small number of people (10–20 people, thus probably 5–10 females) per canoe (see refs. 32 and 33). The conclusion is consistent with more recent work that has tested the possibility of deliberate exploration and migration (16, 19).

Our estimates of the founding female population assume that the samples have no hidden biases favoring any particular haplotype. As discussed earlier (45), the mitochondrial genome is expected to be at least as well dispersed as nuclear alleles in that, in nearly all earlier societies, women transferred much more frequently between groups and settlements than men (46). In addition, adoption of children has been reported to be widespread throughout Polynesia (10, 11), and this also would be expected to reduce any local concentrations in the distribution of particular sequences. It is likely that several women in a canoe would be related maternally, but this too would make it less likely that a small number of founders would give the observed distribution, including the skewed distribution. The presence of older women in canoes also would tend to increase the size of our estimate of founding population size. Overall, our estimate of 50–100 women seems conservative in several respects.

Could there have been earlier settlement by a small population that grew more slowly? A founding population of 50 has been suggested to increase to ≈400 after 200–400 years, allowing for a longer prehistory than proposed (32). There is no good evidence yet for permanent earlier settlement (see the Discussion in ref. 33), and given an abundant food supply for the first settlers and no new indigenous diseases, an initial rapid population expansion may be expected until resources became short (28, 47). The recent discovery of bones of kiore (the Pacific food-rat) dating from 1,800 BP (25) indicates that there may have been early Polynesian contact 1,000 years before permanent human habitations are represented in the archaeological record (24).

However, with the present data, an early settlement combined with a slower population increase makes it even less likely that there was a small founding population—rare alleles are more likely to be lost during a slower increase in population than during a faster increase. This point is shown in row 9 of Table 3, where the founding population was allowed to expand over 40, rather than 30, generations; a greater number of women would be required to give the observed Maori diversity. So, if there were earlier settlement, it would require a founding population of over 100 women, but this would be more likely to leave an archaeological record. Thus, we consider a small, early population to be incompatible with the data.

Several explanations could be proposed for the very unequal frequencies of the four haplotypes (namely 47, 5, 1, 1). These include one or more factors involving chance, continuing migration, and genetic and/or cultural selection as follows:

1.

The group of women setting sail for Aotearoa may, by chance, have had a different frequency of haplotypes from the population in the islands of origin.

2.

Diversity of the mtDNA studied in Maori may reflect truly the diversity found in central eastern Polynesia at the time of migration. However, subsequent migrations from western to eastern Polynesia may have produced the higher diversity currently shown, for example, in the Cook Islands.

3.

Canoes with females with the rare haplotypes may have arrived in New Zealand later, and consequently they expanded less than haplotypes arriving earlier. Maori oral tradition (quoted in ref. 48) records that some canoes arrived after earlier settlement. For example “Wairake had sailed with her father… and she and her descendants had intermarried with ‘the early tribes of Tuhoeland’ to establish a new tribe… . ” Our simulations did not include the effect of differing times of arrival.

4.

Lineages with other mtDNA sequences found in eastern Polynesia may not have survived either the journey to New Zealand or conditions after reaching it. Selection favoring the common haplotype may have occurred. There have been suggestions that certain genetic variants may be advantageous in surviving long canoe voyages, a version of the “thrifty gene” hypothesis (49, 50). In addition, mitochondrial variants have been linked to some forms of diabetes (reviewed in ref. 51), which might play such a role.

5.

High ranking females may have carried the common haplotype in the original population, and there was differential survival of offspring favoring their children. This would give a cultural linkage to survival, rather than a genetic advantage.

Similar explanations could be invoked for the increase in frequency of 9-bp deletions across Polynesia; from Table 1, Maori mtDNA has 98% sequences with the deletion, eastern Polynesia 80%, western Polynesia 89%, and the Melanesian sample 41%.

The extent of genetic variability and admixing of populations is also of interest in relation to exploring genetic contributions and linkages to nonsusceptibility and susceptibility to specific diseases. Maori have a lower incidence than non-Maori for some diseases, for example skin cancer (52), phenylketonuria (53), and cancer of the large bowel. Examples of higher incidence include some infections, inflammatory and immune diseases, cancers, and metabolic diseases (for example, pneumonia, hepatitis, tuberculosis, ear diseases, nephritis, asthma, rheumatic and hypertensive heart disease, cancers of the stomach, liver, lung and cervix, and diabetes, gout, and obesity) (52). In general, it is thought that the higher incidences are caused by socioeconomic factors and that there is negligible contribution from genetic susceptibility (52). However, few genetic investigations have been carried out with Maori samples. These include studies related to the immune system [T cell receptor polymorphism (54) and HLA polymorphism (see, for example, refs. 21, 55, and 56)], globin gene markers (57), hemophilia (58), minisatellites (23), and variable numbers of tandem repeats (59). In general, these studies show genes with restricted diversity and types, combinations, and/or frequencies different from Caucasians. There is as yet little specific information on whether genetic features may influence susceptibilities to major diseases.

As a general comment, we note that genetic data, including interpretation and analysis of DNA sequences, are just one aspect of the multifaceted and profound issues of relationship, kinship, and differences between human groups, their cultures, and perceptions of self-identity. Our data and the interpretation thereof do not include specifics of traditional knowledge and genealogies and therefore offer neither affirmation nor denial. Experts in historical and cultural matters can use, re-interpret, reject, or add to our analysis according to their different perspectives. In particular, we welcome additional information refining our understanding.

Although this work is based on just one section of Polynesia, this region is perhaps the only place at present where hypotheses about the processes of migration of early humans into previously uninhabited regions can be tested quantitatively. For example, models for the founding populations of Australia/New Guinea (60) and the Americas (see, for example, ref. 61) also assume small founding populations and bottlenecks, but these models are not yet quantitative. Polynesia is an excellent region to test models of earlier human expansions.

We thank Professor Paul Harvey for computer facilities for analyses, Z. Velickovic for some 9-bp analyses and for collecting some samples, Elizabeth Watson, John Holloway, and Gina Lento, who assisted with methodology, Elizabeth Watson for supplying comparative data, Professor Mason Durie for initial consultation and interest and for commenting on the final manuscript, and particularly the enthusiastic volunteers who gave samples and expressed strong interest in this study. We acknowledge salary for R.P.M.-M. from the Health Research Council of New Zealand.