Population structure and migration in the Eastern Highlands of Papua New Guinea, a region impacted by the kuru epidemic
Liam Quinn Jerome Whitfield Michael P. Alpers John Collinge 6 Garrett Hellenthal 6 Simon Mead 6
Open Access Published: March 19, 2024
Summary
Populations of the Eastern Highlands of Papua New Guinea (EHPNG, area 11,157 km2) lived in relative isolation from the rest of the world until the mid-20th century, and the region contains a wealth of linguistic and cultural diversity. Notably, several populations of EHPNG were devastated by an epidemic prion disease, kuru, which at its peak in the mid-twentieth century led to some villages being almost depleted of adult women. Until now, population genetic analyses to learn about genetic diversity, migration, admixture, and the impact of the kuru epidemic have been restricted to a small number of variants or samples. Here, we present a population genetic analysis of the region based on genome-wide genotype data of 943 individuals from 21 linguistic groups and 68 villages in EHPNG, including 34 villages in the South Fore linguistic group, the group most affected by kuru. We find a striking degree of genetic population structure in the relatively small region (average FST between linguistic groups 0.024). The genetic population structure correlates well with linguistic grouping, with some noticeable exceptions that reflect the clan system of community organization that has historically existed in EHPNG. We also detect the presence of migrant individuals within the EHPNG region and observe a significant excess of females among migrants compared to among non-migrants in areas of high kuru exposure (p = 0.0145, chi-squared test). This likely reflects the continued practice of patrilocality despite documented fears and strains placed on communities as a result of kuru and its associated skew in female incidence.
Snip…
Discussion
We have presented an in-depth study of the population structure of a remote highland region of PNG, based on a dataset that includes most of the linguistic groups and extensive sampling by village, affording a level of resolution that has not previously been possible. We show that despite covering an area of only 11,157 km2, roughly the size of the island of Jamaica, the genetic differentiation between linguistic groups in the EHPNG region is strikingly high (maximum FST = 0.066, average FST = 0.024, average distance 45 km). This differentiation is comparable to that previously reported for the entire PNG Highlands region,19 of which EHPNG is only a small sub-region. Furthermore, our analyses reveal the presence of complex population structure even at dialect and village level. While of interest for understanding the origins of modern populations, these findings also provide the background for study of the genetic impact of the large-scale prion disease epidemic, kuru.
A key question is what factors have contributed to such strong population structure in such a small region. One possible factor is that while the Western Highlands had optimal conditions for taro cultivation, which originated there, whereas the conditions were suboptimal in EHPNG.10 Indeed, some groups in EHPNG have been described as having been “proto-agriculturalists” retaining elements of hunter-gatherer subsistence in their lifestyles.10 Analogous to other hunter-gatherer groups, most EHPNG linguistic groups have greatly reduced population densities compared to other highland regions1 and lower historical effective population sizes.19 These reduced population densities have likely led to increased effects of genetic drift between groups.
In addition to geography, overall broad-scale population structure correlates with linguistics. For example, the best estimated linguistic tree for these groups analyzed shows the Pawaian and Anga as outgroups (Figure 1C), echoing the genetic analysis. We even see examples of fine-scale parallels between linguistic and genetic differences in the Fore dialect groups. However, the correlation between linguistic groupings and geographical regions makes it difficult to disentangle the relative role of these two factors. When more linguistic data become available for these populations in the future, approaches that quantitatively explore these relationships will add greatly to our understanding of these dynamics.
The colonial era definition of linguistic groups, which has been used in previous genetic studies, does not always satisfactorily describe population structure in the region. For example, linguistic groups are barely distinguishable from one another genetically in the northwest of EHPNG. This relative homogeneity may be due to more intensely practised agriculture and higher population densities in this sub-region,35 where the geography is different, with wide valleys in contrast to the highly dissected terrain in the southeast. Furthermore, in the village-level analyses of the Fore, we found genetic affinity of some villages to be closer to non-Fore groups. This observation is not completely surprising, as it is well understood that the clan structure of political, economic, and social unions that comprised the pre-colonial landscape in EHPNG often spanned linguistic group boundaries. This observed signature (for example as observed with Ilesa in the South Fore, Figure 3) could also be the result of a village founding event when a whole village is uprooted (e.g., due to conflict) and moves considerable distance to new territory, with members acquiring the language of their new residence. Such founding events in the past have been observed in the anthropological record within EHPNG.12
Several analyses revealed the presence of recent long-distance migration in the region. For instance, we found three potential migrants from the Fore linguistic group into the Pawaian (Figure 1A). One individual moved to the Pawaian linguistic area after marrying into a family there. However, marriages across the Fore/Pawaian divide are believed to be very rare or possibly nonexistent in pre-European-contact times, due to the considerable barriers of endemic warfare and extreme terrain. Consistent with this we found that all three observed migrants were born after European contact (although these individuals may be the descendants of migrants), which resulted in a cessation of warfare and the development of transport infrastructure which may have facilitated these movements. Hence it is possible that such long-distance migration is a recent phenomenon, consistent with the two groups being so genetically distant.
A final example demonstrating that analysis of population structure using linguistic group labels is not fully satisfactory was where we observed a clear resolution of village differences based on South Fore dialect spoken within the linguistic group (Figure 3C). This suggests that pooling all South Fore into a single population may not adequately capture population structure, though in this case the small genetic difference does correlate with a small linguistic one. Certainly, it reveals the dynamic and ongoing processes of cultural and demographic change that has been unfolding in the region. Given the small genetic differences between the fineStructure clusters that represent the two dialect groups, the small geographic differences between different dialect villages, and the fact that marital exchange and migration was known to occur between villages across the dialect divide, it is likely that such a split was recent in origin. This echoes oral origin histories held among the Fore that details the expansion and fragmentation of the Fore people into the three distinct dialect groups.
One of the population structure patterns that very clearly follows the linguistic groupings is that the Anga and Pawaian linguistic groups appear highly genetically distant not only from all other groups, but also from each other. Clusters comprising these groups had highest genetic similarity with groups outside of EHPNG, rather than EHPNG neighbors, in contrast to all other EHPNG clusters except a cluster of northwest individuals (EHPNG cluster 10) who were genetically related to the neighboring Chimbu. In the case of the Pawaian, the inferred closest ancestry source was the southern Kiwai, a coastal population more than 300 km from EHPNG. Interestingly, the Pawaian linguistic group is known to live semi-nomadically in forests at lower elevations and lower population densities than the rest of the region,1 which may in itself explain why they have ended up somewhat genetically distinct from the rest of the EHPNG groups. The oral histories held by the Pawaian speak of originating from coastal regions and undertaking long migrations through uninhabited forest regions.26
Our results do not support significant genetic influence on EHPNG populations other than the Anga and Pawaian from outside of the PNG highlands. In particular, unlike in many other regions of the world that have been colonized by countries with people of European descent,36 we found no clear signs of European ancestry in EHPNG individuals (Figure S5). And while we did in our ADMIXTURE analyses observe some signatures that are consistent with a few individuals having a small amount of admixture with people of East Asian ancestry, this could just as well be caused by other PNG populations not being represented in those analyses.19 Our findings of support previously suggested population histories1,19 with an expansion of groups (“neolithic expansion”) emanating from the Western Highlands as a result of the development of taro agriculture and displacement of previous groups that lived there, possibly ancestral to the Anga who have greatly distinct ancestry profiles in our analyses and who now live in the southern fringes of the region. In fact “Anga-like” artifacts, believed to be ancient, have been found as far north as in the Kamano linguistic group, reflecting a more widely dispersed settlement in the region.1
Our observation of recent migrant individuals in multiple analyses allowed the examination of the impact of kuru on migration dynamics in the region. While we observed a higher proportion of females among migrants relative to non-migrants throughout EHPNG, likely due to the general practice of patrilocality, the largest and only significant skew toward female migrants was observed in areas of high kuru incidence. This observation is consistent with accounts from the region that notes during the epidemic the near absence of adult women in villages with high kuru incidence.24 Men would frequently marry multiple times as a result of their previous wives dying from kuru, and strains were also placed on communities as a result of increased child care burden.23 Thus it seems plausible that the need to replace lost adult women and mitigate the impact of kuru on Fore society could have led to an excessive inflow of recent female migrants into villages with high kuru incidence. However, we note that previously the opposite has been reported: that kuru led to a decrease or even a complete cessation of intermarriages between the Fore and neighboring communities, because those communities linked kuru to sorcery, which made them fearful of the Fore.37 In our data, we observed no evidence either for less overall migration into areas with high kuru incidence or for a stop of patrilocal practices in these areas. On the contrary, we observed a significant bias toward females among migrants into high kuru incidence areas. Moreover, the observed difference in the proportion of females among migrants versus migrants is ∼25% higher in the “high” incidence kuru areas relative to the “zero/low” kuru incidence areas. While this difference was not statistically significant using the data here, it may reflect kuru causing a sex bias beyond that driven by patrilocal practices. We note that our approach considers only genetically distinct individuals to be migrants. This means we are also counting migrants that came from greater distances away than would have typically been the case for marital exchange. Hence, it is possible that bachelors within highly kuru-affected communities have sought wives from further afield than usual due to the lack of availability of potential wives more locally and that this led to the observed sex-bias in migration in the high kuru incidence areas. However, additional data and analyses are necessary to validate this. What our current data and analyses do suggest is that there was sex-biased migration into the high-kuru areas despite documented fears and strains placed on communities as a result of kuru.
In summary, our results suggest that the observed population structure is not driven by admixture from outside the highland PNG region, which is consistent with the historical record of the EHPNG region being isolated until recently. Also, while the population structure does to some extent mimic the linguistic groupings in the area, we observe several patterns of population structure that suggest that the different linguistic groups are not entirely genetically distinct and isolated from each other. This is consistent with previous knowledge of clans playing a key role in the cultural grouping in the area and of the presence of cultural features that aided possible migration between neighboring linguistic groups in the region. Finally, we observed signs that long-distance migration has taken place, likely in more recent times. Importantly this, in combination with the understanding of population structure, has permitted an analysis of sex-biased flows of migration that are likely to have been impacted by kuru. This highlights that it is essential to understand the population structure of a region prior to attempting to investigate hypotheses regarding the impact of epidemics on affected populations.
The population structure of EHPNG reveals a complex multi-layered set of factors that have caused high population differentiation, likely including both geographic and cultural factors. Furthermore, it suggests that the current population structure may still be evolving. Overall our results demonstrate that simplistic descriptions of the population structure in regions like EHPNG based on linguistic groupings presumed to be static are likely to neglect the far richer texture of dynamic forces and history that has shaped communities.
Sent from my iPad