East African origins for Madagascan chickens as indicated by mitochondrial DNA

The colonization of Madagascar by Austronesian-speaking people during AD 50–500 represents the most westerly point of the greatest diaspora in prehistory. A range of economically important plants and animals may have accompanied the Austronesians. Domestic chickens (Gallus gallus) are found in Madagascar, but it is unclear how they arrived there. Did they accompany the initial Austronesian-speaking populations that reached Madagascar via the Indian Ocean or were they late arrivals with Arabian and African sea-farers? To address this question, we investigated the mitochondrial DNA control region diversity of modern chickens sampled from around the Indian Ocean rim (Southeast Asia, South Asia, the Arabian Peninsula, East Africa and Madagascar). In contrast to the linguistic and human genetic evidence indicating dual African and Southeast Asian ancestry of the Malagasy people, we find that chickens in Madagascar only share a common ancestor with East Africa, which together are genetically closer to South Asian chickens than to those in Southeast Asia. This suggests that the earliest expansion of Austronesian-speaking people across the Indian Ocean did not successfully introduce chickens to Madagascar. Our results further demonstrate the complexity of the translocation history of introduced domesticates in Madagascar.


Mitochondrial haplogroup distribution patterns
Chickens in Madagascar cluster into two haplogroups: the majority (85%) of samples belong to haplogroup D and the rest belong to haplogroup E (figure 1). Similarly, in East Africa only haplogroup D and E chickens are observed in publically available sequences, with a positive association between haplogroup D frequency and latitude southwards. By contrast, haplogroup D is not observed in chickens from the Arabian Peninsula and western Asia: the haplogroup composition in these regions is dominated by haplogroup E. Within India, haplogroup E chickens are also observed at a high frequency (67%) with haplogroup D (22%) and all other haplogroups (11%) making up the balance. To the east in MSEA, all haplogroups (A to I) are observed, but the frequency of haplogroup D and E are dramatically lower in comparison to other haplogroups (4.8% and 2.4%, respectively). However, chickens from islands further east (in the Pacific Ocean) have a high proportion of haplogroup D (84%).

Population genetic structure
The overall genetic structure in the MDS plot using all haplogroups reveals structuring at a broad geographical scale (figure 2a). Distinctive clustering of populations from ISEA (shown in green) and from the Pacific (in blue) can be seen. African populations, including Madagascar, sit closer to South and West Asian populations, than ISEA populations. Madagascar falls closest to Malawi, which is the closest African population to Madagascar in the dataset. The other continental African populations show broad geographical clines (northern-southern cline follows high-low 'Dim 2' values). When only haplogroup D is used, the Madagascan samples form a distinct cluster with the geographically closest East African populations, Malawi and Zimbabwe (figure 2b).
The AMOVA performed on all chickens from Indonesia, South Asia, East Africa and Madagascar show population structure (table 2). However, the among-group variance components (regional groupings) are highest when only haplogroup D chickens are used. For haplogroup D samples, the among-group variance component is always significant when Indonesia is separated from Madagascar. The only non-significant among-group variance component is when South Asia was isolated from Indonesia, East Africa and Madagascar as a group. The variance components among the groups (regional grouping) are generally low when using only haplogroup E.

Population dynamics and genetic variability
The genetic differentiation observed from the F ST values (  (table 4). However, when using both haplogroup D and E, all consistently deviate from neutrality except Madagascar. These results support a model of demographic expansion of haplogroup D for each of the four regions. Furthermore, the highest level of genetic diversity is observed in South Asia and ISEA. On the other hand, Madagascar has the lowest diversity among the four regional groups being compared, with East Africa next lowest, suggesting a genetic bottleneck effect with every new introduction of chickens into an area.     that the initial chicken populations arrived in Madagascar via the East African coast. What is not known, however, is whether H45 represents the founding lineage, with in situ evolution responsible for the one and two base pair derivations radiating from H45. As the H45 haplotype is currently not observed outside of East Africa/Madagascar and there are no archaeological chicken remains representing the earliest chickens on Madagascar, we cannot tell where H45 originated. The strong phylogeographic signal within haplogroup D suggests that South Asia, rather than Southeast Asia, is the most likely recent source for East African and Madagascan chickens ( figure 3). Therefore, despite the clear linguistic and human genetic associations between Madagascar and Indonesia, the Austronesians do not appear to have successfully translocated chickens directly to Madagascar across the Indian Ocean. However, Austronesians may have transported chickens to Africa indirectly via the Indian subcontinent, rather than via the direct route across the Indian Ocean. Support for this theory can be inferred from the presence of Austronesian speakers in South Asia during the first millennium A.D. [21,58] and the fact that there are no high frequency central nodes in a star-like Indian cluster. Rather, Indian haplotypes belong to two of the high-frequency central nodes in star-like ISEA clusters.
It is interesting that haplogroup D chickens are not observed in the Arabian Peninsula and occur at low frequencies in northeast Africa, suggesting that chickens might have been transported via a direct sea link from India across the Arabian Sea to eastern Africa/Madagascar. Alternatively, Madagascan chickens could also have been transported along a coastal route through the Arabian Peninsula and northeast Africa but with the signal overwritten by subsequent and repeated translocations of haplogroup E chickens.
Haplogroup E is less common in Madagascar compared with haplogroup D. As haplogroup E lacks phylogeographic signature, most probably owing to modern-day translocations (electronic supplementary material, figure S2), it is difficult to derive fine-scale inferences based on haplogroup E other than establishing that it is also most probably South Asian in origin. Furthermore, it is difficult to ascertain whether the arrival of haplogroups D and E was contemporaneous. However, the most parsimonious explanation is that a mixed population of both haplogroup D and E chickens were transported from India to Madagascar via East Africa. Testing this hypothesis is difficult using existing samples. A more deliberate sampling regime, combined with nuclear genetic data and ancient DNA from archaeological samples, would help establish the full history of chicken translocation around the Indian Ocean rim.

Conclusion
Mitochondrial DNA data suggest that chickens were introduced into Madagascar from South Asia via East Africa. A scenario whereby chickens arrived in Madagascar along with the expansion of the Austronesian-speaking people directly across the Indian Ocean is not supported. However, it remains a possibility that Austronesian traders and mariners integrated South Asian chickens during their coastal voyages en route to east Africa and Madagascar. Additional sampling of chickens in Madagascar, increased genomic sequencing of existing samples and accessioning of existing mtDNA data from Mwacharo et al. [31] onto publically available databases will help refine our understanding of the ultimate origins of Madagascan chickens.
Ethics. The sampling of chickens in ISEA and the Pacific for this study was approved by the University of Adelaide