Missing Links On The Journey From Borneo To Madagascar

A linguistically Austronesian population from Borneo Indonesia (with strong genetic ties to Eastern Indonesia as well) settled Madagascar sometime in the vicinity of 2000 to 1000 years ago, split roughly 40% Indonesian-60% East African, with the African language vanishing on Madagascar.

Until now, however, there have been precious few bits of evidence for where they went along the way on this epic journey (there is some indication that some Southeast Indian women were added to the group en route). 

A new paper, using very solid method that fully leverages what we know already, however, changes all of that by finding residual traces of Austronesian genetics in South Arabia and East Africa, both of which are natural stopping points en route to Madagascar. The paper takes the rare step of making high quality use of interdisciplinary research from linguistics, archaeology, anthropology and history to supplement and inform its genetic analysis.

The Austronesian dispersal across the Indonesian Ocean to Madagascar and the Comoros has been well documented, but in an unexplained anomaly, few to no traces have been found of the Austronesian expansion in East Africa or the Arabian Peninsula. To revisit this peculiarity, we surveyed the Western Indian Ocean rim populations to identify potential Austronesian genetic ancestry. We generated full mtDNA genomes and genome-wide genotyping data for these individuals and compared them with the Banjar, the Indonesian source population of the westward Austronesian dispersal. 

We find strong support for Asian genetic contributions to maternal lineages and autosomal variation in modern day Somalia and Yemen. Surprisingly, this input reveals two apparently different geographic origins and timings of admixture for the Austronesian contact; one at a very early phase (likely associated with the early Austronesian dispersals), and a later movement dating to the end of nineteenth century. These Austronesian gene flows come, respectively, from Madagascar and directly from an unidentified location in Island Southeast Asia. This result reveal a far more complex dynamic of Austronesian dispersals through the Western Indian Ocean than has previously been understood, and suggests that Austronesian movements within the Indian Ocean may have been part of a lengthy process, probably continuing well into the modern era.

Nicolas Brucato, et al., "Evidence of Austronesian genetic lineages in East Africa and South Arabia: complex dispersal from Madagascar and Southeast Asia" Genome Biology and Evolution, evz028 (February 4, 2019) https://doi.org/10.1093/gbe/evz028 Full text here.

The new study suggests a presence in what is now Yemen and Somolia, ca. 1500-1800 years ago, followed by an arrival in Comoros and Madagascar ca. 800-1200 years ago. (Other sources suggest an arrival in Madagascar closer to 1500 years ago)  which ties up the dates even more tightly (or perhaps even earlier). (One is tempted to think that the same climate event the helped to cause the Roman empire to collapse may have also motivated this journey in some way, given the increasingly precise timing.)

While the authors don't jump to this conclusion, these intermediate traces of Austronesian DNA due to admixture in the right time period suggest that the Austronesian seafarers who eventually settled in Comoros and Madagascar probably followed something close to a coastal route around the Indian Ocean, rather than traveling straight across the Indian Ocean from Indonesia.

The most direct route that they could have taken which would leave the known genetic traces is illustrated with black arrows (supplied by me) on the map below (and in truth, the initial route demonstrated by the first long black arrow probably actually more closely tracked the coastline). The green arrow from Indonesia to Madagascar is now disfavored.

Of course, this something we could have guessed already based upon the fact that the modern people that the African component in Malagasy people most resemble is East African Bantus (probably close to modern Kenyan Bantus than to modern Somalis, in fact), suggesting yet another stop on the African coast before arriving in Comoros and Madagascar. Another study suggests that the African component may be closer to modern South African Bantus than to East African Bantus, but this might simply reflect the fact that modern East African Bantus have Nilo-Saharan admixture that is absent in modern South African Bantus.

We also know that it is likely that there was a stop in India because the Malagasy people have a small percentage of South Asian admixture, again, reinforcing the coastal journal model.

The study also identifies an individual whose origins appear to derive from a subsequent back migration from Comoros or Madagascar to Yemen in the 1800s, likely corresponding to historically attested trade networks associated to some degree with the ethnicity of Yemenis among whom the individual with the more recent ancestry was found.

More details from the body text are found below the fold.

The body text explains one of the key genetic tracers of this migration (citations omitted at 5-6):

[T]he key genetic marker of the westward Austronesian expansion - [is] the mitochondrial DNA (mtDNA) haplogroup B4a1a1b (Malagasy motif), the Indian Ocean variant of the Polynesian motif (B4a1a1 haplogroup), which has been previously identified in Madagascar and is likely present in the Comoros This lineage is the only Austronesian-specific marker that is both the main Asian maternal lineage found in Madagascar and the main lineage associated with the Austronesian expansion throughout the Indian Ocean and Oceania The Malagasy motif (characterized by polymorphisms C1473T and T3423A) has only been found in Madagascar (ranging from 10 to 50% in frequency), and to date has not been found anywhere else, including within Indonesia and in other regions influenced by the Austronesian dispersal. Its direct precursor, the Polynesian motif (characterized by the polymorphisms A14022G, T16217C, A16247G and C16261T) is largely restricted to the east of the Wallace’s Line, from eastern Indonesia to the Pacific islands, with sporadic occurrences further west. 

Here, we report the first mitogenomes and associated autosomal genome-wide data of individuals from east Africa and the Arabian Peninsula who have been identified to carry the Malagasy motif. To investigate broader questions about Austronesian gene flow within the Indian Ocean, we also present an analysis of the mitogenomes of the Banjar population of Southeast Borneo, who are considered to be descendants of the ancestral population that migrated to Madagascar and Comoros, and explore how this population connects with the newly discovered Malagasy motif lineages in Somalia and Yemen.

The results section then goes on to explain (citations omitted):

The geographic distribution of complete mtDNA sequences affiliated with the Malagasy motif (B4a1a1b haplogroup), its precursor haplogroup B4a1a1 (the Polynesian motif) and related subclades show a clear phylogeographic distribution pattern. On the one hand, the Polynesian motif and all of its sub-clades, with the exception of the Malagasy motif sub-clade (B4a1a1b), are only found east of Wallace’s line (with minor exceptions). On the other hand, the Malagasy motif is only detected in the western part of the Indian Ocean (Madagascar, and now Somalia and Yemen). 

In Island Southeast Asia, the Banjar, who have been shown to be linguistically and genetically the descendants of the ancestors of the Asian background found in Madagascar and the Comoros, carry no maternal lineage affiliated with the Malagasy motif, and only one individual (out of 99) carries a maternal lineage affiliated to the Polynesian motif. The other Banjar mitochondrial haplogroups are similar to the observed haplogroup diversity in other populations in Indonesia.

The phylogenetic tree based on the B4a1a1b high-quality mitogenomes available shows the absence of geographic structuring among these sequences, together with their very low diversity. The two sequences from Yemen and Madagascar are identical to the basal B4a1a1b lineage and the other sequences carry only one additional mutation each, mostly from mutational hotspots in the hypervariable non-coding region (16291T, 16357C, 16368C, 16222T). This low diversity is in agreement with the recent coalescence age estimated for haplogroup B4a1a1b of 1,500-1,800 years BP, a narrower window than previously estimated (1,200- 16,800 years BP). This new age estimate predates, but is consistent with, the earliest Austronesian settlement of the East African offshore islands (Madagascar and Comoros), which is dated to 800-1,200 years BP from recent archaeological and genetic studies.

Genome-wide analysis

To determine whether genome-wide genetic data can shed further light on the history of individuals carrying the Malagasy motif, the genetic ancestries present in the genome-wide dataset were decomposed with ADMIXTURE v.1.3. The lowest cross-validation values were obtained with 26 ancestries. For clarity, using a reduced admixture dataset including representative populations from each region and the population of interest, K=4 also displays the main African, European, South Asia and Southeast Asian ancestries. We observed that the Somali (S25) and two Yemeni (Y270 and Y115) individuals have relatively similar ancestry patterns to their local populations, including a dual Asian ancestry reflected by a South Asian (India) and Southeast Asian (Indonesia) components. Moreover, Yemenese and Somali do not have significantly more Asian ancestry (0.87% and 1.18% respectively) than paired more inland populations (Student’s t test: Yemen versus Saudi Arabia t(54)=–0.67, P=0.51; Somalia versus Ethiopia t(30)=1.9, P=0.067; Somalia versus Egypt t(23)=–0.43; P=0.68). The two Yemeni individuals have one major component (pinky red gradient: 65-80%) shared with other Middle East populations. Other minor ancestries are shared with populations from the horn of Africa (grey gradient: 13%), sub-Saharan Bantu speakers (green gradient: 7-10%), and Europeans (blue gradient 2-5%). The Asian component (yellow gradient) represents just a small component: from 1% in Y115 – similar to those in Yemeni populations, up to 10% in Y270. The Somali individual has one major component (green gradient: 60%) shared with Bantu speakers, and a smaller component shared with populations from the horn of Africa (grey gradient: 25%), the Middle East (pink red gradient: 5%) and very minor components from Asia (yellow gradient <1%).

The admixture scenario from f3 statistics shows that the two Yemeni (Y115 and Y270) and the Somali (S25) samples cannot be modeled as any kind of simple mixture, using any Comoros, Madagascar or Banjar groups as one population source and one of their respective parental or neighboring populations as the other source. To explore their admixture further, the admixture history and Asian component of these individuals was subjected to PCAdmix analysis. This indicated that two individuals (Y115 and S25) have significantly more Malagasy fragments than are observed in their respective parental populations (Z score = 1,572 and 5,556 respectively), thus suggesting a likely Malagasy origin of their Asian ancestry. The other Yemeni individual (Y270) carries more Indonesian fragments than observed in its parental population (i.e., Asian and African Bantu fragments are not associated) (Z score = 8,368), suggesting an origin of its Asian component from a population with limited African Bantu admixture, and thus excluding the Malagasy as a source (in average, Malagasy populations analyzed so far have around 60% African Bantu ancestry and 40% Asian ancestry).

The date of inheritance of the Asian component found in the Yemeni and Somali Malagasy motif carriers can be estimated using an exponential decay function to explore the decline in total genome-wide Asian ancestry. [T]he two individuals with <1% Asian SNPs (Y115 and S25) likely had an Asian ancestor before 5-7 generations ago (lower bound prior to 250-190 years BP) and probably much older. As the Asian ancestry in these individuals is similar to other individuals in the local population who do not carry the Malagasy motif, the inheritance is possibly far earlier. In contrast, the individual with 10% Asian SNPs (Y270) likely had an Asian ancestor 3-4 generations ago (160-130 years BP), and because his total Asian ancestry is much higher than others in his local population (which has similar Asian ancestry to neighboring populations), the inheritance date is probably relatively close to this date estimate. There may therefore be multiple reasons why the Malagasy motif was inherited by these three individuals. . . . 

The absence of B4a1a1b in the Banjar is surprising as it represents the most frequent Asian maternal lineage in Madagascar (22%), and considering the cultural, linguistic and genetic links between the Banjar and Madagascar/Comoros populations, this lineage was expected to be present in the Asian parental population (the Banjar). Its absence may suggests that the main Austronesian maternal lineage (B4a1a1b) found in the western part of the Indian Ocean may not have originated from the Banjar population. However, an origin of the Malagasy motif in situ in Madagascar or the Comoros after the arrival of the Polynesian motif carriers is also unlikely, as it requires that (i) two mutations (1473 and 3423) arose in Madagascar in the last 1300-1000 years, (ii) diffused into population groups across the entire island, and (iii) the Malagasy motif precursor (the Polynesian motif), later disappeared from these populations. 

This ‘missing’ lineage in Southeast Borneo (1) may have been lost in Indonesia due to genetic drift in small island populations, or (2) may remain undetected due to the lower sampling coverage in Borneo compared to other studied regions. As suggested previously, the motif may have emerged in eastern Indonesia – where the Polynesian motif, its immediate precursor, is geographically restricted – and later brought into the Banjar gene pool. This may have occurred during long distance trading activities and admixture among ISEA regional populations favored by the emergence Hindu Malay Kingdoms, such as Śrīvijaya (6th-13th centuries), which developed the Banjarmasin trading post in southeast Borneo. This eastern Indonesian input into the Southeast Borneo Banjar finds some support in the fact that the Banjar are one of the rare populations in western Indonesia, along with the sea nomad Bajo from Borneo, to carry the Polynesian motif, demonstrating their previous connection with eastern Indonesians. 

In the Indian Ocean rim west of Island Southeast Asia (Mainland Southeast Asia and the Indian sub-continent), we note the absence of the Polynesian motif or any of its subclades. This suggests that the Austronesian population that brought the Polynesian Motif subclade B4a1a1b to the western part of the Indian Ocean (e.g., to Madagascar) did not leave detectable genetic traces on the northern Indian Ocean rim, in agreement with an absence of Indonesian gene flow to this region as inferred from recent genome wide data. It is still unclear whether this is due to Indonesian populations restricting their interactions in these regions to trading and/or cultural activities, or because Indonesian traders used a more direct route across the Indian Ocean to cover the 7500 km between Indonesia and Madagascar and the Comoros, which is possible based on ocean current and monsoon weather patterns. . . . 

While the Asian inheritance for two individuals (Y115 and S25) may reflect the development of trading posts by Arab merchants in the East African region, and deportation of African slaves northward to Arabia and South Asia, we cannot reject an older origin from the early Austronesian period. 

The Asian ancestry of the other individual (Y270) represents a different pattern with a likely origin in Island Southeast Asia, which might be explained by the emergence of several sultanates in the 15th-16th centuries AD in East Africa and in ISEA. Diaspora communities spread around the Indian Ocean rim may have played an important role to develop trading activities, and also to maintain contact with their native land. The Hadrami community from Hadramawt in Southeast Yemen - to which the Yemeni individual (Y270) belongs - emerge as one plausible cause to reconcile the different geographic and temporal origins of the Asian inheritance found in the three Yemeni and Somali individuals. The Hadrami diaspora, established in the Comoros at the end of the 1st millennium in South Asia (Gujarat in India) and in Island Southeast Asia (Malacca in Malaysia) from the beginning of the 2nd millennium, was involved in trading activities between Africa, Middle East and Asia. Despite strict intra community marriage rules, admixture with local population was frequent, as the men could choose their bride from among the local African and Asian populations. In the 19th century, some economically successful Hadramis (i.e. from Singapore and Hyderabad, and other places of the diaspora) returned to their native land in the Yemen bringing back their admixed family.