A genome wide pattern of population structure and admixture in peninsular Malaysia Malays

Quantifying genetic distance is the main aspect in population genetic study, especially to characterize population structure and identify substructures (Lao et al. 2006; Rosenberg et al. 2002; Weir et al. 2005 and Tishkoff et al. 2009). Previous studies have also shown the importance of genetic distance assessments, such as in inference of migration patterns (Li et al. 2008; Ramachandran et al. 2005; Deshpande et al. 2009 and Laval et al. 2010) as well as in the need to adjust for population stratification in association studies Bryca et al. 2010; Patterson et al. 2010; Miclaus et al. 2009 and Li et al. 2010). This study was conducted to infer the population structure of Malays that may or may not have shared ancestry from other study populations. Therefore, the genetic distance based on Fst was chosen to measure the genetic variation distributions of studied populations. Fst is a powerful method to show population genetic structure by partitioning genetic variance within populations relative to between populations (Weir and Hill 2002; Weir and Cockerham 1984).

Many studies have shown that genetic distance of global populations correlates with geographic distance between populations, which refers to a situation called Isolation-by-Distance (IBD) (Li et al. 2008; Tishkoff et al. 2009; Ramachandran et al. 2005; Prugnolle et al. 2005 and Gonder et al. 2007). Nevertheless, based on the genetic distance analysis of this study, the IBD model can be applied to a particular group of populations, but not to the Malays group. The close genetic relationship between Melayu Kedah and Melayu Kelantan were mostly reflected to their geographic origin at the northern part of the peninsula, likewise Thai Pattani have smaller value of genetic distance to both of the northern Malays (Table 1). However, in the case of Melayu Bugis and Melayu Jawa, in which both population samples have been collected in the same state at the southern part of the peninsula, the result was in contrast with IBD. As shown in Table 1, the Melayu Bugis were genetically distant from other Malays but closely related to Indonesian Toraja from South Sulawesi, while the Melayu Jawa were significantly closer to Indonesian Jawa from Central Java. In this situation, the genetic distance corresponds largely to shared ancestry between populations, as been discussed in many previous studies (Hatin et al. 2011; Rosenberg et al. 2002; Jorde and Wooding 2004 and Consortium THP-AS 2009).

According to the history of Malay Peninsula and Indonesia, the migration of Indonesian people into mainland peninsula were very common whether due to trading activities or seeking asylum from civil war and colonial invasion (Taylor 2003; Sainuddin 2003 and Munoz 2006). For instance, the modern Melayu Jawa in peninsula are descendants of Jawa people that originated from Java Island. They migrated to the state of Johor and Selangor around the 15^th century to avoid conflicts due to civil war. The large scale migration of Jawa into peninsula was during British colonial era in between of 1880 to 1930 for seeking a better life from harsh invasion of the Dutch. The first influx of Melayu Bugis into Peninsular Malaysia from their origin in South Sulawesi was in the 17^th century. At that time, the Dutch were expanding their trade and political control on Sulawesi. Arising conflicts between the colonial and colonized people were inevitable and most of the Bugis people migrated to Johor state at the southern part of the peninsula. Later on, they also settled in Selangor state for over a few centuries. Since then, the Melayu Bugis and Indonesian Toraja which originated from the same geographic origin were geographically separated for hundreds of years.

It is also known from the history and anthropological evidence, the Melayu Minang were originally Minang kabau people from West Sumatra. They migrated to Malay Peninsula in the 14^th century, long before the arrival of Bugis and Jawa people. They started the colonization of Negeri Sembilan, a state in the middle of the western part of Peninsular Malaysia and at present day are known as Melayu Minang. However, in relation to the other Malays, the Melayu Minang were genetically closer to both of Melayu Kedah and Melayu Kelantan, than to Melayu Jawa (Table 1). Here, the genetic distance between those Malay sub-ethnic groups may have been affected by the presence of genetic admixture in their genetic data from the same mixing populations. In quantifying genetic distance among populations that experienced admixture from the same mixing populations, interpreting the relationships can be tricky as the admixture reduces the average of genetic distance between them (Handley et al. 2007; Halder et al. 2008; Auton et al. 2009 and Moorjani et al. 2011).

It is interesting to see the close genetic relatedness among Malays (except for Melayu Bugis) with the Proto-Malay Temuan, Indonesian Jawa and Chinese Wa from Yunnan, China. This may imply a common origin for all of those populations regardless of their historical migration patterns as been reported (Bellwood 1997 and Omar 2004). But it is more intriguing to see how genetically close the Melayu Kedah and Melayu Kelantan are to the Indian Telugu and Marathi compared to any other populations in this study. This result has conformed to the historical contact between those populations with the Indians (Arasaratnam 1970).

Apart from being close to Proto-Malay Temuan, the Melayu Kedah and Melayu Kelantan also showed relatedness to the earliest aboriginal people of peninsula in view of having the smallest value of genetic distance to both of the Semang sub-groups, Jahai and Kensiu compared with any other Malays. The Indonesian Jawa and Chinese Wa also shared the same lower value to both of the Semang, compared to other group of studied populations. This may imply two probable situations that could explained the genetic relatedness of those populations. Firstly, a common origin for all of those populations with subsequent evolutionary divergence due to pre-historical migration and become closer with subsequent recent migration as been implied by the HUGO-PASNP Consortium (Consortium THP-AS 2009). Secondly, they might have diverse origin from simultaneously pre-historical evolutionary divergence (Cavalli-Sforza et al. 1994) but admixed due to recent migration patterns as been explained by (Andaya 2001). To be certain about which are the most favorable demographic histories among them, further analysis was conducted in model-based approach using ancestral components that will be discussed subsequently.

Meanwhile, the MDS has been widely used for the analysis of proximity data among objects to reveal the hidden structure underlying the data (Steyvers 2002; Borg and Groenen 2005) especially in DNA microarray data (Tzeng et al. 2008). In this study, the genetic structure of Malays showed by Fst was successfully recapitulated in two and three dimensions (2D and 3D) models as shown in Figure 2. The Malay populations are shown explicitly as three sub-clusters on both of the 2D and 3D platforms, signifying an existence of substructure within the Malays. This could be achieved by finding the disposition of studied populations that are compatible with the given genetic distances among them on a map. The Euclidean distance is used to represent the transformed data in such a way that the MDS clustering matches the original data as much as possible even in a smaller number of dimensions (Borg and Groenen 2005). However, the distance-based method could not provide the ancestral membership coefficients of the admixture among the populations. Hence, we implemented further analysis to determine the genetic admixture and ancestry of the Malays.

Previous studies have shown the robustness of the STRUCTURE software in inferring the population structure and ancestry in variety of population data (Xu et al. 2010; Consortium THP-AS 2009; Bamshad et al. 2003; Rosenberg et al. 2005 and Witherspoon et al. 2007). In this study, the assignment of individuals into inferred clusters was in accordance to historical and demographical background of studied populations. At K = 2, all individuals of the 17 pre-defined populations was predominantly separated into African and non-African ancestries, indicating that the modern human dispersal originated from Africa (Cavalli-Sforza et al. 1994; Bowcock et al. 1994). The emergence of cluster that is specific to Orang Asli Semang of Malay Peninsula as early as at K = 3, supports the arrival of the Semang into SEA region via the first wave from Africa as postulated by many researchers (Allen 1879; Carey 1976; Fix 1995; Consortium THP-AS 2009; Kashyap et al. 2003 and Hill et al. 2006). The fact that their cluster occurred before the Indian cluster at K = 4, may be caused by the great effect of genetic drifts in their genetic data due to population bottleneck event and later exhibit the founder effects. As they are extremely geographically isolated and conserved from the outside world, they have preserved their ancestral allele state since the divergence. This is unlike the Indians, who have widespread admixtures with Europeans (Brahmachari et al. 2008). Histories of language shifting are also common in some of the Indian populations, as shown by the Indian Marathi in this analysis.

It is also noted that the ancestral component of Semang could be seen in Indians, Proto-Malays, Thai Pattani and both Melayu Kedah and Melayu Kelantan. However, based on the Q proportions of K = 6 as the ancestral cluster, the component was not significantly high with merely 0.05 in Melayu Kelantan and Proto-Malays, whilst much lower in the other Malays. Interestingly, the admixture coefficients of the Indian ancestral component exist in both of the Melayu Kedah and Melayu Kelantan, as well as in Melayu Minang and Thai Pattani. The highest coefficients were in Melayu Kedah with 0.21, followed by Melayu Kelantan 0.16, Thai Pattani 0.14 and Melayu Minang 0.12.

In population HSAs, the Indians have higher percentage of population private haplotypes than the Chinese and the Malays. This pattern is more compatible with the scenario of population admixture in Indians. However, high level of haplotype diversity is not just expected in an admixture population with divergent ancestries, but also in an ancient population as it has long time to accumulate much more private haplotypes. The HS pattern of Indians with the Chinese and Malays was relatively lower than the HS of those two groups with each other. The reason that could cause to these HS pattern perhaps due to the main separation between Indians and Asian populations dates to about 60,000 years ago (Cavalli-Sforza and Feldman 2003) while the populations in Southeast Asia and East Asia (China) have very close connections in the more recent past, either due to Neolithic expansions from China into mainland Southeast Asia and Island Southeast Asia or somewhat earlier migrations in the late Pleistocene or Early Holocene due to climate change and sea-level changes (Ricaut et al. 2006; Bellwood and Dizon 2008 and Hung 2008).

The STRUCTURE results exhibited very close estimates with the HSAs results, suggesting major contribution of Indian haplotypes in the northern Malays. The centralization of the ancient Indianized kingdoms had occurred in mainland Southeast Asia such as Thailand, Cambodia, and Myanmar for centuries in the early millennium (Tarling 1999; Stark 2006). Although Hinduism also existed in some of the Indonesian islands (eg. Sumatra and Java Island), it was more predominant among the populations in mainland region and the northern part of Malay Peninsula (Shuhaimi 1984; Allen 1997 and Syukri 2002). The existence of Indian ancestral component within these northern Malay populations is relevant to their early historical contacts with the Indians. The long-term historical contacts between Malays and Indians, may explain the higher admixture coefficients in both Melayu Kedah and Melayu Kelantan.

Furthermore, the existence of the Indians haplotypes in the gene pool of the northern Malays may signify that they are the oldest Malay populations in Malay Peninsula as the Indians had been conspicuous in the region very much earlier, since the proto-historic times. The ancient Hindu Malay kingdoms which arose approximately in 100 before common era (BCE) to 7^th century CE such as Chi Tu, Langkasuka and Kadaram have controlled much of the northern Malay Peninsula (Arasaratnam 1970). The Indian influxes continued to expand during the subsequent empires of Srivijaya and Majapahit (Paul 1961). These early Malay states were heavily influenced by concepts of religion, government and arts that were brought by the Indians. The proto-historic of Malay Peninsula ended in the beginning of 15^th century CE with the emergence of Malacca Sultanate. Malacca that encompassed most of modern day Peninsular Malaysia, Singapore and a great portion of eastern Sumatra thrived into the most important entrepôts in Southeast Asia and a hub of Islamic studies, spreading Islam to Malay Archipelago in 16^th century CE. Still, traces of the Indian influence can be found in Malay culture until today (Arasaratnam 1970; Shuhaimi 1984 and Syukri 2002).

Possible admixture between Malays and Indians could also have occurred during the British colonial period from the 19th to the middle of the 20th century. However, the Indians are not a large component of the Kedah and Kelantan population either during or after the British colonial era as most of them reside in the western and north-western regions of Peninsular Malaysia which are the location of the big cities and large urban areas in the country. In the Kelantan state which is the origin of the Melayu Kelantan, the total population is about 1.67 million and the percentage of the Indian community is only 0.2% of the population. In the Kedah state which is the origin of the Melayu Kedah, the total population is about 2.04 million and the percentage of the Indian community is also slightly higher than in Kelantan with 6.6% of the population (Hunley et al. 2009). Moreover, the sampling procedure stringently followed the inclusion and exclusion criteria that emphasized the three generations without any different ethnic admixture rule for an individual to be considered as a valid subject for this study. Hence, we believed that the admixture in both Melayu Kedah and Melayu Kelantan with Indians was ancient and has occurred during the early existence of the Malays.

The ancestral component of Proto-Malay Temuan appeared at K = 5, while the ancestral component of Malays emerged at K = 6 and both of the components also existed in the Chinese individuals, especially in Chinese Wa at respective clusters. In those inferred clusters, the ancestral component of Chinese (yellow component) was predominant in all the Malays, Thais, and Indonesians as well as in Proto-Malays itself. This is in accordance to the historical and anthropological evidences of the migration pattern of Proto-Malays from Yunnan, southern mainland China (Bellwood 1997; Carey 1976). The yellow component also might be related with a large Neolithic input from China into mainland and island Southeast Asia due to the expansion of agriculture and animal domestication (Bellwood and Oxenham 2008; Bellwood 2011). Although the Austronesian dispersal did not originated from the early farming dispersal, but it was a peripheral result of the demographic impetus and technological advancement by the developments of food production in mainland East Asia (Bellwood 2011). Furthermore, the peopling of pacific region started from earlier migration of modern human expansion from Africa throughout much of Southeast Asia during a period of relatively stable climate and sea-level from 45,000 year before present (YBP) to 40,000 YBP. The extreme climate and rapidly changes of sea-level during the Last Glacial Maximum lead to decrease the expansion of human populations from 33,000 YBP – 16,000 YBP (Bird et al. 2004; Forster 2004). Later, the post-glacial expansion in coastal settlement arose concurrently with the development of coastal ecosystems and environments due to the slow rise of the sea-level. However, the sea-level fluctuations inhibited the coastal settlement and the drop in sea level in the mid-Holocene may have caused widespread human expansion throughout Oceania (Bird et al. 2004; Pope and Terrell 2008). The yellow component suggests that despite having great admixture or genetic differences due to genetic drift, all of these populations have a common ancestor, which is referred to as Southern Mongoloid group of races.

In relation to the modern Malays, it is known that Malays have been previously referred to as admixed Deutero-Malays, which are the descendants of the Proto-Malays who had admixture with other populations, such as Arab, Sumatran and Siamese (Sainuddin 2003). Other sources have postulated that the Deutero-Malays originally migrated through the southern part of China, and reached the Malay Peninsula about 1500 to 2000 years ago, after the arrival of the Proto-Malays (Fix 1995). According to Kasimin (Kasimin 1991), compared with the arrival of Indians and Chinese to Malay Peninsula, the Deutero-Malays were the earliest to be settled. Then, the vast and subsequent influxes of other populations to peninsula, mainly due to trading activities had integrated the Deutero-Malays into admixtures. These Deutero-Malays are known as the present day Malays. Given the vague historical facts on the origin of Malays with a fine line to differentiate between Proto-Malays and Deutero-Malays, it is still evident that these populations have profoundly close genetic relationship and shared a common ancestor with the Chinese.

The ancestral clusters or the most biologically sensible number of clusters which captured the structure of the given genotyped data was identified as six clusters (Table 2). In this ancestral clusters (K = 6), any pre-defined populations which shared similar distinctive Q values were merged into the inferred cluster. These clusters associate mainly to self-identified geographic origin or geographic proximity, linguistics family and their ancestry. The first cluster is referred as the cluster of Austronesian speakers and includes all the Malays, Thais and Indonesians. The second cluster was exclusive to Proto-Malays ancestry, which is Temuan in this study. The fact that these Proto-Malays are different from the other Austronesian speakers although belonging to the same linguistic family, perhaps due to evolution forces such as genetic drift or selection pressures that have reshuffled their genetic components. Yet, both these clusters still have conserved great proportions of Chinese ancestral component, testifying their Southern Mongoloid morphological features.

The third cluster belongs to the Austro-Asiatic speakers, which are the Semang Jahai and Kensiu. The fourth cluster associates with Chinese ancestry and belongs to the Chinese Jinuo and Wa from Yunnan, China. However, these Chinese populations speak different languages; Wa speak Austro-Asiatic whilst Jinuo speak Sino-Tibetan language. These two populations are the indigenous populations in Yunnan and as shown by the HUGO-PASNP Consortium (Consortium THP-AS 2009), both populations were clustered among indigenous Thais populations who also speak Austro-Asiatic language. This indicates a history of language shifting in the Chinese Jinuo (Dutton and Tryon 1994). Although sharing the same linguistic family, a great difference between Chinese Wa and the Semang group in Malay Peninsula is likely due to their historical divergence with wide range of demographic histories (Consortium THP-AS 2009). The fifth cluster belongs to Indian ancestry, consist of Marathi and Telugu. Again, despite having different linguistic family, both the Indians were clustered in the same ancestral cluster. Lastly,the sixth cluster consisted of only Yoruba with African ancestry who speak Niger-Congo language.

The population genetic structure of Malays and other studied populations became more apparent at higher number of clusters, but it also might not representative of anything and just residuals of the methodology. Thus, the interpretation must be made carefully as well as must be supported by other evidence, either historical or anthropological. In this analysis, it is clear that whole-genome data of Melayu Bugis were more delineated to Indonesian Melayu and Toraja from Indonesia with very minimal admixture. The Melayu Jawa were similar to the Indonesian Jawa, with significant components of Chinese in their genomes. The Melayu Kelantan and Melayu Kedah resemble more to Thai Pattani, with significant admixture from Indian components. Although the Melayu Minang also shared the Indian ancestral component, they were slightly different at this higher level of clusters as they exhibit almost none of the Chinese ancestral component. This may be due to different demographic histories compared to other Malays as has been reflected by their unique maternally cultural and traditional rule, called “Adat Pepatih” (Reid 2001; Ricklefs 2001). This unique sociological cultural may have had greater effect of pressure selections in the Melayu Minang population. The result of admixture analysis was concordance with the previous study of human leukocyte antigen (HLA) polymorphism and population structure analysis of Malays (Hatin et al. 2011; Edinur et al. 2009).

The resulting Q plot from STRUCTURE and HSAs partition plots did not revealed the relationship among the components in term of evolutionary history. Thus, phylogenetic trees were reconstructed to further refine the analysis. The topology of tree produced by Cavalli DC contradicted most known patterns of historical migrations. Based on historical facts, it is unlikely that Malays and Proto-Malays have a simultaneously historical divergence with the Semang group. Many evidences have shown that the original inhabitants of the Malay Peninsula are the Semang. For instance, the oldest Paleolithic human skeleton estimated about 11,000 years old, was reported to have genetic similarities with the Semang (Majid 2005). The other topology, which was produced by Nei’s DA is more favored to reflect the evolution histories among the study populations. Previous genetic studies have postulated the northwards migration of SEA people to central and eastern Asia (Consortium THP-AS 2009; Su et al. 1999) before gradually migrated back southwards (Rahman et al. 1998; Andaya 2001; Omar 2004; Bellwood 1997 and Fix 1995). The phylogeny analyses of HSAs indicated greater similarity where the Malays are more related to Proto-Malays than to Semang. Although the ancestry line of Malays were traced back to the Proto-Malays and the Chinese, the Indians have contributed more haplotypes to the northern Malays that may resulted to the Melayu Kedah and Melayu Kelantan to be genetically different from the other Malays.