Mitochondrial DNA Polymorphisms of the Saisiyat Indigenous Group of Taiwan, Search for a Negrito Signature

The genetic profile of Negritos of the Philippines differs from the non-Negrito groups with mitochondrial DNA haplogroups B4b1a2, B5, D6a, M, M52a, and N11b. Although Negritos are not seen in Taiwan, the strong genetic affinity between the Philippines and Taiwan Mountain Tribe Aborigines (TwMtA), and Folks tales of TwMtA, Saisiyat and Atayal recounting past contacts with Negritos, warrant the search for a Negrito signature in Taiwan. Material and Method: Discriminant Analysis of Principal Component (DAPC) was used to determine the genetic relationship between TwMtA, Filipino and non-TwMtA groups. Results: The deep coalescence of B4b1a2 in the Philippine Negritos, Saisiyat, Atayal, Island Southeast Asia, and SEA (Southeast Asia) suggested a deeply rooted common ancestry, but could not support a past Negrito presence in Taiwan. Conversely, the sharing of cultural components and mtDNA (mitochondrial DNA) haplogroup D6a2 in Saisiyat, Atayal and Philippine Negritos may characterize a Negrito signature in Taiwan. Although the molecular variation of D6a2 determines its presence in Taiwan back to middle Neolithic, other markers, Y-SNP haplogroups C-M146 and K-M9, warrant further analysis. Conclusion: Most likely, the physical characteristics, languages, and the genetic makeup of the Negritos in Taiwan have been diluted as the result of heavy migration from the mainland in the last 400 years.


Introduction
The Saisiyat tribe is an Austronesian speaking group, a member of the Taiwan Mountain Tribe Aborigines. In the year 2018 April, the Saisiyat numbered 6,607, making them one of the third smallest indigenous groups on the island [1]. Its people live in the North West flank of the Taiwan mountain range, between Hsinchu and Miaoli (Figure 1). The geography of this region, which comprises the Egongji and Hengpingbei Mountains, forced Saisiyat to divide into two groups, the Sai-Kirapa in the north and the Sai-Maghahyobun in the south.
Historically Sai-Kirapa has had significant interaction with the northeastern Atayal tribe while the Sai-Maghahyobun has had more contacts with the Hakka who migrated there from East China in the last 400 years [2][3][4][5][6]. Similarly, Saisyat has two main dialects: the Taai Dialect in the North, and the Tungho Dialect in the south [7]. While Saisiyat has traditional views that mix aspects of ancestor worship and animism where all things are considered being alive and possess a distinct character, other Saisiyat people also practice Christianity [2]. 13 Most TwMtA tribes have kept folktales and myths that relate to past contacts with Negritos. In particular, Saisiyat is the only tribe in Taiwan that has rituals every two years honoring the memory of "the Short People or Pas-ta'ai" [8]. The Short People in these folktales are described as short-statured, dark-skinned and frizzy-haired and have the same anthropometric characteristics as Negritos in the Philippines. Some anthropologists believe these may have been Proto-Australoid people who possibly arrived from Africa during the early Southern Dispersal 60,000 years ago, but to this day, no archeological evidence has ever revealed the past presence of Negritos in Taiwan [9,10]. It now proposed that, instead of a shared ancestral phenotype from an ancient and well-distributed population, the resemblance of Negrito with other Negritos of Asia and Pygmies in Africa is the result of convergent evolution in the different parts of the word under equivalent environmental conditions [11,12]. This is supported by genetic evidence showing that Negritos of different parts of the world region have different genetic structures [13]. Further, other genetic studies observed that there is no simple dichotomy between Negrito and non-Negrito groups of the Philippines [14,15].
Most Negrito groups share genetic variations with neighbor populations while they have more deeply rooted variants that suggest a much earlier arrival in the region, isolation, and admixture with later arrivals [16]. Many studies using mitochondrial DNA (mtDNA) and Ychromosome variation have established a significant common ancestry between the populations of Taiwan and the Philippines, it is therefore expected to find a Negrito signature in Taiwan [14,15,[17][18][19][20][21]. Although Atayal and Saisiyat have a genetic profile that distinguishes them from the southern TwMtA, the polymorphism is homogenously distributed through all the tribes. [2, 3,6,[21][22][23][24].In this study, we analyze the mitochondrial genetic polymorphism of the Saisiyat tribe and search genetic evidence of the speculated presence of Negritos in all Taiwan indigenous groups. The mitochondrial molecular clock is faster than the molecular clock of Non-Recombination Y-chromosome (NRY) haplogroups determined using Single Nucleotide Polymorphisms (SNPs) and slower for NRY haplotypes determined using Short Tandem Repeats (STRs) [25]. The rate of mtDNA is, therefore, most appropriate to measure and trace evolutionary human changes phylogenetically in time and space. Further, its short length (16,569 base pairs), its presence in both males and females, its high polymorphism and the higher concentration than genomic DNA, makes it a most effective material, practically and financially, to use in a small laboratory [26]. However, we will use NRY in our comparative analysis with the Philippines.

Demographic Analysis
A bimodal curve was observed between 11 and 21 basepair differences in the mismatch distribution analysis of Saisiyat for 88 mtDNA sequences (Figure 2). The bimodal mismatch curve may have been the result of admixture. Further, in contrast with the Fu's Fs test, the plot did not support population sudden expansion [28]. Similarly, the hypothesis of sudden expansion was rejected by two demographic indexes, the Sum of Squared Deviation (SSD) test (SSD=0.008, P<0.001) and the raggedness (r=0.011, P<0.001) (Harpending 1994) indicating that the data deviated from the simulation expected under the model of expansion (Figure 2, blue line). The analysis (Number of pairs vs. Base pair differences) was obtained from 88 mtDNA Saisiyat sequences using nucleotide positions (nps) 8,000-9,000, nps 10,000-11,000, and hyper variable segment I (HVS-I) nps 16040-16390 [22,24]. Using the Bayesian Skyline method, patterns of historical demography can also be inferred from estimates of the effective population size over time.

MtDNA Haplogroup Distribution
Out of 20 mitochondrial haplogroups seen in Saisiyat (Figure 4), 15 were uniquely shared with the other the Austronesian speaking groups of Taiwan, and five haplogroups (B5a2a2a2, E1a1a1, F4b1', M7b1a2a and Y2) had a frequency greater than 8% representing 75% of the Saisiyat gene pool. Interestingly, haplogroups D6a2 was only seen in Saisiyat (1.1%), Atayal (4.6%), reported in a single individual of the Pazeh plain tribe, and the Mamanwa Negritos from the Philippines (3.3%) [14,24]. Distribution of shared mtDNA haplogroups of Saisiyat constructed using Taiwan data and other relevant populations [24]. Haplogroups with a shade of grey represent sharing with non-Indigenous groups of Taiwan (Minnan and Hakka). All haplogroups not seen in Saisiyat but present in other groups are represented by "other".

Population Differentiation
Discriminant Analysis of Principal Components ( Figure 5) showed a clear geographic divide along the first Discriminant component (the X axis), which form a separates the Austronesian-speaking groups on the right, from the Han, the TwH, and MSEA (Indochina) on the left. Further, although individuals were not tightly grouped, the Northern and Central TwMtA form a cluster that clearly separates them from a lower cluster encompassing Southern TwMtA and the Philippines. Finally, we note that the Southern Austronesian cluster (Southern TwMtA and the Philippines) shows more admixtures with TwH than the Northern cluster.

Phylogenetic Analysis
A maximum-likelihood tree using mtDNA haplogroup frequencies was inferred with the TreeMix software ( Figure 6A). The tree was consistent with the geographical distribution of populations in Taiwan; it places Saisiyat in a strong relationship with the Northern tribes (Atayal and Taroko). Further, migration arrows were first limited to 20 migration events in the analysis and only the three most significant gene flow events were retained for clarity of ( Figure 6B). The relationship delineated between Saisiyat and the central TwMtA (Thao and Bunun) was expected from the DAPC clustering shown in Figure  5. More interesting, was a significant input from the Philippines to Taiwan, consistent with a previous study [20]. And here seen as gene flow to Saisiyat with a migration weight of ~ 0.65 ( Figure 6). Most likely, this result is the effect of the sharing of haplogroups B4a2a, E2b, M7b1a2a, D6a2, R9c1a and Y2 between Saisiyat and the Philippines (Figure 4).
6A: Maximum-Likelihood tree inferred by TreeMix for all Taiwan and SEA populations assuming 20 significant gene-flow events, only the three most significant gene-flow events are shown for clarity, they are colored according to their weight on a zero to one scale [30]. 6B: Residual fit from A. Shaded colored cells represent Standard error for admixture events across all pairs of populations. Population pairs with a residual above zero are more closely related and more likely to correspond to an admixture event.

Discussion
In the first stage of this work, we examined the genetic variability using mitochondrial nucleotide positions (nps) 8000 to 9000 and nps 10,000 to 11,000 of the coding regions, and the HVS-I segment of the D-loop region to determine the genetic diversity and origin of the Saisiyat tribe of Taiwan. In the second stage, we investigated the general believed that Negrito groups predating the Lower Glacial Maximum period were associated with the first settlers of Sunderland/Island Southeast Asia (ISEA) [9]. Although there are some Chinese accounts of small, dark-skinned people with curly hair in Taiwan, to date, Taiwan has no archeological human remains supporting this hypothesis. Interestingly, most Taiwan Mountain tribes have kept folk tales describing past contacts with Negrito groups, most particularly the Saisiyat tribe of Hsinchu and Miaoli ( Figure 1) who still perform solemn rites to commemorate this past. On the other hand, the presence of Negrito is still existent in the Philippines and other parts of peninsular East Asia, but not in Taiwan.
If there is any truth in these TwMtA folk tales, then a small number of Negrito in Taiwan must have interacted with Neolithic agriculturalist migrants from the Southeast Asian mainland and most likely shaped their genetic diversity. In this regard, it is sensible to expect that these early contacts must have left some genetic traces. Here, our analysis of mtDNA polymorphisms tries to provide new insights into the history of the Saisiyat tribe [31,32].

Genetic Structure of Saisiyat
To investigate the mtDNA structure of Saisiyat, we first generated a mismatch distribution (Figure 3). A bimodal curve was obtained, and the two demographic indexes, the SSD test (SSD=0.008, P<0.001) and the raggedness (r=0.011, P<0.001) were significantly different to expectation. These results did not support expansion. Further, the Bayesian Skyline plot ( Figure 4) along with previous results from Ko's research group did not reveal sufficient structure to visualize a recent population expansion. These results were in contrast with a highly significant Fu's Fs test for the Saisiyat tribe [22,28,33].
It possible these results are revealing a sign that the Saisiyat tribe experienced prehistorically a rapid population growth from an ancestral population with a small effective population size.  Table S1) and were most likely acquired from migrants of Mainland East Asia in the last 400 years.
Most of the remaining haplogroups were commonly seen throughout Taiwan Northern tribes. Among them, B4a2a, E2b, M7b1a2a, R9c1a, and Y2 were also seen in the Philippines. They are thought to represent a plausible signal for a mid-Holocene out-of-Taiwan expansion or a signal for a bidirectional migration between Taiwan and the Philippines [18,20,22,24,[34][35][36][37]. The distribution of haplogroup D6 was intriguing. With a coalescence time of 14,390 years BP (Supplementary Table S1 and Figure S1) haplogroup D6 was seen at low frequency as D6c in East Asia and Southeast Asia (SEA

Principal Component Analysis
To characterize population structure across Saisiyat, other Taiwan groups, and their relationship with neighboring populations in MSEA and the Philippines, we performed a DAPC ( Figure 5). The first component captured a clear geographical divide between Austronesian speaking groups and non-TwA groups including Fujian, Minnan, and Hakka. Component 2 in Figure 5 disclosed a strong affinity between Northern and Central TwMtA, and between the Philippines and the Southern TwMtA, suggesting that all these Austronesian-speaking populations have a common origin. However, the mtDNA composition of the Saisiyat suggested evidence supporting genetic affinity between Saisiyat and Negrito and non-Negrito groups of the Philippines.
Using Delfin mtDNA and our Taiwan data set we constructed a Multidimensional scaling plot to establish this relationship (Supplementary Figure S3). While the relationship between populations was the same as in the DAPC plot, the Aeta and Agta Negrito groups were outbound, most likely because of the conjoint results of drift, the presence of high frequency haplogroups such as P and M52 in Aeta and Agta, and long isolation [14]. On the other hand, haplogroup B4b1a2, E1a1a1, Y2, and D6a in the Mamawan group inferred strong affinity of the Mamanwa Negritos with other Austronesian groups, suggesting a more recent gene-flow of Austronesian-speakers in the Mamawa.

Maximum Likelihood Tree from TreeMix
TreeMix results inferred mtDNA gene-flow events ( Figure 6) potentially summarizing patterns of population in the history of Taiwan such as bottlenecks, isolation, consanguinity within populations, here Saisiyat was characterized as a northern Taiwan tribe. Further, the gene-flow from Saisiyat (or the Northern TwMtA) to the Central TwMtA (Bunun, Thao, and Tsou) was previously foreseen in Figure 4 and 5, and is confirmed in Figure 6 [30]. Moreover, the strong migration event depicted by TreeMix from the Philippines to Saisiyat  Figure S3) and possibly experienced several admixture events in the past. This option is nonetheless not supported by the age estimate of molecular variation obtained for any of the haplogroups of the same clade between Mamanwa and Saisiyat/Atayal. For example, D6a2 dates only to 2600 ± 1500 yrs BP (Supplementary figure S1) and Y2 dates 3956 ± 2455 yrs BP (Supplementary figure S2). One possible way to demonstrate a Negrito ancestry in Taiwan associated to D6 or Y2 would be to find sister branches of these haplogroups in Taiwan and/or the Philippines that would allow a coalescent node in the pre-Holocene period.

Other Gene Systems
Supporting this last observation, our previous Y chromosome analysis observed 4 Y-chromosome single nucleotide polymorphism (Y-SNP) haplogroups out of 24 unrelated Saisiyat males (Supplementary Table  S1). Only one major haplogroup O1a1 * (P203) had a frequency of 87.5%, while all other haplogroups (O1a2 (M50/110, O3a1 * (KL1/122) and O3a2c1a (M133/M7/M134) were seen only once (4.2%) [21]. When compared to the Y haplogroup profile of the Philippines, O1a1 * (P203) was prevalent in all Filipino ethnolinguistisc groups, Negritos and non-Negrito and its presence in any Negrito groups was regarded as an admixture with the former. Most interestingly, Negrito groups in the Philippines invariably possessed, to various levels, haplogroup haplogroups C-RPS4Y/M216, K-M9, and O3-M122 [15]. These haplogroups have not been seen in Saisiyat, but single observations of C-RPS4Y/M216 and K-M9 have been seen in the Taiwan plain tribes and could support a past presence of Negrito in Taiwan [15,21].
Lastly, to our knowledge, no previous studies associating the Filipino Negrito groups and the HLA gene system have yet been published; accordingly, no Negrito HLA inferences could be used for Taiwan. Nonetheless, several HLA * A-B-DRB1 haplotypes were unique to Saisiyat and 1/3 of Saisiyat haplotypes were shared with Atayal (Supplementary Table S1

Summary
This investigation has contributed substantially more insights into the population groups in Taiwan and the Philippines. Further, while the physical appearance of Negritos has never been seen in Taiwan, few Taiwan Mountain tribes, such as the Saisiyat and the Atayal tribes, have conserved folktales inferring prehistoric co-habitation with them, and to this day, still celebrate this period bi-annually. Among the few mtDNA haplogroups shared between Taiwan Northern tribes and the Mamanwa Negritos (B4b1a2, E1a1a1, Y2 and D6a) only D6a may represent a common Negrito genetic legacy of the Saisiyat and Atayal tribes. This finding must be taken with caution, as the mid-Neolithic coalescence age estimate of D6a is too shallow. Further, no support was given from the Y chromosome analysis for Saisiyat and Atayal. Although the apparent affinity between the Taiwan Northern tribes and the Mamanwa Negritos of the southern Philippines could be the result of gene flow brought upon by bidirectional population movements at the time of the out of Taiwan, the presence of C-RPS4Y/M216 and K-M9 in Taiwan were scarce, and warrant more extensive studies of the Taiwan gene pool in the future.

Material and Method Samples
Whole blood or saliva specimens were collected from 2704 unrelated individuals (   (Supplementary Table S2) The whole-mtDNA genome sequencing is available in Supplementary Text File S1. We deposited 2 new whole-mtDNA sequences in GenBank.

Statistical Analysis
In order to test for past population expansion of Saisiyat, we used two statistical tests Tajima's D and Fu's Fs [33,44]. The analyses were implemented in Arlequin 3.5.2.2, and p-values were generated using 1,000 simulations under a model of selective neutrality [45].