Reconstructing the Population Genetic History of the Caribbean

The Caribbean basin is home to some of the most complex interactions in recent history among previously diverged human populations. Here, we investigate the population genetic history of this region by characterizing patterns of genome-wide variation among 330 individuals from three of the Greater Antilles (Cuba, Puerto Rico, Hispaniola), two mainland (Honduras, Colombia), and three Native South American (Yukpa, Bari, and Warao) populations. We combine these data with a unique database of genomic variation in over 3,000 individuals from diverse European, African, and Native American populations. We use local ancestry inference and tract length distributions to test different demographic scenarios for the pre- and post-colonial history of the region. We develop a novel ancestry-specific PCA (ASPCA) method to reconstruct the sub-continental origin of Native American, European, and African haplotypes from admixed genomes. We find that the most likely source of the indigenous ancestry in Caribbean islanders is a Native South American component shared among inland Amazonian tribes, Central America, and the Yucatan peninsula, suggesting extensive gene flow across the Caribbean in pre-Columbian times. We find evidence of two pulses of African migration. The first pulse—which today is reflected by shorter, older ancestry tracts—consists of a genetic component more similar to coastal West African regions involved in early stages of the trans-Atlantic slave trade. The second pulse—reflected by longer, younger tracts—is more similar to present-day West-Central African populations, supporting historical records of later transatlantic deportation. Surprisingly, we also identify a Latino-specific European component that has significantly diverged from its parental Iberian source populations, presumably as a result of small European founder population size. We demonstrate that the ancestral components in admixed genomes can be traced back to distinct sub-continental source populations with far greater resolution than previously thought, even when limited pre-Columbian Caribbean haplotypes have survived.


Introduction
Genomic characterization of diverse human populations is critical to enable multi-ethnic genome-wide association and sequencing studies of complex biomedical traits [1]. The increasing availability of genome-wide data from populations worldwide allows for the reconstruction of population history at finer scales, shedding light on evolutionary processes shaping the genetic composition of peoples with complex demographic histories. This is especially relevant in recently admixed populations from the Americas. Native peoples throughout the American continent suffered a dramatic demographic change triggered by the arrival of Europeans and the subsequent African slave trade. Important progress has been made to characterize genome-wide patterns of these three continental-level ancestral components in admixed populations from the continental landmass [2], and other Hispanic/Latino populations "Latino European" component and it can be seen clearly in Figure 1C ("black" bars represent the Latino European component, "Red" bars represent the "Northern European", and pink the "Mediterranean" or "Southern European" component). At K=8, when the clinal gradient of differentiation between Southern and Northern Europeans appears, the Latino European component is seen only in low proportions in individuals from Portugal and Spain, whereas it is the major European component among Latinos ( Figure 1C, bottom panel).
Colombians and Hondurans show considerably higher proportions of Native Venezuelan components, consistent with their geographic proximity. Both Mesoamerican and Andean Native American samples contain considerable amounts of European ancestry, largely due to post-Columbian admixture. Interestingly, the European component in Native Americans is assigned to the Latino-specific component in Mesoamericans (Nahua/Maya) and to the Mediterraneanspecific European component in Andeans (Aymara/Quechua). The Latino-specific component could be explained as the result of a founder effect driven by early European settlers, hence this pattern would be compatible with an initial introduction of European segments in native Mesoamericans and a later arrival of European chromosomes into the Andean gene pool.
Our data show a strong signature of assortative mating based on genetic ancestry among Caribbean Latinos as suggested by previous studies [17]. In particular, we see a strong correlation between maternal and paternal ancestry proportions ( Figure S4). To assess significance, we compared correlation of ancestry assignments among parent pairs to 100,000 permuted male-female pairs for each continental ancestry. All p-values were highly significant (p <0.00001, Table S2). It should be noted that these tests are not independent since the three components of ancestry by definition must sum to one. Further, apparent assortative mating could be due to random mating within structured sub-populations. To control for this, we performed permutations within sampling localities, and found significant correlations among individuals from every single population, except for Haiti. Although Haitians do show the same trend, with only 2 parent pairs is nearly impossible to assess significance (see Table S2).

Demographic Inference since the onset of admixture
An overview of our analytic strategy for characterizing admixed genomes is presented in Figure 2. Due to meiotic recombination, the correlation in ancestry among founder chromosomes is broken down over time. As a consequence, the length of tracts assigned to distinct ancestries in admixed genomes is informative of the time and mode of migration [18]. To explore the population genetic history of the Caribbean since European colonization, we considered the length distribution of continuous ancestry tracts in each of the six population samples. First, we estimated local ancestry along the genome using an updated version of PCAdmix [19] which was trained using trio-phased data from the admixed individuals and three continental reference populations. Next, we characterized the length distribution of unbroken African, European or Native American ancestry tracts along each chromosome for each population. Finally, we applied the extended space Markov model implemented in Tracts [20] to compare the observed data with predictions from different demographic models considering various migration scenarios.
The simplest model considers a single pulse of migration from each source population, allowing the admixture process to begin with Native American and European chromosomes, followed by the introduction of African chromosomes. In such a scenario each population contributes migrants at a discrete period in time, and the average length of ancestry tracts is expected to decrease with time after admixture, resulting in an exponential decay in the abundance of tracts as a function of tract lengths. Alternative models include a second pulse of either European or African segments migrating into the already admixed gene pool. Allowing for continuous or repeated migration typically results in a concave log-scale distribution, caused by the increase of longer tracts after the second migration event. Table 1 and Figure 3 summarize the results of the best-fitting migration models for each population based on Bayesian Information Criterion (BIC) comparisons, and Figure S5 shows the full results of all models tested. We observed that multiple pulses of admixture offered a better BIC in all cases.
The best-fit model for Colombians and Hondurans involves admixture between Native Americans and Europeans starting 14 generations ago, followed by a second pulse of European ancestry starting 12 and 5 generations ago, respectively. Of note is that between the first and second pulse of migration in Colombians, the proportion of European ancestry increased from 12.5% to 75% in two generations, implying that the European segments in today's Colombians date back to European gene flow happening in a short period of time, thus tracing back their ancestry to a more limited number of founders compared to other Latino populations.
In contrast with mainland population samples, the best-fit model for all four populations from the Caribbean islands involves older time estimates of the initial contact between Native Americans and Europeans. Namely, 17 generations ago for Cubans and 16 generations ago for Puerto Ricans, Dominicans, and Haitians. Historical records state that the first European colonies in the Antilles were set up soon after the initial contact in 1492 [21], that is ~500 years ago or 16.6 generations ago (considering 30 years per generation [22]), in excellent agreement with our time estimates. Another major distinction is that the model involves a second pulse of African ancestry, occurring between 7 to 5 generations ago, with higher migration rates in Haitians and Dominicans, followed by Cubans and Puerto Ricans.

Sub-continental Ancestry of Admixed Genomes
The genomes of admixed populations contain information about both continental and sub-continental population processes. To explore within-continent population structure, we performed PCA on genomic segments of specific continental ancestry. Because the masking out of the other ancestries results in large amounts of missing data, we implemented a novel variation of PCA that allows performing the analysis on the remaining sites alone. Throughout this paper, we refer to this approach as ancestry-specific PCA (ASPCA) and the mathematical details are described in Text S1. We applied this methodology for projecting phased genomic segments of inferred Native American, European, and African continental ancestry onto subcontinental reference panels of parental populations (see diagram in Figure 2). Our implementation is analogous to the subspace PCA (ssPCA) approach by Johnson et al. [23], but it can take advantage of phased data, allowing to accommodate parts of the genome that are heterozygous for ancestry. In the presence of recent admixture, chromosomal ancestry breakpoints dramatically reduce the proportion of the genome that is homozygous for a given ancestry. Therefore, relying on genotypes and restricting to loci estimated to have two copies of a certain ancestry could severely compromise the resolution of the analysis of admixed genomes. Our haplotype-based implementation of the algorithm is packaged into the software PCAmask and details on the samples used are available in Materials and Methods and in Text S1.

Native American Ancestral Components
Our initial structure analysis was based on our high-density dataset (i.e., ~390K SNPs, see Table S1), thus limited to ancestral populations with available Affymetrix SNP array data (i.e., two Mesoamerican, two Andean, and three Venezuelan native populations). To explore possible relationships with additional Native American populations, we expanded our reference panel by combining our data with Illumina 650K data for 493 individuals from 52 indigenous groups from throughout the Americas [11]. Although this analysis has fewer SNPs (i.e., ~30K SNPs), it allows us to resolve within-continent population structure around the Caribbean in much greater geographic detail.
We applied the ASPCA approach described above to project the Native American segments of admixed individuals onto the full reference panel ( Figure 4A, Figure S6). PC1 separates the northernmost populations, such as Canadian Northern Amerind and Na-Dene speakers, from the rest; while Mexican Pima and Central American Cabecar define the extremes of PC2 ( Figure S6). Most Native American haplotypes from the admixed genomes fall along this second axis of variation, and form clearly differentiated population clusters: one cluster is shared among Colombians and most Hondurans, while another one is shared among Cubans, Dominicans, and Puerto Ricans (no Haitian haplotypes were included due to low levels of Native American ancestry). Colombians and Hondurans cluster with Chibchan-Paezan speaking groups from Colombia and Central America, including Kogi, Waunana, and Embera. In contrast, Caribbean islanders cluster with Equatorial-Tucanoan speakers, which is a major linguistic group spread across the Amazonia. With few exceptions, Equatorial-Tucanoans form a rather tight cluster, including Guahibo, Ticuna, Palikur, Karitiana, among others, many of which are settled around fluvial territories of the rainforest. This fact may have facilitated communication from and to the coast, explaining their relationship with Caribbean native components. Interestingly, the indigenous component of insular Caribbean samples seems to be shared across the different islands, indicating gene flow across the Caribbean basin in pre-Columbian times.
To explore this possibility into more detail, we performed a model-based clustering analysis using the full reference panel of 52 Native American populations from Reich et al. [11] in addition to our three native Venezuelan populations. Individual admixture proportions from K=2 through 20 are given in Figure S7. Focusing on Native American components, the first subcontinental signal (at K=4) was accounted for a Chibchan component mainly represented by the Cabecar from Costa Rica and the Bari from Venezuela. Higher order clusters pulled out Amazonian population isolates such as the Surui and Warao, as well as northern populations including the Eskimo-Aleut and Pima, in agreement with the outliers detected in our ASPCA analysis ( Figure S6). Interestingly, from K=5 through 10, the Chibchan component is shared at nearly 100% with the Yukpa sample located near the Venezuelan coast, and at nearly 20% with Mayans from the Yucatan peninsula ( Figure 4B). The presence of considerable proportions of the Chibchan component in the Mayan sample is indicative of possible "back" migrations from Central America and northern South America into the Yucatan peninsula, revealing an active gene flow across the Caribbean, probably following a coastal or maritime route. Moreover, very high order clusters maintain the connection between Mayans and South American components. For example, at K=16 (the model with the lowest cross validation error; Figure S8b), as much as an average of 35% of the genome in Mayans is shared with a mixed Chibchan/Equatorial-Tucanoan component mainly represented by Ticuna, Guahibo, Embera, Waunana, and Arhuaco, among others ( Figure 4C). This observation is in agreement with our ASPCA results and reinforces the notion of a South American expansion of Native American components across the Caribbean basin.

European Ancestral Components
We performed ASPCA analysis by projecting the European segments of admixed individuals onto an extensive reference panel of European source populations, including 1,387 individuals from all over Europe sampled as part of the POPRES project [9], as well as additional Iberian samples from Galicia, Andalusia, and the Basque country in Spain (Rodriguez et al., in revision). The total projection involved 2,882 European haplotypes and 255 haplotypes of European ancestry from the admixed populations. Figure 5 shows the projection of the first two PCs where the background samples recapitulate a PCA map of Europe as reported before [15,24]. While most of the additional Iberian samples cluster together with the POPRES individuals sampled as Portuguese and Spanish, the Basques cluster separately from the centroid of most Iberian samples. The Basques are known for their historical and linguistic isolation, which could explain their genetic differentiation from the main cluster due to drift. Given the known Iberian origin of the first European settlers arriving into the Caribbean and surrounding territories of the New World, one would expect that European blocks derived from admixed Latino populations should cluster with other European haplotypes from present day Iberians. Indeed, our Latino samples aggregate in a well-defined cluster that overlaps with the cluster of samples from the Iberian Peninsula (i.e., Portugal and Spain). However, we observed that the centroid is substantially deviated with respect to the Iberian cluster (bootstrap p-value <10-4, see Materials and Methods), suggesting the possibility of a bottleneck and drift impacting the European haplotypes of Latinos.
Importantly, when we applied ASPCA analysis using the exact same reference panel of European samples but projecting Mexican haplotypes of European ancestry (Moreno-Estrada, Gignoux et al., in preparation), we did not observe a deviated clustering pattern from the Iberian cluster: the effect is much weaker and not significant (bootstrap p-value = 0.099, see Figure S9). Furthermore, the deviation of the European segments of Mexican individuals from the distribution of the rest of Iberian samples is even smaller than the deviation of the Portuguese from the Spanish samples. We further evaluated whether the dispersion of the different subpopulations within the Caribbean cluster follow particular patterns along ASPC2, the axis driving the deviation from the Iberian centroid. We observed that Colombians and Hondurans tend to account for lower (more deviated) ASPC2 values compared to Cubans, Dominicans, and Puerto Ricans (Figure S10), suggesting a mainland versus insular population differentiation. We then performed a Wilcoxon rank test to contrast ASPC2 for mainland (Colombia and Honduras) versus island (Cuba, Dominican Republic and Puerto Rico), resulting in a highly significant pvalue (1.5e-15). Because >25% of European ancestry was required for inclusion in ASPCA analysis, only two Haitian haplotypes were projected, and thus not included in the statistical analysis. Nonetheless, it is noteworthy that one of them clusters with the French, in agreement with historical and linguistic evidence about European settlements in the island (see arrow on Figure 5).
Among European populations, Iberians also have the highest proportion of identical by descent (IBD) segments that are shared with Latino populations, as measured by WELat, a statistic that is informative of the total amount of shared DNA between pairs of populations (see Figure S11 and Text S2). To explore the distribution of IBD sharing within continental groups, we considered Caribbean Latinos and Europeans separately by summing the cumulative amount of DNA shared IBD between each pair of individuals within each group. If European segments from Latino populations derive from a reduced number of European ancestors, then IBD sharing is expected to be increased among Caribbean individuals compared to European source individuals. Indeed, we observed a higher number of pairs sharing larger total IBD segment lengths among Latino individuals than among Europeans ( Figure S12). Within-population endogamy is also compatible with increased IBD sharing. However, this is more likely to occur between individuals from the same subpopulation (e.g., COL-COL) rather than individuals from geographically separated subpopulations (e.g, COL-PUR). For this reason we considered Latinos as a single group as a measure to minimize such possible effect. Yet, we observed an increased proportion of IBD sharing among Latinos, arguing for a shared founder effect.
These results are in agreement with our cluster-based analysis focused on global ancestry proportions, where the European ancestry of Latinos is dominated by a shared Latino-specific component differentiated from both southern and northern European components, although shared to some extent with Spanish and Portuguese ( Figure 1C). Bottlenecked populations may exhibit higher levels of differentiation from their parental gene pool due to loss of prior diversity and shift in allele frequencies. According to F ST estimates between K=8 ancestral clusters (from Figure 1C), the differentiation between southern and northern European components is 0.02. In the absence of drift, a southern-derived Latino component would be expected to show lower F ST values against its closely related southern component. However, the F ST between these two components was 0.021 (Table S3), meaning that the differentiation of the Latino-specific component with respect to southern Europeans is at least as high as the north-south differentiation within Europe. This observation was replicated when including additional Latino and ancestral populations ( Figure S7). Given the increased number of divergent clusters, we focused on K=18 through 20, in which all sub-European components were jointly detected. In this case, the Latino-specific component shows further fragmentation into two components: one predominantly shared among insular Caribbean samples and the other among mainland Latinos. The F ST value for southern versus northern European differentiation was 0.039, while values for southern versus insular (0.041) or mainland Latinos (0.04) were slightly inflated (Table S4), supporting the notion of additional differentiation impacting the European lineages of present day admixed Latinos.

African Ancestral Components
The Caribbean hosts a rich history of population exchange with the African continent as a result of unprincipled slave trade practices during European colonialism. Its proximity with the North Atlantic Ocean facilitated nautical contact with the West African coast and resulted in greater exposure to slave trade routes for the local population and, ultimately, in genetic admixture. The proportion of African ancestry is consistently higher in Caribbean populations compared to individuals from the mainland ( Figure 1C), and this finding is consistent across studies [3,6,25]. To explore the sub-continental composition of African segments derived from Caribbean admixed genomes, we performed ASPCA analysis on individuals with more than 25% of African ancestry using a diverse panel of African populations as potential sources (see Table  S1). Our first approximation showed no dispersion of Afro-Caribbean haplotypes over PCA space. Instead, they form a relatively tight cluster that overlaps with that of the Yoruba sample from southwestern Nigeria ( Figure S13). This is a plausible result, given the extensive historical record supporting a West African origin for the African lineages in the Americas.
However, according to our tract length analysis, there is strong genetic evidence for the occurrence of at least two pulses of African migrants imprinting different genomic signatures in present day admixed Caribbean populations. This poses the question of whether both pulses involved the same source population during the admixture process. If this were the case, it would easily explain our ASPCA results, where all African haplotypes point to a single source.
Alternatively, if more than one source was involved and if enough mixing occurred since the two pulses, it is possible that what we see is the midpoint of the two source populations, causing the difference to remain undetected by our standard ASPCA approach (which gives a point estimate averaging the signature of all African blocks along the genome). Hence we applied a different strategy, in which ASPCA is performed separately for short (thus older) and long (younger) ancestry tracts. For this purpose, we split the African segments of each haploid genome into two categories based on a 50cM length cutoff, and intersected the data with a reference panel of West African populations ( Figure 6A). Then, for each individual, we compute assignment probabilities of coming from each of the putative parental populations based on bivariate normal distributions fitted around each PCA cluster (see Materials and Methods, Figure S14). In Figure 6B we present the scaled mean probabilities for long (>50 cM) versus short (<50 cM) African tracts in Puerto Rican individuals. The pattern that emerges reveals that African haplotypes shorter than 50 cM are more likely to have originated from populations in the coastal Northwest region, such as the Mandenka and Brong; whereas longer haplotypes show higher probabilities of coming from populations closer to the Gulf of Guinea and Equatorial West Africa, including Yoruba, Igbo, Bamoun, Fang, and Kongo (see map on Figure 6A). The significant increase in old, short Mandenka tracts when compared to longer, more recent tracts, was replicated in other insular Caribbean populations, including Cubans and Dominicans. The Brong also seem to have had a greater contribution deeper in the past not only in Puerto Ricans, but also in Dominicans, Hondurans and to a lesser extent in Colombians. In Cubans, the trend is reversed and the Brong seem to have contributed more to long tracts than to short ones ( Figure S15).
One caveat is that short ancestry tracts are more likely to be misassigned. To rule this out as a source of the signal, we added an intermediate block size category (>5 cM and <50 cM) and repeated the size-based ASPCA analysis. We observed that, despite the signal being somewhat weaker due to the lesser amount of overall data, a similar trend was retained after the exclusion of extremely short tracts ( Figure S15). Finally, we gathered additional evidence by running local ancestry estimation on the African blocks alone to distinguish Mandenka vs. Yoruba ancestry tracts (see Materials and Methods). We then binned all segments of inferred Mandenka ancestry into different block sizes and observed that the proportion of the African ancestry called Mandenka is higher within shorter block sizes and decreases as block size increases ( Figure 6C). This gives additional support for the differential origin of African segments and argues that such signal is not driven by the shortest genomic segments alone, but rather characterized by a progressive decay of haplotype length from older migrations as younger segments (of different ancestry) account for the majority of longer African tracts in Caribbean genomes.

Models of admixture for Caribbean and mainland populations
Our results reveal consistent differences regarding the admixture processes occurring across the Caribbean islands as compared to those occurring in neighboring mainland populations. First, our data suggest multiple pulses of African migration contributed significantly to genetic ancestry in the Caribbean, consistent with records of historical slave trade routes. In contrast, we find evidence of a single gene flow event of Native American ancestry into admixed Caribbean populations. Since Native American tracts are shorter, on average, than tracts of African ancestry (and therefore older), this suggests the migration event is the initial founding of admixed populations at the time of European contact. Mainland populations from Colombia and Honduras, on the other hand, are best fit by a model of repeated migration events of European ancestry, consistent with a continuing expansion of Europeans during colonialism. We also find longer Native American tracts than African ancestry tracts in mainland populations, indicating a single pulse of the latter and a greater contribution of Native Americans into admixed continental populations. Admixture timing estimates also show consistent differences between these two groups, with admixture starting around 16-17 generations ago in the islands and 14 generations ago in mainland populations.
Our model shows remarkable agreement with historical records. The earliest European voyage by Christopher Columbus took place in 1492 (i.e., 16-17 generations ago), reaching the Caribbean island of Hispaniola (today's Dominican Republic and Haiti). Later European voyages reached the coasts of Central and South America, so permanent European settlements did not occur in the mainland until the first half of the XVI century, consistent with an approximate difference of 2 generations between the estimated onset of admixture according to our island and mainland models. Here we have focused on Colombians and Hondurans as population samples from mainland territories with coastal access into the Caribbean, but we have previously reported admixture timing estimates for Mexicans as well, namely starting 15 generations ago [5]. The settlement of Europeans in mainland Mexican territory is documented to have occurred between 1519 and 1521 (i.e., 27-29 years apart from the first contact in 1492 in the Caribbean), that is ~1 generation apart between the average estimate of 16 generations for the onset of admixture in the Caribbean compared to 15 generations from our model based on Mexican data. The abundance of historical records about European colonization of the New World is particularly exceptional, facilitating the contrast between written and genetic registries.

South American origin of indigenous components in the Caribbean
In contrast with other regions in the Americas where indigenous peoples are numerous, the genetic characterization of Native American components in the Caribbean required indirect reconstruction via genomic assembly of indigenous ancestry tracts transmitted into extant admixed individuals. By applying ancestry-specific PCA and cluster-based analyses integrating a large number of indigenous groups throughout the Americas, we found that Equatorial-Tucanoan speakers from South America hold the closest relationship with Caribbean indigenous components. This was also observed in a different sample set from the 1000 Genomes Project (Gravel et al., submitted). Despite covering a large geographic area of South America (ranging from northern Colombia and Brazil to southern Bolivia and Paraguay), most Equatorial-Tucanoans cluster together in PCA space, arguing for a common origin. We have intentionally included three additional tribes from the Venezuelan coast since logical candidates for the origin of the ancestors of Caribbean populations include indigenous coastal groups south the Lesser Antilles. However, despite their closer geographic location, none of these groups primarily accounted for the indigenous ancestry of the insular Caribbean samples, pointing to an inland origin rather than a coastal one. Nonetheless, our cluster-based analysis revealed that native Venezuelan components do share membership with several Central American indigenous populations, such as the Costa Rican Cabecar, and, to a lesser extent, with Mayan groups from the Yucatan peninsula of present day Mexico, suggesting substantial gene flow across the Caribbean Sea in pre-Columbian times. In fact, based on the distribution of jade, obsidian, pottery, and other commodities, archaeological evidence supports the existence of maritimebased interaction networks between central Mesoamerica, the Isthmo-Colombian area, and northern Venezuela [26]. Our results demonstrate that such long distance negotiations were accompanied by gene exchange between previously diverged native populations, and give a richer perspective of the dynamics between the inhabitants of the Caribbean basin prior to European contact.
In a recent genomic survey of the relationships between Native American peoples, Reich and colleagues [11] described the Chibchan speakers on both sides of the Panama isthmus as an exception to the simple model involving a southward expansion with sequential population splits and little subsequent gene flow. Instead, Central Americans, such as the Cabecar from Costa Rica, were modeled as a mixture of South and North American ancestry, which the authors reported as evidence for a back-migration from South into Central America. Our findings not only support such interpretation, but also suggest a distant connection between Caribbean Mesoamerica and South American inland territories. Specifically, the fact that Mayans from the Yucatan peninsula share 35% of their genome with Ticuna, Guahibo, and Arhuaco, among other Chibchan and Equatorial-Tucanoan speakers, supports the expansion of an inland South American component across the Caribbean. For context, it is noteworthy that in ASPCA the native ancestry tracts of Colombians and Hondurans cluster with geographically closer indigenous tribes, such as Chibchan speakers from Colombia and Central America. How to account, then, for a shared clustering between more distant Equatorial-Tucanoan speakers (mostly of Amazonian origin), and insular Caribbean haplotypes? One possible explanation is that the fluvial nature of most of these settlements (across the Amazon and Orinoco basins) may have facilitated people movement to the coast, and eventually migrating north through the Lesser Antilles, explaining their relationship with Caribbean native components. In fact, our results are consistent with archaeological records suggesting that the ancestors of the indigenous people that Columbus encountered might have come from populations that migrated from the Lower Orinoco Valley around 3,000 years ago [27,28]. Additionally, our results align with the classification of languages spoken by pre-Columbian inhabitants of the Caribbean. Together with Caribs, Tainos were the major group living in the Greater Antilles and surrounding islands at the moment of European contact. Tainos and insular Caribs spoke Arawakan languages that belong to the Equatorial sub-family, in the Equatorial-Tucanoan family [29]. The geographic distribution of Arawakan languages across northern South America resembles the sampling coverage of the Equatorial-Tucanoan individuals analyzed here (see map in [11]), supporting our findings. Previous genetic studies have also pointed to a South American origin for Tainos [7,30]. Based on mitochondrial haplogroups ascertained from pre-Columbian human remains, Lalueza-Fox and colleagues [30] found that only two of the major mtDNA lineages, namely C and D, were present in their sample (N=27). Given that high frequencies of C and D haplogroups are more common in South American populations, the authors argued for that sub-continent as the homeland of the Taino ancestors.
Overall, our analysis of indigenous ancestry tracts from extant admixed genomes supports previous linguistic, archaeological, and ancient DNA evidence about the peopling of the Caribbean, and goes beyond by pointing to a greater involvement of inland Amazonian populations during the last migration into the Antilles prior to European contact. Earlier migrations may have occurred (e.g., from Mesoamerica or the Florida peninsula), as pre-ceramic archaeological evidence of human presence in the Greater Antilles dates back more than 7,000 years ago [27]. However, the fact that the Equatorial-Tucanoan component is shared among the indigenous haplotypes from different insular and continental populations supports either a single South American origin of Caribbean settlers or a major population replacement involving a more recent migration of agriculturalists from inland South America.

Founder effect in the European lineage of admixed Latinos
We find genomic patterns compatible with the effect of a founder event in the ancestral European population of present day admixed Latinos. Supporting evidence include: 1) a Latinospecific European component revealed by clustering algorithms, which is not assigned to source populations within Europe except Spain and Portugal, and detected at lower order clusters compared to other European and Native American sub-continental components; 2) inflated F ST values between the Latino-specific and southern European components, compared to southern versus northern Europe differentiation; 3) significant deviation of the distribution of European haplotypes from the main cluster of Iberian samples in ASPCA space; and 4) increased IBD sharing among Latino individuals compared to Europeans. Additionally, a similar signature was observed in an independent dataset of Latino samples from the1000 Genomes Project using a combined approach that integrates IBD and local ancestry tracts (Gravel et al., submitted). These findings suggest that early European waves of migration into the New World involved a reduced ancestral population size, mainly composed by Iberians, bearing a subset of the diversity present within the source population, causing the derived admixed populations to diverge from current European populations. Furthermore, we find differences between mainland and insular Caribbean populations including 1) different time estimates for the onset of admixture as revealed by ancestry tract length analysis ( Figure 3); 2) separate memberships in cluster-based analyses ( Figure 4C, Figure S7); and 3) significantly shifted distributions within the Latino cluster in ASPCA projection of European haplotypes ( Figure 5, Figure S10). The fact that mainland Colombians and Hondurans show not only the highest proportions of the Latinospecific European component in ADMIXTURE but also the most extreme deviation from the Iberian cluster in ASPCA, suggests stronger genetic drift in these populations, compatible with a two-stage European settlement involving insular territories at first, and mainland populations subsequently absorbing a subset of migrants from the islands.
There is documented evidence of extensive migration from the islands to the continent throughout the 16 th century [21]. There were only two viceroyalties of the Spanish Empire in the New World until the 18 th century -Viceroyalty of New Spain (capital, Mexico City) and Peru (capital, Lima)-. An additional viceroyalty in South America was created in 1717 with Bogota as capital (Viceroyalty of New Granada), promoting economic and population growth. Interestingly, the estimated time for the second pulse or European migrants into the admixture of present day Colombians (i.e., 12 generations ago) coincides with the creation of the Colombianbased Viceroyalty of New Granada, accounting for the large increase (from 12.5% to 75%) of European ancestry in the model based on tract length distributions. Such small contribution of European ancestry at the onset of admixture in Colombians reinforces the idea that their patterns of European diversity are heavily impacted by a reduced number of founders. In contrast, Mexican-derived European haplotypes do not appear to be impacted by founder events as much as the Caribbean populations analyzed here. A possible explanation is that present day Mexico was the center of the wealthy Viceroyalty of New Spain, constituting one of the largest European settlements under Spanish rule, ensuring continuous exchange with Spain throughout colonial times, resulting in a larger ancestral population size.

Space and time distinction of African migrations into the Caribbean
We find that populations from the insular Caribbean are best modeled as mixtures absorbing two independent waves of African migrants. Assuming a 30-year generation time [22], the estimated average of 15 generations ago for the first pulse (i.e., circa 1550) agrees with the introduction of African slaves soon after European contact in the New World. At first, local natives were used as the source of forced labor, but populations were decimated rapidly, giving rise to the four century long transatlantic slave trade, which is usually divided into two eras. The first one accounted for a small proportion (3-16%) of all Atlantic slave trade, whereas the second Atlantic system peaked in the last two decades of the 18 th century, accounting for more than half of the slave trade. This period of increased activity coincides with the estimated age of the second (and stronger) pulse of African tracts according to our model (e.g., 7 generations ago in Dominicans, one of the major absorbing populations, pointing to late 18 th century). In other words, the estimated time separation between these two pulses (i.e., 8 generations or ~240 years) based on genetic data is in extraordinary agreement with historical records, recapitulating the span between the onset of African slave trade and its period of maximum intensity right before its rapid decline during the 19 th century [31].
To address the question of whether there was also a separation in space between the origins of these two pulses, we relied on the fact that chromosomes from older contributions to admixture have undergone more recombination events, thus leading to shorter continuous African ancestry tracts. By conducting two different but complementary size-based analyses restricted to genomic segments of inferred African ancestry, we provide compelling evidence that short African tracts are enriched with haplotypes from northern coastal West Africa, represented by Mandenka samples from Senegal and Brong from western Ghana, near the Ivory Coast. This is in agreement with documented deportation flows during the 15 th -16 th centuries, where most enslaved Africans were carried off from Senegambia and departed for the Americas from the Gorée Island, near Cape Verde [31]. African slaves were embarked by European traders in ports along the West African coast, but raiding zones extended inland with the involvement of local African kingdoms. The Mandinka Kingdom of Senegambia was part of the Mali Empire, one of the most influential domains in West Africa, spreading its language, laws, and culture along the Niger River. The empire's total area included nearly all the land between the Sahara Desert and coastal forests, and by 1530 reached modern-day Ivory Coast and Ghana, possibly accounting for the shared pattern between Mandenka and Brong with respect to Caribbean's short ancestry tracts. While such interpretation is supported by the fact that Mandenka and Brong are the westernmost population samples of our reference panel, the lack of additional samples from northern West Africa prevent us from determining whether this pattern is shared with other tribes as well. On the other hand, the greater affinity of longer ancestry tracts with the rest of the samples, which cover much of the central West African coast, is compatible with the greater involvement of such regions in the slave trade during the 18 th century.
The volume of captives being embarked from the bights of Benin (e.g., today's Nigeria) and Biafra (e.g., today's Cameroon) was so elevated after 1700 that part of its shore soon became known as the "Slave Coast" [31]. Population samples around this area represented in our reference panel include Yoruba and Igbo from Nigeria, and Bamoun and Fang from Cameroon, all of which show higher probabilities of being assigned as the source for longer African ancestry tracts in the admixed Latino groups analyzed. In fact, together with Brazil, the Caribbean Islands were the major slave import zone during the 18 th century. Later deportation flows in the 19 th century involved ports of origin near the Congo River in West Central Africa. The closest population sample of our reference panel from this region is represented by the Kongo, which also shows higher affinity with longer ancestry tracts, compatible with a later contribution into the admixture of the Caribbean. The 19 th century also saw the abolition of slavery in most parts of the world, but the massive international flow of people it involved, remains as one of the deepest signatures in the genome of descendent populations. While the geographic extension of the regions of origin of African slaves brought to the Americas has been widely documented, it was unclear until now the extent to which particular sub-continental components have shaped the genomic composition of present day Afro-Caribbean descendants. Our ancestry-specific and size-based analyses allowed us to discover that African haplotypes derived from Caribbean populations still retain a signature from the first African ancestors despite the later dominance of African influx from multiple sub-continental components.

Conclusion
Our genome-wide dense genotyping data from six different populations of Caribbean descent, coupled with the availability of large-scale reference panels, allowed us to address longstanding questions regarding the origin and admixture history of the Caribbean Basin. The differences between insular and continental Caribbean populations underscore the importance of characterizing admixed populations at finer scales. We report ancestry-specific recent bottlenecks affecting particular Latino groups, but not others, which may have important implications in the expected relative proportion of deleterious mutations and elevated allele frequencies that can be detected via association studies in theses populations. Finally, the extensive population stratification within sub-continental components implies that medically relevant genetic variants may be geographically restricted, which reinforces the need for sequencing target populations in order to discover local variants that may only be relevant in Latino-specific association studies for disease.

Samples and Data Generation
Generated data and assembled datasets for this study are summarized in Table S1. A total of 251 individuals representing six different Caribbean-descent populations were recruited in South Florida, USA. Participants were required to have at least three grandparents from their countries of origin, thus limited ethnographic and anonymous pedigree information was collected. The majority of pedigrees (94.3%, n=82) had four grandparents from the same country. Only 5 pedigrees (5.7%) had one grandparent from a different country. Informed consent was obtained from all participants under approval by the University of Miami Institutional Review Board. A total of 76 trios, 2 duos, and 19 parents were genotyped using Affymetrix 6.0 SNP arrays, which included: 80 Cubans, 85 Colombians, 34 Dominicans, 27 Puerto Ricans, 19 Hondurans, and 6 Haitians. Out of 173 founders, 18 samples were filtered from structure analyses due to cryptic relatedness as inferred by IBD>10%. Four trios were not considered for trio phasing due to an excess of Mendelian errors (>100K), two trios were removed due to 3 rd or higher degree of relatedness between parents as inferred by IBD, and five trios were filtered due to cryptic relatedness between members of different trios above 10% IBD. After filtering, 65 complete trios remained for haplotype-based analyses. To study population structure and demographic patterns involving relevant ancestral populations, 79 previously collected samples from three native Venezuelan tribes were genotyped using the same array (i.e., 25 Yukpa [aka Yucpa], 29 Bari, and 25 Warao). We combined our data with publicly available genomic resources and assembled a global database incorporating genome-wide SNP array data for 3,042 individuals from which two datasets with different SNP densities were constructed (see Table S1). The high-density dataset included populations with available SNP data from Affymetrix arrays; namely African, European, and Mexican HapMap samples [8], Europeans from POPRES [9], West Africans from Bryc et al. [10], and Native Americans from Mao et al. [32]. After merging and quality control filtering, 389,225 SNPs remained and representative population subsets were used in different analyses as detailed through sections below. Our lower density dataset (30,860 SNPs) resulted from the intersection of our high-density dataset with available SNP data generated on Illumina platform arrays, including 52 additional Native American populations [11], as well as additional Latino populations sampled in New York City [7] and 1000 Genomes Latino samples [6]. The resulting dataset combines genomic data for 1,262 individuals from 80 populations. Full details on the population samples are available in Table S1.

Population Structure
An unsupervised clustering algorithm, ADMIXTURE [12], was run on our high-density dataset to explore global patterns of population structure among a representative subset of 641 samples, including seven Native American, eleven PopRes European, HapMap3 Nigerian Yoruba, HapMap3 Mexican, and our six new Caribbean Latino populations (see Table S1). Fourteen ancestral clusters (K=2 through 15) were successively tested. Log likelihoods and cross-validation errors for each K clusters are available in Figure S3. F ST based on allele frequencies was calculated in ADMIXTURE v1.22 for each identified cluster at K=8 and values are available in Table S3. Our low-density dataset comprising 1,262 samples (detailed in Table  S1) was used to run K=2 through 20. Log likelihoods, cross validation errors and F ST values from ADMIXTURE are available in Figure S8 and Table S4. Principal component analysis (PCA) was applied to both datasets using EIGENSOFT 4.2 [33] and plots were generated using R 2.15.1. Global ancestry estimates from ADMIXTURE at K=3 were used to test the correlation between male and female ancestry proportions considering all trio founders within each Caribbean population as well as within the full set of admixed trios. Linear models and permutations (up to 100,000) were performed using R 2.15.1.

Phasing and Local ancestry assignment
Family trio genotypes from our six Caribbean populations and continental reference samples were phased using BEAGLE 3.0 software [34]. Local ancestry assignment was performed using PCAdmix (http://sites. google.com/site/pcadmix/ [19]) at K=3 ancestral groups. This approach relies on phased data from reference panels and the admixed individuals. To maintain SNP density and maximize phasing accuracy we restricted to a subset of reference samples with available Affymetrix 6.0 trio data, namely 10 YRI, 10 CEU HapMap3 trios, and 10 Native American trios from Kidd et al. [5]. Each chromosome is analyzed independently, and local ancestry assignment is based on loadings from Principal Components Analysis of the three putative ancestral population panels. The scores from the first two PCs were calculated in windows of 70 SNPs for each panel individual (in previous work we have estimated a suitable number of 10,000 windows to break the genome into when inferring local ancestry using PCAdmix, and in this case, after merging Affymetrix 6.0 data from admixed and reference panels, a total of 743,735 SNPs remained /10,000 = window length of ~70 SNPs). For each window, the distribution of individual scores within a population is modeled by fitting a multivariate normal distribution. Given an admixed chromosome, these distributions are used to compute likelihoods of belonging to each panel. These scores are then analyzed in a Hidden Markov Model with transition probabilities as in Bryc et al. [10]. The g (generations) parameter in the HMM transition model was determined iteratively so as to maximize the total likelihood of each analyzed population. Local ancestry assignments were determined using a 0.9 posterior probability threshold for each window using the forward-background algorithm. In analyses that required estimating the length of continuous ancestry tracts, the Viterbi algorithm was used. An assessment of the accuracy of this approach is given in [5].

Tract length analysis
We used the software Tracts [20] to identify the migratory model that best explains the genome-wide distribution of ancestry patterns. Specifically, we considered three migration models, each featuring a panmictic population absorbing migrants from three source populations. The models differ by the number of allowed migration events per population. In the simplest model, the population is founded by Native American and European individuals, and later receives a pulse of African migrants. The initial ancestry proportion and timing, as well as the African migration amplitude and timing, are fitted to the data as described below. The other two models feature an additional input of either European or African migrants; the timing and magnitude of this additional pulse result in two additional parameters that must be fitted to the data. Here, the data consisted of Viterbi calls from PCAdmix (see previous section and Figure 2), that is, the most probable assignment of local ancestry along the genomes. To fit parameters to these data, we tallied the inferred continuous ancestry tracts according to inferred ancestry and tract length using 50 equally spaced length bins per population, and one additional bin to account for full chromosomes. Given a migration model and parameters, Tracts calculates the expected counts per bin. Assuming that counts in each bin are Poisson distributed, it produces a likelihood estimate that is used to fit model parameters. For each population, we report the model with the best Bayesian Information Criterion (BIC) -2 Log(L) + k Log (n), with n=153.

Ancestry-Specific Principal Component Analysis (ASPCA)
To explore within-continent population structure, we applied the following approach for each of the continental ancestries (i.e., Native American, European, and African) of admixed genomes. The general framework is shown in Figure 2. It comprises locus-specific continental ancestry estimation along the genome, followed by PCA analysis restricted to ancestry-specific portions of the genome projected onto sub-continental reference panels of ancestral populations. For this purpose, we used our continental-level local ancestry estimates provided by PCAdmix to partition each genome into ancestral haplotype segments, and retained for subsequent analyses only those haplotypes assigned to the continental ancestry of interest. This is achieved by masking (i.e., setting to missing) all segments from the other two continental ancestries. Because ancestry-specific segments may cover different loci from one individual to another, a large amount of missing data results from scaling this approach to a population level, which limits the resolution of PCA. To overcome this problem, we adapted the subspace PCA (ssPCA) algorithm introduced by Raiko et al.
[35] to implement a novel ancestry-specific PCA (ASPCA) that allows accommodating phased haploid genomes with large amounts of missing data. Our method is analogous to the ssPCA implementation by Johnson et al. [23], which operates on genotype data. In contrast, ASPCA operates on haplotypes, allowing us to use much more of the genome (rather than just the parts estimated to have two copies of a certain ancestry) and to project independently the two haploid genomes of each individual. Finally, ancestry-specific haplotypes derived from admixed individuals are combined with haplotypes derived from putative parental populations and projected together onto PCA space. Details of the ASPCA algorithm and constructed datasets are described in Text S1.

Differentiation of sub-European ancestry components
To measure the observed deviation in ASPCA of European haplotypes derived from admixed Caribbean populations with respect to the cluster of Iberian samples, a bootstrap resampling-based test was performed. The null distribution was generated from comparing bootstraps of Portuguese and Spanish ASPCA values as models of the intrinsic Iberian population structure. We then compared the ASPCA values of the admixed individuals and tested if the observed differences between Iberian ASPCA values and those of the admixed individuals are more extreme than the differences within Iberia. The distance was determined using the chi-squared statistic of Fisher's method combining ASPC1 and ASPC2 t-tests for each bootstrap. We ran 10,000 bootstraps to determine one-tailed p-values. As Iberians we considered: PopRes Spanish, PopRes Portuguese, Andalusians, and Galicians; and as Caribbean Latinos: CUB, PUR, DOM, COL, and HON. Additional tests were performed comparing Portuguese versus the rest of Iberians and between an independent dataset of Mexican individuals analyzed by Moreno-Estrada, Gignoux et al. (in preparation) projected onto ASPCA space using the same reference panel of European populations. A bivariate test was performed to measure the relative deviation from the Iberian cluster of the distribution given by the Caribbean versus the Mexican dataset. To determine whether insular versus mainland Caribbean populations disperse over significantly different ranges in ASPC2, a Wilcoxon rank test was performed between (COL+HON) versus (CUB, PUR, DOM). Haitians were excluded due to low sample size (N=2 haplotypes). Boxplot is available in Figure S10. Population differentiation estimates between clusters inferred with ADMIXTURE were visualized and compared across runs where both the Latino-specific and southern European components were detected. Values are available in Table S3 and Table S4. The analysis of IBD sharing was conducted using our high-density dataset extracting a subset of 203 PopRes European individuals and the founders from the 65 complete admixed trios. We first performed a genome-wide pairwise IBS estimation using PLINK [36] to ensure that the dataset contains no samples with more than 10% IBS with any other sample. Then we used fastIBD [34] to phase the data and estimate segments shared IBD longer than 2 Mb to eliminate false positive IBD matches and assuming that ancestry will be shared among pairwise IBD hits of segments this long. All 2 Mb or greater segments shared IBD between pairs of individuals were summed, and histograms were created for pairwise matches within each group (i.e., PopRes Europeans and Caribbean Latinos).

Size-based ASPCA analyses
Given the evidence from our tract length analysis for a second pulse of African migrants into the admixture of insular Caribbean Latinos, a modified size-based ASPCA analysis was performed. A reference panel was built integrating three different resources [8,10,37] and focusing on putative source populations from along the West African coast, including Mandenka from Senegal, Yoruba and Igbo from Nigeria, Bamoun and Fang from Cameroon, Brong from western Ghana, and Kongo from the Democratic Republic of the Congo. We begin with the continental local ancestry inference from PCAdmix K=3. For each individual we then divide African ancestry tracts into small (0 to 50 cM) and large (> 50 cM) size classes. Given a partition of African ancestry tracts, we take all sites included in one tract class, say short tracts, and run PCA on our sub-continental West African reference populations for only these sites. Using the first two PCs from this analysis, we fit a bivariate normal distribution to each reference population cluster. We then project our test sample into this PCA space, and estimate the probability of it coming from each reference population using the fitted distributions. This procedure is repeated for each tract class, for each individual. For each admixed Caribbean population, we can then estimate the probability that a given class of African ancestry tracts comes from a specific West African source population as the average probability of assignment to this population across all individuals. Finally, under the assumption that a given class of African tracts must come from one of the provided reference populations, we rescale these probabilities to sum to one. Each assignment estimate is also provided with error bars representing the standard error of the mean. We compare the short and long assignment probabilities for each Caribbean population to identify distinct sources for "older" and "younger" West African migratory source populations. Haitians were not included in the analysis due to low sample size (n=4). Due to concerns that shorter tracts have a higher likelihood of misassignment, we added a medium tract size class (5cM to 50 cM) to see if the results were simply due to very short (0 cM to 5 cM) European or Native American tracts being mis-classified as African. We compare the results for short and medium tracts and find that the trends are maintained suggesting the observation that older shorter tracts appear to be primarily from the Mandenka and Brong source populations is not simply due to short tract mis-assignment

Local ancestry estimation within African tracts
To identify likely regions of Yoruba versus Mandenka ancestry in the African component, we modified our implementation of PCAdmix to perform local ancestry deconvolution solely of the African segments of the admixed genomes. The modification is achieved in the final step of the algorithm: whereas the standard approach estimates a single HMM across an entire chromosome, here we fit J disjoint HMMs spanning each of the J blocks of African ancestry in a given chromosome for a given individual. Applying the method, we obtained posterior probabilities for Mandenka versus Yoruba ancestry within the previously inferred African segments. We then selected only those sub-regions that were confidently called as Mandenka or Yoruba, and stratified them by physical size.      Native American populations (gray symbols) from [11] sampled throughout the continent. B) Shared indigenous components across the Caribbean as revealed by cluster-based ADMIXTURE analysis at K=10 of Native American populations from [11], plus three additional Native Venezuelan tribes genotyped for this project. The plot includes individuals from Mesoamerica southwards only (Eskimo-Aleut, Na-Dene, and Canadian North Amerind not shown). Legends on top correspond to linguistic families for samples from [11], and to tribe names for the three additional   Table S1 and regions within Europe are labeled as in [16].   Mexicans and Hondurans distribute between the European and Mesoamerican clusters, whereas Colombians slightly deviate towards the Andean and Venezuelan clusters. Global PCA analysis based on the high-density dataset (~390K SNPs) and thus limited to reference panel populations with available Affymetrix SNP array data (see Table S1 for details).    Table S2).     Figure 4 for more details).    Figure 5 and Figure S9. Population codes as in   . Inset histograms display counts lower than 50 for the same binning categories. The overall count of pairs sharing short segments of total IBD is higher among Europeans, probably as a result of an older shared pool of source haplotypes. In contrast, the higher frequency of longer IBD matches among Latinos is compatible with a recent European founder effect.      Table S4. F ST divergences between estimated populations for K=20 using ADMIXTURE Text S1. Methodology of the Ancestry-Specific PCA (ASPCA) implementation

Text S2. Measuring pairwise IBD between European and Latino populations
Tables: Table 1 Models of Migration into the Caribbean after the advent of admixture  1 Three migration models were tested for each admixed population: a simple model of single pulses of migrants from each source population, beginning with Europeans and Native Americans at T 1 followed by African migrants at T 2 (EUR,NAT +AFR); the simple model followed by an additional pulse of European migrants (EUR,NAT +AFR +EUR); the simple model followed by an additional pulse of African migrants (EUR,NAT +AFR +AFR). Log likelihoods given either model were compared and we present the model with the best Bayesian Information Criterion (log likelihood values in bold). 2 The maximum likelihood estimate of time since admixture initially. We assume prior migration between the populations was zero. Time since migration began is indicated in generations.            1   IE   IE  IE   IE   IE  IE  IE   IE  IE IE   IE   IE   IE  IE  IE   IE  IE IE  IE  IE   IE  IE  IE  IE   IE  IE   IE   IE  IE  IE  IE  IE   IE  IE   IE   IE   IE   IE   IE  IE  IE  IE   IE   IE  IE   IE   IE IE   IE  IE   IE  IE   IE   IE  IE   IE  IE   IE   IE  IE   IE   IE   IE  IE  IE   IE  IE   IE   IE   IE  IE   IE   IE   IE   IE  IE  IE  IE  IE   IE   IE   IE  IE   IE   IE IE  IE  IE  IE   IE  IE   IE  IE   IE IE   IE   IE   IE   IE   IE  IE  IE   IE   IE   IE   IE   IE   IE   IE   IE   IE IE   IE  IE   IE   IE   IE  IE  IE   IE  IE IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT  IT   IT  IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT  IT   IT   IT  IT   IT   IT  IT   IT   IT   IT   IT  IT   IT  IT   IT   IT   IT   IT  IT   IT   IT   IT   IT  IT   IT   IT   IT IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT  IT   IT   IT  IT IT   IT   IT   IT  IT   IT  IT   IT   IT  IT IT   IT IT   IT   IT   IT   IT   IT  IT   IT  IT   IT  IT   IT   IT   IT  IT   IT   IT   IT   IT  IT  IT   IT  IT   IT  IT   IT  IT   IT  IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT  IT   IT   IT   IT  IT   IT  IT   IT   IT IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT  IT   IT  IT   IT   IT  IT   IT  IT  IT  IT   IT   IT   IT  IT  IT   IT   IT  IT  IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT  IT   IT  IT   IT   IT  IT   IT   IT   IT   IT  IT   IT  IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT  IT   IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT  IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   ITIT  IT  IT   IT  IT   IT   IT   IT  IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT IT  IT  IT   IT  IT   IT   IT   IT  IT   IT  IT   IT   IT   IT  IT  IT   IT   IT  IT   IT   IT   IT   IT   IT  IT   IT   IT   IT IT   IT  IT   IT   IT  IT  IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT   IT   IT   IT   IT   IT   IT   IT   IT   IT   IT  IT  IT   IT   IT   IT   IT   IT   IT   IT  IT   IT  IT   IT   IT   IT  IT   IT  IT   IT   IT  IT TSI  TSI  TSI  TSI  TSI TSI TSI TSI  TSI  TSI  TSI  TSI TSI  TSI  TSI TSI TSI  TSI TSI TSI  TSI  TSI TSI TSI  TSI  TSI TSI  TSI  TSI CEU  CEU  CEU  TSI  TSI  TSI  TSI  TSI   TSI   TSI   TSI  TSI   TSI TSI  TSI  TSI TSI  TSI  TSI  TSI  TSI TSI TSI  TSI  TSI TSI   TSI   TSI  TSI  TSI  TSI TSI TSI  YRI  YRI YRI  YRI  YRI Figure S5.