The Analysis of Autosomal STRs Draws the Current Genetic Map and Evolutionary History of Northernmost South America

2. Materials and Methods

2.1. Sample Selection and Classification

This study uses samples collected throughout the country between 2008 and 2017 by the Population Genetics and Identification research group of the Universidad Nacional de Colombia. Sampling was performed in association with local medical centers, and inclusion criteria included adults above 18 years old who were born in the sampling area or resided there for at least two years. Each sample has an informed consent and information from a survey that considered genealogical, demographic, and socio-cultural variables.

Table 1 shows the a priori genealogical classifications based on the collected information. Although some categories are named after the departments, they are not limited to the political-administrative context since they are based not only on birthplace but also on self-Identification of three generations of each participant’s paternal and maternal lineages. Overall, 364 unrelated individuals were grouped in 24 a priori genealogical classifications. Figure 1 shows their geographic distribution.

2.2. DNA Extraction, Amplification, and Genotyping

Samples were processed at the Population Genetics and Identification laboratory of the Universidad Nacional de Colombia. Peripheral blood and/or buccal mucosal epithelial cells were collected on FTA™ WHATMAN® cards, and DNA extraction was performed with Chelex (Walsh et al., 1991) FTA Whatman® buffer, and FTA Elute Whatman® protocols. Multiplex PCR amplification was completed using two different kits on different sample sets: 15 autosomal STR loci with AmpliF/STR Identifiler kit (Applied Biosystems, Warrington, UK); and 16 autosomal STR loci with AmpFLSTR^TM NGM^TM PCR Amplification Kit (ThermoFisher) (the Identifiler and the NGM kits have the same primers in the common markers). Capillary electrophoresis was run on an ABI Prism 310 Genetic Analyzer (Applied Biosystems, Stafford, TX). Data were processed and genotyped using Genemapper ID v.3.2 software (Life Technologies). The samples in this study were collected over more than ten years of research, so they were processed at different times using distinct marker kits. The analysis presented here employed the ten common markers among all data sets: D16S539, D18S51, D19S433, D21S11, D2S1338, D3S1358, D8S1179, FGA, TH01, and vWA, these were included in the analysis.

2.3. Statistical and Population Genetic Analysis

Population genetic analysis was based on the genealogical classifications described above. Allele frequencies, effective alleles, Hardy Weinberg equilibrium, and observed and expected heterozygosities (Nei, 1987) were calculated using Genepop v 4.6 (Rousset, 2017). The Arlequin v. 3.5.2.2 software (Excoffier et al., 1992) was also used for calculating Hardy Weinberg equilibrium, as well as for the analysis of molecular variance (AMOVA) and to perform paired Fst analyses with 10,000 permutations and a significance level of 0.05. A NeighborNet based on the paired Fst matrix was constructed using the Splits Tree4 v. 4.15.1 software (Huson & Bryant, 2006) to graphically represent the relationships between populations.

Subsequently, the Structure 2.3.4 software (Pritchard et al., 2010) provided an individual population assignment analysis under the admixture model and correlated allele frequencies with a burning length of 100,000 and 100,000 MCMC repeats. Five independent simulations were performed for k = 2-10 to obtain the most likely k by determining the modal distribution of ΔK (Evanno et al., 2005) using Structure Harvester (Earl, 2012) to find the number of gene clusters (K) that best fit these data. The results were uploaded to the Clumpak - Cluster Markov Packager Across K software for visualization.

The Neighbour Joining clustering methodology (Nei, 1987) was used to evaluate genetic distances using the Populations 1.2.31 software and TreeView X to plot the results. A multivariate analysis was also performed. Using the default parameters of the MVSP v.3.22 software (Kovach, 1999) the principal component analysis was obtained. Finally, a Mantel test was run to analyze the relationship between geographic and genetic distances and to evaluate independence between both matrices and spatial autocorrelation. This last test used the mantel.test function from the free R development (Team, 2019) employing 10000 permutations and an α = 0.05.

2.4. Classifications Based on Ethnic Belonging

Genetic diversity and Hardy Weinberg equilibrium analyses were also performed for the final ethnic classifications suggested by this study and forensically relevant statistics were calculated using the PowerStats v12.xls tool. Additionally, such classifications were compared with populations from different geographies published by other authors (Table 2). This data set is composed of various ancestry populations. We included samples from Native American and admixed people from Central and South America, as well as from East and West Africa and Europe. We performed principal component analyses using the MVSP 3.22 software (Kovach, 1999). The PCA analyses included data from all populations, as well as independent analyses of those with mainly African and Native American ancestry.

3. Results

3.1. A Priori Classifications Based on Genealogy

The NeighborNet in Figure 2 presents the paired Fst results. This representation shows populations that aggregate according to ethnicity and geographic location. There are three distinct regions: native American, admixed, and afro. There is one extra region associating people with various genetic and cultural histories. At the right side of the graph there are the Colombian afro-descendant communities of Chocó, the Raizales of San Andrés and Providencia and the afro-descendant populations of Guajira, Cesar, and Antioquia. The upper part of the net shows the most admixed populations of our sample: Migrants, Multiple ancestries, Magdalena, Tolima, Atlántico, Antioquia, Huila, and Bogotá. The bottom left part of the NeighborNet groups the Amazonian indigenous communities: Cocama, Ticuna, and Uitoto near the Amazonian population, while the Wayúu are in an independent branch next to the Zenú communities. Finally, at the top left of the net, we find the Arhuacos and Kankuamos (Native American communities from the Caribbean coast mountains) as well as the admixed people from Cordoba and Magdalena (also from the Caribbean coast). Despite being related to the Arhuacos and the Kankuamos, the Mokaná people appear in the mid-section of this graph. The Pijaos, self-determined acculturated natives from the Andean region, are in the middle of the two main native American descent groupings. A neighbor joining dendrogram constructed based on the pairwise Fst values from the 24 populations studied show similar groupings (Figure 1S).

The a priori genealogical categories were also analyzed by an individual assignment to the population in which simulations were performed for K= 2-10 (Figure 2S). According to the Evanno et al. (2005) method, the number of genetic groups (K) that best explains these data is K=5 (Figure 3). The bar charts obtained once again show what appear to be three groups with different ancestral prevalence. At k=5, the afro-descendant (ANT*, CHO, and RAI) and the Guajira populations show assigned individuals represented by a mostly blue genetic component. In the middle band of the graph, multiple ancestry populations (ANC, MIG, BOG, HUI, TOL, ANT, MAG, and CES) together with the Arhuacos, Kankuamos, Mokaná and Pijaos exhibit individual bars that suggest a higher degree of admixture. Finally, from the Amazonians to the Zenú, a pattern dominated by the orange color which we relate with Native American ancestry is observed. In the case of the Córdoba population, a predominantly orange pattern is also evident, which coincides with their grouping with the native populations in Figure 2.

The results from the Nei genetic distance analysis employing the Neighbour Joining clustering methodology support the groupings suggested by previous tests (Figure 3S). Likewise, this dendrogram confirms some previously observed proximities, such as those between the Guajira population and the afro communities, Córdoba and the Zenú, and between the native communities from the Caribbean coast mountains (KAN, ARH, and MOK) with the admixed populations from the departments they inhabit (MAG and ATL). Similarly, the principal component analysis shows that the clouds of data points scattered along the axes aggregate according to their ethnic belonging (Figure 4S).

The Mantel test, a method to assess the relationship between geographic distance and genetic divergence, was used to evaluate the similarity between those two matrices (Figure 5S). In this data set, we found a statistically significant correlation between genetic divergence (Fst) and geographic distances (km) (p-value < 0.0001; α = 0.05; r(AB) = 0.387) that suggests an adherence to the isolation-by-distance model. However, considering their correlation coefficient, these results should be interpreted with caution, and suggest the need for further research.

3.2. Final Population Classifications

After performing population genetics and statistical analysis based on the original classifications, we suggest a new classification that more closely reflects the genetic structure of the Colombian population. Table 3 shows the nine new classifications under the labels of afro-descendant, Native American, and Admixed ethnicity.

3.3. Genetic Diversity Analysis

Given the importance of considering our country’s genetic structure in the forensic practice and hypotheses formulation in future population genetics studies, here we report the allele frequencies (Table S1) and forensically relevant statistics of the nine final classifications (Table S2). Table S3 shows the observed heterozygosity (Ho) and inbreeding coefficient (Fis) values for each of the nine final genealogical classifications. The Wayúu had the lowest observed heterozygosity at 0.756 and the afro admixture classification the highest one at 0.837. These results show a correlation between genetic and geographic distances following the model that suggests heterozygosity losses throughout a series of founder effects experienced by the human population as it moved away from East Africa (Ramachandran et al., 2005).

Regarding the Hardy Weinberg equilibrium, there were some loci scattered throughout the population categories that exhibited p-values below 0.05. There was not a clear pattern of disequilibrium in a single locus throughout the population categories studied, and observed imbalances were solved after adjusting the p-values using the Benjamini-Hochberg method, except for the D19S433 marker of the Native American admixture group (Table S4).

3.4. Genetic Structure Analysis

We evaluated the genetic structure of the whole sample using a four-group AMOVA to test for differentiation levels. The general population exhibited variance mostly among individuals (97.63%) with a p-value of 0.000 (Table S5). Additionally, paired Fst matrices showed no significant genetic differences between the proposed genealogical categories (Table S6).

3.5. Comparative Analysis

Figure 4 shows a comparison between our nine classifications and other populations using a principal component analysis. The two-dimensional representation of the results shows a dispersion along the two axes according to African, European, and indigenous ethnicity, respectively from left to right. Afro populations from Colombia and the admixed population from the Dominican Republic lie between African and European populations, the admixed Latin American populations aggregate at the center of the graph, while the indigenous ones disperse along the two quadrants at the right.

We performed separate PCAs to assess closeness between populations with African and indigenous ancestry (Figures 6S and 7S, respectively). These results show the afro-Pacific sample closer to the admixed Dominicans and the admixed afro-Colombians, while the afro-Insular sample is closer to the Mandinga people from Senegal. The indigenous populations, on the other hand, are scattered through the graph according to their geographic location. The Caribbean admixed indigenous category is at one end of the dispersion.

4. Discussion

The differentiation of Colombia´s current human population goes beyond any artificial grouping based on its political-administrative units. The results presented here, rest on anthropological genetics methodologies and autosomal STR markers that expose diverse patterns of ethnic belonging and relationships between groupings that allude to shared evolutionary histories.

STR markers are large polymorphisms originally used for assessing mechanisms of evolution in population genetics analyses. Before them, other markers sought to unravel evolutionary signals, such as the HLAs that detect genetic drift through polymorphism-level differences between generations and populations, along with the ABO blood system that acts as a clear natural selection marker (Roberts et al., 1980). An example of how informative these classical markers can be, is that of Amerindian communities who used to carry only the O allele. Nowadays, it is possible to evaluate more than 500 years of gene introgression by calculating the frequencies of the O, A, and B alleles to infer the changes they have undergone since the conquest.

Amid a technological race that constantly seeks to employ newer markers and the possibility of exploring the genomics of a population, it should be noted that classical markers continue to be informative of evolutionary processes that answer questions arising in the field of population genetics. Undoubtedly, complete genomes allow for exhaustive analyses of populations, but these technologies are not yet easily accessible to many Latin American research groups that, however, do have access to valuable samples for ancestry studies. Also, there is a correlation between genetic ancestry estimations obtained with ancestry informative markers (AIMs) and high-density data (Adhikari et al., 2017). Such correlation implies that the use of new technologies does not exclude the potential of analyses made with markers whose databases are freely accessible and contain a robust accumulation of information over time, as the sample analyzed in this study. STRs can stratify populations at the continental level, and whenever accompanied by a body of complementary a priori information from the research subjects, they can also be good at doing so at a local scale.

Several works published about Colombia's population ancestry have used the Native American, afro-descendant, and European ancestry categories as if they were immoveable. However, our results show that considering that such categories respond to demographic history, their nature is more dynamic. Here we start with an a priori classification of Colombian population samples representative of several of its regions (Table 1). We proposed such classifications considering personal, genealogical, demographic, and cultural information. However, once we obtained the substructure and ethnic belonging assignment results (Figure 2 and Figure 3), we suggested a reclassification that includes two afro-type populations, three Native American-type populations, and four admixed-type populations (Table 2). Similar groupings were previously proposed based on AIMs (Ossa et al., 2016) (Mogollón Olivares et al., 2020).

The history of the American continent provides a general idea of how the gene pool of human populations reflects their subjection to mechanisms of evolutionary change over time. The settlement of European immigrants, for example, used to coincide with epicenters of the indigenous civilizations they encountered during their colonizing missions, while the settlement of African populations coincided with centers of slave trading and labor demand. Likewise, after the early 19th century independence wars, the degree of admixture in the new world depended on the territory, degree of migration and the legislation of each region (Adhikari et al., 2017).

Similarly, Colombia's current population reflects the historical dynamics of indigenous settlements established before the colony, the arrival of Africans from different geographies to different regions and times, and the subsequent admixture of these two groups with the European population. The genetic trace from Africa varies among territories, as shown here by two differentiated afro ancestry components: Chocó and San Andrés, in addition to the afro admixture category. We also found that some Native American peoples like the Wayúu, Zenú, Ticuna, Huitoto and Cocama have a genetic configuration that remains relatively preserved. Nevertheless, other self-determined indigenous peoples who remained in their ancestral territories, exhibit genetic introgression that is also reflected by their acculturation levels (loss of language) such as the case of the Pijaos, Kankuamos, and Mokaná. Finally, the population classified as European admixture also shows an ancestry component that seems to be more fixed throughout neighboring territories, but whose fluctuation depends on their specific demographic histories.

As mentioned above, microsatellite markers are not ancestry informative markers but make it possible to trace evolutionary change at a more recent scale. As our results show, STRs can be used to assign individuals to specific populations and ethnic belonging. The groupings obtained reflect similar origins, migration, and drift history, maybe not dating as far back as to reveal their ancestry history, but one that arose at a more recent time scale.

The network presented in Figure 2 shows people from Córdoba and Amazonas who did not self-determine as indigenous aggregating to the native American branches from such regions; a reflection of their shared geography that does not imply shared ancestry but recent admixture. It is worth highlighting that these samples come from field sampling at locations with a high degree of ethnic belonging, as reported by the National Administrative Department of Statistics (DANE). Likewise, the close association between Guajira, Cesar, and the afro branch does not imply African descent but might result from introgression of African genomes in the admixed population of the region. A history review unveils migrations and genetic flow caused by the exploitation of pearl banks at the Guajira coast during the 16th century that required enslaved labor force for agriculture, farming, housework, and as diving fishermen for pearls who were brought to the country from Guinea and Angola (Navarrete Peláez, 2003).

The Native American admixture category came up given the reiterated association between native populations from the Caribbean coast mountains and admixed people from Magdalena and Atlántico. As discussed above, these samples were collected at a location with a high degree of ethnic belonging, so even when their genealogical data and self-determination initially suggested independent categories, there is evidence of gene flow in their genomes. In contrast, despite having a small sample size from the native communities from the mountains (Arhuacos, Kankuamos and Mokaná), SNP AIMs analyses have shown that their tendency to associate does stem from a shared ancestry and consequent gene flow (Suárez et al., 2021).

The landscape of peopling processes and gene flow at the gateway of South America extends throughout the subcontinent. Based on SNP data from South America, Homburger et al. (2015) found substructure in populations of European ancestry, depending on whether they came from north or south Europe, as well as in those of Native American ancestry. According to the authors, in Peru there is substructure between the local Andean Quechua and Aymara native groups, while the genomes of indigenous peoples from Argentina and Chile carry a component from the south-center of the subcontinent and another one from the Andes.

The matter of multiple origins of indigenous ancestry in South America has been addressed over time using several genetic markers. The results from this study show a separation between the indigenous communities of the Caribbean and the Amazon (Figure 2). Previous local studies discussed the genetic differences between Amerindian groups in the Amazon and Orinoco regions compared to those from the Andes and the Caribbean coast (Keyeux & Usaquén, 2006). Using autosomal STR markers, a difference in genetic diversity from west to east between Andean and eastern Brazilian tribes was reported as one of the strongest signals of subcontinental genetic differentiation (Wang et al., 2007). Analyses of parental lineages have also reached conclusions about this issue. In terms of the Y chromosome, there is significant geographic structure of genetic variability between the Andean and western populations caused by mechanisms of evolution acting differently throughout South America (Tarazona-Santos et al., 2001). The differential evolutionary model that causes a contrast between east and west, was later upheld by analyses based on mitochondrial DNA (Fuselli et al., 2003).

Several theories about the settlement of the southern part of the continent have attempted to account for the existence of different types of indigenous ancestry. It has been suggested that the first settlers could have entered in a single wave that later separated into three migratory routes that would reach the Andes, the Amazon, and the coastal regions (Rothhammer & Dillehay, 2009) (Reich et al., 2012) or in two independent waves that entered from each side of the Andes, based on the observation that some Amazonian ethnic groups show more similarities with Australasian populations than with Eurasians or current Native Americans (Skoglund et al., 2015). A later study suggested that the signal linking Native Americans from the Brazilian Amazon with "Australasians" may have been the result of a substructure within Siberian populations that gave rise to Native American ancestors (Skoglund & Reich, 2016).

Although they do not exhibit as strong of separation as the two distinct types of indigenous ancestry in our sample, we found differences in association grouping between the African descent populations from mainland Chocó and the island of San Andres. The PCA results that compare our samples with other African or African descent populations (Figure 5S) show the afro Pacific samples closer to the admixed Dominican Republic and the admixed afro Colombian population, while the afro Insular population is closer to the Mandinga from Senegal.

Y chromosome studies of the Colombian population have reported a higher proportion of haplogroup E1b1a. Population analyses from Chocó, Bolívar, and Valle del Cauca reported the presence of such haplogroup in frequencies of around 30% (Rojas et al., 2010) while in populations from San Andrés and Providencia its frequency was 48% (Alonso & Usaquén, 2013). Haplogroup E1b1a reaches its highest frequencies in western, central, eastern, and southern regions of Africa (Trombetta et al., 2011) which is consistent with the fact that the ancestors of the afro Colombian populations were brought and enslaved between 1580 and 1650 from Senegal, Ivory Coast, Mali, and the west coast of Guinea (Navarrete, 2005).

Ultimately, the genetic map of Colombia presented here exhibits a correlation between its ethnic belongings and geography, as is the case in several Central and South American countries. Whole-genome analyses show that ancestry differences among admixed Latin Americans reflect, still today, the genetic imprint of the diversification underwent by pre-Columbian indigenous populations (Wang et al., 2008). Other studies that have employed haplotype-based methods show significant geographic correspondence between native ancestry components in Latin America and the genetic structure of current native groups (Adhikari et al., 2017) (Homburger et al., 2015). For example, the genomes of today´s admixed Mexicans reflect a high degree of fine-scale genomic structure shaped by pre-Columbian population dynamics (Moreno-Estrada et al., 2014). According to the researchers, such continuous geographic distribution of each indigenous component demonstrates a high correlation of individual admixture proportions with geography. In Colombia, SNP-based research has shown that people from the northwest of the country carry ancestry components resembling those of Chibchan-Paezan natives and Central American Mayans, while to the south there is an ancestry resemblance with Peru and Chile that is consistent with the central Andes being highly populated during the peak of the Inca Empire (Chacón-Duque et al., 2018). Finally, based on whole-genome data, it has been shown that Native Americans from Peru exhibit distinct ancestral divisions and that admixed populations have arisen from multiple indigenous communities that occurred before and during the Inca Empire and Spanish rule (Harris et al., 2018). The authors of this study also point to a strong biogeographic signal within the genetic variation of indigenous and admixed populations.

5. Conclusions

In the search for unraveling our country's genetic diversity and evolutionary history, the combined use of STRs, a preconceived sampling strategy aimed at collecting personal, genealogical, demographic, and biological data, as well as robust statistical and population genetics analytical tools, led to notable findings. We present a genetic diversity and ethnic belonging map of Colombia that suggests nine classifications of its populations and reveals traces of evolutionary processes discussed in the light of recent literature based on modern molecular markers. Such classifications indicate a correlation between ancestry and geography, and given their dynamism, emphasize the need to change obsolete colonial labels such as Caucasian or Mestizo. They also lead us to acknowledge the added value that the researcher’s deep knowledge of their study populations provides for obtaining additional layers of information.

The results presented here do not intend to declare STRs as the best alternative for this type of research. They put forward several population classifications supported by the extra information layers provided by anthropological genetics; classifications comparable to those suggested by modern molecular markers. On a local scale, these findings contribute to a deeper understanding of the genetics of a country going through a post-conflict era with an increasingly pressing demand for the human identification of its victims. On a more global scale, they contribute to the field of human population genetics when breakthroughs in molecular biology make it possible to answer increasingly concrete evolutionary questions that require information representative of the genetic variability of human populations from all ancestries and geographies.