Patterns of Genetic Coding Variation in a Native American Population before and after European Contact

Abstract

The effects of European colonization on the genomes of Native Americans may have produced excesses of potentially deleterious features, mainly due to the severe reductions in population size and corresponding losses of genetic diversity. This assumption, however, neither considers actual genomic patterns that existed before colonization nor does it adequately capture the effects of admixture. In this study, we analyze the whole-exome sequences of modern and ancient individuals from a Northwest Coast First Nation, with a demographic history similar to other indigenous populations from the Americas. We show that in approximately ten generations from initial European contact, the modern individuals exhibit reduced levels of novel and low-frequency variants, a lower proportion of potentially deleterious alleles, and decreased heterozygosity when compared to their ancestors. This pattern can be explained by a dramatic population decline, resulting in the loss of potentially damaging low-frequency variants, and subsequent admixture. We also find evidence that the indigenous population was on a steady decline in effective population size for several thousand years before contact, which emphasizes regional demography over the common conception of a uniform expansion after entry into the Americas. This study examines the genomic consequences of colonialism on an indigenous group and describes the continuing role of gene flow among modern populations.

Introduction

The indigenous peoples of the Americas suffered extensive population declines associated with the impact of European colonization. Although the precise extent of this decline is contested^{1, 2} and likely varied with local circumstance,³ these events should have affected the genetic variation within surviving indigenous populations. These effects are, however, further complicated by patterns of gene flow from both native and non-native immigrant groups. Although previous studies have explored the genetic diversity of contemporary Native American populations,^{4, 5, 6, 7, 8} the effects of colonization have not been examined with the aid of studies concerning ancient Native American genetic (autosomal) diversity. Here we examine the effects of colonization by comparing the genome-wide patterns of an indigenous population from two different time frames: before and after European contact.

Broadly speaking, genetic variation among human populations can result from numerous sources, including stochastic (i.e., mutation, recombination, migration, and genetic drift) and deterministic (i.e., natural selection) processes.⁹ With the advent of cost-effective genome-wide sequencing, it has become easier to study genetic patterns across the genome in many individuals simultaneously. Statistical analyses of large datasets can be used to reconstruct key events in human evolutionary history, which are discernable from the distribution of allele frequencies across global populations.^{10, 11, 12} These events include the widely accepted out-of-Africa dispersal, as well as numerous founder effects and population expansions that subsequently occurred as early humans spread throughout the globe.^{13, 14}

Inferring the demographic history of indigenous populations in the Americas has proven difficult due in part to its multi-faceted nature—which likely involved a combination of founder effects, population size changes, and recent admixture.^{5, 8, 15} The short evolutionary timescale of the effects caused by European colonization further complicate the picture since most of the population-level genomic patterns previously identified involve much longer periods of time, spanning thousands instead of hundreds of years.^{16, 17, 18} Many of the statistical methods used to identify these patterns are, accordingly, best suited to identify demographic patterns that emerge over longer periods of evolutionary time. Hence, while early Native American migrations have been explored,^{5, 19} as have the admixture effects of European colonization,^{8, 20, 21} these recent admixture events have not yet been studied with the aid of comprehensive data concerning patterns of ancient genomic diversity in these populations.

In this study, we compare the genomic patterns, offered by previously published whole-exome sequence data,²² of an ancient indigenous population from the Americas (i.e., before any effects of European contact) with the genomic patterns of their modern descendants, the Coast Tsimshian (henceforth, “Tsimshian”). The Tsimshian have occupied Prince Rupert Harbour, British Columbia, since at least 6,000 years before present (BP), as attested to by oral traditions, archeological context, and genetic evidence.^{22, 23, 24} Similar to other indigenous groups of the Americas, the Tsimshian suffered dramatic population declines in response to the effects of European contact, with a culmination of smallpox epidemics in the 1800s.²⁵ The population collapse likely occurred roughly 175 years ago, with a 57% reduction in effective population size.²² After the epidemics, the Tsimshian intermarried with non-Natives, likely individuals of European descent.^{8, 22}

Given the severity of the Tsimshian population collapse, the expectation would be to find a reduction in overall fitness between the modern and ancestral groups, with some evidence of this at both the individual and population levels.²⁶ This expectation is due to the likely effects of population collapses, which lead to decreases in expected heterozygosity due to reductions in effective population size.^{27, 28} Through our comparison, however, we identify changes in genomic patterns that have resulted from multiple demographic processes and paint a more nuanced picture, which include the effects of admixture^{8, 22} (Figure 1). Our study offers an enriched understanding of the genomic impact of the specific demographic factors experienced by the indigenous populations of the Americas.

Figure 1.

Admixture Signal between the Modern Tsimshian and Europeans

TreeMix⁶⁰ graph with a single admixture event. Admixture³⁹ bar plots are shown to the right of each population, assuming K = 8 clusters. In the Prince Rupert Harbour (PRH) Ancients, the oldest individual (939) is depicted by the leftmost bar in the Admixture bar plots and shows a different ancestry pattern than the younger individuals.

Material and Methods

Ethics and Community Engagement

The IRB for this study was approved by the University of Illinois and the protocol was reviewed by the Metlakatla and Lax Kw’alaams tribal governments. Before beginning paleogenomic studies, the partners outlined expectations of the research team and First Nations community. Part of the research team visits the community almost annually to communicate with research participants, Elders, and First Nations government representatives. During these visits, researchers and community members review research goals and discuss the latest results and language to be used in presentations, manuscripts, and press releases. First Nations members also participated in a 2011 workshop, now called the Summer Internship for Indigenous Peoples in Genomics (SING), to learn about the uses and limitations of genomics as well as the ethical, legal, and social implications of genomic data.

Sequence Data

The exome data presented here, from both ancient and modern individuals, were generated and aligned as described in Lindo et al.²² Sample 938²⁴ was extracted and library prepped at the ancient lab facilities at the University of Illinois at Urbana-Champaign (UIUC) following the protocols described in Lindo et al.²² The library underwent single-end shotgun sequencing on two lanes of the Illumina HiSeq 2500 at the Genomics Core of UIUC. After alignment to hg19 using the pipeline described in Lindo et al.,²² the exome coverage for the sample was 1.1. For the heterozygosity analysis, the BAM file was filtered for the exome utilizing the intersect option in BEDTools²⁹ and the bed file corresponding to the Illumina TruSeq exome capture regions. The ancient individual from Alaska, Shuká Káa, who underwent shotgun sequencing,³⁰ was also filtered in the same manner for the heterozygosity analysis.

Heterozygosity

For the estimated heterozygosity through time analysis, we employed high-quality calls at variant and non-variant sites from the exome data released by Lindo et al.²² For these individuals, we considered only sites where C/T or G/A polymorphisms were not observed, to help guard against any damaging transitions. We included this same filter for both modern and ancient individuals, so as to not bias the removal of some regions in the ancient but not the modern individuals. This filter is particularly important, as for this analysis we are analyzing variation at an individual level, unlike the remainder of the analyses in the article. After filtering, observed heterozygosity of an individual was calculated as the fraction of sites that are heterozygous in the individual. Assuming that an individual’s parents are unrelated, this observed heterozygosity provides us with an estimate of expected heterozygosity H in the population.

We augmented our dataset with genome sequences from an additional individual (938) from Lucy Island²⁴ and Shuká Káa³⁰ to expand our sampling into more ancient time frames. To better align these samples with the data of Lindo et al.,²² we considered only the exome region. Furthermore, as these are low-coverage samples, we computed expected heterozygosity H with ANGSD³¹ to account for the uncertainty of genotype calling. We considered only reads with a minimum map quality of 30 and a minimum nucleotide quality of 20. We also adjusted the mapping quality for excessive mismatches by using the −C 50 command. We additionally trimmed the first five and last five bases from each read, as these are locations in which damaging transitions tend to be most prevalent. We further used the –trans option to remove all transitions, to again help guard against post-mortem deamination. We also accounted for European admixture (see Masking) in the modern individuals as an additional analysis, and we computed heterozygosity across non-masked sites. Moreover, to align the analysis of the 48 Tsimshian and ancient individuals with the low-coverage 938 and Shuká Káa samples, we performed a separate analysis by employing a uniform pipeline across all 50 samples using the identical ANGSD pipeline that was applied to 938 and Shuká Káa to account for uncertainty in genotype calling. In all three of these analyses, we observed no significant effect of heterozygosity estimates as a function of sequencing depth at sites in the exome (Pearson correlation with p = 0.13, p ≈ 0.08, and p = 0.21, respectively).

SNP Calling

For all other analyses, which required estimates of low-frequency alleles, the final variants and genotypes, consisting of 62 Mb of captured nucleotides, were called using the ATLAS³² software suite. ATLAS is a useful method for ancient DNA analysis that allows for the incorporation of base quality recalibration and genotype calling that takes into account the specific deamination rates of each ancient individual. This method allowed for low-frequency alleles in the ancient population to be assessed without the need to remove transitions in order to safeguard against variants that may actually be due to DNA damage.³³ The validity of this method was exhibited by the transition/tranversion (Ts/Tv) ratio of the ancient population. With respect to the ancient population, ANGSD³¹ estimated allele frequencies with a ratio of 11.9 and samtools³⁴ mpileup yielded 14.3. Most of the excess transitions were found below a frequency of 0.1 in both sets. However, genotypes called with ATLAS yielded a ratio of 2.4, which is more in line with expectations for the exome and the modern population. To prevent a batch effect with our temporal comparisons, the modern group underwent the same treatment with ATLAS.

The ancient individuals underwent the following treatment with ATLAS. First, damage patterns were estimated with the estimatePMD command, followed by a quality score recalibration based on these estimates (commands: recal for males and BSQR for females). The callMLE feature was used to call genotypes, which is a maximum likelihood approach to determine genotype likelihoods that takes into account post-mortem damage (files generated with estimatePMD) to call the most likely genotype. The resulting genotypes were further filtered for a minimum mapping and genotype quality of 30, minimum nucleotide quality of 30, sites under Hardy-Weinberg equilibrium (10⁻⁶ p value), minimum read depth of 6, a minimum of 10 individuals at each site, and singletons removed. The same method was used for the modern group, except for the deamination estimation step. Furthermore, we do not see C→T and G→A transitions (which are attributable to ancient DNA damage) driving the difference in deleterious categorization between the two groups (Figure 2).

Figure 2.

Predictive Damaging CADD Scores Harbored in C→T and G→A Transitions

Results indicate that transitions due to potential post-mortem deamination cannot account for the differences in CADD scores.

For the low-frequency allele analyses, 24 unrelated individuals (determined via PLINK³⁵ with –rel-cutofff of 0.1875) were used from the modern group and 24 unrelated individuals were used for ancient group.

Masking

The 24 modern Tsimshian individuals underwent two sets of analyses, one masked for European ancestry and the other without. Admixture was determined to be from a European origin in two previous studies.^{8, 22} To perform the masking, sites were first restricted to the coding sites and the dataset was merged with 20 unadmixed Peruvians (i.e., as chosen in Lindo et al.²²), 20 CEU, and 20 CHB from the 1000 Genomes Project.³⁶ We then phased the data using Shapeit2³⁷ and input the resolved haplotypes into RFMix³⁸ to estimate local ancestry in the admixed individuals. Non-native tracts identified by RFMix were then masked in each individual by setting the genotypes to missing. We confirmed that no non-native ancestry signal remains in the indigenous individuals by running ADMIXTURE³⁹ on the masked dataset.

Deleterious SNP Annotation

To determine coding variants, CADD scaled scores, and novel alleles (via 1000 Genomes filter), the annotation suite ANNOVAR⁴⁰ was utilized, using the variant files generated by ATLAS. All population genetic measures were calculated with VCFtools,⁴¹ using the variant files generated by ATLAS.

Results

Through the analysis of whole-exome data (Table 1), we set out to directly investigate how the patterns of genetic variation in a Native American population have been altered before and after European contact. Using previously published data²² from the Tsimshian of Prince Rupert Harbour (British Columbia, Canada), we detected 59,912 high-confidence single-nucleotide polymorphisms (SNPs) from 24 modern individuals (post-European contact) and 62,906 from 24 of the Prince Rupert Harbour and Lucy Island ancient individuals (henceforth, “ancients”) (pre-European contact), derived from 62 megabases (Mb) of targeted regions.

Table 1.

Pre-European Contact Ancient Samples

Sample	Location	Method	Coding Read Depth	Analyses
125	PRH	exome	4	all
158	PRH	exome	10	all
163	PRH	exome	5.6	all
167	PRH	exome	4.8	all
168	PRH	exome	28	all
181	PRH	exome	27	all
300	PRH	exome	7.3	all
302	PRH	exome	90	all
311	PRH	exome	7.4	all
318	PRH	exome	6.3	all
322	PRH	exome	6.8	all
357	PRH	exome	27	all
365	PRH	exome	32	all
386	PRH	exome	3.5	all
406	PRH	exome	5.2	all
413	PRH	exome	17.7	all
443	PRH	exome	76	all
468	PRH	exome	38.9	all
470	PRH	exome	33.6	all
507	PRH	exome	7	all
516	PRH	exome	15	all
525	PRH	exome	5.7	all
532	PRH	exome	10.4	all
939	Lucy	exome	8.5	all
938a	Lucy	shotgun	1.1b	heterozygosity
Shuká Káaa	Alaska	shotgun	2.7b	heterozygosity

Abbreviations: PRH, Prince Rupert Harbour, British Columbia; Lucy, Lucy Island, several miles off of the coast of Prince Rupert Harbour. Shuká Káa was found on Prince Edward Island, Alaska.

^aGenotypes not called due to low coverage and samples used only for heterozygosity analysis. Samples range in age from 10,000 to 1,500 years BP.

^bCoverage calculated over the exome.

Compared with the modern individuals, the ancient group exhibits higher levels of mean observed heterozygosity within coding regions (mean heterozygosity across modern 1.230 × 10⁻⁴ versus ancient 4.935 × 10⁻⁴ individuals) (Table 2). This observation is consistent with expectations for a population that has experienced a recent and dramatic population collapse. We also used heterozygosity (see Material and Methods) as a proxy to estimate changes in effective population size through time (Figure 3). This particular analysis includes the 10,300-year-old Shuká Káa individual from Prince of Wales Island, Alaska, who was found to have a close genetic affinity to the Tsimshian.³⁰ We also included a 5,670-year-old individual, 938, from Lucy Island, off the coast of Prince Rupert Harbour, who was also found to have a close genetic affinity to the Tsimshian.²⁴ Although both samples underwent shotgun sequencing, only regions overlapping with the exome were utilized for consistency with the exome capture data from the 48 Tsimshian and ancient individuals (Table 1). We observe a significant correlation (Figure 3) between increasing heterozygosity and increasing time BP (Pearson correlation p < 9.95 × 10⁻¹¹, with p value averaged across all possible samplings of one modern Tsimshian individual). This trend may reflect the population collapse after European contact and not a general trend of effective population decline before contact, which has been observed in studies of indigenous populations of the Americas utilizing mitochondrial DNA.^{2, 42} Furthermore, when the oldest individual, Shuká Káa, is removed, significance is also maintained (Pearson correlation p < 1.9 × 10⁻⁶, with p value averaged across all possible samplings of one modern Tsimshian individual). We also masked the modern population for European ancestry but still found a significant increasing trend in heterozygosity (Pearson correlation with p < 1.21 × 10⁻⁹ including and p < 1.51 × 10⁻⁵ excluding Shuká Káa, and with p value averaged across all possible samplings of one modern Tsimshian individual). Moreover, rather than calling genotypes, we also considered accounting for uncertainty in genotype calling for the 48 Tsimshian and ancient samples, using the identical pipeline as used for 938 and Shuká Káa. The analyses still maintained a significant increasing trend in heterozygosity (Pearson correlation with p < 2.1 × 10⁻⁸ including and p < 5.61 × 10⁻³ excluding Shuká Káa, and with p value averaged across all possible samplings of one modern Tsimshian individual). There is also variation in heterozygosity across the ancient groups within archaeological sites, which may suggest local demographic factors at play. The modern individuals similarly show variation in heterozygosity (in both masked and non-masked analyses), which may correlate to varying levels of gene flow from other indigenous and non-native populations.⁸

Table 2.

Genetic Measures for the 62 Mb Targeted Regions

	Individuals	Total SNPs	Coding SNPs	Mean Depth (coding only)	Mean Heterozygosity	Tajima’s D	Ts/Tv
Modern	24	59,912	40,311	18.72	1.230 × 10−4	0.471	2.18
Ancient	24	62,906	49,630	18.65	4.935 × 10−4	−0.216	2.42

Depth was calculated per individual, using coding sites, and then averaged across individuals. Heterozygosity was measured per individual at all high-confidence sites, and then averaged across individuals. Tajima’s D was measured on an average across 10 kb windows, with a minimum of 5 segregating sites in each window. Ts/Tv designates the transition/transversion ratio within coding regions; a ratio above 2 is expected.⁶¹

Figure 3.

Estimated Heterozygosity through Time

The ancient individuals are color-coded according to their burial sites. Shuká Káa is an ancient individual from present-day Alaska, which dates to approximately 10,300 calendar years BP.³⁰ The plot also includes individual 938 (Lucy site, dating to approximately 5,670 calendar years BP), which is described in Cui et al.²⁴

To maximize the data available within the ancient population for the analyses described below, while safeguarding against DNA damage, which could artificially increase the proportion of low-frequency variants in the ancient individuals, we employed a genotype calling method that specifically considers deamination patterns when calling bases (see Material and Methods). In addition to various filters described in the methods, we also removed singletons to prevent any further biases introduced from deamination in the ancient samples, as this phenomenon is random and likely segregates at very low frequencies at any given site. The same filters and calling method were applied to the modern group to prevent batch effect differences. Furthermore, we included only the 48 exome capture samples in the preceding analyses for consistency (Table 1).

Next, we examined the derived site frequency spectrum (SFS; Figure 4). The ancient individuals display a significant increase over the modern in variants with a minor allele frequency (MAF) below 0.05 (z-test, p < 10⁻⁴). This result is likely due to two demographic factors: (1) a population expansion after the founder effect from the initial peopling of the Americas in the ancient group and (2) the population collapse after European contact, resulting in the loss of low-frequency variants in the modern group. Further evidence for this early expansion with later contraction arises from the ancient individuals exhibiting a lower Tajima’s D value than the modern individuals (z-test, p = 1.73 × 10⁻⁶; Table 2), where negative values (ancient) could be indicative of a population expansion after a bottleneck or founder effect and positive values (modern) could be indicative of a sudden population collapse or development of population structure.⁴³ We also observe an increase in intermediate frequencies in the modern group over the ancient group, which is likely the effect of low-frequency alleles becoming less abundant than alleles at intermediate frequencies shortly after a population collapse.⁴⁴

Figure 4.

Derived Allele Frequency Spectra

Histograms depict the synonymous (A) and nonsynonymous (B) derived allele frequency spectra of the ancient and modern individuals.

Of the combined SNPs, only a small fraction of functional SNPs is shared between the ancient and modern groups (17.4%), which are enriched in the ancient group for derived alleles with frequencies of less than 0.1 (73% of total potentially functional variants). Similar fractions were identified in a French-Canadian population using an analogous approach.⁴⁵ We also confirmed that the C→T and G→A transitions attributable to ancient DNA damage, which may not have been accounted for by the deamination-based calling method, were not driving the differences between the two groups (see Material and Methods and Figure 2). However, despite seemingly disparate features between the two sampling time frames (i.e., before and after European contact), genome-wide F_ST between the ancient and modern samples remains relatively low (F_ST = 0.022).

Because the differences in low-frequency functional SNPs between the two groups are considerable, we discuss the effects of these variants. We do this, first, by testing for differences in the ratio of nonsynonymous to synonymous changes as a function of minor allele frequency (Figure 5). The nonsynonymous to synonymous ratio of 1.57 in the ancient individuals, for SNPs with a frequency below 0.05, points to a major fraction of potentially deleterious SNPs. The same ratio in the modern is 1.6, which is an increase but not a significant one (p > 0.1, chi-square test). These ratios are higher than those seen in non-Native American populations^{12, 46} but similar trends are also seen in other studies of deleterious genomic features, such as runs of homozygosity.^{47, 48} Significant differences between the two groups are found between frequencies of 0.05 and 0.15, where the ancient individuals display a higher ratio compared to the modern individuals (p < 1 × 10⁻⁵, chi-square test) (Figure 5A). For the most common variants (above a frequency of 0.25), the two groups again show a non-significant difference between ratios, 1.1 for the ancient and 1.06 for the modern (p > 0.09, chi-square test).

Figure 5.

Excess of Potentially Functional Variants in the Ancient Individuals

(A) Ratio of nonsynonymous to synonymous changes in the ancient and modern individuals for variants grouped by minor allele frequency.

(B) Mean scaled CADD score of the functional changes for each frequency class in the ancient and modern individuals.

We also tested for the effects of admixture in the modern group by masking European ancestry. The estimated European admixture fraction in the modern group is approximately 30%.^{8, 22} With the masking, the difference in the ratio for low-frequency alleles becomes significant (p < 6 × 10⁻³, chi-square test), with the modern showing a marked increase over the ancient at 1.72 versus 1.57, respectively. Three other frequency bins also show increases of the masked modern over the ancient, with variants between frequencies 0.30 and 0.35 showing a significant difference (p < 0.01, chi-square test).

Second, because previous studies predict that low-frequency nonsynonymous variants tend to be deleterious,^{11, 14, 49} we consider whether this difference in nonsynonymous variants correlate to a shift in the burden of potentially damaging alleles. To assess this possibility, we examine the predicted effects of nonsynonymous variants using the Combined Annotation Dependent Depletion (CADD) scaled scores.⁵⁰ CADD is a powerful method because it integrates various forms of information, which include conservation, allelic diversity, pathogenicity, and experimentally measured regulatory effects. Scaled scores range from 1 to 99 and are based on the rank of each variant compared to the 8.6 billion mutational positions in the human reference genome (hg19). Scores higher than 10 represent the top 10%, higher than 20 the top 1%, and higher than 30 the top 0.1%, all with respect to potential deleteriousness. Using this combined measure, the ancient individuals demonstrate evidence for an excess of potentially damaging mutations below a frequency of 0.2, when compared with the modern individuals (Figure 5B). This trend is affected by masking European admixture in the modern group, where the masked individuals exhibit higher means than the unmasked and overtake the ancient group on several frequency bins above 0.15 (Figure 5B). Examining only functional sites, the ancient individuals exhibit a larger proportion of scores above 20 than the modern individuals (Figure 6). However, when the modern group is masked for European ancestry, the modern group exhibits a larger proportion of scores above 20, which approaches that of the ancient group. This may indicate that admixture has contributed alleles to the population that are potentially less damaging.

Figure 6.

CADD Classification of Functional SNPs in Each Population

The modern individuals show a reduced proportion of possibly damaging sites (CADD > 20), when compared to the ancient. However, a marked increase occurs in possibly damaging sites when the modern individuals are masked for European ancestry.

Third, we examine the number of novel variants in both the ancient and modern individuals. We define novel variants as those that are not found in the populations represented in the 1000 Genomes Project phase 3 release.³⁶ The ancient individuals show an excess of novel alleles when compared with the modern individuals (Figure 7A). This enrichment is likely due in part to the overall loss of genetic variation in the modern individuals caused by the population collapse associated with European colonization, as well as subsequent admixture with non-indigenous populations. The modern individuals exhibit, however, a significant decrease in the portion of the most likely damaging sites within these novel alleles (above the top 0.1%), when compared with the ancient individuals (Figure 7B; modern fraction 0.091, ancient fraction 0.103, z-test p < 9 × 10⁻³). Previous studies have found that rare and novel alleles tend to be of a deleterious nature^{11, 51, 52} and an expansion following bottlenecks or founder effects may create a proportionally greater number of these variants in a population. However, these time frames are on the order of thousands of years and it has been less than 150 (only several generations) since the modern group’s population collapse from the effects of European contact.⁵³

Figure 7.

Novel Potentially Functional Variants

Novel variants are defined as variants not found in populations represented in the 1000 Genomes Project phase 3 release.³⁶

(A) The modern group exhibits reduced levels of novel nonsynonymous alleles with respect to the ancient.

(B) The novel nonsynonymous variants are measured via CADD prediction scores of possibly damaging sites, where scores greater than 30 reflect potential deleterious changes that rank above the top 0.1% (red line). The ancient individuals show a larger distribution of potentially damaging alleles when compared to the modern.

Discussion

The impact of European colonization has altered the genomes of Native Americans in multiple and dynamic ways. The data discussed in this study suggest that within approximately nine generations since the time of European contact, the modern group have significantly fewer low-frequency and potentially damaging alleles than their ancient ancestors. The differences between the sampling periods can be partially explained by the population expansion that increased the number of low-frequency alleles in the ancient individuals, following the initial peopling of the Americas.⁵⁴ Although the genomic signatures of a human population expansion following a founder effect have been explored in other populations,^{11, 12, 55} the genomic consequences of some more recent and severe demographic events seems to carry additional impacts.

The modern individuals have experienced a relatively recent and severe population decline after European contact. This may partially explain the loss of low-frequency and novel alleles when contrasted with the ancient individuals and, in turn, the decrease in the number of potentially deleterious alleles. The effects of such a recent collapse, coupled with slow recovery, is, however, typically expected to amplify certain alleles to higher frequency within a population, due to the distortion of allele frequencies by genetic drift.⁴⁴ But these expectations are not borne out in our data as a result of two primary factors: first, the relatively short evolutionary timescale within which these events occurred; and, second, the recent admixture with both indigenous and non-indigenous populations, which may have increased genetic diversity and countered the deleterious effects of reduced population size.⁵⁶ Even at very low levels, gene flow from an admixture event has been observed to increase genetic diversity stemming from the connectivity between two populations.^{57, 58} The Coast Tsimshian may have established this increased variation through population connectivity, from admixture with both native and non-native groups,^{8, 59} and we believe recent and rapid genomic changes like these need further study in a broader range of cases.

Population collapses can also have a different type of effect in terms of the expected heterozygosity of a population, which is typically expected to decrease due to the associated loss of alleles.⁴⁴ This decrease is substantially influenced by the magnitude of the population collapse. Despite the relatively short amount of time, the modern individuals show variance in the associated heterozygosity, some reaching the level of their ancestors (Figure 3). However, European admixture does not seem to be a factor here, as variation is seen in both masked and non-masked individuals and is in line with the variation seen in the ancient population through time.

It should also be noted that the term “deleterious” to describe the potential consequence of an allele is problematic when discussing a population due to the term’s lack of sensitivity. Here we assess the potential damaging effects of an allele in various ways, including the strength of conservation and the probability that the protein function itself will be altered. These predictions are, however, mainly a tool used by biomedical research to identify a putative disease-causing variant in a population context, which are then explored further via other approaches. Given the uncertainty that any allele marked as “deleterious” has a disease outcome, we utilized the most extreme predictions (i.e., those ranked above the 99^th percentile) to examine changes in genomic patterns between two time frames. We do not conclude that ancient Native Americans harbored large reservoirs of deleterious alleles but instead demonstrate a fluctuation of a particular class of alleles through time. The role of novel variants in a population is largely environment dependent and a negative categorization, especially related to the ancient past, is misleading.

Native American evolutionary history is complex and involves a maelstrom of demographic processes. Some of these complexities are reflected in the genomic patterns observed here in a single Native American population, before and after European contact. Our study suggests that while the ancient individuals exhibit the trademark genomic patterns of a rapid population expansion following a founder effect, their modern descendants have more nuanced genetic patterns. Although the effects of the recent population collapse, associated with European contact, seem to have removed a large portion of low-frequency alleles from the population, this reduction is not accompanied by the same expected increase of potentially deleterious genomic features. For reasons discussed, this pattern may represent the ameliorating effects of allele introgression caused by admixture. We also find a trend of population decline in the ancient group before contact, which could be related to substructure after populations were established regionally in North America. It is also possible that instead of a steady expansion after the entry into the Americas, population size varied by region and was potentially linked to the many cultures and environments of the Americas. These factors could have contributed to the steady population decline in this particular region before European contact.

In conclusion, with the use of ancient DNA, we uncovered the genomic patterns of a population representative of the indigenous peoples of the Americas and show how the effects of gene flow have reshaped these patterns in subtle ways. As human populations continue to intermingle through the effects of globalization, this study highlights the impacts of increased gene flow on the genomic patterns between populations with both similar and divergent evolutionary histories.

Published: April 26, 2018

Accession Numbers

The ancient data and sample 938 are available from NCBI Sequence Read Archive, accession number PRJNA288803. The data from modern individuals are available via a data access agreement with R.S.M. at the University of Illinois. All other data are available from the authors on reasonable request.

Web Resources

• Sequence Read Archive (SRA), http://www.ncbi.nlm.nih.gov/sra