Genome-wide evidence for the local ancestry of the Agotes from Navarra, Spain

Summary

The Agotes are among the most enigmatic minority social groups in the Iberian Peninsula, historically subjected to long-standing social marginalization by their non-Agote neighbors. Nowadays, all Agotes live in the small town of Bozate (Baztan valley, Navarra, Spain), but their origins remain disputed, with different hypotheses. In this study, we analyze genome-wide genotype data from 125 newly sampled individuals from Bozate, the Baztan valley, and Navarra. Our findings indicate that the Agotes have a local genetic origin and exhibit high levels of homozygosity. Their isolation is likely due to socio-cultural reasons. Additionally, we find fine-scale genetic substructure within the broader Basque-like landscape, detecting the Agote and Baztan components as part of these microstructures, due to social or geographic isolation, respectively. Overall, our findings challenge the long-standing historical perception of the Agotes as a foreign group and suggest a shared genetic ancestry with the local populations.

Graphical abstract

Highlights

• •Agotes have local origins and high levels of homozygosity due to marginalization
• •Their origins could be related to Cathar theory
• •Agotes shape a genetic substructure within the Basques driven by social isolation
• •Baztan population also substructures within Basques due to geographical barriers

Genomics; Human genetics; Evolutionary biology; Human geography

Introduction

The genetics of the Iberian Peninsula have been characterized in both ancient and present-day populations, providing insights into their overall genetic diversity and complex historical heterogeneity within western Europe.^1,2 Within this broad context, several groups stand out for their genetic isolation, such as the Basques³ and the Spanish Roma.⁴ However, the Iberian Peninsula hosts other singular and lesser-known groups. Some of them have already been genetically analyzed, including the Chuetas,⁵ the Maragatos,⁶ the Pasiegos⁷ and the Vaqueiros d’Alzada.⁸ Nonetheless, many other groups remain understudied from a genomic perspective.

The Agotes are a minority social group currently residing in the neighborhood of Bozate, located in the Baztan valley (Municipality of Arizkun, Navarra, Spain), a Basque-speaking area, where Basque culture is predominant (Figure 1). They stand out due to the lack of information available about their ancestral origins. Moreover, the history of this group has been profoundly marked by persistent social and religious marginalization imposed by neighboring populations.

Figure 1.

Geographical location of Bozate

The Bozate neighborhood (red diamond) in the Baztan valley in Navarra (Spain) is the only population nucleus where the Agote community or their descendants are still present. Blue-colored areas correspond to locations where genetic samples were obtained for this study (sample sizes are shown in brackets), while the gray areas correspond to collection points for reference genetic samples corresponding to the nearest geographical areas to the Bozate neighborhood.

The precise origin of the Agotes is still a matter of debate. Some authors attribute the first reference to them in the cartulary of the abbey of Luc-de-Bearn (France) between CE 950 and 1000,⁹ while others claim that their first unequivocal mention is in 1413.^10,11 Historically, their geographic distribution spanned both sides of the western and central Pyrenees. On the French side, they inhabited the provinces of Béarn, Gascogne, Guienne, Comté de Foix, and the southwestern region of Languedoc-Roussillon. On the Spanish side, their settlements were limited to the provinces of Guipuzcoa, Navarra, Aragón, and western Catalonia.¹² Over time, their territory contracted to mainly Bozate, where some family groups still live today.

The Agotes have been historically mistreated and ostracized facing social and religious exclusion persisting until the mid-20th century. In the past, Agotes were required to wear a distinctive badge on their clothes and ring a bell as they passed, alerting non-Agote people to avoid them. They were also prohibited from social celebrations and had to abstain from certain jobs, particularly those involving direct contact with food, such as agriculture and livestock, limiting their work to activities such as woodworking and stonework. Religious restrictions included their banning from entering churches through the main door, were banned from marrying non-Agote individuals and had specially designated sections in cemeteries. Superstitious beliefs regarding the Agotes by the non-Agote individuals have been documented. The reasons behind this discriminatory behavior remain unknown and are particularly baffling given that, there are no physical or cultural traits that differentiate them from non-Agote populations.¹³

Numerous hypotheses have been proposed regarding the origins of the Agotes, ranging from mythological to historical events based accounts. Some are rooted in legends, such as the Jewish origin hypothesis, which proposes that they are descendants of the builders of Solomon’s Temple (10th century BCE), expelled from Jerusalem for alleged inefficiency as workers.^14,15 Other hypotheses are based on specific historical events. According to the Saracen origin hypothesis, the Agotes would be North African soldiers who dispersed along the French-Spanish border after the Muslim defeat at the battle of Poitiers in 732.¹⁶ The Viking origin hypothesis references documents attesting to Viking explorers reaching the Cantabrian coast and Euskadi in the mid-9th century,¹⁷ with settlements in the Mundaka estuary influencing the cultural and demographic landscape of the Basque Country and Navarra.^18,19,20 According to the Gothic origin hypothesis, the Agotes are descendants of the Goths, a Germanic or northern European political elite who ruled over parts of France and the Iberian Peninsula during the 5th and 6th centuries. The Goths practiced Arianism and opposed the Catholicism of the time.^9,21,22,23 Under this hypothesis, the combination of a distant ancestry and heretical religious practices might explain the historical rejection of the Agotes. In fact, the term “Agote” (Cagots in France) is believed to derive from the Latin derogatory term cani Godi (“Gothic dog”). Finally, the Cathar origin hypothesis suggests that the Agotes were Cathar heretics fleeing southern France during the Albigensian Crusade in the 13th century.²⁴ Some scholars also propose an origin linked to leprosy, considering the spread of the disease throughout Europe during the Middle Ages and its perception as a marker of social and religious exclusion.²⁵

The Agotes have been described and characterized by many French and Spanish ethnographers.^{9,12,13,26,27,28,29,30,31,32} To date, however, only one study³³ has sought to explore their ancestry through anthropometric and seroanthropological observations and no genetic based study has been reported.

Given the lack of genetic studies on the geographical ancestry of the Agotes, we obtained whole genome genotype data from DNA samples from 125 individuals belonging to Bozate and surrounding populations, including Agotes, Agotes of mixed ancestry, residents of the Baztan valley and other regions of Navarra, to generate high-throughput genotype datasets, including uniparental markers. Our goals were to investigate the genetic ancestry, population history, and genetic diversity of the Agotes. To do so, we first integrated our genotyped data with data from other geographically and historically relevant reference populations, representing hypothesized ancestral components. Then, we (i) provided a comprehensive genetic characterization of the Agotes in the broader genetic landscape of the Iberian Peninsula and globally; (ii) investigated their genetic isolation; (iii) explored their population history and origins; and (iv) contextualized the population at a microgeographic level.

Our findings demonstrate that, genetically, the Agotes are most closely related to other local populations, effectively ruling out hypotheses of a distant or remote ancestry. Our findings suggest that the most plausible hypothesis is related to Catharism. We also observed a high degree of genetic consanguinity, consistent with a history of social exclusion. Additionally, our analyses reveal a signal of population substructure within the Basque genetic landscape, with the Agotes and other inhabitants of the Baztan valley shaping some of these microstructures. Importantly, this work aims to increase the visibility of the Agotes as a historically marginalized group and contribute to a better understanding of the high genetic diversity within the Iberian Peninsula and western Europe.

Results

Overview of the analysis

Using genome-wide genotype data from the Agotes and neighboring populations near Bozate (Figure 1; Table S1), along with other reference populations that could represent potential sources of the Agote population (Table S1), we first performed quality control and kinship relatedness analyses.

Then, we analyzed the genetic structure and global ancestry using methods assuming independence between genetic markers. Specifically, principal component analysis (PCA) and ADMIXTURE analyses were conducted at two complementary levels to address different objectives of the study. On the one hand, to assess the genetic origin of the Agotes, we used the hypotheses dataset (Table S1), which includes populations that represent the potential ancestry profiles according to different origin hypotheses. On the other hand, aiming at investigating fine-scale genetic structure, we employed the microphotographic (Table S1) and the Microgeographic_WO_PIR datasets, the latter being restricted to populations from the immediate geographic surroundings of Bozate without the Pyrenean one, specifically: Baztan, Navarra, Spanish and French Basques, southern French populations, including southwestern groups (Bearn, Bigorre, Chalosse, Landes, and Hautes-Pyrénées) and southeastern groups (Haute-Garonne), and northern Spanish populations (Burgos, Cantabria, northern Aragón, Rioja, and western Bizkaia). To assess genetic affinities and gene flow among Agotes and any population, we performed f₃ and f₄ statistics based on the microphotographic dataset.

The data from the Hypotheses dataset, after removing non-European populations (hereinafter referred to as the Hypotheses_European dataset), were then phased to enable haplotype-based analyses, allowing us to further investigate the origin of the Agotes at a more detailed scale.

Principal component analysis

After confirming with KING³⁴ that detected outliers by smartPCA³⁵ were due to high relatedness for the Agotes and Agotes of mixed ancestry individuals, we reran the PCA projecting these outliers onto the main plots.

PCA results (PC1 vs. PC2) for the Hypotheses dataset, which includes populations from around the globe to test all proposed hypotheses for the Agote ancestry, Figure 2A left reveals three distinct groups along PC1 and PC2, corresponding to European, North African, and Middle Eastern populations. The Agotes and Agotes of mixed ancestry cluster with western European populations. This genetic pattern is also evident when plotting other principal components (Figures S1A and S1B). Given that the Agotes unambiguously clump with other European populations and separate from North African and Middle Eastern groups, we performed a further PCA using Hypotheses_European dataset that includes data from European populations only (Figure 2A, right). This PCA showed a tight cluster of individuals of Spanish ancestry not including Agotes, Agotes of mixed ancestry and individuals from the Baztan valley and Navarra. Instead, these groups were plotted closer to individuals of Spanish and French Basque ancestry, as well as those from the south of France, clearly distinguished from the Spanish populations. Notably, individuals from the Baztan valley are plotted at the distal end of this group of Spanish and French populations, indicating that they are the most genetically differentiated, likely due to genetic drift based on their history of geographic isolation within the Baztan valley. Similar genetic patterns were observed when plotting other principal components (Figures S2A and S2B).

Figure 2.

Global ancestry and population structure of the hypotheses dataset (N = 651; 427,044 SNPs)

(A) PCA results for the two levels of analyses. On the left, we show PCA results using all populations that the Agotes could potentially have originated from (N = 651). On the right, we show plot results for the Hypotheses_European dataset (N = 435). The acronyms from the plots correspond to the following populations: MDE (Middle East), NAF (North Africa), EUR (Europe), AGM (Agotes of mixed ancestry), AGO (Agotes), FIN (Finland), NOR (Norway), CEU (central Europe), FRN (France), SPN (Spain), SFR (south of France), BASF (French Basques), BAS (Spanish Basques), NAV (Navarra), BAZ (Baztan).

(B) ADMIXTURE results for K = 4 based on the Hypotheses dataset (Table S1).

Overall, our PCA results suggest a European ancestry for the Agotes with higher affinity to Spanish and French Basques, consistent with a local and not a geographically remote origin.

Regarding the analysis at a regional level, the PCA results based on the microphotographic dataset with (Figure S3) and without (Figure S4) Middle Eastern and North African genetic samples are consistent with those previously obtained (Figure 2A), confirming the overlap of the Agotes with local populations. The PCA based on a larger number of single nucleotide polymorphisms (SNPs) (Microgeographic_WO_PIR dataset), including only populations surrounding Bozate (Figure S5), reveals a modest genetic differentiation between the Agotes and the southeastern French population, as well as a closer genetic affinity with groups from southwestern France and Basque populations.

Model-based clustering analysis

Using a complementary approach, we used the unsupervised clustering ADMIXTURE³⁶ method based on both the Hypotheses and Hypotheses_European datasets and the microphotographic and Microgeographic_WO_PIR datasets. We estimated proportions for each cluster from K = 2 to K = 10 in all datasets. We used ADMIXTURE projection mode to project onto the main clusters the outliers previously detected using smartPCA. Cross-validation (CV) errors before and after projection are shown in Table S3.

For the Hypotheses dataset ADMIXTURE analysis, focusing on K = 4 (the K with the lowest CV-error value), we can roughly interpret mixture profiles resulting from Middle Eastern, North African, European, and Basque-like ancestry components (Figure 2B). The estimated proportion of Basque-like ancestry (purple) component predominates in populations surrounding and including the Agote and Agotes of mixed ancestry. In particular, this component accounts for the highest proportion in the Baztan valley (on average: 95.2%), French Basques (87.7%), Spanish Basques (76%), individuals from Navarra (73%), Agotes (72.3%), Agotes of mixed ancestry (66%), and individuals from the south of France (58%). The lower estimated proportion of Agotes of mixed ancestry is consistent with the self-reported ancestry of the individuals. The non-Basque European-like (cyan) component is the second most prominent across these populations and shows an inverse relationship with the Basque-like component. Its proportion ranges from highest to lowest as follows: individuals from the south of France (38.8%), Agotes of mixed ancestry (26%), Agotes (22.5%), Spanish Basques (22.5%), individuals from Navarra (22.2%), French Basques (11.5%), and individuals from the Baztan valley (3.9%). Regarding the Middle Eastern-like (orange) component, the highest proportions are observed in the Agotes of mixed ancestry (4.4%) and Agotes (3.4%), followed by individuals from Navarra (2.6%), individuals from the south of France (1.8%), Spanish Basques (1.1%), French Basques (0.8%), and Baztan valley (0.6%). Finally, the North African-like (brown) component is absent in individuals from the Baztan valley, as well as Spanish and French Basques. It is present in low frequencies among Agotes of mixed ancestry (3.6%), followed by individuals from Navarra (2.2%), Agotes (1.8%), and individuals from the south of France (1.3%). Regarding other values of K, results follow a similar pattern (Figure S6). These results suggest a certain level of gene flow from North African populations into specific Basque-related groups, most likely mediated by gene flow between the latter and populations from other parts of Spain.

Within the Hypotheses_European dataset alone (Figure S7), the distinctive genetic signature of the Basque-like populations becomes evident even for small values of K. For K = 2, the Basque-like ancestry (purple) is differentiated from a European-like ancestry (navy blue). For K = 3, a new cluster (cyan) emerges that further subdivides the European populations, distinguishing northern groups from the rest. For K = 4, a Finnish-like (gray) component appears. The additional component observed for K = 5 (dark blue) is predominant in Iberian samples, but does not have a definitive geographic attribution, reflecting an unidentified ancestral component. Finally, for K = 6, a yellow component emerges within the Basque-like substructure, reaching its highest levels in some individuals from the Baztan valley and is also present in high proportions in Agotes and Agotes of mixed ancestry. This component may represent a Baztan-specific ancestry.

Taken together, these findings highlight that the Agote population has a global ancestry profile similar to that of its surrounding populations, albeit with subtle differences when directly compared to Basque populations. The overall genetic profile of the Agotes is characterized by a predominant Basque-like component, followed mainly by the European-like component.

For the ADMIXTURE results based on the microphotographic dataset (Figure S8), a specific component (pink) that reaches its maximum proportion among individuals from the Pyrenees (on average: 38.5%), emerges at K = 8. However, its proportion in Agotes (5.6%) and Agotes of mixed ancestry (3.5%) is similar or lower to that observed in the nearest populations: Basques from France (3.6%), Baztan (5.1%), Basques from Spain (8%), Navarra (10.7%), southwest of France (18%), southeast of France (19.8%), and northern Spain (21.8%).

The analyses to explore population structure for the Microgeographic_WO_PIR dataset (increasing the number of SNPs and only for the geographically closest populations to Bozate) are shown in Figure S9. At K = 2, which corresponded to the lowest CV error, two ancestral components were identified. These components were associated with Basque (purple) and European (pale green) ancestries, with the latter maximizing in southeastern France. Increasing to K = 3 revealed a novel (light lavender) component that may explain a sub-structure within the Basque cluster, as it maximizes in Spanish Basque populations. For K values greater than 4, additional components emerged, reflecting further sub-structure within the Navarra, French and Spanish Basque populations. Finally, it is worth noting that at K = 6, a (green) component associated with the Agotes population becomes apparent, and at K = 9, a Baztan-like (yellow) component emerges.

Population affinity and gene flow: f₃ and f₄ statistics

To determine which extant population is most closely related to the Agotes, we performed outgroup f₃-statistics³⁷ analyses, comparing the Agotes against all groups included in the Microgeographic_WO_PIR dataset. This dataset was selected because it encompasses a broader range of populations, both geographically distant and proximate, while also capturing the maximum number of SNPs. The African hunter-gatherer Mbuti population was used as the outgroup in two distinct testing configurations: f₃(Agotes, X; Mbuti) and f₃(Agotes_Admixed, X; Mbuti), where X denotes each of the populations under comparison. In both tests (Figures S10 and S11), the results were highly consistent, with the Baztan population emerging as the closest relative to both the Agotes (Figure S10) and the Agotes of mixed ancestry group (Figure S11), followed by French Basques, Spanish Basques, Navarra, southwest France, and northern Spain, essentially, the geographically closest populations.

To investigate potential gene flow among populations distributed along the Pyrenean axis, we conducted f₄-statistics³⁷ analyses using the microphotographic dataset. The following configurations were employed: f₄(Agotes, Agotes_Admixed; X, Mbuti) and f₄(Agotes, Baztan; X, Mbuti). They were designed to ensure that the related populations (i.e., A and B) used were as closely related as possible to the Agotes. In the first case, the Agotes of mixed ancestry group was used due to its shared genetic background with the target population. In the second case, the Baztan population was selected, as it was identified as the closest relative to the Agotes based on the results of the f₃-statistics tests. The results obtained from the first configuration (Figure S12) revealed no significant gene flow from any of the populations analyzed. In contrast, the second configuration (Figure S13) identified significant gene flow between the Agotes and the Agotes of mixed ancestry group (Z score >0), as well as significant gene flow between the French Basques and the Baztan population (Z score <0).

Using a larger number of SNPs (Microgeographic_WO_PIR dataset) and the f₄(Agotes, Agotes of mixed ancestry; X, Mbuti) configuration, where X corresponds to either Baztan or Navarra, we do find significant and positive results for both testings (Figure S14), demonstrating gene flow between Agotes and Baztan, as well as between Agotes and Navarra.

Taken together, these findings align with our PCA and ADMIXTURE results supporting a local origin for the Agotes, pointing to the Baztan valley or, more broadly, the surrounding Basque genetic and cultural landscape as the most likely source population.

Haplotype-based clustering methods

Given that above results do not support a genetic affinity between the Agotes and North African or Middle Eastern populations, we further investigated their origin using methods assuming marker independence. Then, we employed haplotype-based approaches restricted to populations potentially representative of the hypotheses based on a European origin (Hypotheses_European dataset). Specifically, we performed haplotype-based clustering methods using the CHROMOPAINTER and fineSTRUCTURE³⁸ softwares. These analyses aimed both to elucidate the haplotype-based relationships among the samples and to define genetically homogeneous clusters for the subsequent haplotype-based analyses.

The results of fineSTRUCTURE are presented as a simplified dendrogram (Figure 3A), in which the leaves represent the inferred genetic clusters (hereinafter referred to as “clusters”). These clusters are organized into clades of varying depths (hereinafter referred to as “major clusters”). The composition of each cluster based on the geographic origin of the individuals is shown in Figure 3B. The full fineSTRUCTURE dendrogram and the associated coancestry matrix are shown in Figure S15 and the detailed composition of each inferred cluster is shown in Table S4.

Figure 3.

Haplotype-based fineSTRUCTURE results based on the Hypotheses_European dataset (N = 435; 427,044 SNPs)

(A) Summary dendrogram obtained by applying fineSTRUCTURE (fS) showing the inferred genetic clusters and the relations among them. From the top to the bottom: PYRfS (“Pyrenees”; N = 23 individuals: 47.8% are from Navarra; 26.1% are Basques from Spain; 17.4% are French from south France and the remaining 8.7% are 2 individuals from Baztan and Bozate); BASfS (“Basque”; N = 47: 48.9% are Basques from France; 31.9% are Navarreses and 19.2% are Basques from Spain); AGOfS (“Agote”; N = 12: 58.3% are from Bozate; 25% from Navarra and 16.7% from Baztan); BAZfS (“Baztan”; N = 25: 76% are from Baztan; 20% are from Navarra and 4% Basques from France); NFRfS (“North France”; N = 17: all of them being from the north of France); CEUfS (“Central Europe”; N = 40: 97.5% are from central Europe, except from one individual, being from Norway); NEUfS (“North Europe”; N = 20: 55% are from Norway and 45% from Finland); UNK (“Unknown”; N = 16 removed individuals with unspecified geographic origin); SPN1fS (“Spain1”; N = 45: 86.7% are from Spain; 6.6% are from Navarra and 6.6% from south of France); CFRfS (“Central France”; N = 33: 63.3% are from central France and 36.4% are from north of France); and SPN2fS (“Spain2”; N = 96: 99% are from Spain and the remaining 1% from central France). The full fineSTRUCTURE dendrogram along with the coancestry matrix was included in Figure S15.

(B) Composition of each fS cluster based on the geographic origin of the individuals (see Table S4 for a more detailed description).

From the eight non-outlier (i.e., unrelated) Agotes (Agotes + Agotes of mixed ancestry) that entered the analysis, seven are included and form the predominant population (58%), in a cluster labeled as “Agote” (AGOfS; N = 12), which also includes individuals from Baztan (N = 2) and Navarra (N = 3). The closest cluster to “Agote”, labeled “Baztan” (BAZfS; N = 25) mainly consists of individuals from the Baztan valley (85%) along with Basques from France and individuals from Navarra. These two clusters further group together with a broader “Basque” cluster (BASfS; N = 47), predominantly composed of French and Spanish Basques (68% in total) and individuals from Navarra. This highlights the hierarchical fine-scale genetic substructure within the larger Basque genetic landscape. Moving higher in the dendrogram, the major “Basque” cluster groups together with a cluster labeled “Pyrenees” (PYRfS; N = 22), primarily including individuals from Navarra (50%), but also individuals from populations from both sides of the western French-Spanish border (aligned with the Pyrenean mountain range) and one Agote of mixed ancestry. The four clusters (AGOfS, BAZfS, BASfS and PYRfS) form a major “Basque-like” cluster, which is distinct from the major “non-Basque-like European” cluster, containing all other European clusters. This separation highlights the genetic differentiation of the Basque-like populations within the broader European context.

The major “non-Basque-like European” cluster encompasses all remaining European individuals, grouped into two major clusters: “Northeast Europe” and “Southwest Europe.” “Northeast Europe” comprises three clusters: “North France” (NFRfS; N = 17), consisting solely of individuals from the North of France: “North Europe” (NEUfS; N = 20), which groups individuals mainly from Norway (55%) and Finland; and “Central Europe” (CEUfS; N = 40), predominantly composed (97.5%) of individuals from various parts of central Europe. “Southwest Europe” also contains three clusters: “Spain1” (SPN1fS; N = 45) and “Spain2” (SPN2fS; N = 96), primarily composed of individuals from Spain (86.7% and 99%, respectively); and “Central France” (CFRfS; N = 33), where individuals are mainly from central France (63.3%). Within this major “Southwest Europe” cluster, a geographically unclassifiable “Unknown” (UNK; N = 16) cluster, consisting of individuals from Spain, central France and Bulgaria, was identified but excluded from downstream haplotype-based analyses due to its lack of clear geographic coherence.

It is important to note that, regardless of the fineSTRUCTURE classification, we retained Agotes and Agotes of mixed ancestry individuals in their original population designations (i.e., Agotes; N = 11 and Agotes of mixed ancestry; N = 57) for the subsequent haplotype-based analyses.

Historical patterns of effective population size

We estimated the effective population size (N_e) over the past 50 generations for the combined Agotes and Agotes of mixed ancestry samples using IBDNe.³⁹

The main results of the demographic inference for the Agotes are shown in Figure 4A, also including geographically proximate groups (BAZfs and BASfs). However, to first provide a general context beforehand, plots of different combinations with other European groups were generated (Figures S16 and S17). When comparing the Agotes N_e curve with those of all European genetic clusters inferred using fineSTRUCTURE (Figure S16), all curves initially share a common starting point 50 generations ago. However, over time, the curves diverge. The European-like clusters maintain a similar pattern, while the Agotes, Baztan valley and Basque ones follow a distinct demographic trajectory, reinforcing their shared Basque origin. Focusing on the clusters within the major “Basque-like” cluster (i.e., BAZfS, BASfS, and PYRfS fineSTRUCTURE clusters) (Figure S17), the N_e curves for the Agotes, Baztan valley, and Basques differ markedly from the Pyrenees cluster. The N_e trajectory of the latter aligns more closely with those for the European clusters. These curves suggest three demographic scenarios: exponential growth among Pyrenean populations around 20 generations ago, in contrast to a relative stability of the Basque and Baztan valley populations for the same period, and a declining trend in the Agotes that began 11 generations ago, consistent with their potential genetic isolation.

Figure 4.

Demographic patterns based on the Hypotheses_European dataset (N = 435; 427,044 SNPs)

(A) Effective population sizes (N_e) over the last 50 generations inferred using IBDNe analysis for three Basque-like groups: Agotes (original Agotes + Agotes of mixed ancestry); BAZfS and BASfS clusters from fineSTRUCTURE. The lines represent the mean N_e values for each population, whereas the shaded areas correspond to the 95% confidence intervals calculated by bootstrap as implemented in IBDNe. The vertical gray line indicates the chronological period for the Albigensian Crusade against the Cathars.⁴⁰ The red dot corresponds to the time of admixture date for the Agotes of mixed ancestry, inferred from the fastGLOBETROTTER analysis (B).

(B) fastGLOBETROTTER results for one-date admixture event for Agotes of mixed ancestry: dates (it is shown the inferred admixture date with its corresponding confidence interval: 95% CI = [3.25–7.03]), major sources and proportions.

Focusing specifically on the major “Basque” cluster (i.e., BAZfS and BASfS fineSTRUCTURE clusters) (Figure 4A), we observe that N_e trajectories for the Agotes follow a markedly different pattern compared to the other two groups, which show more similar demographic histories from 30 to 50 generations ago. The Agotes show an initial population decline around 50 generations ago (N_e = 9,010), reaching a local minimum at about 20 generations ago (N_e = 2,500). This was followed by a period of population growth reaching a local maximum 11 generations ago (N_e = 7,650), before a sharp decline, which has reached its lowest point today (N_e = 18.7). The N_e curves for the Baztan valley and Basque clusters are also characterized by a decline followed by a period of increasing N_e, which is more pronounced in the Basque cluster. However, currently, the trajectories of these two clusters diverge, with the Basque one showing a decreasing trend and the Baztan valley an increasing one.

The above N_e trajectories reveal that, while the Agotes exhibit patterns broadly similar to other regional populations, they experience a sharp and ongoing decline in population size probably due to their genetic isolation, which distinguishes them from their neighbors, reflecting a unique demographic history.

Dating admixture events

To infer admixture proportions and identify the most supported model and timing for the admixture event in the Agotes of mixed ancestry, we treated these individuals as the “target” population, with all inferred clusters from fineSTRUCTURE, as well as the original Agotes individuals (N = 11), as “surrogate” populations in a subsequent analysis using fastGLOBETROTTER.⁴¹ The analysis revealed that the best-supported model involves a single admixture event between two sources (Figure 4B). The estimated date of this event is 6.13 generations ago (95% CI = [3.25–7.03]), which roughly corresponds to the year 1867 (95% CI = [1939–1844]), assuming 25 years per generation. Regarding the two sources involved in this admixture event, the major source, contributing approximately 70% to the profile of the Agotes of mixed ancestry, is predominantly composed of an Agote-like component (59%), with additional contributions from a Baztan-like (27%), and a Pyrenees-like (14%) component. The minor source (dubbed Baztan-like) accounts for roughly 30% and mainly consists of a Baztan-like component (57%), followed by the north France-like (26%), central European-like (12%), and north European-like (5%) components. Considering the 70%–30% source ratio, the resulting admixture proportions for the Agotes of mixed ancestry are approximately 41.3% Agote-like, 36% Baztan-like, 9.8% Pyrenees-like, 7.8% north France-like, 3.3% central Europe-like, and 1.5% northern Europe-like (Table S7).

Historical patterns of identity-by-descent segment sharing

Analyzing identity-by-descent (IBD) segments of specific lengths, we mean pairwise IBD-sharing matrices across three IBD length categories (Figures S18–S20).

The observed pattern indicates that for the shortest segments, there is greater IBD sharing between the Agotes and surrounding populations (BAZfS, BASfS, and PYRfS), and to a lesser extent, with groups from the Iberian Peninsula (SPN1fS and SPN2fS), reflecting shared ancestry among all these groups. As the analysis progresses to longer segments, reflective of more recent temporal processes, it becomes evident that IBD sharing is concentrated among Agotes and individuals of mixed Agote ancestry, pointing to recent consanguinity within the group.

Homozygosity and genetic consanguinity

Based on historical knowledge of their status as an isolated and marginalized small population, and in light of the high levels of consanguinity previously detected using KING,³⁴ we investigated runs of homozygosity (ROH) patterns⁴² based on the Hypotheses_European dataset.

Attending to the genomic inbreeding coefficient (FROH) (Figure 5A), the population exhibiting the highest values of FROH corresponds to the Agotes (0.0425 ± 0.0255, on average), followed by Baztan (0.0299 ± 0.0141, on average). As could be expected, given their history of admixture, the group of Agotes of mixed ancestry presents lower values of FROH (0.0179 ± 0.00550, on average), similar to those of the non-Basque European populations (ranging: 0.0133–0.0173, on average).

Figure 5.

Patterns of ROH estimated using the Hypotheses_European dataset (N = 435; 427,044 SNPs)

(A) Violin plots of individual FROH estimated for each population.

(B) Mean total SROH versus the mean NROH estimated for each population.

To distinguish whether the observed homozygosity stems from historical genetic drift or recent inbreeding, two scatterplots of the total sum of ROH length (SROH) versus the number of ROH (NROH) were generated: one for population mean values (Figure 5B) and another for individual values (Figure S21). Regarding the former, the analysis reveals that the Agote population exhibits the highest levels of both mean SROH and mean NROH, indicating both shared parental ancestry resulting from isolation and consanguinity over multiple generations, as well as recent inbreeding. This pattern is consistently observed when examining individual-level SROH versus NROH values, rather than population means.

Subsequently, we plotted the mean total SROH across different ROH length categories for the nearest populations to Bozate (Figure S22). For the 1–2 and 2–4 centiMorgan (cM) categories, the mean total SROH of the Agotes is not higher than the rest of the other populations, reflecting the genetic drift background common to all Basque and near Basque populations. However, for other categories related to more recent times (i.e., 4–6 cM, 6–10 cM), the Agotes show the highest mean total SROH, according to a recent history of inbreeding, which agrees with the previous results (Figure 5B).

Uniparental marker characterization

The samples were further characterized using uniparental markers: mitochondrial DNA (mtDNA) and Y chromosome (Y-chr), from which population haplogroup frequencies, diversity indices and haplotype networks were obtained. The results for the mtDNA analyses are shown in Table S5.

Frequencies of mtDNA haplogroups (Figure S23; Table S5, spreadsheet: mtDNA haplogroup frequencies) show that the Agotes have the lowest haplogroup diversity among the studied populations, whereas Navarra has the highest. The most common mtDNA haplogroup in all populations is H, reaching the highest frequency in the Agotes (70.60%), followed by Baztan (45.83%), Navarra (30.77%), and others (26.67%). The next most common haplogroups in the Agote population are J (17.65%), T and U (both reaching 5.88%), which are also present in all the remaining populations. Notably, the results obtained for others cannot be interpreted as a population (see STAR Methods section), being any of their maternal ancestors from outside Bozate, Baztan, or Navarra. Here, we find haplogroups A and B that are not present in Agotes, Baztan or Navarra. Genetic diversity analyses (Table S5; spreadsheet: diversity indices) show that the highest value of haplotype gene diversity is found in Navarra (H = 0.995) and the lowest in the Agote population (H = 0.860). In the case of the nucleotide diversity (π) and the mean number of pairwise differences (M), the Agotes show the highest value (π = 0.0083, M = 10.132) and Navarra the lowest (π = 0.0073, M = 8.920). As before, although the group classified as others is of interest for understanding population movements in recent generations, it does not provide substantial information when exploring the genetic characterization and diversity of the Agotes and the surrounding populations. Only common haplogroups among populations of study (H, T, and U) have been represented as haplotype networks (see STAR Methods). Notably, it was not possible to construct the haplotype network representation for J, as the diversity within this haplogroup was limited (only two distinct sequences). In the H haplotype network (Figure S25), the nearest node (two changes) to the Agote haplotype is from Baztan and their inferred divergence time is 961.5 (±961.5) years ago (y.a.). Regarding the T haplotype network (Figure S26), the Agote haplotype emerges from an inferred common node shared with a haplotype from Navarra (two changes each one from the common node) and their inferred divergence is 1,923 (±1,923) y.a. Finally, in the U5b network (Figure S27) we find that the Agote haplotype is near a node shaped by individuals from Baztan and Navarra, whose divergence time is 120.19 (±120.19) y.a.

The results of the Y-chr analyses are shown in Table S6. Haplogroup frequencies (Figure S24; Table S6, spreadsheet: Y-chr haplogroup frequencies) revealed a pattern comparable with the mtDNA. The most common haplogroup in the Agotes is R1b (80.00%), followed by I2a1a (20.00%), both being also present in Navarra (R1b: 85.71% and I2a1a: 4.76%) and Baztan (R1b: 71.43% and I2a1a: 28.57%). Y-STR diversity indices (Table S6; spreadsheet: diversity indices) were still highest in Navarra (H = 0.9862) but lowest in Baztan (H = 0.8660). Regarding the mean number of pairwise differences and average gene diversity over loci, the Agotes have the highest values (M = 3.4923; π = 0.4989), while Baztan has the lowest (M = 3.1340; π = 0.4477). Notably, the patterns of diversity indices differ between the two uniparental marker systems, suggesting sex-biased mobility patterns in the region. Finally, haplotype networks were represented only for the common Y-chr haplogroups (R1b and I2a1a) among the three populations. The R1b network (Figure S28) does not show a central reference node from which the Agote haplotypes derive; therefore, no divergence times were calculated. The I2a1a network (Figure S29) shows a central star-like inferred node from which two Agote haplotypes emerge, with three and four changes, respectively. The respective inferred divergence times are 1,260 ± 890.95 y.a. and 630 ± 630 y.a.

Notably, the standard deviation values for the inferred divergence times are wide for the two uniparental markers, likely due to the few mutational changes that occur in their haplotype networks.

Discussion

Bozate, located in the Baztan valley in Navarra, Spain (Figure 1), is the only place where the descendants of the enigmatic population of Agotes can still be found. Their story is special within the context of western Europe, having experienced a long-standing high degree of marginalization of unknown motivation lasting up until the mid-20th century from their non-Agote neighbors.¹³ This study provides the first insights into the origins and demographic history of the Agotes and aims to provide an interpretation in the context of their long-standing documented marginalization.

With regards to the genetic relationship of the Agotes to other geographically and historically relevant groups, as shown in our PCA results using a broad comparative context (Figure 2A, left), the Agotes overlap with individuals from European populations and remain clearly distinct from non-European groups. ADMIXTURE analyses confirm these results (Figures 2B and S6), as they show virtually no North African or Middle Eastern genetic contributions to the Agote population. This result shows that the Jewish and Muslim hypotheses for their origin⁹ have no genetic support. Nonetheless, the process of religious conversion that took place in the Iberian Peninsula should be considered. The fact that the Agotes do not present a genetic profile characteristic of Jewish or Muslim populations, with origins in the Middle East or North Africa, respectively, does not entirely rule out the possibility of conversion to these religious practices, followed by systematic marginalization as a consequence. Even so, this hypothesis remains unlikely. On the one hand, one would expect to find cultural traces of these religions within the Agote community. On the other hand, historical texts explicitly distinguish between Christian, Jewish, Muslim, and Agote groups,¹¹ suggesting that the Agotes were perceived as a separate category rather than as converts from another faith.

In both PCA and ADMIXTURE analyses based on European comparative context (Figure 2A, right; Figure S7), the Agotes seem to have greater genetic affinity with the Basque and near Basque populations (Navarra and south of France). Consequently, the hypotheses that propose a distant European origin (i.e., Viking and Gothic origins) are also not supported. The greater genetic affinity of the Agotes with their surrounding populations, rather than with distant populations, is further supported by the results of the f₃-statistics (Figures S10 and S11). These findings confirm a close genetic affinity between the Agotes and the Basques, particularly with the population from the Baztan valley. The results reveal a strong relationship with neighboring populations, which gradually diminishes with increasing geographic distance from the Bozate settlement.

The dendrogram inferred using fineSTRUCTURE (Figure 3A) further supports a local origin of the Agotes, as it reveals that the Agotes clustered within the broader major “Basque-like” cluster that is separated early from the broader European one. This major Basque-like cluster shows a deep split between individuals from the broader Basque region and those from the western Pyrenees (i.e., individuals of non-Basque origin inhabiting the Basque cultural area). This result is consistent with the idea that Basque culture historically acted as an effective genetic filter in shaping the region’s genetic structure.³ Further down in the major Basque cluster, a distinction emerges between Spanish and French Basques, followed by Baztan valley and Agotes splitting out and clustering together, thus differentiating them from Basques.

The IBD sharing results based on short segments (Figure S18) show greater sharing with surrounding groups, suggesting a shared common ancestry with these populations. This, together with other findings, supports a local origin for the Agotes.

Because the Agotes seem to fit within the genetic landscape of their broader geographic region, it is unlikely that their isolation was driven solely by hostility toward migrants. Instead, their exclusion may have resulted from other cultural factors. While genomic approaches alone face methodological challenges in addressing these cultural questions, combining genetic data with historical evidence can help refine the remaining hypotheses.

Historically, the Agotes were present from the Atlantic coast to the central Pyrenees, inhabiting regions both north and south of the Pyrenean axis (i.e., French vs. Spanish side). Their widespread distribution suggests that the factors driving their exclusion likely operated on a macro-regional scale. If the Agotes once belonged to a larger population and, for reasons still to be determined, became isolated or split off, then both groups would be expected to share a similar demographic history.⁴¹ Comparing the N_e trajectory of the Agotes with that of their closest genetically related population (Baztan valley) (Figure 4A), a shared demographic pattern is shown until the 12th century. From that point onward, the Baztan valley population grew while the Agote population declined. This finding places the most recent bound on the split of the Agotes in the 12th century.

The 12th century was a time of significant instability in the western Pyrenees, characterized by profound political and religious upheaval,⁴³ which may have facilitated the development of isolated communities. This period also witnessed the persecution of various dissident religious groups, most notably the Cathars, a Christian movement that originated in western Europe in the 10th century, condemned as heretical by the Catholic Church.^44,45,46,47 The last Cathars were forced to take refuge in the western Pyrenees, where they were decimated during the last Albigensian Crusade between 1209 and 1244.⁴⁰ Survivors dispersed throughout the western Pyrenees, reaching different regions in the south. Given this history, the Cathars hypothesis meets many criteria to be considered the origin of the Agotes. However, this hypothesis predicts that the Agotes would be genetically related to the populations of the western Pyrenees, the main origin of the converts to Catharism, and not to the Basque populations.

An alternative hypothesis proposes that the Agotes were not refugees from the Albigensian Crusade, but rather local Pyrenean communities that converted to Catharism.^48,49 This could explain their widespread distribution across both sides of the Pyrenees without requiring large-scale migration. The Treaty of Meaux-Paris (1229), which marked the end of the Albigensian Crusade, included provisions for the suppression of Cathar communities across Pyrenean territories, corresponding to regions where the Agotes historically resided (i.e., Navarra, Aragón, and Catalonia).^40,48,49

A final hypothesis associates the Agotes with leprosy, a disease caused by the bacterium Mycobacterium leprae, which was widespread during the Middle Ages.⁵⁰ In the 12th century, leprosaria were established in the western Pyrenees to isolate individuals afflicted with the disease.⁵¹ According to this view, the Agotes may be descendants of these isolated populations, segregated by epidemiological policies aimed at containing disease transmission. However, the lack of archaeological or historical evidence for leprosaria in the Baztan valley or other regions historically inhabited by the Agotes, along with the absence of leprosy traces in 17th-century autopsies of Agote individuals,⁹ suggests that their origin does not stem from this scenario. Nonetheless, this hypothesis cannot be completely dismissed, and further research is required. Selection studies could be employed to identify markers under selective pressure related to leprosy. Additionally, metagenomic analyses could be conducted to detect the presence of M. leprae DNA, providing further evidence to support or refute this hypothesis.

In sum, although the results have made it possible to rule out most hypotheses, they do not provide solid evidence in favor of any single one. Based on the available genetic data, the Cathar hypothesis remains the most plausible, although the possibility of multifactorial causes cannot be entirely excluded.

Social isolation, such as that experienced by the Agotes, can have lasting traceable effects on both the demography and genetic structure of a population, as evidenced by their historical N_e patterns (Figure 4A). The high degree of consanguinity experienced by the Agotes, is likely a consequence of their social isolation, consistent with the historical sources.³³ The Agotes show the highest homozygosity levels among the surrounding populations (Figure 5A), both due to genetic inbreeding background and recent consanguinity (Figure 5B), which is also evident for IBD sharing for longest segments (Figures S19 and S20). This group is followed by individuals from the Baztan valley, which likely reflects geographic isolation.

In recent times the Agotes began to marry outside their group,³³ particularly from the 19th century onwards, as exhibited by the low homozygosity levels found in the Agotes of mixed ancestry (Figure 5A). This phenomenon is also detected by our fastGLOBETROTTER analysis (Figure 4B; Table S7). The current gene pool of the Agotes of mixed ancestry comprises approximately 70% ancestry specific to the group, with the remaining 30% primarily deriving from the Baztan valley and, to a lesser extent, neighboring populations. Gene flow likely occurred gradually over the course of the 19th century rather than as a single event. Although gene flow is expected to increase N_e, this pattern has not yet been observed in the case of the Agotes (Figure 4A). Their recent decline in N_e seems to match the small census size of Bozate (N = 93 according to the 2020 national census) and the enduring social isolation. However, it is possible that N_e may increase in the future.

As indicated by the results of the f₄-statistics (Figure S13), this recent opening of the population has been confirmed, with gene flow between Agotes and individuals of Agotes of mixed ancestry, as would be expected. In contrast, with this analysis no gene flow was detected between the Agotes and any other neighboring population (Figures S12 and S13), likely due to the low number of SNPs used. To improve this analysis, we repeated it using a dataset with more SNPs, this time finding gene flow between the Agotes and Baztan (Figure S14), which is consistent with the results obtained with fastGLOBETROTTER. Notably, despite the history of isolation of the population, these findings show that any human population is potentially permeable to gene flow.

The study of the Agote population has also provided valuable insights into the genetic structure of the broader Basque population.

The fineSTRUCTURE results (Figure 3A) further support this complexity. The major “Basque” cluster splits into three closely related groups, with the general Basque group branching out first, and individuals from the Baztan valley and the Agotes clustering together, reflecting higher genetic affinity, though there are slight genetic differences between those groups, being also considered different genetic substrates by fastGLOBETROTTER results (Figure 4B).

PCA and ADMIXTURE analyses based on datasets including populations at a regional level provide complementary information. Regarding the PCA results (Figures S3 and S4), the Agotes, along with Baztan and Navarra, continue to fall within the diversity of the surrounding populations. However, they diverge from the Catalan Pyrenean groups, which is consistent with the ADMIXTURE results (Figure S8), where the Pyrenean-like component (pink) for K = 8 is almost absent in the Agotes. After removing this latter population to increase the resolution of inter-population relationships, the PCA (Figure S5), shows greater affinity between the Agotes and populations from southwestern France, as well as with French and Spanish Basques, while appearing more distant from southeastern French populations. It should be noted, however, that the diversity captured in this analysis is relatively limited, and therefore, interpretations should be taken with caution. Based on the results obtained with ADMIXTURE (Figure S9), different microstructures can be observed within the population context. Specifically, Baztan-like and Agote-like components are observed for K = 4 and K = 6, respectively. This result shows the complex population dynamics characteristic of the region.

Moreover, N_e patterns reveal a demographic shift between the broader Basque groups and the Baztan valley population around the 11th century. The political changes in the Kingdom of Navarra between 1004 and 1134, including territorial gains and losses, internal conflicts or dynastic changes,^52,53 may have led to socio-cultural changes that contributed to the varying demographic patterns observed among these populations.

Apart from the autosomal analyses, we explored evidence from the uniparental marker systems, with concordant results. Uniparental haplogroup composition of the Agote population seems to be similar to those from the surrounding regions. The most common mtDNA haplogroups were H, J, T, and U5b (Figure S23; Table S5; spreadsheet: mtDNA haplogroup frequencies), all of them were also found in Baztan, Navarra, Basque populations⁵⁴ and in the broad geographic context of the study population.⁵⁵ Notably, haplogroups of Native American origin (A and B) were identified in certain individuals (Others), consistent with their reported maternal Native American ancestry.⁵⁶ This result reflects the documented migration⁵⁷ of Agote men to the Americas and consecutive mating with local women. Only two Y-chr haplogroups were detected: I2a1a and R1b (Figure S24; Table S6; spreadsheet: Y-chr haplogroup frequencies), both also present in western Europe,⁵⁸ including the Basque Country⁵⁹ and the surrounding populations.⁶⁰ However, frequencies of both mtDNA and Y-chr vary between the Agotes and surrounding populations likely due to their founder effect and population isolation, a phenomenon that is also remarked in our found diversity indices (Tables S5 and S6; spreadsheet: diversity indices).

In summary, this study presents the genomic characterization of the Agotes. The new findings show that this population is part of the local genetic pool, showing high levels of consanguinity, likely driven by their long-standing social isolation. The study was able to unambiguously refute most of the previous hypotheses regarding the Agotes’ origins, being the one linked to Catharism the most plausible. Our results complement historical sources and address a long-standing debate about the origin of the Agotes, showing how historical and cultural events can shape the genetic structure of human populations. Therefore, this study highlights the complex diversity of the Basque population and includes the Agotes as a distinct population in the Basque genomic landscape.

Limitations of the study

Although this study makes some initial progress in understanding the origins and history of the Agote population, there are some limitations to consider when reconstructing its complete population history. First, determining the exact triggers that led to and sustained the severe social marginalization of this population over centuries is outside the scope of the present study. Due to the limitations and nature of our genome-wide genotype data and uniparental markers from present day individuals, some hypotheses about the Agotes origin, such as leprosy, remained unexplored. Having whole-genome sequences, a selection test could be applied to search for variants related to leprosy immunity. It would also be interesting to trace the possible presence of M. leprae DNA in Agote funerary remains through metagenomics, as it has been performed elsewhere.⁶¹ Ancient Agote DNA could give insights about the genetic composition of this population in past times.⁶² While this could be explored in future studies, this would not be a simple task, since the Agotes are sensitive regarding the burials of their ancestors. In addition, given that only a handful of Agotes survive in an isolated location, this study is based on a very small sample size with a high number of biological close relatives, which could have reduced the accuracy of certain analyses. However, achieving larger sample sizes in future studies is unlikely since we sampled in Bozate every single individual willing to participate in this study, which represents 73% of the total population (the current census of Bozate is 93 inhabitants). In addition, we understand that the Agotes from Bozate may not be representative of the demographic and genetic history of all the Agotes (they occupied Béarn, Gascogne, Guienne, Comté de Foix, the southwestern region of Languedoc-Roussillon, Guipuzcoa, and Aragón in the past). In that sense, the history that we present in this study is a partial one. Finally, and regarding the Methods used in this study, we recognize that the haplotype-based method fastGLOBETROTTER for the inference of admixture dates, requires additional consideration. It did not recover the known genealogical admixture for the Agotes of mixed ancestry within the last 1–2 generations. This may be due to different and likely complementary causes. First, although the estimated date of the admixture event is 1867, the confidence interval is broader (95% CI = [1,939–1,844]), which may reflect continuous gene flow across the entire temporal range. When sampling, priority was given to the oldest members of the same family, which would be consistent with admixture events in their parents or grandparents. However, although admixture dating methods are highly sensitive, they infer only approximate dates. In addition, it should be noted that a standard average generation time of 25 years per generation was used, which affects the conversion of inferred generations into calendar years. On the other hand, it could be possible that the non-Agote grandparents may themselves have Agote ancestry or that some individuals classified as Agotes may be previously admixed. It is also worth noting that, although the haplotype-based software fineSTRUCTURE assigns Agote and Baztan individuals to two distinct clusters, it is plausible that these groups are not genetically differentiated enough to allow the precise inference of admixture events between them.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Antonio González-Martín (antonio@ucm.es).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The human genetic data reported in this study are deposited in the European Genome–Phenome Archive (EGA; https://ega-archive.org) under accession numbers: Agotes Project: https://ega-archive.org/search/agotes; Study: EGAS00001008390 (https://ega-archive.org/studies/EGAS00001008390); Dataset: EGAD00010002839 (https://ega-archive.org/datasets/EGAD00010002839); Agotes Project Data Access Committee: EGAC00001003624 (https://ega-archive.org/dacs/EGAC00001003624); Genome_Data_Agotes Policy: EGAP00001003707. Due to the sensitive nature of human genetic data and the ethical requirements governing their use, these data are held under controlled access rather than being publicly available. Access requests must be submitted directly through the EGA platform using the accession codes provided above and will be evaluated by the relevant Data Access Committee to ensure compliance with the ethical, legal, and data protection frameworks governing human genetic research. All other data reported in the manuscript will be shared by the lead contact upon request.
• •This study does not report original code.
• •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.

Acknowledgments

We sincerely thank all the voluntary donors for providing the samples that made this study possible, with special gratitude to the residents of the Bozate community for their commitment and generosity. We are also particularly grateful to Aitziber Cebrián Bidegain and Joseba Moreno Bidegain for their support, sensitivity, and dedication to this project; without their collaboration, it would not have been feasible. We extend our sincere thanks to the Government of Navarra and the City Council of the Baztan valley for their financial support and for facilitating and enabling the development of this study. C.A.F.-L. was funded by the Bertil Lundman’s Foundation and the Royal Physiographic Society in Lund (Nilsson-Ehle Endowments). G.A. was supported by the Generalitat de Catalunya via a Consolidated Research Group grant (2021-SGR-00706).

Author contributions

Conceptualization, A.G.-M. and A.G.; data curation, A.P.-E. and A.G.-M.; formal analysis, A.P.-E., A.P.A.-A., C.A.F.-L., E.d.C., G.A., and A.G.-M.; funding acquisition, A.G.-M.; methodology, A.P.-E., A.P.A.-A., C.A.F.-L., E.d.C., and G.A.; project administration, A.G.-M.; resources, A.G.-M., A.G., and M.T.d.G.; software, A.P.-E., A.P.A.-A., C.A.F.-L., E.d.C., and G.A.; supervision, A.G.-M., G.A., C.A.F.-L., and A.O.U.; validation; writing – original draft, A.P.-E. and A.G.-M.; writing – review and editing, A.P.-E., A.P.A.-A., C.A.F.-L., E.d.C., A.G., M.T.d.G., A.O.U., G.A., and A.G.-M.

Declaration of interests

The authors declare that they have no conflicts of interest.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Genetic data have been deposited at the European Genome–Phenome Archive (EGA; https://ega-archive.org) with the following information and accession codes: Agotes Project: https://ega-archive.org/search/agotes; Study: EGAS00001008390 (https://ega-archive.org/studies/EGAS00001008390); Dataset: EGAD00010002839 (https://ega-archive.org/datasets/EGAD00010002839); Agotes Project Data Access Committee: EGAC00001003624 (https://ega-archive.org/dacs/EGAC00001003624); Genome_Data_Agotes Policy: EGAP00001003707.	This paper	EGAS00001008390 EGAD00010002839 EGAC00001003624 EGAP00001003707
Software and algorithms
VCFtools (0.1.16)	Danecek et al.63	https://vcftools.sourceforge.net/; RRID: SCR_001235
BCFtools 1.22-3-g15746941	Danecek et al.64	https://samtools.github.io/bcftools/bcftools.html; RRID: SCR_005227
PLINK v1.9	Purcell et al.42	https://zzz.bwh.harvard.edu/plink/; RRID: SCR_001757
EIGENSOFT	Patterson et al.35	https://github.com/DReichLab/EIG; RRID: SCR_004965
KING v2.3.2	Manichaikul et al.34	https://www.kingrelatedness.com/; RRID: SCR_009251
ADMIXTURE v1.3.0	Alexander et al.36	https://dalexander.github.io/admixture/download.html; RRID: SCR_001263
PONG v1.5	Behr et al.65	https://github.com/ramachandran-lab/pong
ADMIXTOOLS (qp3Pop, qpDstat, qpWave, qpAdm)	Patterson et al.37	https://github.com/DReichLab/AdmixTools/; RRID: SCR_018495
SHAPEIT v2	O’Connell et al.66	https://mathgen.stats.ox.ac.uk/genetics_software/shapeit/shapeit.html
CHROMOPAINTER	Lawson at al.38	https://people.maths.bris.ac.uk/∼madjl/finestructure-old/chromopainter_info.html; RRID: SCR_026062
fineSTRCUTURE	Lawson at al.38	https://people.maths.bris.ac.uk/∼madjl/finestructure-old/finestructure_info.html; RRID: SCR_018170
iTOL v6	Letunic and Bork67	https://itol.embl.de/; RRID: SCR_018174
hap-IBD	Zhou et al.68	https://github.com/browning-lab/hap-ibd
IBDNe	Browning and Browning39	https://faculty.washington.edu/browning/ibdne.html; RRID: SCR_015064
fastGLOBETROTTER	Wangkumhang et al.41	https://github.com/hellenthal-group-UCL/fastGLOBETROTTER
Haplogrep 3 v3.2.1	Schönherr et al.69	https://github.com/genepi/haplogrep3
EMMA	Röck et al.70	https://empop.online/haplotypes#query_form
NevGen	Willuweit and Roewer71	https://www.nevgen.org/
Arlequin v3.5.2.2	Excoffier and Lischer72	RRID: SCR_009051
Fluxus Network 4.6	Bandelt et al.73	RRID: SCR_008517
Rho dating method	Saillard et al.74

Experimental model and study participant details

This study is based on the genomic analysis of 125 oral swab samples from adult human volunteer donors, of whom 56 (44.8%) are male and 69 (55.2%) are female. The genotype data are publicly available in the EGA repository (see resource availability section, data and code availability subsection), along with the specific population origin of each individual: Bozate (11 Agotes - AGO- and 57 Agotes of mixed ancestry - AGM), the Baztan Valley -BAZ- (N = 21), and Navarra -NAV- (N = 36). The project began by contacting the Agote population through a municipal council member from the Baztan Valley, who belongs to this community. The study was designed in accordance with the ethical principles outlined in the Declaration of Helsinki and received approval from the Department of Universities, Innovation and Digital Transformation of the Government of Navarra (National Subsidy Database reference number: 594556). All donors were provided a written informed consent, in which they voluntarily agreed to participate in the research project entitled “Genomic approaches to the biological history of the population of the Bozate neighborhood (Arizkun, Navarra).” The document specifies that donors have been clearly informed by the researcher about the purpose and scope of the study, and that all data and information collected will be used exclusively for this research. Donors are also given the opportunity to ask any questions and are free to withdraw from the study at any time without consequence. Furthermore, the consent form states that the study will comply at all times with the applicable national and European Union legal regulations. The community was also promised to share the study results once it was finished. Along with the samples, surveys containing information about the origin of donors, their parents, and four grandparents, as well as family relationships among donors, were also conducted. This information was used to reconstruct genealogies up to two generations for each participant, which was used as a criterion for group classification. Specifically, an individual was classified as “Agote” only if all members of their reconstructed genealogy (both parents and all four grandparents) were consistently identified as belonging to the Agote community. If any individual within this genealogy was not identified as an Agote, the donor was classified under the “Agotes of mixed ancestry” group.

Method details

DNA extraction

The 125 samples were intended to be used to obtain two types of genetic data: autosomal genotypes and uniparental markers. For that reason, two DNA extraction protocols were carried out. For microarray genotyping, DNA was extracted via salt precipitation, quantified by fluorimetry (Qubit/Nanodrop 3300) and assessed for sample integrity with Genomic DNA ScreenTape (Tapestation 2200). For the uniparental markers, the salting-out method was used.

Genotype datasets

The 125 new DNA samples were genotyped on the Axiom™ Genome-Wide Human Origins array,³⁷ obtaining a dataset (hereinafter referred to as the Agotes dataset, see Figure 1; Table S1) containing 619,036 single nucleotide polymorphisms (SNPs).

To provide genomic context and test prevailing hypotheses about the origin of the Agotes, we incorporated additional reference populations, typed on the same microarray. From the Allen Ancient DNA Resource (AADR),⁷⁵ we retrieved data for a total of 543 modern individuals (hereinafter referred to as the Reference dataset, see Table S1) from regions including the Middle East, North Africa, North Europe, Central Europe, and Western Europe, specifically populations in the geographical proximity of Bozate (Figure 1), such as Basques from Spain, Basques from France and French populations from South France. We also included data from the Mbuti population to use it as an outgroup for f₃ and f₄ statistics analyses.

The primary objective of this study was to determine the genetic origin of the Agotes. However, we sought to contextualize the Agotes within their broader geographic and fine-scale genetic landscape. To this end, we performed additional analyses using new genotype datasets with 881 reference genetic samples representative of neighboring regions (hereinafter referred to as the Reference 2 dataset, see Table S1), comprising 190 individuals from Basque and peri-Basque populations,³ 256 individuals from French populations⁷⁶ and 435 individuals from populations located in the Catalan Pyrenees.⁷⁷

Uniparental genetic datasets

For the characterization of mitochondrial DNA (mtDNA), samples with no maternal relationship based on genealogies were selected to reduce redundant data. Therefore, out of the 125 total samples, 98 were analyzed. The control region of the mitochondrial genome was amplified for all samples⁷⁸ using the high-throughput strategy.⁷⁹ For PCR amplification of the complete control region, three overlapping primer pairs were used (F15878/R16410; F16190/R285; and F15/R649).^80,81 Post-PCR, unincorporated oligonucleotides and dNTPs were enzymatically removed with ExoSAP-IT (USB, Cleveland, OH, USA). Cycle sequencing of the PCR amplicons was performed using BigDye Terminator version 1.1 (Thermo Fisher Scientific, formerly Life Technologies/Applied Biosystems, Foster City, CA, USA). Unincorporated fluorescent ddNTPs were removed using Performa V3 96-well short plates (Edge Biosystems, Gaithersburg, MD, USA). Sequence data were generated on either the Applied Biosystems 3130xl Genetic Analyzer using POP-6 polymer or the 3730 Genetic Analyzer using POP-7 polymer. Sequences were verified independently by two analysts, and an additional quality control check was performed by EMPOP.⁷⁷ Among these, complete D-loop region sequences could not be obtained for three samples. Since information on the maternal lineage up to two generations prior to the donor was available for most cases, this was used to assign mtDNA origin: 17 Agotes (AGO), 24 Baztan (BAZ), 39 Navarra (NAV), and 15 Others (Table S5). The last category corresponds to individuals whose maternal lineages come from a region other than Bozate, Baztan, or Navarra. Specifically, the Basque Country, Castile and Leon and the American continent.

For the Y chromosome (Y-chr) Y-STR characterization, male samples with no paternal relationship based on genealogies were selected. Hence, 46 out of the 125 total samples were included in the analysis. The amplification kit used was AmpFLSTR™ Yfiler™ PCR Amplification Kit (Applied Biosystems). All relevant controls were included in the STR PCR set-up to verify the validity of the results generated. Following PCR the amplified products were run on the SeqStudio Genetic Analyzer (ThermoFisher). Once the run was completed, the data were analyzed via GeneMapper™ ID-X v1.6 software (ThermoFisher Scientific) using the analysis parameters previously imported for this Y-STR system. Finally, 16 Y-STR were characterized, including both DYS385a and DYS385b. For most samples, data on the paternal lineage spanning up to two generations were available. As for the mtDNA, this information was utilized to assign the Y-chr origin, yielding 15 Agotes (AGO), 7 Baztan (BAZ), 21 Navarra (NAV), and 3 Others (Table S6), whose paternal lineages come from other regions from Spain, specifically, Aragón and Andalucia.

Quantification and statistical analysis

Quality control

We constructed several datasets to address the study objectives. Specifically, to investigate the origin of the Agotes, we employed the Agotes and Reference datasets. For each one, we applied the same quality control (QC) procedures. This included selecting only autosomal SNPs with VCFtools (0.1.16) [vcftools --chr 1-22],⁶³ removing invariant sites with BCFtools 1.22-3-g15746941 [bcftools view -c 1]⁶⁴ and filtering SNPs and individuals based on both per locus and per individual genotype missingness with PLINK v1.9 [--geno 0.05 --mind 0.1].⁴² After QC, the Agotes dataset retained 461,261 SNPs and 125 individuals, while the Reference dataset retained 549,078 SNPs, with 7 Moroccan individuals removed. The two datasets were then merged with PLINK v1.9, resulting in a final dataset (hereinafter referred to as Hypotheses dataset, see Table S1) containing 427,179 common SNPs, which underwent additional QC, ultimately yielding a merged dataset with 427,044 SNPs and 651 individuals.

To examine population structure on a broader geographic scale, the Agotes, Reference and Reference 2 datasets were merged. Following QC, 56,820 SNPs and 1542 individuals remained (hereinafter referred to as the Microgeographic dataset, see Table S1). Notably, the dataset containing populations from the Pyrenees⁷⁷ was genotyped using a different microarray, resulting in limited SNP overlap. To obtain more robust results we followed the same dataset merging, this time excluding the Pyrenean populations. After QC, the resulting dataset (hereinafter referred to as the Microgeographic_WO_PIR dataset) contained 487,008 SNPs and 1107 individuals, thereby increasing the number of overlapping markers. The Microgeographic and Microgeographic_WO_PIR datasets were then used to perform Principal Component Analysis (PCA), ADMIXTURE, and f₃ and f₄ statistical analyses.

Genetic relatedness

Extreme kinship relatedness within a dataset can misguide downstream analyses. To address this, we first computed kinship coefficients among individuals from the after QCed Agotes dataset using KING v2.3.2³⁴ (Table S2, spreadsheet: Kinship coefficients). The kinship coefficient threshold was set at 0.0442 to remove individuals with third-degree or higher relatedness, in accordance with KING software guidelines. KING was re-implemented using the parameters --unrelated --degree 3, identifying 50 related individuals (7 Agotes, 34 Agotes of mixed ancestry, 5 from Baztan, and 4 from Navarra). This resulted in a final set of 75 unrelated individuals (4 Agotes, 23 Agotes of mixed ancestry, 16 from Baztan, and 32 from Navarra) (Table S2, spreadsheet: Related_unrelated individuals). It should be noted that the removal of related individuals from this dataset leads to a substantial reduction in population sample sizes, especially within the Agote and Agote of mixed-ancestry groups. Therefore, to reduce bias due to consanguinity in analyses sensitive to this phenomenon, while preserving the largest possible number of samples, different strategies were implemented, as detailed in the following sections.

Principal component analysis and ADMIXTURE

We examined patterns of population structure at two levels. First, to assess the origin of the Agotes in light of existing hypotheses, we conducted analyses involving all samples from the Hypotheses dataset. We also performed analyses restricted to European populations from the Hypotheses dataset (hereinafter referred to as the Hypotheses_European dataset, N = 435). Second, we carried out additional analyses at a microgeographic level using the Microgeographic and Microgeographic_WO_PIR datasets. Results based on the latter includes only populations from the immediate surroundings of Bozate (N = 404): Baztan, Navarra, Spanish and French Basques, southern French populations, including southwestern groups (Bearn, Bigorre, Chalosse, Landes, and Hautes-Pyrénées) and southeastern groups (Haute-Garonne), as well as northern Spanish populations (Burgos, Cantabria, Northern Aragón, Rioja, and western Bizkaia).

PCA was performed with smartPCA from the EIGENSOFT v5.0 package³⁵ under default parameters to identify potential outliers that might bias the results. Following smartPCA documentation, cryptically related individuals are typically discarded as outliers. Accordingly, we performed a double-check by inferring kinship coefficients for the outlier individuals using KING software. The results (Table S2, spreadsheet: Outliers) revealed a high degree of relatedness, exceeding 0.0422 in most cases among individuals from the Agotes dataset, specifically for Agotes and Agotes of mixed ancestry. Specifically, for the Hypotheses dataset, we identified 73 outliers: 7 Agotes, 44 Agotes of mixed ancestry, 2 from Navarra, and 1 from the Baztan Valley. Notably, 34 of 37 pairs (91.9%) involved Agotes or Agotes of mixed ancestry exhibiting high kinship coefficients (≥ 0.177, indicating first-degree relatives according to KING) with either another outlier or a non-outlier individual from the dataset. In addition, one pair showed a kinship coefficient between 0.0884 and 0.177 (second-degree relatives); another between 0.0442 and 0.0884 (third-degree relatives); and one pair had a coefficient below 0.0442 (specifically 0.0365), corresponding to a third-degree relationship based on prior KING inferences. Hence, this individual was also classified as related. Because outlier detection is conditioned by genetic context, we repeated the outlier detection for the Hypotheses_European dataset. This analysis identified 66 outliers: 10 Agotes; 50 Agotes of mixed ancestry, 2 from Navarra and 1 from the Baztan Valley. Among these, 33 of 43 pairs (76.74%) exhibited kinship coefficients ≥ 0.177 between Agotes or Agotes of mixed ancestry and another outlier or non-outlier individual. There were 5 pairs (11.63%) of second-degree relatives, 3 pairs (6.98%) of third-degree relatives, and only 2 pairs with coefficients below 0.0442. Specifically, one pair of Agotes showed a kinship coefficient of 0.0378, corresponding to a third-degree relationship, and another pair from Navarra had a coefficient of 0.0125. Although this latter value would typically be considered unrelated according to previously reported KING results, one individual from this pair was still retained as an outlier. The same procedures were implemented for the Microgeographic and the Microgeographic_WO_PIR datasets, identifying the same kinship patterns for outliers. Overall, the outliers identified with smartPCA were attributable to close kinship among Agotes and Agotes of mixed ancestry. The reasons underlying the outlier status of other individuals from reference populations fall outside the scope of this study.

Once outliers were identified using smartPCA, PCA itself was performed and ADMIXTURE v1.3.0³⁶ was run for K = 2 to 10 groups. Both analyses were performed in projection mode ('-P' option), whereby PCA and ADMIXTURE were initially applied to non-outlier individuals, followed by projection of outliers belonging from the Agotes dataset. For each K value in the ADMIXTURE analysis, a cross-validation test was performed to identify the model with the highest likelihood. ADMIXTURE results were visualized using PONG v1.5.⁶⁵

Runs of homozygosity

Runs of Homozygosity (ROHs) were identified for the Hypotheses_European dataset using PLINK v1.9. Because genotypes were not pruned for linkage disequilibrium (LD), the minimum ROH length was set to 1 Mb to reduce LD effects,^82,83 with all other parameters configured as in Dixit.⁸² The following ROH parameters were extracted: total length of ROH per individual and number of ROH segments per individual. First, we calculated individual FROH indices, which reflect the proportion between the sum of ROHs (SROH) and the total genome length. To decompose this measure into a genetic drift versus a consanguinity component, we examined the relationship between the sum (S) and the number (N) of ROH. This approach reflects demographic history by inferring the demographic origin of ROH segments, in comparison to the expected 1:1 relationship between both parameters.⁸⁴ These analyses were conducted at both the individual and population levels. To this end, we first computed the mean SROH and the mean NROH within each population.

To explore consanguinity at different ROH length categories for the Agotes and their closest neighboring populations (i.e. Baztan Valley, Navarra, Spanish Basques, French Basques and South France), ROH segments were categorized into different length classes: 1-2 centiMorgans (cM), 2-4 cM, 4-6 cM and 6-10 cM. It should be noted that the aim of this analysis is not to date population events, but rather to understand the patterns related to isolation and/or consanguinity within the Agote group.

f₃ and f₄-statistics

Three-population (f₃) statistics were calculated using the qp3Pop package from AdmixTools v2³⁷ with the Mbuti population as the outgroup. The tested configuration was f₃(PopA, PopB; Outgroup), aiming to assess the amount of shared drift between populations PopA and PopB. A f₃ positive value indicates shared genetic drift proportional to the length of the shared population history relative to the outgroup.⁷⁷ We conducted two outgroup f₃-statistic analyses for the Microgeographic_WO_PIR dataset, which contains the largest number of populations and SNPs. First, we compared Agotes (PopA) against all the populations (PopB). Then, the analysis was performed comparing Agotes of mixed ancestry (PopA) against all the populations (PopB).

The f₄-statistics were calculated based on the Microgeographic dataset, to explore potential gene flow with any population, including the Pyrenees. We used the qpDstat package of ADMIXTOOLS v.7.0.2³⁷ with “f₄mode: YES” parameter. Standard errors were calculated with the default block jacknife. The f₄-statistic f₄(A, B; C, D), measures the average correlation of allele frequencies in the two pairs of populations (i.e., (pA −pB) ∗ (pC −pD), for allele frequencies p, typically averaged over many biallelic SNPs).⁸⁵ For f₄-statistics, the Mbuti population was assigned as population D to serve as highly divergent group. To test cladal relationships between Agotes and Agotes of mixed ancestry the model implemented was f₄(Agotes, Agotes admixed; X, Mbuti). To detect whether Agotes show a distinct or enriched Pyrenean signal the test f₄(Agotes, Baztan; X, Mbuti) was performed. For further support, we repeated the analysis using a larger SNP dataset (Microgeographic_WO_PIR dataset) with the following configuration: f₄(Agotes, Agotes of mixed ancestry; X, Mbuti), where X is either Baztan or Navarra.

Haplotype phasing

To further test the European origin hypotheses at a deeper level, we applied haplotype-based methods. We first phased the genotypes from the Hypotheses_European dataset using SHAPEIT v2.⁶⁶ The 1000 Genomes Project Phase 3 haplotypes served as a reference panel, and the HapMapII GRCh37 genetic map was used for genetic positioning. A total of 2004 SNPs were removed due to mismatches with reference panel SNPs and Mendelian errors.

Haplotype-based definition of target and reference clusters

We applied haplotype-based methods such as CHROMOPAINTER and fineSTRUCTURE³⁸ for two main purposes: (i) to determine the phylogenetic position of the unrelated Agotes and Agotes of mixed ancestry within the European genetic landscape, and (ii) to group the rest of the European individuals into genetically homogeneous clusters and use them in subsequent haplotype-based methods.

Using CHROMOPAINTER in the ‘All Vs All’ configuration (-a 0 0), we calculated haplotype sharing between individuals. Previously identified outliers were removed to prevent bias from high kinship, resulting in a dataset of N = 369 individuals. As a first step, we estimated global mutation (emission) probability (M) and switch rate (n) parameters using four chromosomes (1, 4, 15 and 22) and 10% of the total individuals across 10 iterations of the expected-maximization (EM) algorithm (-i 10 -in -iM). We then averaged n and M values across chromosomes (weighting by the number of SNPs per chromosome) and individuals to obtain the global n and M parameters (864.16 and 0.000328, respectively). These parameters were used to run CHROMOPAINTER across all individuals and chromosomes, with outputs for each chromosome combined using ChromoCombine.³⁸

The resulting coancestry matrix of shared combined chunkcounts served as an input for fineSTRUCTURE, which we used in a two-step process to group our samples in a dendrogram. First, we ran it to obtain the clusters with 2 million sample iterations for Markov Chain Monte Carlo (MCMC) (fineSTRUCTURE command line option: -x 2000000), 2 million burn-in iterations for MCMC (-y 2000000), sampling values every 10,000 iterations (-z 10000) and start with all individuals assigned to 1 cluster, then split from there (-I 1). We then used the output of the first step to build the dendrogram with 10,000 additional hill climbing iterations before inferring trees (-x 10000), tree mode (-m T), changing the tree building algorithm to maintain ordering and likelihood (-k 2) and, at each level of the tree, comparing 1 million pairs of current populations to identify which pair to merge (-t 1000000). The inferred dendrogram was visualized with iTOL v6.⁶⁷

Clusters were defined based on two criteria: (i) being separated at the same branch level, and (ii) showing geographical coherence. We retained the original population labels for Agotes and Agotes of mixed ancestry (containing the original individuals, including the previously identified outliers), with them being the focus of our work. The rest of European outliers were not reclassified.

Patterns of shared ancestry and inferred demographic histories

We calculated shared Identity-by-Descent (IBD) segments using hap-IBD⁶⁸ with default parameters (minimum IBD length > 2 cM) for all Agotes and Agotes of mixed ancestry, as well as the previously inferred fineSTRUCTURE genetic clusters.

Assuming a relationship between IBD segment lengths and temporal distance,⁸⁶ we examined different groups of IBD segment lengths. The aim was not so much to provide specific dates, but rather to broadly explore IBD sharing patterns and shared ancestry over time. The groups were defined as follows: (i) short IBD segments (≤ 3.67 cM); (ii) medium IBD segments (> 3.67 and ≤ 11.7 cM); and (iii) large IBD segments (> 11.7 cM). Also, the hap-IBD output was used to infer historical N_e patterns and demographic history over the last 50 generations before present (BP) with IBDNe.³⁹ These results were compared with N_e trajectories derived from the genetic clusters inferred with fineSTRUCTURE.

Patterns of haplotype-based admixture

To assess admixture patterns in the Agotes of mixed ancestry, we reran CHROMOPAINTER in the All Vs All mode (-a 0 0) for all individuals, including outliers (N = 435), using the previously estimated global n and M estimated parameters. The outputs from all chromosomes were combined with ChromoCombine to obtain a single chunk lengths file, representing total haplotype sharing among samples. We also ran CHROMOPAINTER, setting the Agotes of mixed ancestry (N = 57) as recipients, and the original Agotes, as well as the fineSTRUCTURE clusters (N = 356) as donors for the haplotype-based mixture analysis.

We then used fastGLOBETROTTER⁴¹ with the same recipient-donor configuration to infer admixture parameters (proportion, timing and sources of admixture). The analysis was conducted in two steps: first, with null.ind: 1 over 5 iterations (num.mixing.iterations: 5) and 100 bootstrap re-samples (bootstrap.num: 100) to estimate p-values for admixture evidence; second, with null.ind: 0 using the same parameters to estimate proportions, dates, admixture sources, and 95% confidence intervals (CI) for admixture dates. To convert the estimated times in years, we assumed a fixed generation time of 25 years per generation.

Uniparental markers analyses

Haplogroups were inferred using Haplogrep 3 v3.2.1⁶⁹ and EMMA⁷⁰ for mitochondrial DNA (mtDNA), though we based our posterior analyses on EMMA predictions. NevGen⁷¹ was used for the Y chromosome (Y-chr) haplogroup inference. For both types of uniparental markers (mtDNA and Y-chr), the following analyses were conducted in parallel. Diversity indices (H, haplotype gene diversity; M, mean number of pairwise differences; and π, nucleotide diversity) were calculated using Arlequin v3.5.2.2.⁷² Finally, graphical representations of the haplotype networks allow the reconstruction of haplogroup relationships according to the geographic origin of each of them. For this aim, we also included uniparental data from Basque populations, both for the mtDNA⁵⁹ and Y-chr⁵⁴ when possible. These networks were generated using Fluxus Network 4.6,⁷³ only for the haplogroups both present in the Agotes and in the remaining populations, to infer divergence times among them. Divergence times were calculated using the rho dating method⁷⁴ which requires inputting the mutation rate in years/mutation. In order to calculate how many years it takes for a mutation to occur we used the following formula: $y=\frac{G}{n\times \mu }$ , where y = years/mutation; G = years/generation (30); n = number of segments or loci (HVI and HVII segments for mtDNA and 17 Y-STR including DYS385a/b which are considered as two separate loci); μ = mutation rate (7.8 × 10^-3 mut/segment/gen for mtDNA⁸⁷ and 2.8 × 10^-3 mut/locus/gen for⁸⁸ the Y-chr). We estimated one mutation every 1923 years for mtDNA and one mutation every 630 years for Y-chr.

Supplemental information

Table S1. Samples used in this study

Samples in Agotes (this study), reference (from the Allen Ancient DNA Resource - AADR)⁷⁵ and Reference 2 datasets.^3,75,77 Agotes + reference datasets were merged, giving rise to the hypotheses dataset. Merging of Agotes + reference + Reference 2 datasets gave rise to the microgeographic dataset.

Table S2. Kinship relatedness

Spreadsheet 1: kinship coefficients. Kinship coefficient among individuals from the Agotes dataset (N = 125); spreadsheet 2: related/unrelated individuals. Related (kinship coeff. ≥ 0.0442) and unrelated (kinship coeff. < 0.0442) individuals from the Agotes dataset (N = 125) inferred by KING software; spreadsheet 3: outliers. Maximum kinship coefficient (KING) for outliers inferred by smartPCA. Outliers are shown in green.

Table S3. Cross-validation errors

CV error for ADMIXTURE analyses (K = 2–10) before and after outliers projection. Lowest CV errors are shown in green. Related to Figure 2B.

Table S4. fineSTRUCTURE clusters

Composition by individuals of genetically homogeneous clusters after running fineSTRUCTURE. Related to Figure 3.

Table S5. mtDNA characterization

Spreadsheet 1: D-Loop. mtDNA polymorphisms from the D-Loop region of 95 out of the 125 samples collected. Haplogroups were inferred using Haplogrep and EMMA; spreadsheet 2: mtDNA haplogroup frequencies. Haplogroup frequencies in analyzed populations based on EMMA predictions; Spreadsheet 3: Diversity indices. Diversity indices for mtDNA. H, haplotype gene diversity; M, mean number of pairwise differences and; π, nucleotide diversity.

Table S6. Y-chr characterization

Spreadsheet 1: Y-STRs. Y-STR characterization for 47 out of 125 samples. Haplogroups were inferred using NevGen; Spreadsheet 2: Y-chr haplogroup frequencies. Haplogroup frequencies in analyzed populations; Spreadsheet 3: Diversity indices. Diversity indices for Y-chr. H, haplotype gene diversity; M, mean number of pairwise differences and; π, nucleotide diversity.

Table S7. Proportions of admixture from fastGLOBETROTTER

Major and minor sources are included, related to Figure 4B.