Introduction

The genetic diversity of European populations has been an intensely researched topic because of its implications in the reconstruction of the history of the human species.1 In the last decade, many efforts have been dedicated to the search for patterns that could be related to known prehistoric events; for example, migrations during the Last Glacial Maximum or the Neolithic demic diffusion.2 Uniparental markers, such as mitochondrial DNA (mtDNA) and the Y-chromosome, have proven increasingly useful for delving into these patterns, allowing for a mostly accepted chronology of key events for European demography to be established throughout the years.3 Nevertheless, discussion of some of the inferences related to this chronology has caused controversies that remain open, including the presumed repopulation of most of Western Europe during the Magdalenian (ca.17–12 ka).4

The most common interpretation of the phylogeography of European mtDNA haplogroups currently supports this theory, as there is a strong geographical southwestern–northeastern cline in the frequencies of haplogroups with postglacial coalescence times, such as H1, H3 and V.5 Frequency peaks for those lineages can all be found in the northern Iberian Peninsula and the southwestern part of France, in a territory which formed a glacial refuge called the ‘Franco-Cantabria’.6 Molecular dissection studies and intensive samplings have shown these clines to be consistent and compatible with a postglacial (formerly late-glacial) resettlement of most European regions, which would be centered in this refuge.7, 8

Lately, this idea has been criticized with the argument that while the frequencies of European postglacial haplogroups are the highest in Franco-Cantabrian populations, they also show their lowest molecular diversities there.9 Furthermore, marked heterogeneities in haplogroup composition appear even between geographically close populations of this area, which are probably related to genetic drift due the steep relief of northern Iberia. Both evidences point to the possibility that the Franco-Cantabrian genetic makeup may have evolved discretely and differently from the rest of Europe, and may not be adequate for inferring continental patterns. Following this view, the proposed repopulation should be centered in other refuges, such as those located in Southern France,10 Italy,11 the Balkan Peninsula12, 13 or the Near East.14

But these low diversities can be interpreted in another way. It is known that human population size in Europe declined from the beginning of the Upper Paleolithic until the Bølling-Allerød interstadial, because of a dietary breadth contraction caused by prey extinction.15 A likely effect of this would have been an enhancement of genetic drift coupled with recurrent population bottlenecks, which are the major factors in the shaping of genetic diversity.16 During the Last Glacial Maximum, conditions would have been worsened, as paleoecological modeling has shown the maximum range of the glacial refuges to be small and constrained by permafrost areas.17 In this situation, carrying capacity of the Franco-Cantabrian refuge would not have been able to sustain a large, expanding population, and a series of fragmented populations would have been more likely, as has been inferred to have occurred in regions of the refuge in a previous ancient DNA study.18 In such a situation, incoming migration promotes increased competition for resources in the area, which in turn hinders population growth; meanwhile, fragmentation promotes ‘refuge effects’19 in which local populations can be entirely monomorphic, but the majority of lineages remain somewhere in the metapopulation.20 Such heterogeneity can be indirectly seen by the diversification and coexistence of Epigravettian and Solutrean cultures during glacial times.17, 21 After the improvement of climatic conditions, a major increase of carrying capacity would occur outside the refuge, promoting a range expansion of the refugee population and migration towards the newly available area. This would be paired with an increase of the metapopulation size, especially in the margins of the inhabited area, as has been inferred by archeological findings for this period.22 After migration, demographic expansions would follow, and if gene flow with the former refugia is not interrupted, genetic diversity will increase.23 This is compatible with the neutrality statistics and mismatch distributions of most European populations, which bear the signal of a population expansion.8 As for the parent population, still in the refuge, it will only expand if the carrying capacity of the area rises, which can be related with the fact that Franco-Cantabrian populations have been inferred to begin their expansion in the Neolithic (after the introduction of agriculture), whereas populations from Southern-Central Europe show earlier tardiglacial expansion times.24 In these conditions, it becomes possible that a glacial refuge, due to a long-term small carrying capacity and internal extinction–recolonization processes,25 might nowadays show low genetic diversity values in comparison with outside areas repopulated from that refuge.

If our hypothesis is true, three variables should positively correlate with the diversity of expanded lineages: geographical distance to the refuge (as northern territories would have been more densely repopulated),4 effective population size (which should be regarded as an historical mean)26 and the number of private lineages of each population (which would also increase due to mutation and the assimilation of the remnants of indigenous populations that survived in cryptic refugia).12

As for the heterogeneity found even in close extant Franco-Cantabrian populations, it is not surprising in the context of the Iberian Peninsula, which is a crossroad between mainland Europe, Northern Africa, the Atlantic Ocean and the Mediterranean Sea.27 The varied Iberian geography favors the existence of barriers and isolated populations all throughout the territory, which calls for detailed studies to account for localized historical processes.28 These have to be paired with extensive sampling efforts to have a chance of identifying the historic and demographic processes behind the distribution of mtDNA lineages.29 On that basis, one of the most understudied regions of the peninsula is the Principality of Asturias, which is located precisely in the western edge of the Franco-Cantabrian refuge (Figure 1).

Figure 1
figure 1

Map of the Iberian Peninsula showing the location of the Principality of Asturias (black). The Franco-Cantabrian region is shaded in gray. Names are given of populations mentioned in the text.

Full size image

So far, mtDNA studies in Asturias have been clinically oriented and based on classical restriction fragment-length polymorphisms analyses30, 31 or included only a few samples as a part of broader studies.32, 33 To our knowledge, the only other sizeable population genetic study is the one from Garcia et al.,9 which used 89 samples collected in Asturias. They noted the aforementioned pattern of local drift as a possible cause for the differences between Asturias and other regions of the Franco-Cantabrian area, but a greater number of samples is likely needed to achieve an unbiased representation of Asturian mtDNA lineages.34 Recent immigration should also be taken into account to avoid admixture biases in the sample, at least to some extent. Thus, we decided to perform an extensive ancestry-restricted population-wide study. The aim of the study was to perform spatial analyses that can be used to test the hypothesis of a postglacial range expansion compatible with the present-day genetic makeup of Western Europe. For that matter, new mtDNA sequences would be obtained from a population in the border of the presumed center of the expansion, which would be used to obtain picture of the pattern of local heterogeneities and drift processes.

Materials and methods

Study area providing new data within the Franco-Cantabrian refuge

Asturias is a Spanish autonomous community of around one million inhabitants, unequally distributed between the densely populated cities of the center and many smaller scattered towns. Its territory, spanning 10 000 km2, is the most mountainous region of Spain and one of the most mountainous region of Europe, with political limits that have existed since the Roman times.35 After the Roman invasion of the peninsula (29–19 BC), no important population movements into the territory were historically recorded, as its difficult terrain kept it outside the main routes of migration into Iberia for many centuries.36 In fact, in medieval times, Asturias was described as an isolated and impoverished province on the northern margin of the Spanish kingdom, a status that hardly changed even after the Industrial Revolution.37 Only in the years following 1950, the development of coal-mining facilities in the region improved the economy and favored the immigration of entire families from other Spanish regions, shaping the Asturian demography as it is today.38

Subjects

Samples of buccal cells were obtained from 429 volunteers recruited in various cities and towns of Asturias during the period 2009–2011, who confirmed having at least two generations of autochthonous maternal ancestry (i.e., grandmothers born in Asturias). No volunteers less than 18 years of age were allowed in the study, informed consent was obtained from each sample donor, and the study conformed to the Spanish Law for Biomedical Research (Law 14/2007—3 July). Sampling was performed using swabs with Dacron tips (Deltalab, Barcelona, Spain), and a protocol based on the Chelex-100 resin (Bio-Rad, Hercules, CA, USA) was used for DNA extraction.39

Sequence amplification and analyses

From each sample, the complete hypervariable segment I (HVS-I) of the control region of the mtDNA was amplified by PCR using the primers L15900 and H00016.40 Two independent amplifications were performed to reduce the possibility of sequence ambiguities derived from low-quality bases. Sequencing was carried out in an ABI 3730xl DNA Analyzer by the company Macrogen (Seoul, Korea).

Sequences were analyzed using the software Geneious v4.85,41 in which the highest quality consensuses were obtained and aligned with the revised Cambridge Reference Sequence.42 This alignment was checked manually for phylogenetic consistence following the recommendations of Bandelt and Parson43 and the EMPOP online tool.44 Haplogroup assignment of the samples was performed with the software HaploGrep,45 using version 13 of the mtDNA phylogenetic tree as a reference.46 Samples were included in paragroups when various equally probable assignments were possible due to the absence of diagnostic mutations. All samples that obtained a HaploGrep quality index lower than 80% and all samples allocated to the HV* paragroup were analyzed further by genotyping the coding-region single-nucleotide polymorphisms adequate for restriction fragment-length polymorphism assays,47 to assure the accuracy of their phylogenetic classification. Also, all samples allocated to the H* paragroup were restriction fragment-length polymorphism-tested for the presence of diagnostic mutations for the H1 and H3 haplogroups.48

Additional data acquisition

Asturian sequences were compared with an HVS-I data set of 9539 sequences from 45 European and North African populations (Supplementary Table 1). To include the maximum number of sequences in this data set, all of them were trimmed to a length of 342 nucleotides between positions 16 024 and 16 366 of the mitochondrial genome. This range is sufficient for our analyses, as it includes the most polymorphic interval of the entire hypervariable region.49

Data for the diversity of the haplogroups H1, H3, K, T2b, W, HV0 and V were computed from the Asturian haplotypes and obtained from the Garcia et al. study9 for nine European regions and Northern Africa, including the Franco-Cantabrian area. Values for the θK and KP parameters were also computed for these regions using the most approximate populations of our HVS-I data set (Supplementary Table S3), with the aim of performing model fitting on the diversity data. For Central and Eastern European populations not present in our data set, estimates previously computed by Helgason et al.50 were used.

Statistical analyses

The software Arlequin v3.51251 was used to obtain estimates of the haplotype and nucleotide diversities described by Nei52 and Tajima,53 respectively. Also, the number of private lineages found in each population was obtained, defined as the number of lineages that are only found in a given population when this is compared with all the others of our HVS-I data set. Comparison of the haplotype diversities of the pairs of populations was performed with the Test_h_diff software,54 which implements a Bayesian procedure. Differentiation between populations was assessed in terms of ΦST pairwise values, which were computed using the Tamura and Nei55 distance between sequences, with an unequal transition/transversion weighting of 4/5.56 To depict population relationships in a simpler manner, Slatkin57 generalized distances that were computed from the pairwise ΦST values were plotted into a two-dimensional multidimensional scaling, with the algorithm ALSCAL included in the software SPSS v17.58

The Watterson59 estimator θK (also called ‘population-mutation rate parameter’) was also computed according to the Ewens sampling formula.60 This value is an approximation, based on the number of haplotypes, to the widely known composite 2Nef·m, in which Nef is the effective number of females of the population and m is the mutation rate per nucleotide of the HVS-I region. Previous studies have defined the existence of a positive, significant correlation between θK and the proportion of private lineages of a group of populations.50, 61 This is because private lineages arise by mutation of a previously existing haplotype, which increases the θK estimates under the theoretical framework of an infinite-allele mutation model.

To properly compare different populations on the basis of their diversity patterns, sampling coverage estimates (i.e., the proportion of the true number of different haplotypes that have actually been sampled) in each population were calculated with the R statistical framework62 using the unseen2 package, which incorporates the formulae developed by Egeland and Salas63 for this matter. This same framework was also used to fit linear models and to compute Moran autocorrelation indexes for some statistics, the latter using package ape.64 Spatial mappings of population genetic statistics were performed with the Quantum GIS v1.73 software,65 whereas interpolation routines shown in the graphics were performed with the Kriging algorithm of the SDA4PP plugin.66 Distances between two populations, when needed, were computed from geographical coordinates using GDMG v1.23.67

Estimation of migration rates between populations was performed with the software MIGRATE-N v3.216.68 For each model that we tested, we used 80 random sequences from each included population and executed 5 separate runs under the Bayesian implementation. Each analysis involved two long Markov chains of 10 million generations, with a standard uniform slice priors on the M and θ parameters. A static heating scheme with four chains and standard temperatures was also used. Model likelihood was obtained by a thermodynamic integration with Bezier approximation, as implemented in the software. Finally, the median parameter values of each run were compared with the 95% credibility intervals (95% CI) of all other runs to ensure their fit and remove outliers if any.

Results

Haplogroup composition and diversity of the Asturian population

The HVS-I sequences from the Asturian volunteers, along with their haplogroup assignment, can be found in Supplementary Table 2. A total of 63 different haplogroups were identified in the sample, and their frequencies are shown in Table 1. The haplogroup composition of Asturias is dominated by nearly a 50% of the haplogroup H, as could be expected given a Southern European Atlantic location. A previous study highlighted the high frequency of haplogroup J in the region, which amounts to a 10.51% in our results (95% CI 7.85–13.71) and seems to be a local peak in respect to neighboring populations.9 About other haplogroups commonly encountered in the Franco-Cantabrian area, another peak appears in the frequency of T2b, which is a 4.66% of the sample (95% CI 2.94–7.04). Interestingly, this frequency is higher than those reported on the neighboring regions of Galicia, Leon and Cantabria, nearly reaching the 5%, which is common in the Mediterranean.69 In contrast, H1, HV, V and X haplogroups seem to experience local minimums in Asturias.

Table 1 Haplogroups found on the Asturias sample and their frequency
Full size table

Some non-European haplogroups, of Near Eastern, African and Asian affinities or origins, were also found: M1, N1b, L3f1b4a and D4g1. Although finding such lineages is rare outside of their proposed areas of origin, they have been reported in other populations from the Iberian Peninsula28 and appear sporadically in European surveys in frequencies around 1% or less, as it is the case for Asturias.70

More detailed information can be inferred by observing the raw Slatkin ΦST matrix, which is in Supplementary Table 5. Regarding Asturias, this index is statistically significant for its comparisons with all other Spanish populations, with the exception of the Balearic Islands, but this could be an effect of their small sample size. It has to be noted that the regions immediately to the west and east of Asturias, Galicia and Cantabria, respectively (Figure 1), have low ΦST values of 0.003 and 0.004, whereas the value for the southern neighboring region, Leon (Figure 1), reaches 0.008. This is an even higher value than those estimated between Asturias and every region of the British Isles, with the exception of Orkney and the Hebrides. Also, a significant difference in haplotype diversity (P=0.002) between Asturias and Leon was found, which did not appear in comparisons with Galicia, Cantabria or even England. This suggests a relationship with the north (British regions), which can be seen also in the case of Galicia and Cantabria, without the steep differences with the south (Leon) in terms of diversity or ΦST. Graphically, the pattern can also be seen in the multidimensional scaling plot (Figure 2), in which the Asturian sample (ES-AS, marked with an arrow) is located in the edge of the Iberian group and farthest from the African group, surrounded by British and Scandinavian samples. As a sidenote, in this study, the term ‘Iberian’ includes the Balearic Islands and the French Province of Béarn (Figure 1) because of their deep historical relationship.71, 72

Figure 2
figure 2

Multidimensional scaling plot of European and African populations based on their population pairwise Slatkin linearized ΦST distances. The Asturian population (ES-AS) is marked with a black arrow. Other population identifiers are given in Supplementary Table 1.

Full size image

Effective population sizes and private lineages in Iberia

Estimates of sampling coverage, nucleotide and haplotype diversity, the θK parameter and the percentage of private lineages (KP) for each population of the HVS-I data set can be found in Table 2. We obtained a positive and significant correlation r=0.683 (P<0.001) between θK and the number of private lineages for the whole HVS-I data set and r=0.649 (P=0.006) for only the Iberian regions. The latter is graphically shown in Figure 3 along with the 95% confidence interval of the linear regression line, to highlight the different population groups that exist in territory of the Iberian Peninsula. Two of the populations of the Franco-Cantabrian refuge, Cantabria and Navarra, are outside the 95% CI because of a scarcity of private lineages, which is typical of areas in which both inward and outward migrations have been historically high.50 Asturias, on the contrary, is in the upper half, but outside the regression 95% CI for having an excess of private lineages relative to its effective number of females. It is, in fact, a population remarkable for having the highest proportion of private lineages of all our Iberian data set.

Table 2 Diversity indices obtained from the populations of the HVS-I database
Full size table
Figure 3
figure 3

Scatterplot of Iberian populations based on their θK values and proportions of private lineages. The linear regression line and its 95% confidence interval (shaded) corresponds to r=0.649.

Full size image

Spatial analysis of European populations

Spatial maps of the θK and KP values of the populations of our HVS-I data set can be found in Figures 4 and 5. As it can be seen, both statistics show a cline with the lowest values in the Franco-Cantabrian refuge area. As the θK and KP values were correlated (r2=0.458; P-value=0.018), only the most explanatory variable was used in the models to avoid collinearity. Diversity of the H1 haplogroup was fitted into a linear model that used the distance to the Franco-Cantabrian refuge and the θK value as predictors (r2=0.569; P-value=0.022). The same two predictors were adequate for fitting the diversity of the K haplogroup (r2=0.587; P-value=0.012). Diversity of the H3 haplogroup was fitted by using the distance to the refuge and the percentage of private lineages as predictors (r2=0.491; P-value=0.039). As for the V haplogroup, diversity in Asturias was much higher than in other Franco-Cantabrian areas (0.806 vs 0.653), which hindered the fit of any model that considered it. After removing this outlier value, diversity of the V haplogroup for the rest of the data set could be fitted to a high goodness by using a quadratic model that considered the distance to the refuge and the percentage of private lineages (r2=0.883; P-value=0.009). Diversities of the T2b, W and HV0 haplogroups could not be significantly fitted to a model containing any of the aforementioned variables. Plots of these models can be found in Supplementary Figures 1 to 4.

Figure 4
figure 4

Map of θK values in Europe and Northern Africa, with the locations of the HVS-I data set populations shown as asterisks. Lighter colors indicate lower values, whereas darker colors indicate higher values. Moran’s I spatial autocorrelation analyses were not significant (P=0.089). Iberian genetic isolates (Pasiegos and Maragatos) were not considered in the analysis.

Full size image
Figure 5
figure 5

Map of percentage of private lineages in Europe and Northern Africa, with the locations of the HVS-I data set populations shown as asterisks. Lighter colors indicate lower values, whereas darker colors indicate higher values. Moran’s I spatial autocorrelation analyses were statistically significant (P=0.006). Iberian genetic isolates (Pasiegos and Maragatos) were not considered in the analysis.

Full size image

To further assess the population dynamics of the Franco-Cantabrian populations, we computed a MIGRATE analysis on seven Iberian, four French and one British population. All of them were inside or near the Franco-Cantabrian area. First, the likelihood of five models with differing levels of migration and panmixia were calculated, including a full model to serve as a reference (Supplementary Table 4). With the likelihood values, a model was chosen according to the Bayes factor criterion.73 This model supported the division of the Franco-Cantabrian refuge in two different areas: Asturias–Cantabria and Basques–Navarra–Bèarn, with panmictic populations inside each area. The migration matrix was constrained so that migration was only possible between directly adjacent areas, and this can be graphically seen in Figure 6. Reported values of migration parameters are their averages over the five runs, as no outlier results were found. They were rounded to integers, as they represent the effective number of migrants per generation (Nef·m). In all the regions tested, inward–outward gene flow was nearly symmetrical with the exception of trans-Pyrenean migrations, which were greater towards France. Migration from the Basque populations was nearly six times higher towards Catalonia and France (60) than towards Asturias and Cantabria (11).

Figure 6
figure 6

Results of the MIGRATE analysis. Values indicate the effective number of migrants per generation (Nef·m). Populations with the same number belong to the same panmictic cluster. Identifiers are: 1: Galicia; 2a: Asturias; 2b: Cantabria; 3a: Basque Country; 3b: Navarra; 3c: Bèarn; 4: Catalonia; 5a: Languedoc; 5b: Poitou; 5c: Maine-Anjou; 5d: Brittany; 6: England.

Full size image

Discussion

Population dynamics around the Franco-Cantabrian refuge

Both θK and KP calculated from our HVS-I data set show a cline in Europe (Figures 4 and 5), in which the lower values are coincident with the Franco-Cantabrian populations. This pattern is opposite to the frequency clines shown by haplogroups H1, H3 and V in earlier studies.7, 74, 75

In our analysis, diversities of H1, H3 and K can be explained with simple linear models using only two of the aforementioned covariates. The fit is hindered by some outlier populations (Supplementary Figures 1–3), but this fact can be related to local drift processes, such as those previously inferred in some regions from France76 or Scandinavia,77 and does not disrupt the general trend. Also, the historical migration pattern that enhanced the diversity of these haplogroups during the range expansion can be seen in the results of the MIGRATE analysis (Figure 6). The result of haplogroup K is particularly noteworthy, as its coalescence in Europe has been dated to the Initial Upper Paleolithic and not to postglacial times, as is the case for the other haplogroups.4 Nevertheless, its involvement in a postglacial repopulation was already proposed on the basis of the star-like mutational pattern of its common haplotypes.78

As for the diversity of the V haplogroup, a very good fit was obtained using a quadratic model (not accounting for the Asturian data), where increases of predictors are correlated with increases of diversity up to a point. This accounts for the founder effects that constrained the diversity of distant northern European populations,79 and is consistent with our interpretation of the refuge effect in Southern-Central Europe (Supplementary Figure 4). The outlier result of Asturias is probably a reflection of the heterogeneity of extant northern Iberian populations, which is discussed in the in-depth analysis of its sample. As for the high diversity of southern Iberian populations, it could be due to long-term contacts with Berber populations across the Gibraltar Strait,80 or be a signal of the proposed Iberian origin of this haplogroup.5

The Asturian edge of the Franco-Cantabrian area

In the genetic landscape of the northwestern Iberian Peninsula, a clear outlier in terms of effective size and percentage of private lineages can be found in Asturias (Figure 3). Such an outlier result could be explained by at least three basic causes:50 demographic isolation, significant gene flow from unsampled regions and/or higher sampling coverage than other regions. Before discussing the first two, it should be said that the latter is unlikely, given the sampling coverage estimated for the different Iberian regions, as most populations have been equally or more thoroughly sampled than Asturias, without finding a higher proportion of private lineages in them (Table 2).

The historical isolation of this region, coupled with its complex orography, is well documented. Consequently, it is reasonable to deduce such feature as the most probable cause of its great proportion of private lineages, probably because of internal genetic structuring or microdifferentiation, in a manner similar to cryptically subdivided populations.81 Also, particularly high inbreeding values for the region were inferred in a previous isonymy study.82 For the female case, ΦST pairwise P-values show that the mitochondrial pool of Asturians is statistically different from most of the other Iberian regions. However, the greater neighboring divergence, not counting the isolated Pasiegos and Maragatos, happens with Leon at the south. Significant differences in haplotype diversities between these two populations are consistent with a geographical–genetical structuring process, as it indicates that most of the differences between them would be related to mutational distances between non-shared haplotypes.83 Such a process could be promoted by a genetic drift, aided by a barrier that inhibits, at least partially, the flow of haplotypes after their apparition. A candidate barrier, in this case, would be the Cantabrian Mountain Range of the southern part of Asturias, a loftier and most intricate mountainous system called the Asturian Massif, which is the political border between the Province of Asturias and the autonomous community of Castilla–Leon.84

Our migration analysis also suggests restricted gene flow from the Basque population group. This is reminiscent of the distinct and fragmented mitochondrial pools that have been found in populations from different timeframes throughout the Franco-Cantabrian area.18 Nevertheless, there is no clear east–west spatial barrier in the geography of northern Spain; and archeological, climatological and paleontological contexts of Asturias, Cantabria and the Basque Country were all very similar during the Upper Paleolithic.85, 86 Alternative explanations for this low historical migration rate should be sought with a more in-depth regional study, in which relatively recent timeframes should be explicitly considered.87

The second explanation for an excess of private lineages in Asturias, gene flow with unsampled regions, cannot be discarded but does not seem to be the main cause, as previous studies did not show significant relationships between Spain and regions not present in our database, such as the eastern countries.88 In any case, lineages that reached Iberia by ancient southward migrations should be more or less homogenously distributed throughout it, so any bias in our results due to unrecognized shared haplotypes is probably very small.

Regarding similarities with other European regions, the ΦST patterns observed in the multidimensional scaling plot are consistent with the results of the migration analysis and suggest a close relationship between the northwestern Iberian coast and the British Isles, in accordance to the genetic makeup of the Atlantic façade evidenced in other studies.89 Neolithic coastal dispersions could be a parsimonious explanation,4 without excluding some input of ‘British Celtic’ markers like the haplogroup J2b1a (2.10% of Asturian sample; 95% CI 1.05–3.88), which is also shared with Mediterranean populations.90

Conclusions

The spatial patterns of classical population genetic statistics seem to correlate well with the diversity of haplogroups which were proposed to participate in the repopulation of Europe during the Magdalenian period. These diversities are low in the former territory of the Franco-Cantabrian refuge, which can be interpreted in the context of a spatially restricted refuge with small carrying capacity. A recolonization process in which the populations greatly expanded after arriving to the new territories would explain the positive association of diversity and distance to the refuge. In this way, diversity measures can be reconciled with previous interpretations based on haplogroup frequency distributions and paleontological findings. Nevertheless, any model that can be proposed based on extant DNA data will have to be tested by the analysis of ancient DNA of Franco-Cantabrian populations. Only such a direct approach can be used to find if the now common European haplogroups were actually present in the gene pool of Paleolithic Europeans and clarify their spatial distribution. In this regard, we have to highlight the recent study of Hervella et al.,18 which has been the first report of haplogroup H among Magdalenian Iberians.

Also, this study was a first attempt to characterize the mitochondrial gene pool of the Asturian population, by performing direct sequencing of the hypervariable HVS-I of the mtDNA control region. Regional peculiarities highlight a partial isolation of the Asturian mitochondrial pool that differentiates it from the rest of Franco-Cantabrian populations, with an unexpected high number of private lineages that would be compatible with a certain degree of genetic structuring inside the population. It seems that the complex history of the Franco-Cantabrian region might have left signals in the genetic makeup of its populations that have not been entirely uncovered yet.