Abstract

Background

Processes shaping the formation of the present-day population structure in highly urbanized Northern Europe are still poorly understood. Gaps remain in our understanding of when and how currently observable regional differences emerged and what impact city growth, migration, and disease pandemics during and after the Middle Ages had on these processes.

Results

We perform low-coverage sequencing of the genomes of 338 individuals spanning the eighth to the eighteenth centuries in the city of Sint-Truiden in Flanders, in the northern part of Belgium. The early/high medieval Sint-Truiden population was more heterogeneous, having received migrants from Scotland or Ireland, and displayed less genetic relatedness than observed today between individuals in present-day Flanders. We find differences in gene variants associated with high vitamin D blood levels between individuals with Gaulish or Germanic ancestry. Although we find evidence of a Yersinia pestis infection in 5 of the 58 late medieval burials, we were unable to detect a major population-scale impact of the second plague pandemic on genetic diversity or on the elevated differentiation of immunity genes.

Conclusions

This study reveals that the genetic homogenization process in a medieval city population in the Low Countries was protracted for centuries. Over time, the Sint-Truiden population became more similar to the current population of the surrounding Limburg province, likely as a result of reduced long-distance migration after the high medieval period, and the continuous process of local admixture of Germanic and Gaulish ancestries which formed the genetic cline observable today in the Low Countries.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13059-025-03580-z.

Keywords: Urbanization, Palaeo-genomics, Low countries, Medieval, Migration, Plague, Flanders

Background

Compared to other world regions, Europe is genetically highly homogenous today, yet it displays fine-scale geographically correlated variation [1, 2] that can, if uncorrected, confound complex trait analyses [3]. The current interdisciplinary synthesis of evidence derived from archeology and genome-scale studies of ancient human remains explains allele frequency similarities across Europe by multiple prehistoric episodes of continent-wide population movements and admixture in the Neolithic and Bronze Age resulting in a more stable population structure in most parts of Europe since the Iron Age [4]. While the medieval period witnessed less gene flow to Europe from outside sources than previous periods, it was a time of major political changes and migration within the continental boundaries. The North Sea region, for example, witnessed major population movements in the Early Middle Ages (EMA) that led to massive changes in local genetic ancestries [5, 6]. In the same millennium, population movements in the Baltic Sea region contributed to the formation of regional population structure seen in the region today [7, 8]. On the other hand, evidence from Britain shows that local population structure continued to undergo extensive changes in the later medieval and post-medieval periods [9, 10]. The impact of demographic changes in continental western Europe during the last two millennia is less known due to the lack of relevant transect of time studies.

The Low Countries, consisting of Belgium, the Netherlands, and Luxembourg, have been since the medieval period one of the most densely populated and highly urbanized regions in Europe. Despite the homogenizing effects of population growth and migration in the last centuries, genetic differences between subregional populations are presently still clearly pronounced in the Low Countries, with an underlying north to south genetic ancestry cline that is likely to derive from admixture events dating at least to the eleventh century AD [11, 12]. In the absence of other obvious geographical boundaries, these regional patterns may have been shaped by the major rivers. Without direct genetic evidence with ancient DNA samples from a time transect through the Middle Ages, however, it remains unknown when and which historical events and ancestry sources have contributed to the formation of present-day population structure in the Low Countries.

The medieval history of the Low Countries, particularly in the now densely populated Flanders region of northern Belgium, is marked by extensive socio-political and economic developments that may have shaped the genetic landscape observed today. In this context, the town of Sint-Truiden, located in the modern province of Limburg, provides a particularly valuable case study as its history is well documented. The town’s origin is situated along two major Roman roads [13]. In the seventh century, Saint Trudo founded an abbey at the settlement called Sarchinium, which, after his death in 693, became a highly popular pilgrimage destination [14]. The influx of pilgrims played a crucial role in the growth of the settlement around the Benedictine abbey. By the twelfth c., this settlement had evolved into a thriving town named after its founder, Sint-Truiden. The town experienced strong expansion and population growth in the thirteenth c., largely due to its strategic location in the fertile agricultural region of Haspengouw and its specialization in the production of cloth and beer, which were traded internationally [13]. While historical scholarship has now acknowledged that Yersinia pestis circulated in the Low Countries from the Black Death (1346–1352) onwards, the chronicles of the abbey, Gesta abbatum Trudonensium, and the city’s administrative records do not mention any impact of the plague in the fourteenth c. [15, 16]. In the fifteenth c., historical evidence does refer to the town suffering from excess mortality and other problems likely caused by the plague [17]. The town was subject to intense political and military upheaval since the fourteenth c. and its decline was further exacerbated by the conquest of the town in 1467 during the conflict between the Burgundian State and the Prince-Bishopric of Liège [13]. Since then, Sint-Truiden has served as a regional center with a current population size of approximately 24,000 individuals within the city.

A recent archeological excavation for the redevelopment of the city squares in the center of Sint-Truiden conducted between 2018 and 2020 unearthed a cemetery area that had been in use for over a thousand years, from the seventh c. until the eighteenth c. [18, 19]. The excavations were carried out at two adjacent locations (Fig. 1). The first—and the smallest—location was on the present-day Trudoplein square. This was part of a burial ground outside the abbey, near the tower of the Abbey’s church, used from the seventh to fourteenth century. The second location formed a large part of the present-day Groenmarkt square which was a burial ground from the seventh to the eighteenth c., extending before the eleventh c. possibly to the Trudoplein. From the eleventh to the eighteenth c., the eastern part of the Groenmarkt was the parish cemetery of Our Lady. From 1286, the burials continued only in zones 1 and 2, within the boundaries of the churchyard (Fig. 1). Over 3000 inhumation burials were uncovered, of which a subset of 404, selected by their skeletal completeness and availability of teeth, was included for ancient genomic analysis, representing a spatial distribution across the site and the broad temporal range. The availability of a large number of individuals from one site and the long use of the cemetery offers a unique opportunity to study the microgeography of the burials in time within a single urban site in relation to genetic ancestry, genetic relatedness, and metagenomic findings. The large number of genomes sequenced and the long-term use of the cemetery makes it also suitable as a model for the study of the development of a medieval European town, local population structure formation, and the influence of historically attested events, such as pandemics on genetic/population history and health.

Fig. 1.

Sint-Truiden city center cemetery map. Two main burial groups at Trudoplein and Groenmarkt are highlighted by pink and green, respectively. Timeline of the main events mentioned in the text is shown to the right

Results

We generated whole-genome shotgun sequence data from 404 human skeletal remains from Sint-Truiden, Belgium, dating from eighth to eighteenth c. AD, keeping for further analyses 372 genomes sequenced to mean coverage higher than 0.01x (median coverage 0.69x) and contamination rate < 0.05, including 332 with coverage > 0.1 × that were used in imputation-based analyses (Table 1, Additional file 1: Table S1). Radiocarbon dating and dietary isotope analyses of 114 individuals included in the genetic study, enabled probabilistic time group assignment of individuals by their burial context and showed a trend (r = 0.49) of higher δ15N values in time (Additional file 2: Fig. S1). Nitrogen isotope values reflect dietary behaviors, specifically related to protein intake, including terrestrial and aquatic protein-rich plants and animals [20]. The increasing δ15N values in time could either reflect rising marine fish consumption in inland Flanders in the late Middle Ages and in the post-medieval period [21] and/or generally improved nutritional status of the later phase individuals from the parish group. On the basis of radiocarbon dates and archeological context, the burials were assigned to broader time groups: I—700–1000 AD, Early Middle Ages; II—1000–1286 AD, High Middle Ages; III—1286–1600, later medieval; 1600–1800 AD, post-medieval (Additional file 1: Table S1). 1286 AD, as the time point separating time groups II and III, refers to the date at which there was a rearrangement of the current Groenmarkt [18]. Individuals from distinct burial zones (Fig. 1) and time groups were grouped together for downstream analyses.

Table 1.

Ancient genomes analyzed in this study

Number of individuals archaeological site, province, country period, date range source total ≥0.01x imputed Sint-Truiden, Limburg, Belgium 1 404 372 332 Time group I: Early Medieval, 700-1000 42 40 32 Time group I/II 45 37 31 Time group II: High Medieval, 1000-1286 207 191 171 Time group III: later Medieval, 1286-1600 58 55 52 Time group III/IV 22 21 19 Time group IV: post-medieval, 1600+ 30 28 27 Ypres, W-Flanders, Belgium High Medieval 1 15 15 8 Koksijde, W-Flanders, Belgium Merovingian, 7th to 8th cc 2 20 Wulpen, W-Flanders, Belgium High to Late Medieval, 11th to 14th cc 2 5 France Late Iron Age (LIA), 5th to 1st cc BC 3 18 England Late Iron Age/Roman 4 to 6 30 England Early Medieval, 5th to 9th cc 7 152 Ireland Early to Late Medieval 7th to 13th cc 7 21 Netherlands Early Medieval, 4th to 11th cc 7 22 Alt-Inden, N-Rhine, Germany Merovingian, 5th to 8th cc 7 9 Lower Saxony, Germany Early Medieval, 5th to 10th cc 7 26 Rathausmarkt, Schleswig, Germany High Medieval, 11th to 12th cc 7 11 Denmark High Medieval, 11th to 13th cc 7 7 Denmark Viking Age, 8th to 11th cc 8 73 Norway Viking Age, 9th to 11th cc 8 28 Orkney, Scotland Viking Age, 9th to 11th cc 8 33

≥ 0.01 × —number of genomes sequenced to coverage ≥ 0.01 × and contamination rate < 0.05, imputed—number of genomes imputed in this study; all date ranges except for the Late Iron Age France are AD. Sources: 1—this study, 2—Sasso et al. 2024 [6], 3—Fischer et al. 2022 [22], 4—Scheib et al. 2023 [23], 5—Schiffels et al. 2016 [24], 6—Martiniano et al. 2016 [25], 7—Gretzinger et al. 2022 [5] and 8—Margaryan et al. 2020 [8]

In order to provide regional reference data from Sint-Truiden, we also sequenced the genomes of 15 individuals from the Hooge Siecken site in Ypres, in the modern province of West-Flanders, to coverage > 0.01x, including 8 to coverage > 0.1x (Table 1, Additional file 1: Table S1).

Analyses of genetic ancestry and population structure

To explore the genetic ancestry of individuals buried at the cemetery of Sint-Truiden city center (Fig. 1), we first performed principal component analysis (PCA) on imputed genomes in the context of a broader reference set of genomic data. Analyses of ancient Sint-Truiden genomes in context of global reference populations showed that all individuals clustered tightly with those from present-day Northwest Europe (Additional file 2: Fig. S2). PCA focused on modern and ancient reference data from Northwest Europe (Fig. 2A) confirms the north–south structure observed previously in present-day data of the Low Countries [11], showing that the majority of the ancient Sint-Truiden individuals are placed on a PC1-defined cline between the Scandinavian and Dutch populations, on one end, and, modern and ancient genomes from France on the other (Fig. 2B, C). During the Early (EMA) and High (HMA) Middle Ages (time groups I–II), the Sint-Truiden population appears to have been more heterogeneous and spread out across this cline than it was during the later medieval and post-medieval period, being also more diverse in EMA/HMA than the whole Flemish population today (Fig. 2B). The second PC separates ancient and modern individuals from the British Isles from continental western Europe. Five individuals from the early (I/II) phase of Sint-Truiden appear as outliers by their PC2 values, clustering with the medieval and modern Irish and Scottish (Fig. 2B).

Fig. 2.

Sampling locations and genetic ancestry of the studied populations. A A map of the North Sea region with Early (EMA) and High (HMA) Middle Ages, Roman, and Late Iron Age (LIA) archeological sites used in data analyses, including genomes from Sint-Truiden (red), this study, and available reference data. PCA of selected modern (B) and ancient (C) genomes from Europe. PCA was run, after excluding closely related individuals, with FlashPCA2 without projection. B Modern population sources: 400 English, 191 French, 200 Irish, 190 Spanish, 443 Scandinavian (Danish, Norwegian, Swedish) and 400 Scottish/Welsh genomes from the UK Biobank; 112 Flemish and 195 Dutch genomes from the MinE consortium data [26, 27]. C 329 medieval/early modern imputed Sint-Truiden genomes and 8 Ypres individuals from this study, 20 Koksijde and 5 Wulpen individuals from Sasso et al. [6], 18 Iron Age French Gaul individuals from Fischer et al. 2022 [22], and 190 Early medieval genomes from Gretzinger et al. 2022 [5]

To further study the composition and temporal dynamics of the main ancestry components of the Sint-Truiden population, we used qpAdm [28]. As the PCA placed the majority of the Sint-Truiden genomes between genomes from Late Iron Age (LIA) France [22] and Early medieval (EMA) genomes from the Netherlands [5], we used these sources as proxies for Gaulish and Germanic ancestry, respectively. We found that people buried in Sint-Truiden had predominantly Gaulish ancestry (63% on average) since the Early Middle Ages with a minor (37%) Germanic component and that this ancestry composition has been retained for centuries largely unchanged, up to the population of present-day Limburg province of Flanders (Fig. 3, Additional file 1: Table S2). The higher Gaulish ancestry, ranging from 52 to 69% in different provinces of present-day Flanders, is part of a broader northeast to southwest cline of decreasing Germanic ancestry in the Low Countries [11]. In contrast to the transect of time in West Flanders, where the EMA and HMA genomes had lower (< 40%) than currently (> 60%) observed proportion of Gaulish ancestry, we find high temporal stability of the average Gaulish ancestry (60–72%) in the Sint-Truiden time series since the EMA period. Although we observe little change in the average ancestry composition, we find, consistent with the results of PCA, higher (p = 0.043, F-test) individual variance of ancestry in the Early/High Middle Ages than in the Late/Post-medieval period (Additional file 1: Table S3, Fig. 3). These results are also supported by analyses of supervised ADMIXTURE (p = 0.004, F-test), which reveal somewhat higher overall Gaulish ancestry (0.68 on average) across Sint-Truiden time groups than by qpAdm (Additional file 1: Table S2), and suggest that while the main ancestry sources were already in place in Sint-Truiden in EMA, the admixture process continued locally until at least the later medieval period. Considering the wide range of individual ancestry proportions in modern Flemish genomes (Fig. 3), this admixture process is likely still ongoing.

Fig. 3.

qpAdm based Gaulish and Germanic ancestry estimates in Flemish and Dutch provinces. A Presented Gaulish (blue) and Germanic (orange) ancestry proportions represent the best fit qpAdm model (Additional file 1: Table S3) for each population tested. B Violin plot comparing Gaulish ancestry proportions between Groenmarkt and Trudoplein burials. C Violin plot comparing Gaulish ancestry proportions between early (pre-1286 AD) and late (post-1286 AD) phase burials in Groenmarkt. Light gray background in panels A–C refers to early phase burials, predating 1286, and dark gray background to later phase burials postdating 1286. D Map showing average Gaulish (blue) and Germanic (orange) ancestry proportions by the provinces of the Netherlands and Flanders of Belgium. West-Flanders and Limburg provinces of present-day Flanders for which we present a transect of time data are highlighted with a darker yellow background. EMA—Early Middle Ages, HMA—High Middle Ages, LMA—Late Middle Ages, PMA—post-Middle-Ages, MOD—modern controls from the MinE project data

Two-way qpAdm admixture analyses of Sint-Truiden burials revealed a higher average proportion of Gaulish ancestry (71.9%, Fig. 3, Additional file 1: Table S2) in individuals buried in the Trudoplein near the St Trudo’s Abbey than in contemporary (time groups I and II) burials from the Groenmarkt (p = 0.029, 2-tailed t-test), supported also by the supervised ADMIXTURE results (p = 0.0006, 2-tailed t-test). Within Groenmarkt, we observed no difference in ancestry proportions between the burial zones or time groups (Figs. 1 and 3).

To test if the five Sint-Truiden samples that clustered with Irish and Scottish genomes on the PCA share allele frequency affinity specifically with any ancient or modern population, we applied outgroup f3-statistics on the five outliers and compared the results with the rest of the Sint-Truiden genomes (Additional file 1: Table S4). We found that the five outliers share more drift than the main group with Early Medieval Ireland and Viking Age Orkney than the remaining Sint-Truiden individuals. We found that both Viking Age Orkney and Early Medieval Ireland [5] can be used as ancestry proxies to model the outlier group, both with a two-way admixture model with Early Medieval (EMA) Netherlands as the alternative ancestry source. In a one-way admixture model scenario, with either Viking Age Orkney or Early Medieval Ireland as the only ancestry source, we could model the group as sharing 100% ancestry with Viking Age Orkney only. Importantly, such assignment does not indicate a direct link with the Viking Age Orkney population specifically, but confirms the placement within the PC showing similarity with genomes from Scotland/Ireland, and little to no Gaulish ancestry. At the individual level, all five outliers could be modeled as 100% derived from an ancient source in Scotland/Ireland, two of these outliers could also be described with 100% Germanic ancestry (Additional file 1: Table S5).

Analyses of mtDNA and Y chromosome haplogroups showed, similarly to autosomal PCA and qpAdm results, the predominance of west European ancestry components in Sint-Truiden (Additional file 1: Tables S1, S6-S7). Two individuals (ST1016 and ST2420) with mtDNA haplogroups typical to African populations (L2a and M1a) had no autosomal evidence of African ancestry and the observed mitochondrial sub-clades (L2a1k, formerly known as L2a1a, and M1a3a) appear to be found at low frequency today in Central and Southern Europe [29], reflecting likely prehistoric gene flow from Africa to Europe. We found no evidence (p = 0.38, 1-tailed t-test) of increased Germanic autosomal ancestry among 40 Sint-Truiden individuals with mtDNA haplogroups observed more commonly in Germanic source areas (match with > 2 individuals from EMA England, Netherlands, and Scandinavia) than in Celtic areas (match with < 2 individuals from Bronze/Iron Age France, Netherlands, or Britain, Additional file 1: Table S6). The basal Y-haplogroup frequencies of EMA and HMA Sint-Truiden were similar to those of populations from surrounding regions and have remained relatively unchanged in modern Flemish population [30], with no specific outliers indicative of inter-continental migration (Additional file 1: Table S7, Additional file 2: Fig. S3). All four male outliers identified with PCA belonged to the R1b2-L21 (R1b1a1b1a1a2c1a) clade (Additional file 2: Fig. S4) that was uniquely frequent in Bronze and Iron Age Britain [31], being still the most common haplogroup found in present-day Scotland and Ireland. As the four outliers belong to distinct subclades of R1b2-L21 they are not closely related to each other in their patrilines.

A previous study by Byrne et al. [11] had observed high levels of genomic differentiation between regional groups in the present-day Netherlands, some of the genetic distances between local groups being greater than differences between distant regional populations of present-day Europe. To assess regional continuity over time and compare the extent of modern population stratification with the extent of regional genetic differentiation in medieval Flanders, we measured genetic distances between the ancient and modern populations from the Low Countries and West Europe with Fst and found that the genetic distance between the Sint-Truiden population and the population of Merovingian Koksijde in the province of West Flanders in the first millennium (EMA period) as well as that of a Merovingian site in North Rhine-Westphalia in Germany were nearly as high as the distance between modern Belgian Limburg and Spain (Fig. 4), while the distance with the early medieval populations of The Netherlands and England was smaller. However, by the HMA period, the Fst differences between Sint-Truiden and the two high medieval populations from West Flanders had become similarly low and comparable to the distances among the modern populations of all Flemish provinces. The genetic distances between present-day Limburg and the HMA West Flanders, on the one hand, and the Sint-Truiden population sampled at different time points, on the other, remained low over time, suggesting that the unusually high allele frequency differentiation in the EMA period is likely due to the atypical ancestry profile of the Merovingian Koksijde community [6].

Fig. 4.

Genetic distances between the ancient and modern populations of the region. Genetic distances are presented as Fst × 1000000. Sint-Truiden time groups I = 675–999 AD (~ EMA), II = 1000–1286 (~ HMA), III/IV = 1287–1775, LIA = Late Iron Age, EMA = Early Middle Ages, HMA = High Middle Ages. BE-Limburg = Belgian Limburg, NL-Limburg = Dutch Limburg. Data sources: Sint-Truiden, this study; Koksijde, Sasso et al. [6]; West-Flanders, this study and Sasso et al. [6]; Early Medieval Netherlands, England and North Rhine-Westphalia, Germany – Gretzinger et al. 2022 [5]; Late Iron Age (LIA) France, Fischer et al. 2022 [22]; present-day Belgian and Dutch Limburg, the MiNE Consortium data [26, 27]; France, England and Spain, the UK Biobank data [32]

To further study the spatial and temporal dynamics of regional population structure in the Low Countries in the last 2000 years, we estimated the probabilities of inter-individual sharing of > 7 cM IBD segments for Sint-Truiden and reference genomes with IBIS [33]. We found contrasting patterns of IBD sharing for the Sint-Truiden main group of individuals and the five outliers identified with PCA (Fig. 5A). The main group shows the highest affinity to present-day individuals from the Belgian provinces of Limburg, Antwerp, and Flemish Brabant and to Early Medieval genomes from Germany, the Netherlands, and England. The outliers, in contrast, show on average low sharing probabilities with these groups and instead show the highest affinity to Early Medieval and modern populations from Scotland and Ireland. Notably, the average probability of finding individual pairs sharing > 7 cM IBD segments among the Sint-Truiden main group of burials (0.053) is lower than in sets of MinE consortium [26, 27] controls whose four grandparents were born in present-day Limburg (0.105) or Flemish Brabant (0.076) (Fig. 5A). This suggests genetic heterogeneity of the main group of burials beyond the identified five outliers. The comparative analysis of IBD sharing probabilities between genomes from present-day provinces of the Netherlands and Belgium on the one hand, and ancient genomes sampled at different time points in West Flanders and Limburg provinces of Belgium on the other (Fig. 5B) revealed that EMA genomes from Koksijde and Sint-Truiden had only broad geographic affinity to present-day Low Countries and that the finer regional province-specific patterns of interindividual connectedness emerged from HMA onwards, with the post-medieval (time group IV) population of Sint-Truiden showing clearly the highest IBD sharing probability with the present-day genomes from its surrounding Belgian Limburg province.

Fig. 5.

Probability of individual connectedness (PiC) with modern and ancient populations. A Heatmap of probabilities (× 1000) of between individual sharing of > 7 cM segments within and among populations. Ancient population data (shown red) include data for Sint-Truiden and Ypres from this study, Early medieval (EMA) data from Gretzinger et al. 2022 [5], Sasso et al. [6], Late Iron Age (LIA) genomes from Fischer et al. 2022 [22], Viking Age (VA) data from Margaryan et al. 2020 [8], and Roman period data from Scheib et al. 2023 [23], Martiniano et al. 2016 [25] and Schiffels et al. 2016 [24]. Modern reference data include populations from the UK Biobank [32] and MinE Project data [26, 27]. ST-main – 318 genomes from the Sint-Truiden main group; ST-out—five outliers identified by PCA (Fig. 2). B. PiC scores estimated between present-day provinces of the Netherlands and Belgium, MinE project data [26, 27] and transect of time data from West Flanders ([6], this study) and Sint-Truiden (this study). Red stars on each map indicate the geographic location of the archeological sites with data

Genetic relatedness analysis

High rates of genetic relatedness in a sample can influence the results of population genetic analyses but they can also be informative of demographic and social aspects of the burial population. Bigger communities or those with high rates of immigration would be expected to show lower rates of relatedness than local communities with small population size. High proportion of inter-individual relationships in cemeteries can be expected if the burial ground was used by the same small community continuously over a period of time. To determine the extent of genetic relatedness among the Sint-Truiden burials, we first performed a screening of closely related pairs in 372 genomes with coverage > 0.01 × with KIN [34] and READ2 [35]. Among the 70,790 individual pairs with more than 30,000 overlapping SNPs, we detected 17 cases of 1st–3rd degree relatedness (Additional file 1: Table S8), including six pairs with 1st degree relationships and five with 2nd degree relationships. The average probability of 1st–3rd degree relatedness (0.00024) of the entire set of Sint-Truiden burials is 79-fold lower (p < 0.00001) than the rate observed in medieval parish cemeteries in Cambridge [9]. A similarly low rate (0.00023) is observed among the 12,936 pairs of the biggest subset of time group II (1000–1286 AD) burials from the Groenmarkt. We find, however, notable differences in the rates of close degree relationships between locations and time periods: in the early phase (before 1286, time groups I and II), Trudoplein burials show a more than tenfold higher rate than Groenmarkt burials of 1st–3rd degree relationships, and time group III and IV burials in Groenmarkt show a more than fivefold higher rate than earlier burials in the same location (Fig. 6). These results of relatively higher rates of genetic relatedness in Trudoplein and later burials of Groenmarkt than among the earlier burials of Groenmarkt are further corroborated by the screens of cases of 1st–6th degree relatedness (Fig. 6 right) with IBIS [33]. We observed a 13-fold higher rate of 4–6th degree relationships among Trudoplein than among Groenmarkt burials. The low probability of 4–6th degree relatedness in time group I and II burials of Groenmarkt (0.0005) is comparable to the rate (0.0003) observed in 183 MinE consortium controls sampled across Belgium while being lower than the rate within individual provinces. The 99 individuals with at least one 1st–6th degree relationship show higher (p = 0.005, 2-tailed t-test) proportion of Gaulish ancestry (0.68) than individuals with no relationship detected (0.60). This difference, however, is linked with the higher Gaulish ancestry in the Trudoplein group as the proportion of Gaulish ancestry in Groenmarkt individuals (0.65) is not significantly (p = 0.11) higher than observed in Groenmarkt individuals with no relationships (0.60).

Fig. 6.

Probabilities to observe genetic relatedness between Sint-Truiden burials. The probabilities of 1st–3rd degree relatedness are expressed as the ratio of observed relationships in the burial place and time with KIN and READ2 (Additional file 1: Table S8) and the total number of individual pairs with minimum aggregate coverage to detect them. The probability of 4–6th degree relatedness is expressed as the ratio of observed relationships with IBIS (Additional file 1: Table S9) and the total number of pairs of imputed individual genomes available from the burial place and time for the analyses

Low rate of genetic genetic relatedness among Groenmarkt burials could reflect either large effective population size or large catchment area of rural to urban migration potentially including migrants from sources with low effective population size. To distinguish between these possibilities we screened the Sint-Truiden genomes for evidence of runs of homozygosity (ROH), which are indicative of parental relatedness, and tested the relationship between heterozygosity at variants with MAF > 0.05 and genetic ancestry. Using the pseudo-haploid model of hapROH [36], we find high correlation (r = 0.99991) between imputed and non-imputed pseudo-haplodized results for ROH segments exceeding 8 cM (Additional file 2: Fig. S5, Additional file 1: Table S10), which encourages us to use imputation-based results with more individuals. Our ROH data provides evidence for 14 individuals with ROH > 8 cM, interestingly all of which were buried in Groenmarkt. Overall, the prevalence of individuals with ROH > 8 cM tracks in Sint-Truiden (~ 4%) is similar to the ~ 5% rate observed in late medieval Cambridge [9]. None of the individuals in our dataset showed evidence of “long ROH,” the sum of ROH > 20 cM lengths exceeding 50 cM [36], a threshold typically indicative of inbreeding between first or second cousins. However, we do find two individuals (ST1186 and ST1233) with ROH > 20 cM, suggesting parental relatedness at the level of fourth cousins or closer [37]. Notably, one of these individuals (ST1233) is among the five outliers with Scottish or Irish ancestry. We find a weakly positive correlation between heterozygosity and Gaulish ancestry, which is stronger (r = 0.28) in individuals from Trudoplein (Additional file 2: Fig. S6). In combination with the differences in genetic relatedness probabilities, these results suggest that the catchment area of the Groenmarkt burials was different from Trudoplein burials, likely including different external communities.

The observed low rate of relatedness is notable also in light of the presence of at least 80 multiple burials, one from Trudoplein and all others from the Groenmarkt, of which twelve graves included multiple individuals whose genome we had sequenced. Not a single case of 1st–7th degree relatedness was detected between individuals (including child–adult pairs) co-buried in 12 graves from which we had sampled multiple individuals (Additional file 1: Table S11) suggesting that time of death may have been more important than genetic relatedness in defining co-burials. On the other hand, we are likely to underestimate genetic relatedness, particularly in co-burials with children, given our low rate of success of retrieving DNA from the teeth of sub-adults. However, we do find a pair of 3rd degree related adult males (ST2000 and ST2228) with identical mitochondrial DNA buried in different graves with unrelated individuals in a co-burial context, both in zone 2 and layer 7 in Groenmarkt. Furthermore, six of the individuals from co-burials had a distant 4–6th degree relationship with individuals from single burials, including ST2000 who had four such relationships. While the higher concentration of genetic relatedness findings in Trudoplein could signify the importance of the proximity of the Abbey for selected groups of families of the city, the sporadic cases of genetic relatedness between but not within Groenmarkt co-burials further confirm that genetic relatedness did not play a primary role in determining where precisely in the cemetery individuals were buried.

To specify autosomally detected genetic relatedness relationships and to explore sex-specific differences in relatedness, we determined which pairs shared mtDNA and Y chromosome haplogroups. Analyses of mtDNA revealed 248 distinct haplogroups, including 171 detected in only a single individual, and 77 haplogroups shared by 214 individuals (Additional file 1: Tables S1, S6). Among the 332 individuals that were included in IBD analyses, only seven pairs of individuals out of 158 with shared mtDNA haplotype showed evidence of autosomal relatedness. In the majority of the cases (151 pairs, 95.6%) sharing of mtDNA haplotype was not indicative of 1st–6th degree of relatedness. Unsurprisingly, considering the decreasing likelihood to share either matrilineal or patrilineal relatives with increasing degree of relatedness, we found the majority of the matches, five in mtDNA and one in Y chromosome, among the nine cases of the first and second degree pairs (Additional file 1: Table S9). In contrast, among the 64 pairs of 3rd–6th degree, only two mtDNA and a single Y chromosome match was detected. We observed no sex difference (p = 0.22, 2-tailed t-test) in the number of relationships among the 99 individuals identified with 1st–6th degree relationships: 51 were male including 12 with multiple 1st–6th degree relationships in the data, while among 48 females we found 17 individuals with multiple relationships.

Changes in effective population size

Effective population size is often estimated from genetic data as a proxy of the number of individuals involved in reproduction, being widely used for inferences about the demographic, social, and cultural history of populations [38, 39]. To determine that of the Sint-Truiden city center population, we independently ran the software HapNe-LD [40] on each time group, removing closely related individuals and the five outliers identified by PCA (Methods). With its regularization mechanism, HapNe avoids spurious inferences of demographic fluctuations and favors smoother curves instead. As foreseen by the authors of the method in the case of insufficient demographic signal in the data, effective population size was thus always inferred to be constant (Additional file 2: Fig. S7). To get around this limitation, we plotted each time group’s mean Ne over the past 4 generations in order to obtain a broad overview of changes in effective population size between the eighth and eighteenth centuries (Fig. 7). We note an increase in effective population size from the time groups I to II, followed by a drop reaching its minimum with time group IV. This tendency is similar to the inferred historical Ne trajectory Byrne et al. [11] obtained using modern DNA from various Dutch provinces, including that of Limburg. Our results also corroborate the higher diversity noted in the earlier periods through PCA (Fig. 2) and qpAdm (Additional file 1: Table S3, Fig. 3).

Fig. 7.

Effective population size (Ne) over time. Effective population size (Ne) of the Sint-Truiden city center population between the eighth and the eighteenth century inferred by plotting the mean Ne estimated for each time group over its past 4 generations. The dark blue dotted line joins the different mean Ne estimates and the 95% confidence intervals obtained with bootstrap quantiles are represented by the light blue error bars

Health and phenotype-informative variation and metagenomic findings

The study of the Early Medieval site Koksijde in Belgium revealed major allele frequency differences between groups of Gaulish and Germanic ancestry in variants associated with pigmentation and dietary phenotypes [6]. Firstly, to gain broader insights into phenotype-related changes through time, we focused on a shortlist of 112 diet, immunity, and pigmentation-related variants including those previously highlighted as targets of selection (Additional file 1: Table S12). We compared the allele frequencies of these variants in Early/High medieval Sint-Truiden with modern Flemish genomes from the MinE dataset [26, 27]. Similarly to the results of Sasso et al. [6], none of the examined variants retained significance after multiple test corrections (Additional file 1: Tables S7, S12–S13). These results reflect the stability of average allele frequencies through time, observed in our genome-wide analyses (Fig. 3). While we found no evidence of major temporal changes in tested phenotype-informative markers, we observed significant differences related to ancestry. We found that individuals carrying at least one of the five red hair causing alleles in the MCR1 gene had higher proportion of Germanic ancestry, while individuals carrying the rs7944926-G allele in the DHCR7 gene, which is associated with higher levels of vitamin D precursor 25(OH)D3 (calcidiol) in the blood [41], had higher proportion of Gaulish ancestry (Additional file 2: Fig. S8). Notably, the increased 25(OH)D3 levels have also been observed in redheaded people [42].

Variation in immunity genes

The second pandemic of plague with its high mortality is likely to have affected allele frequencies of immune genes associated with infectious disease vulnerability. Reports of significant enrichment of high Fst variants among immunity genes in comparisons of pre- and post-Black Death cohorts [43] have been contested by subsequent studies [9, 44]. Our estimates of Fst between 239 early (time groups I and II) and 46 late (time groups III/IV and IV) phase imputed Sint-Truiden genomes revealed no enrichment of highly differentiated (F**ST > 99th percentile) variants related to innate immunity either before or after the LD pruning step (Additional file 1: Table S14). We observed a cluster of highly differentiated variants in the 22:22,481,208–22,725,114 region of the IGL locus behind the enrichment signal of highly differentiated variants related to adaptive immunity, which disappeared after the LD pruning step (Additional file 1: Table S14).

Genome-wide screens of Fst outliers

Despite not finding significant enrichment signals of immunity, we carried out a systematic screening of the genome for variants that most highly differentiated between the early and late phase genomes. Our genome-wide screens of 37,403 independent SNPs retained after LD pruning revealed 17 variants as Fst outliers in all comparisons made between the early (time group II) and later phase time groups (Additional file 2: Fig. S9). Eight of these variants map to genic regions and are all intronic variants (Additional file 1: Table S15). The variant with the highest average Fst among all groups is located in the gene PLCE1 (Additional file 2: Fig. S10), involved in T-cell migration to inflamed skin [45] and mediating macrophage activation [46]. The gene itself is associated with several phenotypes including leukocyte and monocyte counts [47].

Metagenomic analyses

The metagenomic screening of 372 Sint-Truiden genomes for reads matching human-associated microbial pathogens revealed probable findings in 35 individuals. In most cases, identifications are based on low levels of DNA and can thus not be fully validated (Additional file 1: Table S16). The findings included 10 individuals with likely cases of hepatitis B virus (HBV), five with Yersinia pestis, six with Borrelia recurrentis, three with human betaherpesvirus 6A (HHV-6A), one with human betaherpesvirus 6B (HHV-6B), three with Leptospira interrogans, and two with herpes simplex virus 1 (HSV-1) as well as eight other viral or bacterial findings in single individuals (Additional file 1: Table S16). One individual (ST1319) showed evidence for co-infections of Yersinia pestis and HBV. This is in line with previous studies reporting the presence of Y. pestis with a chronic infection [44, 45]. Capture-based analyses of individuals with Y. pestis reads failed to produce enough sequence coverage for strain-specific assignments in six individuals while in case of SK1516 two substitutions were detected that are not found in Black Death strains while present in all strains in the main Second Plague Pandemic lineage (Supplementary Information, Additional file 1: Tables S17-18, Additional file 2: Fig. S11).

Further, more detailed strain-level analyses of full HHV-6AB genomes findings will be presented in a separate study [46]. Radiocarbon dating of 34 individuals with pathogen findings showed a wide temporal range for most pathogens, except for Y. pestis findings, all of which dated to the fourteenth c. (Fig. 8). Two adult males with Y. pestis findings, buried apart in different burial zones of the Groenmarkt, were found to be second degree related and to share the same mtDNA haplogroup.

Fig. 8.

Temporal distribution of the metagenomic findings. Individuals with pathogen findings and radiocarbon dates are presented for the subset of pathogens detected at least twice (Additional file 1: Table S16). HBV—hepatitis B virus, HHV—human betaherpesviruses. The mean value of the 95% range of the radiocarbon date of each individual is shown on the y axis, x axis values showing δ15N values

Discussion

Homogenization during urbanization

Our analyses of genetic variation in Sint-Truiden population over a time transect spanning more than a thousand years revealed a process of homogenization of two distinct genetic ancestries during the intensive urbanization phase in the Low Countries. In the Early and High Middle Ages, the population of Sint-Truiden was more heterogeneous, with approximately 10% of the individuals clustering outside the range of the Late/Post medieval and modern genomes from Flanders in the PCA (Fig. 2). This indicates that the genetic diversity of the EMA/HMA population of Sint-Truiden was remarkably higher than the modern genomes from present-day Flanders, a pattern also observed in the EMA population of Koksijde [6]. Moreover, the genetic differences between Sint-Truiden and Koksijde during the EMA were larger than those between present-day Spain and Flanders (Fig. 4). Over time, the genetic variation decreased, and the ancient genomes increasingly resembled the present-day population in the region around Sint-Truiden (Fig. 5).

Contrary to what might be expected based on the abbey’s international connections, as noted in its chronicles [15, 48, 49], and the long-distance trade of products from Sint-Truiden, no long-distance migrants from regions outside Northwest Europe were identified among the genomes we studied. Most individuals in Sint-Truiden, based on their clustering with modern-day genomes from Flanders, likely had local origins in the region surrounding Sint-Truiden. This contrasts with findings from the late medieval Sint-Rombout’s parish cemetery in the city Mechelen, 54 km from Sint-Truiden, where several individuals within a small sample showed Mediterranean ancestry [50]. The only exception in our sample to the predominantly local ancestry profile typical to the Low Countries is a distinct group of five individuals with profiles similar to those from Ireland and Scotland (Figs. 2 and 5). Four of the outliers date to 1000–1286 and one to Early Middle Ages, and they were buried separately at the cemetery site. All male individuals belonged to the Y chromosome haplogroup R1b2-L21 clade, whose ancestry is uniquely related to the British Isles [31]. Despite two of the outliers being juveniles, none of the individuals shared close genetic relationships with each other or anyone else sampled from Sint-Truiden, and all four Y chromosomes belonged to different subclades, ruling out close patrilineal relatedness. It is plausible that these persons were connected to the Benedictine abbey, although the abbey’s chronicles make no mention of connections to Ireland or Scotland [15, 48, 49]. Therefore, they may have been specialized craftsmen or pilgrims during the abbey’s period of growth when a large church and other parts of the abbey were constructed [51].

Admixture of Gaulish versus Germanic ancestry

Amidst the higher inter-individual diversity observed in the EMA and HMA, we noticed temporal stability in the main ancestry components in Sint-Truiden over the span of a thousand years, with higher Gaulish and lower Germanic ancestry based on the qpAdm analysis (Fig. 3). These ancestry components were present from at least the Early Medieval period and did not show notable changes over time in Sint-Truiden and the surrounding province of Limburg. This suggests that the admixture of Germanic and Gaulish ancestries was a prolonged process over centuries, rather than an instant event. Consequently, the admixture dates, e.g., as estimated by Byrne et al. [11], could potentially be separated from the initial migration events that led to the geographic distribution of ancestry components by a considerable amount of time.

The prolonged nature of Germanic and Gaulish admixture is particularly evident in West-Flanders. Our results indicate that Germanic ancestry was the dominant component in EMA and HMA, while Gaulish ancestr