The Newfoundland and Labrador mosaic founder population descends from an Irish and British diaspora from 300 years ago

Abstract

The founder population of Newfoundland and Labrador (NL) is a unique genetic resource, in part due to its geographic and cultural isolation, where historical records describe a migration of European settlers, primarily from Ireland and England, to NL in the 18th and 19th centuries. Whilst its historical isolation, and increased prevalence of certain monogenic disorders are well appreciated, details of the fine-scale genetic structure and ancestry of the population are lacking. Understanding the genetic origins and background of functional, disease causing, genetic variants would aid genetic mapping efforts in the Province. Here, we leverage dense genome-wide SNP data on 1,807 NL individuals to reveal fine-scale genetic structure in NL that is clustered around coastal communities and correlated with Christian denomination. We show that the majority of NL European ancestry can be traced back to the south-east and south-west of Ireland and England, respectively. We date a substantial population size bottleneck approximately 10-15 generations ago in NL, associated with increased haplotype sharing and autozygosity. Our results reveal insights into the population history of NL and demonstrate evidence of a population conducive to further genetic studies and biomarker discovery.

Fig. 1. Genetic structure of Newfoundland and Labrador.

a Dendrogram of the fineSTRUCTURE final MAP state, showing 22 summarising clusters of NL membership. Solid branches indicate the 22 cluster branches, with grey branch-lines showing merged clusters within each of the 22 clusters. fineSTRUCTURE clusters are colour and shaped coded to reflect grouping on adjacent branches, with labels reflecting common geographic birthplace of members’ grandparents. Cluster sizes shown to the right of cluster labels. b The proportions of genetic ancestry groups and religious background in each of the 22 clusters, shown in the order of A. c The second versus the third principal components of the ChromoPainter co-ancestry matrix, with individual colour and shape coded according to fineSTRUCTURE cluster membership. d The second versus the third principal components of the ChromoPainter co-ancestry matrix, with individual points colour coded to religious background; red indicating Christian Catholic, blue indicating Christian Protestant, and grey indicating other or unknown. All panels were plotted using the statistical computing language R and the packages ggplot2 and rworldxtra.

Fig. 2. Genetic landscape of Newfoundland and Labrador.

Map of the grandparental birthplaces of individuals with colour and shape coded according to fineSTRUCTURE cluster. A small jitter has been introduced to aid legibility and preserve anonymity. An insert shows individual details of the Trinity and Conception Bays. Panel was plotted within photoshop, with geography boundary data sourced from Tableau.

Fig. 3. Irish and British ancestry in Newfoundland and Labrador.

a Map of Irish and British individuals placed according to recent mean ancestral birthplace, colour and shape coded according to IBD-network clustering. Colours are used to group hierarchically related clusters together (see “Methods”). A slight transparency for individual points has been introduced proportional to the maximal proportion that each cluster contribute to any one NL cluster, with great contributors having less transparency. b The summed proportions of the estimated ancestry proportions calculated from Irish or British IBD-segments shared with NL. Contributions from individual IBD clusters are summed together in groups according to Irish or British region. c Individual Irish or British IBD cluster contributions to NL ancestry proportions, showing only clusters that contribute >5% to any one NL cluster. All panels were plotted using the statistical computing language R and the packages ggplot2 and rworldxtra.

Fig. 4. Evidence of bottleneck with Newfoundland and Labrador.

a The estimated historical N_e for Irish and British regions using IBD-segments and the IBDNe tool. Shading shows the 95% confidence intervals. b The estimated historical N_e for NL clusters >100 individuals in membership size. Shading indicates 95% confidence intervals. c The individual mean total length of Identity-by-Descent (IBD) segments shared with another individual placed in the same cluster, versus the mean number of IBD segments shared with individuals placed in the same cluster. Clusters are colour and shaped coded in the same system as Fig. 1. Irish and British clusters are plotted separately from NL clusters for legibility, and grey points indicate non-Irish and British (or non-NL) clusters. Error bars show the 95% confidence intervals. d The average total length of ROH within each NL or Irish and British cluster. The mean total length of ROH > 1.5 Mb for each cluster is shown by a hollow white circle (left y axis), or the proportion of the genome in ROH > 1.5 Mb in length (F_ROH—right y axis). Also shown are average total lengths of ROH in four length bins: (1) 4 cM ≤ROH < 8 cM, (2) 8 cM ≤ROH < 12 cM, (3) 12 cM ≤ROH < 20 cM, (4) 20 cM ≤ROH < 300 cM. Groups of clusters are coloured together, and error bars shows 95% confidence intervals. All panels were plotted using the statistical computing language R and the package ggplot2.

Fig. 5. Evidence of NL specific genetic drift from Irish and English source.

a Patterson’s D statistic testing the allele sharing between NL clusters and English and Irish population references. Positive values indicate excess sharing within NL clusters to Ireland, and negative indicate excess sharing to England. YRI indicates the outgroup for these tests (Yoruban sampled in the 1000 Genomes Project Phase 3). b Testing NL-specific drift with Patterson’s D statistic. X-axis tests if each NL cluster presents an excess of Irish (negative D value) or NL (positive D value) alleles, and Y axis tests if each NL cluster presents an excess of English (negative D value) or NL (positive D value) alleles. All panels were plotted using the statistical computing language R and the package ggplot2, with error bars showing 1 standard error.