SARS-CoV-2 introductions to the island of Ireland: a phylogenetic and geospatiotemporal study of infection dynamics

Abstract

Background: Ireland's COVID-19 response combined extensive SARS-CoV-2 testing to estimate incidence, with whole genome sequencing (WGS) for genome surveillance. As an island with two political jurisdictions-Northern Ireland (NI) and Republic of Ireland (RoI)-and access to detailed passenger travel data, Ireland provides a unique setting to study virus introductions and evaluate public health measures. Using a substantial Irish genomic dataset alongside global data from GISAID, this study aimed to trace the introduction and spread of SARS-CoV-2 across the island.

Methods: We recursively searched for 29,518 SARS-CoV-2 genome sequences collected in Ireland from March 2020 to June 2022 within the global SARS-CoV-2 phylogenetic tree and identified clusters based on shared last common non-Irish ancestors. A maximum parsimony approach was used to assign a likely country of origin to each cluster. The geographic locations and collection dates of the samples in each introduction cluster were used to map the spread of the virus across Ireland. Downsampling was used to model the impact of varying levels of sequencing and normalisation for population permitted comparison between jurisdictions.

Results: Six periods spanning the early introductions and the emergence of Alpha, Delta, and Omicron variants were studied in detail. Among 4439 SARS-CoV-2 introductions to Ireland, 2535 originated in England, with additional cases largely from the rest of Great Britain, United States of America, and Northwestern Europe. Introduction clusters ranged in size from a single to thousands of cases. Introductions were concentrated in the densely populated Dublin and Belfast areas, with many clusters spreading islandwide. Genetic phylogeny was able to effectively trace localised transmission patterns. Introduction rates were similar in NI and RoI for most variants, except for Delta, which was more frequently introduced to NI.

Conclusions: Tracking individual introduction events enables detailed modelling of virus spread patterns and clearer assessment of the effectiveness of control measures. Stricter travel restrictions in RoI likely reduced Delta introductions but not infection rates, which were similar across jurisdictions. Local and global sequencing levels influence the information available from phylogenomic analyses and we describe an approach to assess the ability of a chosen WGS level to detect virus introductions.

Keywords: COVID-19; Genomic surveillance; Phylogenomics; Public health; SARS-CoV-2; Transmission dynamics; Variant tracking; Viral introductions; Viral phylogeography; Whole genome sequencing (WGS).

Fig. 1

Schematic phylogeny. An abstract representation of a phylogeny illustrating an Irish introduction cluster as analysed in this study. The ancestral node to the introduction cluster is emphasised with a red square (indicating inferred origin from England) and is marked with an ‘A’. An internal node of ambiguous origin is depicted upstream of the ancestral node to the introduction cluster (‘A’), to which it could not be discerned whether the internal node descends from Scotland or England (see blue and red square)

Fig. 2

SARS-CoV-2 timeline in Northern Ireland and Republic of Ireland. A Seven day rolling daily mean of confirmed SARS-CoV-2 cases per 100,000 population. Data from coronavirus (COVID-19) in the UK dashboard and Ireland’s COVID-19 Data Hub. B Monthly histogram by collection date of SARS-CoV-2 genome sequences deposited to GISAID for Northern Ireland (NI) and Republic of Ireland (RoI). C Pango lineage of genome sequences deposited in GISAID for NI and RoI. Major lineages highlighted, all other lineages grouped under ‘other’. D Analysis periods used in this study to identify SARS-CoV-2 introductions to the island of Ireland. E Maritime and aviation passenger arrival volumes to NI and RoI per month. Lockdowns in Ireland are represented by three periods (L1-3) that occurred from late March to mid May 2020, late October to December 2020, and late December 2020 to February 2021, respectively. Although there were slight variations in the particular measures implemented between NI and RoI, these dates were selected to demonstrate a general agreement on lockdown periods between the two regions. The starts ($⊳$) and the ends (|) of the periods studied are indicated

Fig. 3

SARS-CoV-2 introductions. Left: for each period (A–F), a world map depicting the ratio of introductions to Ireland given the proportion of samples in the global phylogeny. World maps were generated with the public domain Natural Earth map dataset [122] using the GeoPandas v0.11.1 library [109, 110]. Note that the colour bar scales differ across the time periods examined. Furthermore, countries displayed in grey indicate absence in the GISAID-derived global phylogeny. Relative sequencing enrichment is defined: $\text{Enrichment}=\frac{\text{proportion of introductions to Ireland}}{\text{proportion of tips in the global phylogeny}}$ Right: for each period (A–F), the proportion of introductions originating from each country

Fig. 4

Irish samples per introduction. Number of Irish samples per introduction grouped by cluster size 1–9+ for periods A–F. Bar height shows the proportion of total clusters for each cluster size within a period

Fig. 5

Effect of downsampling English samples. Ten replicates for each downsampled proportion. The black point is the number of observed introductions for each period using all data. Dashed grey lines represents a linear relationship between zero and observed introductions. The smoothed lines are fit to the downsampled data points using local polynomial regression fitting (loess). A Proportion of introductions to Ireland observed from England when English samples are randomly downsampled by 10% to 90%. B Proportion of introductions to Ireland observed from England when English samples are randomly downsampled and the proportion of samples in the phylogenetic tree that are from England. Three replicates for each downsampled proportion. Dashed grey line represents a linear relationship between 0% and 100%

Fig. 6

Effect of downsampling Irish samples (A and B), corresponding extrapolation to estimate total introductions to Ireland (C and D), and population normalised introduction and sequencing rates for NI and RoI (E). A Number of introductions to Ireland observed when Irish samples are randomly downsampled by 10% to 90%. One replicate for each downsampled proportion. B Average cluster size of Irish introductions when Irish samples are randomly downsampled by 10% to 90%. C The numerical derivatives of the downsampled data (A) were assessed and fit to a straight line to approximate the rate additional sequencing will have on finding additional introduction clusters. Estimated additional introductions at regular intervals are indicated by the symbol ‘+’. D The data is extrapolated from C revealing predicted sequencing saturation levels, where additional sequencing is unlikely to detect novel introduction events. Data using extrapolation are indicated by the symbol ‘+’. E Comparison of the number of introductions to NI and RoI for each period studied. Downsampling was performed independently for samples from each jurisdiction. To account for the difference in total populations, both the sets of downsampled sequences and identified introductions were normalised per 100k population

Fig. 7

Bivariate sequential choropleth maps (9-class) depicting relationships between population density, deprivation, and SARS-CoV-2 introductions and spread for all local government districts in Ireland. A The association between population density (cyan palette) and deprivation (magenta palette) in Ireland. It is assumed that this relationship remains constant throughout the periods studied. B The correlations between population density (cyan palette) and the proportion of samples linked to the introduction and spread of SARS-CoV-2 (dark yellow palette). The three maps labelled i–iii correspond to periods D–F, respectively. C The correlations between deprivation (magenta palette) and the proportion of samples linked to the introduction and spread of SARS-CoV-2 (dark yellow palette). The three maps labelled i–iii correspond to periods D–F, respectively

Fig. 8

Geospatial tracking of the spread of samples from introduction events within Ireland. A A map of Ireland created with GeoPandas v0.11.1 ([109, 110]) using the Esri ’World Imagery’ basemap (Sources: Esri, DigitalGlobe, GeoEye, i-cubed, USDA FSA, USGS, AEX, Getmapping, Aerogrid, IGN, IGP, swisstopo, and the GIS User Community) as retrieved using contextily v1.2.0 [124]. Red points designate the locations of population centres in each government district with the size of each point scaled according to recent population estimates. Refer to Additional file 1: Tab. S1 for the corresponding metadata for region abbreviations and population centres. B-i Delta SARS-CoV-2 from Scotland begins a cluster in NI (Causeway Coast and Glens) on June 2, 2021. B-ii The cluster spreads to adjacent districts in NI by June 10, 2021. B-iii The cluster spreads to County Roscommon and County Dublin in RoI and adjacent districts in NI by June 18, 2021. B-iv The cluster reaches the rest of NI, County Donegal in RoI, and County Limerick in RoI by June 26, 2021. B-v Spread continues within NI and to additional regions in RoI by July 8, 2021. B-vi Spread across Ireland continues, affecting at least 960 individuals by July 31, 2021. C-i BA.1 SARS-CoV-2 from England is detected in County Dublin and County Kildare in RoI on November 30, 2022. C-ii The cluster spreads to County Offaly and County Louth by January 6, 2022. C-iii The cluster rapidly spreads to adjacent districts in RoI, two districts in NI, and County Mayo in RoI by December 10, 2021. C-iv The cluster continues spreading, reaching several districts in NI and RoI by December 15, 2021. C-v Infections linked to this cluster reach all districts in Ireland by January 17, 2022. C-vi Spread across Ireland continues until the last sequenced case on January 31, 2022, afflicting at least 1274 individuals over 62 days

Fig. 9

Population-wide SARS-CoV-2 genomic trends in Ireland. Genome coverage for all samples included were at least 99.9% upon alignment to the Wuhan-Hu-1 SARS-CoV-2 reference (GISAID: EPI_ISL_402125). A Observed substitutions from the Wuhan-Hu-1 SARS-CoV-2 reference per sequence. B Apparent substitution rates per major SARS-CoV-2 lineage related to each studied introduction period. See Additional file 1: Fig. S14 for a visualisation of each linear regression, and Additional file 1: Tab. S14 for the corresponding statistics. C Genome C content per sequence. D Genome T content per sequence