Spatiotemporal Reconstruction of the Introduction of Hepatitis C Virus into Scotland and Its Subsequent Regional Transmission

DISCUSSION

Bayesian MCMC methods were used to reconstruct the introduction of HCV Gt1a and Gt3a extant sequences into Scotland from the global pool using the Bayesian skyline plot (BSP) coalescent model (16) which allows for variable population sizes over time. A Gt1a global data set containing 506 sequences, the largest study of its type thus far, was used to calculate the evolutionary rate of this genotype (1.88 ×10⁻³ s/s/y), which was higher than previously determined. Calculations of evolutionary rates can be adversely affected by analyzing small numbers of sequences, and lower evolutionary rates (1.48 × 10⁻³, 1.0 × 10⁻³, and 9.05 × 10⁻⁴ s/s/y) have been calculated in studies with smaller data sets (n = 334 [17], n = 111 [18], and n = 176 [19], respectively). Our results suggest that Gt1a coalesced in 1888 (1857 to 1914), which confirms a previous report (1900 [1802 to 1957]) (18), but predates another comparable study (1920 [19]).

Gt1a lineage expansion principally occurred between 1940 and 1965, and subsequent sequence clustering displayed geographical specificity. This time period coincides with the mass population mixing of World War II, the subsequent return of individuals to their countries of origin, and the first large-scale parenteral treatments. It can be speculated that these global-scale events were instrumental in the worldwide mixing and spread of a large pool of strains which then disseminated locally as a consequence of commonplace parenteral treatment and IDU of the 1960s. Examination of the Scottish data set within the global context also supports this theory. The corresponding coalescence dates of the Scottish and global data sets, the intercalation of Scottish Gt1a clusters among RoW sequences, and the similarity in dates between the MRCA of the Scottish clusters and global lineage expansion suggests that extant lineages were introduced into Scotland from multiple independent sources concurrently with the diversification and dissemination of HCV worldwide.

Subsequent expansion of a proportion of lineages occurred within Scotland, and there is no evidence of onward transmission or new introductions with the exception of the Aberdeen-specific clade C4, which was first detected in this study in the 1990s. Interestingly, the MRCA of C4 with the RoW pool dates to 1928, predating the emergence of other Scottish clusters from the global pool; it may have remained unobserved earlier due to the small sample numbers from Aberdeen, or more likely, this strain originates from a country not included in the RoW data set. The lack of onward transmission of this strain to other Scottish regions suggests that Aberdeen may operate within a different network, perhaps linked more closely to international HCV transmission networks. Although it is a small city, Aberdeen is a major oil-producing center with consequent international population links, including connections to regions where information on HCV is sparse, such as central Asia.

The RoW data set for Gt3a consisted of only 44 sequences. Although more NS5B sequences with data on sampling date and geographic origin were available from international databases, these sequences were short in length (approximately 300 bp) and/or only partially overlapped the genome region used in the current study. As short sequences can adversely affect the calculation of important parameters in MCMC analyses, they were not included in this analysis. The evolutionary rate of the Gt3a sequences from the global data set was 1.65 × 10⁻³ s/s/y which is slightly higher than previously noted (1.3 × 10⁻³ s/s/y [20]). These sequences coalesced in 1899 (1865 to 1932) similar to previous studies (1905 [1851 to 1932] [21]; 1920s [22]), but another study has estimated that extant global Gt3a sequences coalesced much earlier, approximately 300 years ago (20). In the latter study, however, three sequences from Pakistan formed a distinct cluster which was phylogenetically distant from the remainder of the Pakistani sequences and sequences from other countries; without this outlying cluster, the remaining global Gt3a pool coalesces approximately 80 years ago, supporting our findings.

Although robust contextualization of the Scottish Gt3a sequences within the global setting was not possible due to the excess of Scottish compared to RoW sequences within the data set, the Scottish sequences did however form distinct clusters. Gt3a lineage expansion occurred in two periods of exponential growth, peaking in 1940 and 1960. The earlier time period parallels the MRCAs of the Scottish Gt3a clusters (1926 to 1942) and suggests that lineage expansion during World War II was responsible for the introduction of HCV Gt3a into Scotland, similar to Gt1a. The second peak of Gt3a lineage expansion (1951 to 1975) suggests that migration from the Indian subcontinent may be the source of currently circulating Gt3a sequences in Scotland. Approximately 1% of the Scottish population originates from Pakistan, where epidemic Gt3a transmission has been shown to have occurred earlier than in any other country (22). Extant Gt3a strains in the United Kingdom are considered to originate from the Indian subcontinent where HCV is endemic (23) via migration which peaked in the 1950s and 1960s, and subsequently, these sequences progressed into and expanded within the PWID community (20). Resolving which of the two time periods led to the extant pool of Scottish Gt3a strains would require an increase in global Gt3a sequences, particularly from regions of the world where HCV is endemic.

The transmission of HCV Gt1a and Gt3a within Scotland's PWID community was reconstructed. Since the predominant strains differed in the different Scottish regions, it is likely that they represent separate introductory events from the global pool. An alternative hypothesis for the regional specificity of HCV strains is the variability in the genetic makeup of the host population in Scotland, as revealed by a recent study which shows separate genetic clustering of individuals from Aberdeenshire and WCS (24). The earliest HCV sequences in the study were from individuals living in Glasgow (Gt3a) or Edinburgh (Gt1a), suggesting that these two major cities constituted the initial hubs of infection. Although index Gt1a sequences occurred in Edinburgh, they were mainly transmitted within that region alone; in contrast, widespread Gt1a transmission in Glasgow occurred 10 years later but constituted the dominant cluster that subsequently spread to surrounding regions. Taken together, these findings suggest that Glasgow is the key region driving HCV transmission in Scotland. An analysis of HCV infection hot spots in Glasgow was performed at a finer geographical precision than previously attempted. A disproportional amount of sequences were derived from individuals residing in a few PC sectors in a cluster-specific manner, suggesting that these regions represent key individual networks that play a central role in HCV transmission within the city. Targeting intervention and treatment initiatives to these regions could aid their overall effectiveness.

It is curious that despite the cocirculation of both genotypes within Glasgow, Scotland, coinfection with multiple HCV genotypes is relatively rare in Scotland (25), with an overall rate of 4.1% in the current study. This is supported by a recent modeling study (26) suggesting that HCV reinfections frequently result in spontaneous clearance and by empirical data from a number of studies showing reinfection rates in PWID of 2 to 9% (27–30). Other studies have suggested that reinfection is much more common in PWID (20 to 39% [31–33]), and differences may be due to needle exchange programs in the regions studied or sensitivities of the assays used. If reinfection is a rare occurrence (26), it is unlikely that sequences from individuals infected before the nationwide dispersal of strains would be supplanted by a recently imported strain from a different region, and newly infected individuals would become the main source of such introduced strains. Since the 1990s, only a few new lineages have been imported from the global pool that have undergone onward transmission, notably the Aberdeen-specific Gt1a cluster C4. It is interesting to speculate whether the continued geographical restriction of this clade is due to a later introduction into Scotland subsequent to the full implementation of clean needle strategies or to separate networks operating in Aberdeen.

The combination of epidemiology data and molecular phylogenetics has been previously used to construct injecting social networks for the investigation of HCV transmission dynamics (10). Researchers established an association between reported injecting relationship and HCV phylogeny but not between genetic and social distance. A disadvantage of this study type is that transmission inference is based on phylogenetic data derived from strains of long-standing infection combined with information on current injecting partners, similar to the contemporaneous residential data used in the current study. Other investigators have taken phylogenetic clustering outcome as direct evidence of individual membership in a contact network (34–38).

Spatiotemporal analysis based on combined phylogenetic clustering data, EYIs, and residential PC information for each sample was used in this study. There are limitations associated with utilizing these data for time-correlated phylogeographic reconstruction of chronic viruses such as HCV. First, diagnosis normally occurs many years subsequent to infection. To overcome this drawback, we calculated the average age individuals commenced IDU and surmised HCV infection within the first year of injecting. Studies have suggested that new initiates to injecting are particularly vulnerable to HCV infection, seroconverting in approximately 4 months (39). As well as new infections, PWID may experience multiple exposures to HCV, with the possibility that a new strain may infect and supplant the original strain, thus affecting the EYI. Recent modeling studies, however, suggest that HCV reinfection generally results in spontaneous clearance of the newly infecting strain (26), and this is supported by experimental data indicating a low incidence of coinfection with more than one HCV genotype (25). The second limitation arises from combining current residential information with historical EYI data. However, studies suggest that PWID in the United Kingdom are not a highly mobile group of individuals (41, 42). Nevertheless. it is likely that the discernment of transmission networks at a finer temporal and spatial resolution could be achieved by either the availability of residential data from the EYI or, alternatively, obtaining samples from newly infected individuals in concert with current residential information.

Although we observed a decade of increased transmission commencing in the mid-1980s, this observation may be subject to a degree of sample bias. The average age of HCV-infected individuals was 40 years, equivalent to an EYI in the mid-1990s. To enrich for samples from PWID, sequences from subjects with an EYI before 1989, the year widespread testing for HCV in blood products commenced in the United Kingdom, were excluded unless they specifically confirmed IDU; this has likely resulted in an augmentation of samples from patients born after 1969. Individuals born in the 1950s (EYI in the 1970s) are subject to naturally higher, age-associated mortality rates, whereas more recently acquired HCV strains may not be captured in this cohort of hospital-diagnosed individuals due to the extensive asymptomatic period of HCV infection. Previous epidemiological studies have, however, shown a similar pattern of peak HCV transmission occurring in Glasgow, Scotland, in the 1980s (4, 40).

Through this research, we reconstructed the dispersal of HCV into and throughout Scotland and successfully identified transmission hot spots that could enable health care and outreach workers to effectively target intervention and treatment as prevention strategies and monitor the effectiveness of these methods. The accuracy of this strategy could be further improved by increasing the sample size from individual cities and analyzing sequences from recently infected individuals with current residential data. Another tool that could help improve HCV transmission inference studies is next-generation sequencing (NGS) which has recently been investigated for analyzing small-scale transmission events and within-patient connections (43–45). Minor sequence variants in one individual may predominate in a different individual, as illustrated by studies of liver transplants in patients with HCV (46, 47), and NGS may enable the identification of connections involving minority variants which are not apparent using traditional sequencing techniques. We now intend to utilize NGS methodology to expand this study, investigating HCV transmission on a United Kingdom-wide basis.

In summary, heterochronous HCV sequences were used in this study to determine the timeline of introduction of extant Gt1a and Gt3a into Scotland from the global pool and to infer the spatiotemporal distribution of these sequences in the regional setting. Most Scottish sequences formed discrete clusters interspersed between RoW sequences, and the MRCAs of these Scottish clusters were dated to the period of global exponential lineage expansion. Although the origin of the two epidemic HCV genotypes in Scotland differed (Edinburgh for Gt1a and Glasgow for Gt3a), transmission to other regions occurred predominantly from Glasgow. Geographical specificity of transmitting clusters was apparent, particularly for Gt1a. In Glasgow, individual clusters formed different transmission zones initially, but these networks subsequently overlapped, suggesting that cocirculation of genotypes and clusters has occurred throughout the city since the 1990s. Hubs of infection were detected, and these hubs of infection likely play a key role in HCV transmission throughout the city; targeting intervention strategies to these regions could assist their effectiveness.