We use the large scale population database of HIV sequences maintained by the UK Collaborative Group on HIV Drug Resistance to estimate the patterns of HIV transmission among different communities in the UK. The structure of the sexual contact network is a key issue in the epidemiology of sexually transmitted infections. As HIV is only transmitted with low efficiency compared to many STIs, the transmission network structure is more readily reconstructed from the viral genotypes than from interview data. Using the approach of molecular phylodynamics to analyse anonymized HIV genotypes from MSM (men who have sex with men) in a London clinic, we originally found that 25% of patients with a link to any other were linked to 6 or more individuals. In these clusters, almost 25% of transmissions occurred within 6 months of first infection (Lewis et al. 2008).
Moving to HIV-1 subtypes A and C, which in the UK are predominantly associated with heterosexual transmission, the picture was quite different. Large clusters were far less frequent and there was very little evidence of transmission in acute infection (Hughes et al. 2009).
The early analyses were labour intensive and to accommodate increasing dataset size, Sam Lycett and Manon Ragonnet automated the process by developing the Cluster Picker tool (Ragonnet et al, 2013). Applying this tool, Manon has found that the few large clusters that were previously found in these subtypes have arisen through "crossover" of these strains into MSM (Ragonnet-Cronin et al. 2015).
Our collaborator Stephane Hue at UCL has recently led on another joint project which applied the Cluster Picker to investigate heterosexual transmission of subtype B in the UK. This has led to the important discovery that among HIV-positive individuals who identified as heterosexual and were clustered with MSM, 2/3 were male: amounting to 6% of all heterosexual males with this subtype. This misidentification of risk group could only have been revealed by the phylogenetic analysis performed here (Hue et al. 2014).
We have extended this work by using the phylodynamic approach to estimate the parameters of the network structure within which HIV is spreading among MSM, exploring the well-known "power law" effect in greater detail (Leigh Brown et al. 2011). We have found that the distribution of cluster size ("degree distribution") is such that a randomly distributed intervention would never stop the epidemic. This level of detailed knowledge can provide important insights into delivery of interventions such as pre-exposure prophylaxis.
Our group has had a long association with the MRC/UVRI Uganda Research Unit on AIDS, which has consolidated in recent years with a collaborative publication bringing forward the estimated age of the epidemic in that country (Yebra et al. 2015). Now, together with UCL, Imperial College, the Sanger Institute and the Africa Centre we are partners in an exciting initiative "Pangea_HIV" to characterise epidemics in sub-Saharan Africa in the same depth as we have been able to for the UK. This is particularly relevant as the approaches used to quantify transmission patterns from analysis of viral sequence data have the potential to reveal the impact of interventions. There are 4 partner sites in the Pangea-HIV consortium, 2 in Uganda (MRC/UVRI and Rakai), one in Botswana, and one at the Africa Centre in KwaZulu Natal, and the programme is funded by the Bill and Melinda Gates Foundation.
One of the main goals of PANGEA_HIV is to use phylogenetic and molecular epidemiology techniques to better characterise HIV epidemics in sub-Saharan Africa. In order to evaluate the performance of current phylogenetic analyses at estimating epidemiological parameters, a comparison exercise was devised, based on computer simulations of HIV evolution within epidemics.
Using two separate models to simulate realistic HIV epidemics in an African setting, phylogenetic and sequence data was simulated according to a variety of different parameters, such as the number of infections imported from surrounding villages, the infectiousness during the acute stage, and the speed of treatment roll-out. This dataset was made available for research groups to analyse using their chosen method, and the resulting estimates were compared against the true values from the simulation.
The Leigh Brown group's current focus as part of PANGEA_HIV has been to provide one set of simulated HIV data for the comparison exercise. Samantha Lycett's stochastic, agent based model, the Discrete Spatial Phylo
Simulator (DSPS), has been extensively modified by Emma Hodcroft to enable it to simulated realistic
HIV epidemics (DSPS-HIV). The model calculates disease progression and transmission risk based on viral
load, population growth has been incorporated, and contact networks and treatment are highly customizable.
The DSPS-HIV was used to generate thirteen 'village-like' datasets for the comparison exercise. Five teams from three countries participated in the exercise and submitted estimates of incidence, change in incidence, infectiousness during the acute stage, and change in infectiousness during the acute stage over differing levels of sampling coverage, imported infections, and treatment roll-out speed, which were presented at the international 22nd HIV Dynamics & Evolution conference in Budapest in May 2015.
Datasets generated with the DSPS-HIV have also been used by Manon Ragonnet to evaluate the performance of the ClusterPicker tool at selecting closely related 'clusters' of HIV sequences, which may represent transmission chains.
There has been significant recent debate about the role of the viral genome in determining the rate of progression to AIDS and death. This has been studied using plasma viral load which provides a convenient and robust surrogate marker, leading to the claim that the "heritability" of virulence is high. Emma Hodcroft has been addressing this question using the sequences collected through the UK Collaborative Group on HIV Drug Resistance together with linked viral load data collected through the UK Collaborative HIV Cohort Study (UK CHIC). Emma has exploited the expertise on quantitative genetics available in Edinburgh to apply a novel approach to the question, allowing her to simultaneously analyse the genetic contribution of over 8000 viral genotypes to plasma viral load based on the relationships revealed by phylogenetic analysis of their sequences. This work, published in PLoS Pathogens (Hodcroft et al, 2014) showed that in fact viral genotype contributes relatively little to the variation in plasma viral load among infected individuals.
Since publishing her study Emma took part in the 2014 "3 Minute Thesis" challenge, and won the Edinburgh University competition from which she went on to take part in the UK national final and the Universitas 21 international final. Emma's 3 Minute Thesis presentation can be seen here.
||Manon and Emma hard at work
a. Evolutionary, antigenic and epidemiological dynamics of human influenza virus
Influenza viruses are another fast evolving group which results in their ability to continually adapt to existing host immunity.
Thus influenza has been one of the most consistent global burdens over the last century.
Influenza viruses have genes which are arranged on 8 RNA molecules (“segments”).
However, the exact characteristics of each type of segment varies according to the mutations it carries,
and different combinations (reassortments) of segment types yield influenza strains with different pathogenicities, transmissibilities and drug resistances.
We exploit the large body of full-genome influenza sequences now available using statistical phylogenetic tools on high performance computational platforms. In this way, Lu Lu has determined the origin of a highly pathogenic strain of avian influenza which caused multiple outbreaks in poultry, with some human cases, in Mexico in 2012-13. She found this strain arose following several segment exchanges among viruses from wild birds, including some from different N. American flyways (routes taken by migrating birds), previously thought not to exchange viruses frequently. The closest wild strain to the pathogenic Mexican viruses was isolated earlier from a green-winged teal (Lu et al. 2014b)
We have therefore also been interested in quantifying the rates and patterns of reassortment events, especially in swine and avian influenza (Lu et al. 2014), as well as detecting and characterising intra and inter segment interactions.
This work was funded by the Wellcome Trust.
b. Pandemic Influenza
The 2009 H1N1 pandemic strain was identified in Mexico in February 2009 and first detected in Scotland in late April. It was significantly less pathogenic than the previous pandemic (and, unlike Asian 'flu, never featured in a pop song). As is typical, it disappeared in summer and reappeared in autumn.
It was unknown whether initial cases in the summer arose from multiple imports of the virus from other parts of the UK or elsewhere,
and if cases in the winter represented persistent Scottish lineages or new imports of the virus into Scotland.
In this project we obtained viral genome sequences of scottish isolates from a representative set of samples and analysed them
with additional UK and worldwide sequences to infer the pattern of transmission between countries and within the UK.
Using Bayesian Phylogeography we were able to detect different epidemic patterns in the summer and winter phase and showed the epidemic in Scotland spread from England, despite several local introductions from overseas (Lycett et al. 2012).
This work was supported by Scotland’s Chief Scientist Office,
the University of Edinburgh Interdisciplinary Centre for Human and Avian Influenza Research
and the BBSRC.
c. Swine Influenza in Swine
Swine are often considered as a mixing vessel for different influenza strains, since
they may become infected with viruses with avian-like or human-like sialic acid binding receptors.
The co-circulation in pigs of a North American H3N2 strain (itself a reassortant between avian, human and swine viruses)
with an Eurasian avian-like swine H1N1 strain,
led to the production of a novel reassortant virus (presumably in swine) which caused the 2009 human pandemic
(Smith et al 2009).
As part of the Combating Swine Influenza Initiative consortium
(under the lead of Ian H. Brown, Veterinary Laboratories Agency-Weybridge, UK,
and James Wood, University of Cambridge, UK),
we and our collaborators obtained whole genome sequences from a retrospective swine influenza surveillance study.
The aim of this work was to characterise the reassortants present in European swine and
estimate the rate of reassortment of subtypes in swine using time resolved bayesian phylogenetics and it was published in 2012 (Lycett et al. 2012)