Lorcan Carnegie: Phylodynamic approaches to studying avian influenza virus
These viruses belong to the species Alphainfluenzavirus influenzae (previously known as Influenza A virus (IAV)) (Lefkowitz et al., Citation2018; ICTV, Citation2022), and have single-stranded, negative-sense, and eight-segmented RNA genomes (Seiler et al., Citation2018; Rimi et al., Citation2019; Wille & Holmes, Citation2020). AIVs are subtyped into “HxNy” based on the antigenicity and genetic diversity of the two surface glycoproteins: haemagglutinin (H1 – H16 in birds) and neuraminidase (N1 – N9 in birds) (Yoon et al., Citation2014; Blaurock et al., Citation2020; Verhagen, Eriksson, et al., Citation2021). Wild aquatic birds, particularly Anseriformes (e.g. geese, ducks) and Charadriiformes (e.g. gulls, shorebirds), are the primary reservoirs of AIVs (Olsen et al., Citation2006). However, AIVs can spill over to cause sporadic infection or sustained transmission within domestic avian hosts (Mostafa et al., Citation2018; Lycett et al., Citation2019).
We can categorize viruses as low pathogenic avian influenza viruses (LPAIVs) and highly pathogenic avian influenza viruses (HPAIVs) based on their pathogenicity in chickens and the presence of insertions in the HA cleavage site (More et al., Citation2017). LPAIVs cause asymptomatic infection or mild disease in domestic birds, thereby harming the poultry industry via decreased egg or meat production, higher vaccination expenses, and trade restrictions (Busani et al., Citation2007; Gonzales et al., Citation2021; Ripa et al., Citation2021).
Only LPAIVs of H5 and H7 subtypes are known to evolve into HPAIVs, a process that involves the insertion of multiple basic amino acids in the HA cleavage site (Rott, Citation1992; Alexander, Citation2007). HPAIVs cause severe disease and fatalities in both domestic birds (Verhagen et al., Citation2014; Nuñez & Ross, Citation2019) and wild birds, respectively harming the economy and conservation efforts (Kleyheeg et al., Citation2017; Banyard et al., Citation2022; Lean et al., Citation2022). Several AIV strains can also infect humans and other mammalian species (e.g. swine, seals) and, thus, pose a potential pandemic threat (Ren et al., Citation2016; Nuñez & Ross, Citation2019; Blagodatski et al., Citation2021; Agüero et al., Citation2023; Puryear et al., Citation2023; Vreman et al., Citation2023).
Phylodynamics studies how epidemiological, immunological, and evolutionary processes shape viral genetic diversity. Approaches developed within this framework can help recover viral dispersal patterns and evolutionary processes, even when virus genomic data is sampled relatively sparsely from an infected population (Grenfell et al., Citation2004; Volz et al., Citation2013; Rife et al., Citation2017). Time-scaled phylogenies can be inferred using molecular clock models, which quantify the rate of genetic change over time and therefore enable phylogenetic branch lengths to be expressed as units of time rather than as nucleotide substitutions per site (Drummond et al., Citation2006; Pybus & Rambaut, Citation2009) (Box 1). RNA viruses, including AIVs, typically have exceptionally short generation times, high evolutionary rates, and large population sizes (Duffy, Citation2018; Wille & Holmes, Citation2020). Consequently, genetic substitutions in viral genomes often occur on similar time scales as transmission events between hosts. Hence, it is possible to reconstruct outbreak dynamics from time-scaled phylogenies, as they contain a “molecular footprint” of viral spread (Grenfell et al., Citation2004; Lemey et al., Citation2009; Pybus & Rambaut, Citation2009). When genome sampling location is available, we can use phylodynamic techniques to reconstruct the geographical distribution of viral lineages (“phylogeography”), thereby revealing valuable information about viral spread and factors associated with faster or more frequent viral lineage movement events (Lemey et al., Citation2010; Faria et al., Citation2011; Gill et al., Citation2016).
The most common tools for phylodynamic analyses employ a Bayesian Markov Chain Monte Carlo (MCMC) framework to efficiently explore highly complex models involving many different parameters (Drummond & Rambaut, Citation2007; Bouckaert et al., Citation2014). A Bayesian framework has several advantages compared to the maximum likelihood or parsimony-based approaches. Firstly, Bayesian approaches allow for the incorporation of multiple sources of data or prior knowledge (e.g. divergence times, substitution rates) (Alfaro & Holder, Citation2006; Baele et al., Citation2017; Chakraborty et al., Citation2021). Perhaps more importantly, such approaches generate posterior distributions of phylogenetic trees, thus allowing uncertainty in parameter estimates to be captured (this has been reviewed extensively elsewhere, (e.g. Faria et al., Citation2011; Volz et al., Citation2013; Gill et al., Citation2016; Rasmussen & Grünwald, Citation2020; Dellicour et al., Citation2021). Maximum likelihood (ML)-based methods are more limited in scope of possible analyses, but are often less computationally intensive than the more popular Bayesian approaches (Baele et al., Citation2018; Sagulenko et al., Citation2018; Ishikawa et al., Citation2019). This can be beneficial when dealing with large datasets, limited computed resources, or when faster but lower complexity models are appropriate to inform emergency responses. Such ML phylodynamic methods typically use a single ML tree, enabling faster time-to-answer compared to Bayesian phylodynamic inference.
Recent trending decreases in the cost and time required to generate and analyse virus genetic data have led to rapid innovation within the field of phylodynamics and a subsequent increase in popularity (Gill et al., Citation2016; Rife et al., Citation2017; Grubaugh et al., Citation2019; Cardona-Ospina et al., Citation2021). This review highlights how phylodynamic methods have and may continue to aid the study of AIV spatiotemporal dispersal. We first explore recent studies that use phylodynamics to infer AIV dispersal dynamics within or between wild birds and domestic poultry populations at various geographical scales and discuss factors that can complicate the generation of reliable conclusions. Specifically, we discuss the inference of factors associated with AIV transmission, the order and timing of transmission and lineage dispersal events during an outbreak, and how viral lineages can move between different regions and sectors of poultry production systems. We then consider future challenges and opportunities for using phylodynamic approaches within AIV research.