Performing Department
Population Health and Pathobiology
Non Technical Summary
This proposal leverages innovative phylodynamic methods to elucidate the influence of environmental factors and animal movement on the spread of PRRSv. Thus, the main goal of this project is to improve our epidemiological understanding of PRRSv, improve outbreak responses and guide efforts towards better-informed disease control strategies. The overall impact of our project will be in enhancing specific regional capacities for the control of PRRSv spread among all farm types, optimizing PRRSv management interventions.
Animal Health Component
100%
Research Effort Categories
Basic
100%
Applied
(N/A)
Developmental
(N/A)
Goals / Objectives
Porcine reproductive and respiratory syndrome virus (PRRSv) causes one of the most economically relevant endemic swine diseases in the U.S. (Newman et al., 2005; Pileri and Mateu 2016), costing approximately $1 billion/year (Holtkamp et al., 2012; 2013). The high mutation rate of PRRSv hinders the effectivity of the pigs' immune system and the differentiation between resident and newly introduced variants, especially in farms with recurrent outbreaks. We propose to develop phylogenetic models to identify factors facilitating local PRRSv spread and investigate the influence of new virus introduction in the overall incidence in North Carolina. Specific Aim-I. Identify the current circulating PRRSv strains in North Carolina and reconstruct their evolutionary history. We hypothesize that some regions present higher variability of PRRSv strains, forming clusters around key farms. We will evaluate this hypothesis through Bayesian phylodynamic methods that integrate genetic and geographic data at herd-level. This aim will map hotspot areas of PRRSv and will characterize how its spread occurs. Specific Aim-II. Determine the most influential factors for PRRSv spread. The hypothesis for this aim is that PRRSv spreads mainly via animal movements. We will assess this hypothesis by combining ecology and evolution to identify factors (e.g. biosecurity, environment, animal movement) which may facilitate or restrict PRRSv spread between farms. We will visualize these results by ranking these factors by its level of importance. Specific Aim-III. Determine the contribution of local transmission versus external virus introductions to the overall incidence of PRRSv. We hypothesize that the introduction of novel virus variants to local populations plays a key role in sustaining high PRRSv incidence and the occurrence of outbreaks. To achieve this aim, we will apply a phylodynamic approach developed by our team (Rasmussen 2017, Rasmussen et al., 2018). We will track the movement of viral lineages between populations to quantify the contribution of local transmission versus external (new PRSSv variants) introduction to the overall PRRSv incidence.
Project Methods
We will use data from the MSHMP, which is an academic-industry partnership in which participating swine production companies voluntarily report changes in PRRSv status of breeding herds on a weekly basis, as well as other epidemiologically relevant metadata (see Dr. Corzo-MSHMP letter of support). Data will be input into a comprehensive genetic and geographic database that gathers information from about 90% of the swine farms in the state of North Carolina. That data is composed of PRRSv genetic information (ORF5 sequences), of all PRRSv outbreaks from 2015 until 2018, farm type, between farms animal movement (pigs and batches), location, herd size and information on whether these farms commingle animals (farm receives groups of pigs from several farms). The environmental/landscape factors will include: vegetation, land cover, wetlands area, global relief (altitude), global topography (slope), average solar radiation, average precipitation and temperature, topsoil pH, major soil group classification, mean wind speed (Arruda et al., 2017; Machado et al., 2019). Most of the environmental and climatic information are freely available at a daily or weekly time resolution; we will average all information to the month to match with the phylogenic bit of this project. All algorithms used to automatic download and prepare these data will be available at the PI's GitHub repository. In addition, we will also include hog density as potential factors influencing PRRSv spread, this data is also freely available (visit http://www.fao.org/livestock-systems/en/). The original layers (environmental factors) will consist of 0.62 mi2 grid resolution to allow the assessment of land surface biophysical patterns at high-resolution. Additionally, data on feral pig occurrences, their geographic location and PRRSv status (serology) will be available from the USDA-APHIS-Wildlife Services (see letter of support). Finally, we will combine the genetic sequence data with farm-level information, environmental/climatic data and feral pigs information into one database. We will then develop the proposed multistage research project through three key aims. B. Phylogenetic relationship between PRRSv strains, tree is colored according to different farm types, C. Most influential factors on PRRSv spread, each factor was considered as a conductance (C) or resistance (R) variable depending on the positive or negative effect that it exhibited on PRRSv spread, pie chart represents the level of contribution of each factor and D. PRRSv phylodynamic model, representing the total incidence (gray) and incidence attributable to external introductions.Aim 1: Identify the current circulating PRRSv strains in North Carolina and reconstruct their evolutionary history.Phylogeographic analysis: All PRRSv sequences will be aligned using Mega X, available at www.megasoftware.net/ (Kumar et al., 2008). Phylogenetic trees will be reconstructed via Bayesian inference, simultaneously with the ancestral location of each lineage in BEAST v2.5.0 (Bouckaert et al., 2014).Spatiotemporal epidemiological statistics: Using the R package 'SERAPHIM' (Dellicour, et al., 2016), we will extract the start and end geographic location of each branch in the tree, which constitutes its spatiotemporal information. This will provide views of the spatial movement of the lineages along with the start and end dates, allowing us to compute PRRSv diffusion rates (Dellicour et al., 2018; Faria et al., 2018). Dispersal velocity will be visualized. In this analysis, we will account for the time (in months) in which each specific lineage has moved from its epidemic origin to its current local spread. The information of each lineage with its associated locations will be also used to assess the invasion velocity of each strain, via the calculation of the linear diffusion distance, described by Pybus et al., (2012). Finally, to visualize the spatiotemporal diffusion of each PRRSv lineage, we will use Spatial Phylogenetic Reconstruction of Evolutionary Dynamics using Data-Driven Documents (D3) SPREAD3 software (Bielejec et al., 2016).Aim 2: To determine the most influential factors for PRRSv spread.Drivers of PRRSv spread: We will determine the most important factors associated with the presence of PRRSv (see data collection section). To assess the effect of different factors in the spread of PRRSv, we will use the analytical framework developed by Dellicour et al. (2016). The statistical significance between the phylogenetic and all potential factors: biosecurity (here represented by farm type, e.g. finisher farms are considered to have low biosecurity and sow farms are considered to have strong biosecurity), farm level information (herd size, commingle pigs), environment, animal movement, and feral pigs data, will be tested using 100 trees generated before (aim 1) and expressed in the form of Bayes Factors (BF) (Dellicour et al., 2016). We will approximate a BF value for each combination of data set, path model, and potential factors. For the interpretation of BF values, i) all potential factors will be treated either as a conductance (variables that promote the spread of the disease) or a resistance factor (variables that impede its spread) and ii) we will determine the magnitude of impact of each factor on the spread of PRRSv by following Jeffreys index (Jeffreys, 1961), where BF values between 3.6-10 are considered as "substantial", values between 10-31.62 "strong" and between 31.62-100 "very strong". This analysis will be performed using R package SERAPHIM (Dellicour et al., 2016).Aim 3: Determine the contribution of local transmission versus external virus introductions to the overall incidence of PRRSv.We will apply a phylodynamic method developed by our team member Rasmussen et al. (2018) to quantify the contribution of external virus introductions versus local PRRSv transmission. This model will allow us to quantify the fraction of new infections attributable to local transmission versus external virus introductions from other geographic locations. First, we will group all farms located within a radius of 10 km, which will be considered "local populations", this radius is based in previous literature that indicated the possibility of airborne spread of PRRSv (Dee, S. et al. 2009), and all farms outside this range will be classified as "external populations". Within the local population, we will apply a SIR-type epidemiological model where the number of susceptible (S), infected (I) or removed (R) individuals can change over time due to changes in risk, treatment or other factors (Rasmussen et al., 2018). The sources of transmission affecting the studied population will be classified as i) transmission from infected farms within the local population and ii) transmission from an individual coming from an external infected population. These methods have been implemented in BEAST 2 (Bouckaert et al. 2014) as an add-on package named Marula (Rasmussen 2017, freely available at https://github.com/davidrasm/Marula. These analyses will allow us to estimate both the overall epidemic dynamics and the incidence attributable to local farms versus external introductions of PRRSv.