The mean pairwise genetic distance among PRRSV strains was considerable (11.6%), which is almost the double of the 6.5% genetic diversity obtained for 55 sequences submitted by 48 farms from the Midwest of United States . The maximum genetic distance also exceeded the 12% mentioned in another report using sequences from 62 individual farms belonging to a single pork-producing company in South Dakota . These latter studies were conducted on a small number of sites and it was not reported whether vaccine-like strains were included in their computations which could have contributed to lower the mean and extent of genetic diversity. Surprisingly, for the small geographical and temporal frames, our results were rather closed to the 12.5% average pairwise diversity obtained in phylogenetic analyses of 8624 ORF5 sequences gathered world-wide from the North American, Asian and European continents and over more than 15 years . The extent of genetic diversity certainly adds to the complexity of PRRSV management . However, its impact on PRRSV regional control feasibility is difficult to assess since previous initiatives did not report such diversity within their targeted zone [34–36].
A correlation between genetic distance and ownership was observed, which to the best of our knowledge has never been explored before. About 16 times more pairwise combinations of strains sharing ≥98% homology were observed for sites having the same owner compared to those having different owners (Table 2). Results were also corroborated by the Mantel tests showing that sites managed by same owner globally had more often ≥98% homology, even when adjusting for Euclidean distance (Tables 3 and 4). It suggests a common source of animals or semen [9–11]. Common employees or personnel from technical services could also convey the virus between sites on different fomites such as boots, coveralls or vehicles [15, 23, 24]. As the exact source of pigs for each site was not available, the variable ownership was our best asset to examine the correlations but it resulted in difficulties in identifying precisely the mechanism(s) involved behind the concept of ownership. For independent producers, multi-site production system designed for early segregated weaning  would fit the assumption of a common animal source. In contrast, the use of ownership oversimplified the pyramidal structure of integrated systems.
Similarly to other results , we did not observe any correlation between genetic and Euclidean distance with continuous matrices. However, a significant correlation was found with binary distances. As the Mantel test is based on a linear correlation, a nonlinear relationship may be lost using that technique . Therefore, a significant correlation limited to a particular scale might be diluted by inclusion of other distances leading to an overall absence of relationship. With binary distances, two to three times more combinations of strains having ≥98% homology were observed in pairs of sites located at ≤5 km from each other compared to others with distances >5 km, and this association was supported by the Mantel using binary distances (Tables 2
3 and 4). Area spread which is possibly involved behind this relationship represents between-herd transmission occurring without any pig contact, but rather through several pathways of transmission: aerosols (up to 9.1 km), houseflies (up to 2.3 km), mosquitoes and potentially, other mammalian or avian mechanical vectors [11, 20–22, 39–41]. Vehicles and inert fomites such as boots, coveralls or other equipments conveyed by people combined with absence of biosecurity measures could contribute as well to area spread of PRRSV [15, 17, 24]. Area spread has been frequently suspected among herds having similar strains in Quebec as well as in different states in United States [11, 25, 39]. Results from sensitivity analysis supported the 5 km threshold previously chosen; correlation between genetic and Euclidean distances decreasing rapidly and almost linearly up to 5 km, to become not significant thereafter (Figure 1). Virus survival might explain the decrease of the strength as greater Euclidean distances favour longer exposure to adverse environmental conditions. Vulnerability of PRRSV to ultraviolet light, level of humidity and high temperature is demonstrated [42–44].
The positive correlation observed between genetic and temporal distances is compatible with genetic evolution of the virus in the area over a two year period of observation and was also reported by others . Interestingly, when examining the temporal correlogram (Figure 2), the relationship was more linear than the one for genetic and Euclidean distances, which might explain that correlations were significant for both continuous and binary matrices of distances. According to the ability of the virus to mutate over time and considering that new viruses can be introduced into the area through animal sources or transportation, a certain level of evolution of the viruses present in the area is expected. The correlation between genetic and time decreased slowly and was observed up to 11 months, indicating that virus populations are changing over time but that certain viruses might persist on the territory over about 1 year. The time frame for which sequences were obtained was quite large. It could partially explain the low frequency of homologous PRRSV strains observed on the territory. Indeed, compared to short time interval between sampling time, large intervals could have lead to an increased possibility of new virus introduction or to mutation of the virus within herds. Because of the cross-sectional design, the sampling time did not necessarily correspond to the moment of virus introduction on the site and consequently the real temporal process of transmission between farms could not be assessed.
This study attempted to include all sites located in a restricted geographical scale to obtain more precision in exploring area spread compared to other studies [25, 26]. However, some producers did not want to participate in the survey whereas others could not be contacted. The fairly good participation rate (77%) improved the internal validity of the study, but total absence of selection bias was impossible to assess. Moreover, underestimation of our participation is possible as some unreached producers might have been out of business. Even if the sampling strategy used on sites with PRRS history was performed in order to maximize the probability of identifying a PRRSV strain, it was not possible to identify a sequence for 25% of the PRRSV positive sites. Whether the exclusion of the latter sites influences the estimate of genetic diversity cannot be determined. Furthermore, only one PRRSV strain was identified per site, assuming the existence of a sole viral strain or at least obtaining the predominant strain. Although not frequently reported, more than one strain can co-exist on a site [11, 45]. Analyses were performed using ORF5 gene based on its high genomic variability and its widely use for molecular epidemiology studies on North American PRRSV strains or in studies assessing the extent of genetic diversity [25, 46, 47]. However, ORF5 gene is only a part of the whole 15 000 kb genome  and it could be interesting to compare results when using the entire genome rather than just ORF5 for this kind of analyses. Also, more than 50% of strains gathered in this restricted geographic area did not show any homology of ≥98% with other strains. These findings suggest that additional investigations using different geographical boundaries would allow the study of other mechanisms of transmission acting on a larger scale and involving human intervention such as pig transportation.