This study has produced original data on PRRSV transmission and dynamics of infection which increases our understanding of PRRS epidemiology in pig populations. The biological results obtained in inoculated pigs showed that the kinetics of blood viral genome load was consistent with descriptions available in the literature . After a rapid increase in genome load, the peak was reached two weeks after infection followed by a gradual decrease until 90 dpi. In our study, a further increase of viral genome load was observed at 56 dpi, which has not been previously described in the literature, to the best of our knowledge. However, as there were no statistically significant differences between the viral genome loads at 42, 49 and 56 dpi, these results mainly suggest that there was still a significant genome load at 56 dpi. This observation merits confirmation in further investigations over a similar period using more animals. However, the results from the contact groups CD56 and CD63 suggested that this supposed new increase in virus activity was not sufficient for the inoculated pigs to infect other animals. In addition, the amount of shed viral genome (nasal swabs) at that time was very low. In our study, the viral genome load shed by inoculated piglets increased very rapidly, in agreement with previous studies , reaching a maximum at 2 dpi and then decreasing steadily until 48 dpi when it was negligible.
Seroconversion was observed in some pigs at 7 dpi and a week later in all animals, which was consistent with previous research conducted on this virus [32–34]. The amount of antibody increased until 56 dpi and subsequently stabilized. In contrast to other studies in which a late appearance of neutralizing antibodies (NAbs) was observed [33–35], NAbs were detected as early as the second week after inoculation in some pigs and in all pigs at 21 dpi (except for one that did not have any neutralizing antibodies before 28 dpi). In another study, neutralizing antibodies were only detected from 56 dpi onwards . The neutralizing antibody titer was relatively low (maximum titer in log10 of 1.8 at 77 dpi) in agreement with previous findings . No significant difference was observed between the IC and IWC groups for the virological and serological parameters studied. Hence, the stress due to the successive contact trials with different animals did not affect the serological response of inoculated pigs or promote the replication of virus and viral shedding.
The principle of this study was based on 7 successive contact trials between groups of susceptible and inoculated pigs at different times post-inoculation. This approach differs from other experiments on PRRS virus transmission which all used the final size algorithm to estimate transmission parameters [36, 37]. Most of them gave disappointing results, particularly when the duration of the experiment was too short compared to the actual duration of the infectious process or when all the contact animals were found infected at the end of the trial. In our study, the number of primary infected animals increased up to 87.5% (susceptible animals infected at 14 dpi) then decreased and no contact animal was found infected at 42 dpi. Since a primary infection took place at 28 dpi, running a contact trial at 35 dpi would have probably been informative. The results would certainly have helped to determine more precisely the date when pigs no longer shed infectious particles in a sufficient quantity to infect susceptible individuals, especially as the results of PCR on nasal swabs did not indicate a total termination of viral genome shedding before 48 dpi. It cannot be excluded that a primary infection could have occurred in a contact at 35 dpi. Various data (results from serology, viral genome shedding, viral genome load in blood and gross lesions observed at necropsy) were analyzed to differentiate primary from secondary and tertiary infections. However, the most informative results were the analyses of nasal swabs collected every 3 days which provided the most precise date from which the pigs started to shed virus. Indeed, the first blood sample was taken only 10 days after the first day of contact and failed to differentiate primary from secondary infections in groups CD7 and CD14, based on blood genome load or serological results, since all the animals already tested positive at that sampling time. Because PRRSV has a very short incubation period, it is not sure whether additional blood samples at 3 dpi and 7 dpi would have been able to discriminate between primary and secondary infections. Some authors detected the appearance of genome load in blood at 3 or 7 dpi but only in some pigs [38, 39] and seroconversion usually occurs between 5 and 14 dpi [32–34]. In addition, taking blood samples very frequently increases the risk of hematoma at the collection site, which is not ethically acceptable. Since all inoculated animals shed viral genome at 2 dpi and the contact rate between inoculated and contact pigs could reasonably be assumed to be uniform, it was assumed that primary infections corresponded only to animals shedding the virus from their third day post-contact. The only method that would have allowed us to be absolutely certain of the number of primary infections in the groups would have involved separating individual animals from each contact group after the contact. This was not feasible given the number of animals used for the study. Each room contained two pens of four pigs with a polyethylene separation between the pens. However, the group CD28 had only one primary infection (in one pen) and pigs that were not located in that pen still ended up being infected. This suggests the possibility of cross-contamination between pens that were 40 cm apart and did not allow the pigs to have direct contact.
By adapting the algorithm used by Andraud et al.  to the numbers of primary infections in each contact group, we were able to estimate the infectiousness of PRRSV infected animals with time since infection. We obtained a distribution of infectiousness with an average duration of infectiousness of 14.8 days, peak infectivity at 9 dpi and a negligible probability of transmission beyond 48 dpi. These results suggest that the infectious period was shorter than that obtained by other authors who reported up to 56 dpi  or 62 dpi . The major differences compared to the above studies are due to detailed characterization of the evolution of this infectious potential with time, which definitely cannot be considered constant. Even though a viral genome load could be detected beyond 48 dpi, the results from contact trials (which are the direct evidence of virus transmission) showed that pigs were not able to infect others at that time. These estimates are specific to the strain we used (genotype I, subtype 1), and significant differences might be expected with strains having different pathogenic characteristics and/or a different host immune response, such as genotype II strains or even other subtypes of genotype I strains . The correlations established between virological parameters and transmission characteristics could however be used to assess the transmission features of other strains according to virological and immunological data. Even if this strain is representative of our regional field situation, differences might be expected in other areas where a larger panel of strains can be involved. In addition, further work should also be carried out to assess the impact of multiple infections by different strains on transmission.
In our study, the rapid increase in infectiousness from 1 to 9 dpi was related to the increase of blood viral genome load, and to a lesser extent with the shed viral genome load which started earlier. Infectiousness was better correlated with viral genome load in the blood, over the entire period of the study, than with the shed genome load (corr = 0.82 versus 0.59 respectively). However when considering only the decreasing part of the infectiousness curve, the correlation was stronger with the viral genome shed in nasal swabs which ended simultaneously to transmission termination. Several assumptions can be made about the observed delay between viral genome shedding and significant increase of the infectious potential. First, it is possible that infected animals shed defective interfering viral particles at the early times (spontaneous generation of mutant viruses in which an essential part of the viral genome had been removed, making it non-functional) . Another hypothesis is that the infectiousness of the shed viral particles differs over time with particles shed at 9 dpi being more infectious than those shed at 2 dpi. It is also possible that at the peak of infectiousness, the virus was shed by other routes in addition to nasal secretions such as feces, saliva or urine . The decrease in infectiousness at 9 dpi also corresponds to the time when the amount of total antibody and neutralizing antibody increased. The decrease in infectiousness was slightly less correlated with the amount of neutralizing antibodies than with the amount of total antibodies (corr = −0.47 compared with −0.53 respectively). However, this might be due to the fewer dates when neutralizing antibodies could be titrated. The role of neutralizing antibodies in protection against PRRSV is controversial. Even if a link between the development of neutralizing antibodies and viral clearance has been described , most studies have shown that the protection provided by neutralizing antibodies after experimental infection is very low and irregular with great individual variability [32, 33]. The results from our study suggest that neutralizing antibodies could help to control infectiousness even though a significant viral genome load persisted in the blood in their presence. Further work has to be done to assess the contribution of the cell-mediated immune response on infectiousness control.
The estimated R0 in this study is a theoretical parameter, indicating how the virus might spread in a fully susceptible and large population. Given the structure of pig farms, the results now need to be applied to a model describing population structure and dynamics (variability between individuals, contact structures, culling and replacement of the animals). The estimated basic reproduction ratio of PRRSV was 2.6, which is quite low compared to the R0 estimates available for other viral infections in pigs (100 for Classical Swine Fever virus , 5.9 for the PCV2 virus , 8.8 for hepatitis E ). The relatively small R0 value for PRRSV together with the relatively long duration of infectiousness can be explained by the huge variations in transmission rate with time-since-infection, as shown in the present study. However, the value obtained is significantly higher than 1, which suggests that the virus is gradually but efficiently spread and maintained in a structured population.