This study presented for the first time an extensive landscape genetic study carried out on a main vector species of interest for livestock diseases, C. obsoletus. Bayesian clustering analysis revealed a very weak genetic structure of C. obsoletus in France and a low level of inter-individual genetic differentiation was observed. We assessed the impact of a wide range of environmental factors on this pattern. Univariate and multivariate analyses highlighted an absence, or a weak impact of most of the tested environmental factors on inter-individual measures of pairwise genetic differentiation. No isolation-by-distance pattern was detected at the French scale with the set of inter-individual genetic distances used. However, an anisotropic isolation-by-distance analysis revealed a significant distance isolation on a North-South axis.
Important gene flows between C. obsoletus populations
The low overall inter-individual genetic dissimilarity highlighted by our study reflected the important level of gene flow for C. obsoletus and comes to reinforce what has been described previously for other Culicoides species. In particular, important level of gene flows have already been observed in France for C. imicola [29] and in Australia for C. brevitarsis [82]. Genetic studies on C. imicola in Europe revealed an important gene flow as reflected by the inference of two large genetic clusters: a “Central Mediterranean cluster” including Algeria, Sardinia, Corsica, Pyrénées-Orientales and Var French departments of France, and a “Western Mediterranean cluster” including Morocco, Spain, Portugal and Majorca [26]. In North America, similar results were also found in C. stellifer. No barriers to gene flow could be identified in this species in the southeast United States [83]. In addition, no isolation-by-distance has been observed at the French scale. Identifying a genetic structure pattern is a real challenge when inter-individual genetic dissimilarity is low. Although STRUCTURE performs well at low levels of population differentiation (0.02 < FST < 0.10) by using prior population and correlated allele frequency models [84], when the differentiation is weaker, such as in the case of highly dispersed organisms, the latter may encounter difficulties [85]. This underlines the importance of using complementary tools of visualization approaches like MAPI. In addition, the use of an individual approach avoids to artificially consider individuals grouped into populations solely due to the effect of the trapping sites [2, 86]. Although our method of analysis seems complementary and coherent to detect a genetic structure, C. obsoletus is not genetically or geographically structured at the French scale.
Livestock densities as potential drivers of C. obsoletus gene flow
It is undeniable that large gene flows homogenize the genetic diversity of Culicoides. This can be explained by active dispersal and host-seeking movements, as suggested by our results. While the unique contribution of all environmental factors tested was very small, the one of cattle density is the strongest we have detected. Host density as a conductance factor for Culicoides is obviously a factor consistent with the biology of the species. These results are in line with those obtained in landscape genetics studies of the BTV, which had identified distributions of cattle and sheep as a key factors in BTV dispersal [13]. In addition, previous studies showed that dairy cattle density is negatively correlated with BTV spread. Although paradoxical with the previous conclusions, this could be explained by the fact that dairy cattle are clustered around the milking parlour and move little, forming a fixed feed source, limiting the spread of Culicoides. On the contrary, beef cattle disperse much more in the pastures and by this phenomenon could encourage the diffusion of Culicoides in search of a source of blood [87]. In view of these converging results, of the marked preference of certain species of Culicoides vector for cattle [88], and of BTV emergence or reemergence events in cattle in the Netherlands in 2006 and in France in 2015 [89, 90], it seems crucial to closely monitor the surveillance of Culicoides in the vicinity of beef cattle farms [88].
In addition, it has been shown that BTV spread is facilitated at low elevation, up to 300 m [38, 87]. Altitude, an environmental factor tested in this study, failed to explain the inter-individual genetic differentiation of C. obsoletus. The altitude therefore does not act as a barrier between sampling sites located at relatively low altitude. It would thus be interesting to integrate in future studies sampling sites with a high altitude above 1,000 m, which is possible for C. obsoletus highlighting a large altitudinal range [91]. However, it is possible that the altitude does not directly impact the dispersal activity of Culicoides, but only the replication or viral infection of BTV due to low temperature. Culicoides obsoletus is extremely generalist and can take a blood meal on a wide range of hosts [92]. The study of phylogenetically close but ecologically very different species of Culicoides, such as C. chiopterus [93], might then allow the identification of very different dispersal patterns and an importance of bovine density, the latter being exclusively dependent on it for egg-laying [94].
Long-dispersal of C. obsoletus
The most genetically dissimilar individuals are mainly located in the southernmost populations of the sampling area. Multiple non-exclusive lines of argument might explain the significant anisotropic isolation-by-distance observed on the North/South axis in France.
First, the significant anisotropic isolation-by-distance could be explained by the wind dispersal. Dispersion phenomena caused by wind currents are already established for Culicoides, mainly over the seas [3, 18, 21, 22, 23, 24, 25, 29]. The reflecting dispersal events can be described as passive and active because recapture beam has been carried out downwind and upwind of the prevailing wind direction [19]. The map of the average wind direction in France over the last 10 years shows potential answers. It can be seen that the southernmost sampling sites (with the most dissimilar individuals) are in an area where the wind direction is different from the rest of France. It should be noted that the diffusion of the BTV, and thus the dispersion of the Culicoides, has already been associated with the wind direction. For example, 2% of BTV infections occurred at distances greater than 31 km [95, 96] during the 2006 epizootic. The South of France was initially sampled, however a poor conservation of Culicoides DNA did not allow us to perform barcoding and microsatellite genotyping of these populations. The study of this geographical area could thus complete this study and potentially identify a stronger genetic structure. This will make it possible to decide whether this premise of differentiation observed in this study is due to an older phylogeographic structure, a different wind dispersion in this geographical area, or a random pattern. It is also questionable whether the very low indications of genetic structure could be due to differences in genetic diversity between the south and the north. A lower genetic diversity in the north could then be a sign of an expansion process that would contrast with the interpretation of a large-scale dispersion [97, 98]. Indeed the spatial expansion of populations is generally accompanied by a gradients of reduction in the genetic diversity of the population during the expansion process, caused by serial founder events, creating a genetic bottleneck through a founding effect [99, 100]. Population sampling at the same scale could allow comparison and estimation of genetic diversity.
Second, anisotropic isolation-by-distance may be due to an artifact due to sampling methods. Indeed, if the extent of the sampling varies depending on the directions, the distances projected from the angles may represent different distance distributions and lead to the over-representation of strong genetic metrics value in the direction of the scatterplot and the hit of positive correlation signals. Moreover, the absence of correlation does not necessarily mean more gene flow but an absence of isolation-by-distance which can also result from a strong drift and thus less gene flow, a drift which depends on both dispersion and population sizes. It is therefore essential to use as a complement, as we have done here and come up with a similar result, an approach that weights the geographical distances between populations according to their orientation with respect to a given angle axis passing through the barycenter, as implemented in PASSAGE.
Consideration for future work
Our results underline the importance of methodological development of wind dispersal models of Culicoides on land and not only over water. However, studies of landscape genetics remain indispensable and complementary in order to improve the accuracy of predictive models for Culicoides dispersal over land through integration of meteorological, landscape and activity-based parameters previously tested and validated in the landscape genetics work flow. In addition, population sampling of the same species at European level, i.e. a very large proportion of the known range of the species, would be necessary to observe a more marked structuring at European level and to estimate more precisely the gene flow. For example, C. imicola in Europe is structured in two large genetic clusters, “Europe central cluster” and “Europe western cluster” [26]. Moreover, the question of the phylogeographic history of C. obsoletus is still very little explored in view of its geographical distribution throughout Europe and North America. On the contrary, the study of the phylogeography of C. imicola, the main Afrotropical vector, has shown its distribution range out from the northern part of sub-Saharan Africa to the Mediterranean basin [38]. This type of study would make it possible to estimate the effective size of the populations, a key factor in the dispersal of C. obsoletus. In future study, the use of High-throughput sequencing approaches using markers such as ddRadseq (Double-digest restriction-site associated DNA sequencing) can provide greater resolution in view of the large number of SNPs revealed (single nucleotide polymorphism) at a local scale and improve our understanding of the active and passive dispersal of Culicoides. It could also be relevant to look at on scale with finer genetic information. This could include more microsatellite markers or SNP to improve genetic resolution and observe the matching and assignment of each individual.