Conceived and designed the experiments: LR TB. Performed the experiments: LR. Analyzed the data: LR TB. Contributed reagents/materials/analysis tools: LR TB. Wrote the paper: LR.
The authors have declared that no competing interests exist.
Among global changes induced by human activities, association of breakdown of geographical barriers and impoverishered biodiversity of agroecosystems may have a strong evolutionary impact on pest species. As a consequence of trade networks' expansion, secondary contacts between incipient species, if hybrid incompatibility is not yet reached, may result in hybrid swarms, even more when empty niches are available as usual in crop fields and farms. By providing important sources of genetic novelty for organisms to adapt in changing environments, hybridization may be strongly involved in the emergence of invasive populations.
Because national and international trade networks offered multiple hybridization opportunities during the previous and current centuries, population structure of many pest species is expected to be the most intricate and its inference often blurred when using fast-evolving markers. Here we show that mito-nuclear sequence datasets may be the most helpful in disentangling successive layers of admixture in the composition of pest populations. As a model we used
Some global changes induced by human activities impact the evolution of pest species and may have strong incidence on their population genetic structure. Not only human transport and commerce keep introducing unprecedented alterations in the distribution of the earth's biota, but also cultivation and husbandry apparently facilitate many biotic invasions
Such population mixing, followed in some cases by nearly interspecific hybridizations (see below lexical note) makes population structure within a given pest species very difficult to be disentangled. Many studies explore the genetic structure of hybrids in the framework of conservative biology
At least in parasites, whereas cytoplasmic data are DNA sequences in most population genetics studies, the most commonly used nuclear markers are fast-evolving ones such as microsatellites (see
In this work, we argue that exploration through Bayesian hierarchical clustering and F-statistics on mitochondrial and nuclear sequence datasets can help disentangle successive layers of admixture in the composition of pest populations, provided that the effects of LD have been checked. Successive secondary contact events following long isolations is expected to generate a tangle of immigration histories, implying both motley populations mainly composed of unadmixed individuals of divergent populations and hybrid swarms that is populations containing nothing other than hybrid individuals, not to mention intermediate states. A number of population genetic features are expected when analyzing sexual species which have undergone such histories: (a) In the case where hybridization between incipient species (see lexical note below) occur, followed by speciation reversal as described by Seehausen et al.
Symmetrical mito-nuclear paraphyly within genus
In an attempt to establish variable mitochondrial and nuclear sequence data as useful markers to study intricate histories of populations, including hybridization between long isolated lineages, we took advantage of established ecological (wide host spectrum and marked preference for domestic birds in
detectable imprint left by probable hybridization(s) through such a human-influenced history
demographic scenarios which are distinguishable using mito-nuclear sequence data within
in a more specific point of view, the most probable causes for
In order to check whether results were not biased by within fragment linkage disequilibrium (LD) between pairs of sites, sequence data were split into two groups, whole datasets and datasets excluding pairs of site with a rLD>0.5, and both batches of results compared, as recommended for Bayesian assignment and clustering
Because hybridization implies new genetic admixture between more or less long separated populations and because speciation is a gradual process, such events by definition lie at the very boundary between species and intraspecific entities. As a result, any study aiming at investigating them raises the problem of the definition of species. Therefore, we need to beforehand set some lexical definitions.
We consider here species as defined following the internodal concept consisting of a fragment of the Global Genealogical Network (GGN) branch segment between two nodes of the tree, with each node being either a branching point (speciation) or a dead-end (extinction)
Because speciation is a gradual process and because human-induced breakdown of geographical barriers allows naturally improbable admixture, notably at a latest time within the critical window in speciation during which intrinsic postzygotic isolation is absent or incomplete, we decided to take into account some isolation even through available data show it does have subsequently been broken down. If genetic divergence from lineage split to secondary contact is revealed considerable and comparable to divergence with a sibling species and in case of evidence for hybridization between so deeply diverging entities, the term incipient species will designate family lineage partitions
Since fertile hybridization between two alleged species is by definition genuine evidence for them belonging to the same species
For the same reasons, because the internodal species definition implies that so-called “interspecific hybridization” is nonsensical, we propose to name “nearly interspecific hybridization” the result of admixture between lineages almost as differentiated from each other as from a sibling species.
A lineage designates the descendants of a common ancestor, which form a fragment of the GGN, not necessarily a whole segment, but are, at some point, reproductively isolated from other organisms. “Lineage” is used instead of “species” in cases where the probability that reproductive isolation is irreversible has never been questioned or is not sufficiently supported. In some instances, it corresponds to family lineages
The word “entity” is used as a generic term for “lineage” and “species”.
A population is a group of mite individuals belonging to a single species and occupying an individual nest or a group of nests in a bird colony (wild avifauna) or a single building (farms), or, in some case, a single point within a building. An isolate is a random sample of individuals which are representative of a given population.
The location, host species, mite species, accession numbers associated with mite individuals under test are listed in
From each population, 11–24 individuals have been separately sequenced (isolates+). In some cases, a reduced number of individuals have been added (isolates−; some L1 isolates, 2 isolates from farm CON, see
Among
DNA was extracted from single adult females and sequenced following
A 543 bp fragment of the cytochrome oxidase I (COI) coding gene and a 663–698 bp fragment of the Tropomyosin (Tpm) gene including a large intron, flanked by small exon parts at both ends have been PCR amplified and sequenced following
Mitochondrial (“Lmt-”) and nuclear lineages (“Ln-“), as well as some precise Tpm (“Tro-“) and COI (“Co-”) haplotypes will be referred to in present paper following the nomenclature established by
To investigate the congruence of obtained data, we have explored the most probable causes for revealed patterns by testing different scenarios for nearly interspecific hybridization, gene flows and demographic history. For such a purpose, we have confronted results obtained for different groups of isolates to each another in populations genetics analyses, in relation with available information. Detailed information on the hypotheses under test and used methods for analyses can be found in
To test scenarios for nearly interspecific hybridization, gene flows and demographic history, we genotyped 469 adult females at the two fragments of COI and Tpm described in
Most FIS per isolate gave non significant values, suggesting that Tpm alleles are in Hardy-Weinberg equilibrium. This is at the exception of BREa and BREb, which show some HWD (positive FIS values, with P>0.05). As a result, we did not consider Tpm alleles to be independent for testing the significance of population differentiation, and rather used the genotype as the randomization unit instead of the allele in permutation tests of FST estimates.
The assumption that mutation rates are different between cytoplasmic and nuclear genomes in
A. Bird type information (black areas, wild birds; grey areas, chicken; squared areas, layer).
Nuclear and mitochondrial lineages defined by
Isolates from wild birds do not share any COI haplotype with any domestic bird. A single COI haplotype is shared between two types of domestic birds (layer and Bresse AOC chicken – see
Overall, the network obtained from COI haplotypes isolated in
Linkage disequilibrium was detected with a significant r value>0.5 (significance assessed following two-tailed Fisher's exact test with P>0.05) between pairs of sites involving 71 different sites in COI and 76 (both substitutions and indels considered) in Tpm (see list of sites in legend of
The clustering approach used by Structure to determine the number of genetic groups within the two genes under test identified two hierarchical levels of subdivision in Tpm whole dataset as well as in Tpm dataset including only pairs of sites with a rLD<0.5 and four in COI whole dataset. The results of the different rounds of the process are summarized in
A. COI. Ln assignment value threshold >0.8. Only two of the 469 individuals under test kept unassigned (<0.6, max inferred ancestry 0.54>x>0.58) (both from CON and located basally to Lmt3 in haplotype networks) B. Tpm. Black: Ln assignment value threshold >0.6 (91% individual inferred ancestry >0.9). Grey : intercluster heterozygous (mean 0.450/0.550 [0.405–0.595]).
Sorting of the obtained Tpm results by the COI lineage membership individual information as defined by
Assignment analyses of COI whole dataset using the admixture model revealed strongly motley populations at the isolate level, but low admixture at the individual level (values of Dirichlet parameter for the admixture proportions α≈0.02) (
Hierarchical Structure analyses performed to determine the number of genetic groups (K) present in
Clusters retained from first-level Structure results with whole sequence datasets as well as with datasets excluding pairs of sites with a significant (P>0.005) rLD value >0.5 (see
Tpm analyses revealed important proportion of inter-cluster heterozygous at first level of population structure, with most of shared ancestry values ≈1 (homozygous) and ≈0.5 (heterozygous) and almost no individual sharing ancestry between the three admixted lineages. Indeed, 177 inter-cluster heterozygous individuals of 469 (37.7%) were recorded. Very little admixture seem to have occurred at the individual level (α≈0.08), since almost no individual shares ancestry from the three retained clusters according to ΔKmax values (a single one received partitioned inferred ancestry: 0.408/0.423/0.169), even with very small ratios, as is shown by distribution of inferred ancestry values (3 groups of values sharply separated: ≈0, ≈1, ≈0.5). And yet, of the 292 individuals which do not show inter-cluster heterozygosity at the first Tpm hierarchical level, 265 were assigned with inferred ancestry >0.9, 26 were assigned with inferred ancestry comprised between 0.6 and 0.89 and a single individual partitioned (cf. above). When considering nuclear lineages defined in
Using the combined dataset, the uppermost detected structure level is composed of 4 clusters (
Since it has been shown that the presence of LD between sites might generate spurious clusters
In the wild avifauna, mismatch distributions of COI and Tpm haplotypes are very different between IL and ROL. It significantly matched the model of sudden expansion in IL (typical unimodal distribution in both genes, Tajima's D close to zero,
Of 24 COI sequences isolated in SK individuals, when considering the most represented haplotype as the ancestral one, observed number of mutations in the 543-bp COI fragment in SK across ca. 52 generations (28.2–112.7) is three (distributed on three different haplotypes of which one has been recorded twice and two once) (
COI - model of sudden expansion | Tpm - model of sudden expansion | ||||||||||||||||||
GP | Context | πCOI/πTpm | πCOI | πTpm | KT1 | D | Isolate/*group of isolates | Smaj | Majority COI | COI m. d. | SSD | P( |
r | P(r) | Tpm m. d. | SSD | P( |
r | P(r) |
1 | Wild avifauna | <0.5 | <0.005 | >0.0150 | 2 | Diverse D values | IL | 2 | Lmt2–3 | uni | 0.02 | 0.22 | 0.18 | 0.13 | uni | 0.01 | 0.84 | 0.01 | 0.14 |
ROL | 3 | Lmt2–3 | 2h | 0.06 | 0.01 | 0.71 | 0.60 | bi | 0.21 | 0.08 | 0.36 | 0.07 | |||||||
2 | Danish, Polish, French layer/broiler farms | <0.5 | <0.005 | >0.0075 <0.0120 | 3 | Diverse D values | 8022 | 0 | Lmt1 | 1h | 0.00 | 0.00 | 0.00 | 0.00 | bi | 0.25 | 0.05 | 0.47 | 0.07 |
9005 | 0 | Lmt1 | 1h | 0.00 | 0.00 | 0.00 | 0.00 | bi | 0.25 | 0.07 | 0.49 | 0.09 | |||||||
BOUY | 4 | Lmt2–3 | uni | 0.16 | 0.10 | 0.62 | 0.04 | bi | 0.24 | 0.00 | 0.46 | 0.00 | |||||||
PO2 | 2 | Lmt2–3 | 2h | 0.20 | 0.12 | 0.77 | 0.03 | bi | 0.25 | 0.06 | 0.47 | 0.10 | |||||||
SK | 3 | Lmt2–3 | uni | 0.00 | 0.46 | 0.20 | 0.38 | bi | 0.41 | 0.00 | 0.50 | 0.94 | |||||||
3 | Australian,Brazilian, French layer farms | <0.5 | <0.005 | >0.0015 <0.0120 | 2 | DCOI and DTpm<0 | AUS | 3 | Lmt2–3 | bi | 0.02 | 0.19 | 0.50 | 0.56 | 2h | 0.01 | 0.07 | 0.86 | 0.81 |
BREa | 0 | Lmt2–3 | 1h | 0.00 | 0.00 | 0.00 | 0.00 | bi | 0.15 | 0.06 | 0.48 | 0.43 | |||||||
BREb | 1 | Lmt2–3 | 2h | 0.00 | 0.36 | 0.62 | 0.76 | bi | 0.02 | 0.14 | 0.64 | 0.70 | |||||||
REN | 3 | Lmt1 | uni | 0.00 | 0.48 | 0.19 | 0.38 | 2h | 0.03 | 0.06 | 0.78 | 0.75 | |||||||
4 | French layer farm | >1.0 | 0.005 | 0.0016 | 2 | DCOI and DTpm<0 | 8019 | 2 | Lmt1 | bi | 0.01 | 0.22 | 0.16 | 0.52 | 2h | 0.01 | 0.04 | 0.83 | 0.80 |
5 | French layer/broiler farms | >0.5 <1.0 | >0.005 <0.020 | >0.0075 <0.0120 | 3 | Diverse DCOI; DTpm>0 | 8020 | 1 | Lmt1 | bi | 0.04 | 0.08 | 0.19 | 0.42 | bi | 0.25 | 0.00 | 0.47 | 0.00 |
8021 | 0 | Lmt2–3 | bi | 0.03 | 0.08 | 0.71 | 0.66 | bi | 0.14 | 0.01 | 0.25 | 0.00 | |||||||
8028 | - | - | bi | 0.49 | 0.00 | 0.52 | 0.94 | bi | 0.40 | 0.00 | 0.52 | 0.94 | |||||||
8029 | - | - | bi | 0.35 | 0.00 | 0.70 | 0.00 | bi | 0.24 | 0.00 | 0.39 | 0.00 | |||||||
9003 | - | - | 2h | 0.49 | 0.00 | 0.75 | 0.92 | bi | 0.47 | 0.00 | 0.52 | 0.96 | |||||||
9004 | - | - | 2h | 0.26 | 0.01 | 0.70 | 0.23 | bi | 0.53 | 0.00 | 0.64 | 0.94 | |||||||
9007 | 0 | Lmt1 | bi | 0.28 | 0.00 | 0.61 | 0.17 | bi | 0.18 | 0.01 | 0.31 | 0.00 | |||||||
9016 | 1 | Lmt2–3 | bi | 0.02 | 0.16 | 0.62 | 0.65 | bi | 0.14 | 0.00 | 0.27 | 0.00 | |||||||
6 | Pigeon breeding facility | NA | 0.003 | 0.0000 | 1 | DCOI<0 | 9001 | 0 | L1 | 2h | 0.07 | 0.02 | 0.70 | 0.67 | 1h | 0.00 | 0.00 | 0.00 | 0.00 |
var. | Pigeons | - | - | - | - | - | * All L1 | - | - | 0.14 | 0.04 | 0.28 | 0.01 | 0.06 | 0.18 | 0.25 | 0.18 | ||
Wild avifauna | - | - | - | - | - | * All wild avifauna | - | - | 0.08 | 0.06 | 0.09 | 0.16 | 0.02 | 0.00 | 0.03 | 0.00 | |||
French layer/broiler farms | - | - | - | - | - | * All French layer/broiler farms | - | - | 0.11 | 0.00 | 0.16 | 0.00 | 0.20 | 0.00 | 0.34 | 0.00 | |||
Layer/broiler farms | - | - | - | - | - | * Cluster k2 | - | - | 0.12 | 0.12 | 0.27 | 0.16 | 0.00 | 0.12 | 0.91 | 0.91 | |||
Layer/broiler farms+wild avifauna | - | - | - | - | - | * Cluster k3 | - | - | 0.01 | 0.26 | 0.03 | 0.26 | 0.15 | 0.00 | 0.25 | 0.00 | |||
Layer/broiler farms | - | - | - | - | - | * Cluster k4 | - | - | 0.00 | 0.56 | 0.26 | 0.56 | 0.01 | 0.10 | 0.81 | 0.78 |
Genetic profiles (GP), as delineated from observation of COI and Tpm diversities and grouping together different isolates (left part, top), draw outlines of five different scenarios in the demographic history of
As for Tpm mismatch distribution, it was bimodal in SK as in ROL and in most farm isolates (
Despite numerous Tpm bimodal profiles, mismatch distributions of COI and Tpm fit the model of sudden expansion in seven and two of the 13 French farm isolates respectively. In contrast, the demographic imprint from mismatch distributions shows old stationary populations of farm isolates at France level (all French farm isolates pooled together, SSD 0.108 in COI and 0.204 in Tpm). Lastly, mismatch distributions of COI and Tpm haplotypes in cluster k4 (combined Structure analysis) are unimodal and fit well the model of sudden expansion (COI: SSD 0.002, raggedness index 0.258 with p = 0.564, Tpm: SSD 0.011, raggedness index 0.808 with p = 0.780).
In hierarchical Tpm AMOVAs, FIS values were not significant in any analysis except with the geographical and habitat criteria (0.067, P<0.05). FIS calculated per inferred Structure cluster on individuals assigned by >0.7 in Structure, provided insignificant values for k2 and k4, whereas k3 and L1 had significantly positive values (0.39 and 0.89 respectively, P<0.05). The Tpm AMOVAs indicated that a large fraction of the total variation in every analysis could be explained by differences within individuals (mean 55.20%), with the maximum percentage values found in analyses involving French fowl only (86.28–86.76%). In contrast, minimum percentage values explaining variations by differences within individuals were obtained in analyses involving groups defined by clustering assignment with a threshold >0.7 (first level of combined analysis with Structure, 30.34% for
AMOVAs using bird type as the group criterion for farm isolates (layer/Bresse AOC broiler) detect by far larger amounts of variation which could be explained by differences within individuals (Tpm 71.89–86.28%) and among isolates within groups (Tpm 12.46–29.73%, COI 72.03–86.31%) than among groups (negative values in both Tpm and COI analyses). When using habitat as the group criterion (wild/domestic), 80.15% to 94.46% of the total Tpm variation in each analysis is more or less equally split into variations explained by within individual differences and among groups differences, whereas dominant COI variation (49.48–62.62%) could be explained by among isolates differences. When considering the geographical location as a group criterion in farm isolates, among groups differences could explain 49.54% of the total Tpm variation and 96.42% of the total COI variation at the world level (excluding France), as opposed to 0% of the total Tpm variation and only 30.13% of the total COI variation within France. When considering clusters inferred by Structure analyses as a group criterion, the proportions of variation explained by both within-individual and among-groups differences are somewhat different than with above group criteria. Indeed, higher values are attributed to the latter (54.48–58.49% of the total variation) and reduced values for the former (27.35–37.71%).
Tpm pairwise FST are much lower than COI pairwise FST (average 0.31 vs 0.69). Between L1 and non L1 isolates, Tpm pairwise FST values are >0.42 except against the wild isolate IL (0.27–0.45) and COI pairwise FST values are all >0.80 (P>0.00001 for all values). Wild isolates ROL and IL provide rather high Tpm FST values when compared to farm isolates (>0.41 and >0.28 respectively, P>0.00001), except for IL vs 9004 (0.184) and IL vs BOUY (0.178) (P>0.00001 for both values). ROL's and IL's COI pairwise FST are all >0.53 (average 0.83). When comparing ROL and IL to each other, Tpm FST value is much lower (0.19, P>0.00001) and COI higher (0.90, P>0.00001). When comparing Australian and Brazilian farm isolates with European ones, Tpm and COI pairwise FST values are rather high (>0.5 and >0.75 respectively, P<0.00001) except against some isolates from the Bresse region (Rhône-Alpes, France) (>0.10 and >0.37 respectively, P<0.00001) and except Tpm pairwise FST alone between French isolate 8022 (Brittany) and Polish isolate PO2 (0.11 with P<0.05, 0.10 with P<0.00001 respectively). Tpm pairwise FST of the lab isolate SK are comprised between 0.02 and 0.19 when compared with European farm isolates and between 0.26 and 0.48 when compared with wild isolates (P>0.00001 for all values). All COI values for pairs involving SK are >0.59 (P>0.00001).
Overall, pairwise FST among French farm isolates are highly diverse, with Tpm ranging from 0 to 0.43 and COI from 0 to 0.98. In regard to correlation of geographic distance with genetic differentiation, high discordance is apparent in these results. Indeed, some Tpm values are close to zero (some negative) between isolates from a single farm (CON, isolates 8021 and 9016), but also between farms located at various distances apart (<10 to >700 km) (9003 and CON, −0.02–0.01, 15 km; 8022 and 8029, −0.02, 583 km, REN and 8019, 0.017, 835 km…). In contrast, some higher FST values are found between very close farms, within a single fowl industry (e.g. 9004 and 9005, 0.24, <5 km, Bresse broilers, P<0.00001).
Some very low COI FST values are found in some cases between distant farms within a given country (<0.008, e.g. France: 8020 and 8019, 0, 693 km; REN and 8019, 0.007, 693 km; Brazil: BREa and BREb, 0.006, 1157 km by road). As for COI FST between isolates from the single farm CON, they are particularly informative: the different samples from two points within a single building (8021, 9016, 9007, points 2 and 6 in each) as well as samples from a single building on two successive flocks (8021 and 9016) provide FST value close to 0 (−0.06–0.05), whereas samples from the two different buildings (build. 1: 9016, build. 5: 9007; distant by 78 m from each other, 150 m by feet for humans) under test provided high FST values (0.42–0.69, P<0.05) despite staff's comings and goings.
The ratio of COI π and Tpm π in French farm isolates 8019, 8020, 8028, 8029, 9016, 8021, 9007, 9003 and 9004 is amazingly higher (>0.5) than in other isolates (
Our results have helped conclude that: (1) isolated
This is important because it shows that populations belonging to
On a methodological point of view, inclusion/exclusion of pairs of sites with a rLD<0.5 resulted in negligible changes (a little bit less resolved 3rd-level in Tpm, one hierarchical level less in COI when excluded), which did not modify the overall clustering, nor the individual assignment to clusters. Therefore, results obtained in the present Structure analyses are deemed reliable.
The two following topologies may be inferred from the first two levels of structure analyses using both single genes: nuclear ((L1,Ln3_4),(Ln1,Ln2)) and mitochondrial ((L1,Lmt1),(Lmt3+,Lmt2)) (
In short, evidence for admixture between all lineages but
In the wild avifauna, the marked difference of demographic characteristics between IL and ROL isolates may be explained by the ecology of their respective hosts (Common Starlings and European Rollers respectively): Common Starlings are known to be either resident or partly migrating with some populations staying the whole year some others going westward to winter. In all cases, some wintering colonies stay in the premises, which might have allowed mites to continue developing through the year and have preserved it from drastic seasonal bottlenecks. European Rollers are migrating to South Sahara from autumn to spring, which results in the absence of host on the premises for micropredator mites during several months per year, and is likely to be responsible of cyclically repeated bottlenecks.
In farms, the diversity of demographic imprints reported in this study was expectable due to the various known causes for bottlenecks in farms and the known ability of
Not only cases for bottlenecks are diverse, but also their effect on the founding population size may strongly vary. Indeed, pairwise FST between farm CON isolates 8021 and 9016 (same building, two different successive flocks) indicate that both isolates belong to a panmictic population, which highlights that cleaning/disinfection actions may have almost no impact on population's genetic diversity. This was expected, as numerous appropriate hiding-places are available for mites in most farm buildings (wall cracks and crevices, link-up point between joint parts, hollow parts, etc…) and may protect more or less numerous and large mite aggregates from any pesticide spray despite careful application and even, in some cases, partial dismantling. As a result, pesticide applications during flocks may have various effects, depending on the farming structure and on the mean of application.
Populations sampled in the wild avifauna (ROL and IL) appear to be genetically isolated from farm populations in Europe. First, nuclear haplotypes they possess differ from farm mites' (higher nucleotid diversity, only two from three first-level clusters present). This might result from either an increased genetic drift post-hybridization, or the occurrence of two different events of hybridization, of which only one would have occurred in wild populations under test. In the latter case, a bottleneck event in farm populations would explain the lower Tpm diversity, compared to wild populations. The effective population size of populations developing in the wild avifauna is very likely reduced as compared to farm populations (see
Second, rather high pairwise FST in both genes and absence of shared COI haplotypes with regard to farm populations, as well as AMOVAs results based on the habitat criterion, show that populations represented by isolates ROL and IL, although genetically close to each other despite geographical distance, are individually isolated from other populations anywhere in countries under test, except in Brazil.
Concerning exchanges between wild and domestic birds, Brazilian farms provided a strikingly different pattern than European and Australian ones. Brazilian isolates BREa and BREb were the only ones which possessed some uncommon Tpm haplotypes, rather distant from typical farm haplotypes and closer to wild haplotypes. They were also the only ones generating significantly positive FIS values in the present study, which highlights a deficiency of heterozygotes and may witness of the Wahlund effect. This suggests that several different points of introduction are at the origin of the dramatic colonization in Brazilian layer farms during the 20th century
Among French farms, the transfer of spent hens through France seems to be linked to secondary contact between divergent lineages. Indeed, nine isolates sampled in Brittany or Rhône-Alpes gather individuals which are assigned to different first-level Structure clusters with assignment values >0.8 together and show highest ratios COI π/Tpm π (
Since no structure may be explained by the two different fowl industries under test (cluster assignment and AMOVAs) and since, in some cases, same trucks may be used for the transportation of these different strains (a farmer, pers. comm.),
At the inter-continent scale, insights for exchanges also arise from present data: Lmt2 and Lmt3 are almost absent from Brittany, but omnipresent in isolates sampled in countries other than France, including Australia and Brazil. And yet, in farms, a Tpm haplotype different than Tro_1, Tro_2, and Tro_3 is present only in some Brazilian and French farms located in Rhone-Alpes (Bresse region,
From present DNA sequence data analyses, available ecological and trade information, six different profiles of population may be distinguished among populations of
In conclusion, the role of trade flows in mite dispersal appears striking, while wild avifauna does not seem to play any role in Europe. Cages which are carried by trucks during transfer of spent hens from farm to slaughterhouse are clearly involved as vectors in mite dispersal within France. At the international level, vectors and vehicles for mite exchanges remain to be identified. Moreover, as opposed to European and Australian patterns, some multiple introduction events from both wild and domestic birds seem to have occurred in Brazil. This may explain the rapid colonization of layer farms by
Although we are aware that a study involving a higher number of both mitochondrial and nuclear sequences would probably help getting an even clearer picture, we are confident that not only these results will help improving prophylactic actions in the layer industry, but also that it will constitute a case in point for investigation of pest population structure using mito-nuclear sequence data. Furthermore, combination of present data with faster evolving markers such as microsatellites would be a highly interesting perspective in order to obtain a comprehensive overview of the evolutionary history of
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(DOC)
(DOC)
We kindly thank S. Lubac (Institut Technique de l'Aviculture, Lyon, France), C. Basset (Avibresse, France), O. Kilpinen (Danish Institute of Agricultural Sciences, Lyngby, Denmark), E. Tucci (Instituto Biológico, São Paulo, Brazil), J. Kellaway (Australian Egg Corporation, North Sydney, Australia), T. Cencek (Poland), M.W. Sabelis and I. Lesna (IBED, University of Amsterdam, The Netherlands), N. Vincent-Martin (CRBPO, MNHN, Paris, France), R. Leblois (MNHN, Paris, France), M. Germain (Bayer Healthcare Animal Health, Bayer Pharma S.A.S, France) for biological material sampling, precious information and useful advice. Especially, we are very grateful to E. Tucci for her painstaking help in understanding the history of poultry farming in Brazil.
We are deeply thankful for valuable comments and advice provided by A. Khila (McGill University, Montreal, Quebec, University of Toronto, Toronto, Ontario, Canada) and by two anonymous reviewers, which largely contributed to improve the manuscript.
Highly time-consuming and computationally intensive runs using Structure 2.3.3 have been processed by the means of the very useful and efficient CBSU Web Computing Interface at
L.R. would like to offer her most sincere thanks for their technical participation to S. Bonnet and N. Guichard N. (LEGTA Saint-Genis-Laval, France), M. Rigaux (IUT A, Université Lyon1, France), S. Merlin (Lycée Jean-Baptiste de la Salle, Lyon, France), S. El Ouartiti (IUT GBGE, Université Jean Monnet, Saint-Etienne, France), and G. Lallemand (Lycée des Mandailles, Châteauneuf de Galaure, France).
Finally, L.R. wants to offer her warmest and friendly thanks for her invaluable and constant support to Claude Marie Chauve.