Typing A. fumigatus isolates may help to improve the understanding of the distribution of this major pathogen in different situations and environments, including susceptible birds in poultry farms. This molecular approach may also give a deeper understanding of the colonization pattern of putative hosts. To date, it is still a matter of controversy whether certain isolates are more virulent and genetically distinct from other isolates, or whether infection by A. fumigatus is simply a matter of contracting infection from any environmental source.
The choice of a specific typing technique depends on the scientific questions and the equipment of the laboratory. Many different techniques have already been described for A. fumigatus: Random Amplified Polymorphic DNA (RAPD) , Restriction Enzyme Analysis (REA) , Restriction Fragment Length Polymorphism (RFLP) , Amplified Fragment Length Polymorphism (AFLP) , Microsatellite Length Polymorphism (MLP) [23–27] and Multilocus Sequence Typing (MLST) . CSP typing is a recently developed typing strategy that involves DNA sequence typing of a repetitive region of the A. fumigatus AFUA_3G08990 gene coding for a Cell Surface Protein, designated the CSP locus [29, 30]. All of these typing techniques were developed in order to resolve closely related isolates for the purposes of outbreak investigation in hospitals and disease surveillance in humans. RFLP (with Afut1 probe) and MLP typing methods were proved to be highly discriminant. Furthermore MLP showed high reproducibility because of the unambiguous data. For these reasons, MLP method is now considered as the gold standard for the analysis of epidemiological relationships between large amounts of A. fumigatus isolates over a long period of time in hospitals. Another method with high reproducibility is MLST, but the loci described so far for A. fumigatus are probably not discriminant enough to identify the source of an outbreak situation. The RAPD method was used in many investigations probably because it requires simple equipment and no genomic sequence information, but it suffered from limited discriminatory power and reproducibility.
In the present study, a molecular typing method for A. fumigatus based on the study of 10 VNTR markers with repeat size larger than 9 bp was developed and further applied to 277 isolates from birds or from the environment. The MLVA typing method proved highly discriminant with a Simpson's diversity index of 0.9994. This value was obtained with unrelated isolates from animals or humans and was exactly the same as that obtained with isolates from humans using microsatellite markers . Size differences between alleles of the 10 selected VNTRs were large enough to allow efficient sizing on agarose gel. This makes the present MLVA scheme easy to implement in laboratories possessing basic molecular biology equipment. The method showed a good reproducibility, which could be increased by the production of an internal ladder (including an example of each allele amplicon size) or the use of capillary electrophoresis . The MLVA was shown to be rapid and very discriminant. Performing monoplex amplifications, like in the present study, leads to more effort than using multiplex amplifications. In future development of the MLVA technique, the combination of two or more VNTR amplifications in a single reaction tube should be tested.
For the clustering analysis of VNTR profiles, we used a graphing algorithm termed minimum spanning tree (MST). This method was introduced to improve analysis of VNTR profiles . Similar to maximum-parsimony phylogenetic tree reconstruction methods, MST constructs a tree that connects all the genetic profiles in such a way that the summed genetic distance of all branches is minimized. The differences in mathematical approach between MST and UPGMA methods may account for the changes in isolates clustering. Thus, MST allowed to group A. fumigatus isolates which were unclustered with UPGMA. A first cluster included most of the isolates from birds in France whereas the second included most of the isolates from birds in China (Figure 2). The third cluster included most of the environmental isolates collected in a hatchery in France. As a consequence, MST results clearly reflected the geographic origin of the isolates. However, the clustering did not allow the separation of isolates collected from birds living in two different farms in the same department (in France) or province in China. This suggests that geographic clustering occurs at the scale of large areas. The distance between the two farms in Sarthe department in France or in Guangxi province was 20 km and 30 km, respectively. The mean distance between the two farms in Sarthe department and the hatchery in Maine-et-Loire was 120 km. To confirm the geographic clustering and evaluate the minimum size of geographic clusters, additional samples from other origins should be included. We should also collect environmental isolates near the poultry farms in Sarthe department or Guangxi province and avian isolates near the hatchery in Maine-et-Loire department. Geographic clustering of A. fumigatus isolates using repeat sequence analysis with the CSP method, was suggested by Balajee in 2007 . Recently, another study using the AFLP method showed a geographic structuration of A. fumigatus isolates .