Arbuscular mycorrhizal fungi (AMF) are widespread in terrestrial ecosystems and form obligatory symbiotic relationships with most land plants . These symbioses are not host specific but are to some extent host-preferential . AMF may have a biogeographical distribution partially because of their soil-borne life form, host preference, and limited dispersal resulting from geographic isolation . Increasingly, studies have shown that AMF are cosmopolitan at the genus or higher taxonomic level but have limited distributions at the species level. Global-scale studies showed that AMF have distinct distribution patterns  and that habitat filtering or dispersal limitation affects these patterns . At the continental scale, a single widespread host plant was found primarily to harbor geographically generalist AMF . At the regional scale, a typical distance-decay distribution of AMF suggested that geographic distance and environmental heterogeneity determined the patterns of spatial scaling . At the local scale, AMF distribution is closely related to host identity and habitat . At the fine scale, AMF diversity has a patchy distribution . Studies have also shown that AMF distribution could be influenced by climate [4, 9], habitat , geographical isolation, soil conditions, and anthropogenic activities .
Several studies have reported high selectivity between host plants and AMF, which may affect AMF distribution [10–12]. Bever et al  reported host-specific differences in the population growth rates of AMF. Helgason et al  found physical and functional selectivity in AMF, and Zhang et al.  recently documented that the invasive plant Solidago canadensis promoted the growth of the most beneficial AMF and thereby increased its own competitiveness. Distinctive AMF communities were associated with coexisting plant species, including grasses [13, 14], forbs [15, 16], and trees [17, 18]. Even plants of the same species that differed in age harbored distinctive AMF [19, 20]. Thus, it seems likely that plants can actively select with AMF to associate. Some studies have suggested that plants have evolved an ability to recognize beneficial AMF. For example, Bever et al  found that the host plant preferentially allocated carbon to beneficial symbionts. Kiers et al , who used qPCR and stable isotope probing methods, documented the existence of a reciprocal rewards mechanism that stabilizes the cooperation between plants and AMF. Because host plants have distinct geographical distributions , selection by host plants may result in distinct geographical distributions for AMF . However, whether and how host selection affects the pattern of AMF distribution at large scales has not been well studied.
Recently, two studies have revealed global patterns of AMF by using molecular approach. One of these studies  has developed a database (MaarjAM) and demonstrated ecosystemic and global patterns for AMF distribution. This study also documented that AMF had high specificity of host plant superorders and wider distribution with larger breadth of hosts. The other study  also found a high endemism of AMF at global scales that could be affected by geographic distance, soil condition and plant community. Although both of these studies indicated that host plants related to AMF distribution at a global scale, whether and how host plants affect AMF distribution remain unknown. Here, we hypothesized that selective pressure from host plants at different scales (taxonomic order and functional group) may be the main factor affecting the distribution of AMF at large scales, and that the effects of biogeographical, ecosystemic and climatic factors may be mediated by host plants.
ITS (Internal Transcribed Spacer), SSU (Small Sub-Unit) and LSU (Large Sub-Unit) are the three widely used marker genes in molecular diversity of AMF. SSU was utilized widely while LSU was relatively infrequent compared with ITS (based on Kivlin's comprehensive publication survey  and genus search tool of emerencia  as well as personal survey from GenBank). As relative conserved variability of sequences among species, SSU is widely used in molecular diversity of AMF. However, SSU has a limitation to resolve species in genera of Ambispora, Diversispora and Scutellospora [27–29]. This limitation can affect the designation of phylotype and further influence the analysis on the distribution of the whole AMF . For LSU, although it had relative higher intra-specific variability than SSU and can also well resolve AMF species, the LSU based data in GenBank are limited for the study of AMF distribution at global scale. To avoid the limitations of SSU and LSU, we selected ITS as marker gene for our analysis. Although ITS has a high intra- and inter-specific variability and is not suitable for operational approaches, it can provide sufficient phylogenetic signal that fit well in the phylogenetic methods. Moreover, the complete ITS region could separate almost all AMF species except Glomus intraradices and its close relatives . In addition, abundant ITS sequences are stored in public databases that support our present study. Also, the online genus search tool emerencia  provides great convenience for extracting all the AMF ITS sequences from GenBank.
In this study, we used data mining to explore the diversity and distribution of AMF based on ITS sequences in public databases. Our objectives were: 1) to explore the global distribution patterns of AMF; 2) to test whether selectivity for host at scales of taxonomic order and functional group is the main factor affecting the current distribution patterns of AMF.