High-quality sequence alignments were performed using NAST. Sequence clustering by CD-hit and OUT delineation by DOTUR were performed as described previously (Zhang et al., 2012a (link), 2012b (link)). The representative sequences of operational taxonomy units (OTUs) with their relative abundance were used to calculate rarefaction analysis and Shannon diversity index by QIIME (Caporaso et al., 2010 (link)). In addition, the representative sequences were inserted into a pre-established phylogenetic tree of the full-length 16S rRNA gene sequences in ARB (Ludwig et al., 2004 (link)). Then, the phylogenetic tree and the relative abundance table of representative sequences of OTUs were used for UniFrac principal coordinate analysis (PCoA) (Lozupone and Knight, 2005 (link)). The statistical significance between different groups was assessed by multivariate analysis of variance in MATLAB 2010b (The MathWorks Inc., Natick, MA, USA). Redundancy analysis was performed using CANOCO for Windows 4.5 (Microcomputer Power, Ithaca, NY, USA) according to the manufacturer's instructions (Braak and Smilauer, 2002 ). Statistical significance was assessed by MCPP with 499 random permutations under the full model. Ribosomal Database Project Classifier was used to assess the amounts of different genera by taxonomic assignment of all sequences.