Quality of barcoded pyrosequencing reads.

Total number of raw reads was 143,487.
aValid sequences of nifH, archaeal amoA, bacterial amoA and nosZ genes were defined as high quality sequences with correct barcode and primer (at 5′-end), length >350 bp and that did not have frameshifts and chimeric structure. The possible sequencing errors causing frameshifts and chimeras were removed based on the BLASTX result. Sequences with different regions matching the same sequence in the database but with different frame positions were considered to be frameshifts. Sequences that matched two or more different origin sequences were classified as chimeras. Valid sequences of 16S rRNA gene were sequences with correct barcode and primer, length >350 bp and passed the chimeric check program in Greengenes with the Bellerophon method. The sequence numbers for each sample are listed in Table S3.