The ARB software package was used to analyze sequence data and construct trees (version 2.5b; O. Strunk and W. Ludwig, Technische Universitat München, Munich, Germany).

The raw tag sequences were processed with QIIME 1.8.0 (Caporaso et al., 2010). Sequences were excluded from analysis if they had a mean quality score < 25, were either < 200 or > 1000 bp in length, contained ambiguous nucleotides or had any mismatches in the forward and reverse primers. The sequences were assigned to individual samples by their adapter tags and the 16S rRNA primers were removed prior to analysis. The resulting data sets contained 254 bp of the bacterial V4 region. Potential chimeras were identified and removed using ChimeraSlayer. Trimmed sequences were then classified with RDP classifier. Operational taxonomic units (OTUs) were clustered at 97% similarity based on a distance matrix generated with the program QIIME.

Phylogenetic trees were constructed from aligned clone sequences and closely related environmental clones and cultures using the ARB software package (version 5.5; O. Strunk and W. Ludwig, Technische Universität Munchen, Munich, Germany). Tag sequences found at high abundances (>50 tags detected per sample), as well as those that were identical to a clone sequence or those that represent a unique lineage were inserted into bootstrapped trees using parsimony insertion tool. Alpha and beta diversity estimates were calculated using QIIME. After trimming each sample to an equal number of tags, bacterial diversity was estimated with N obs (observed richness), Chao1 (nonparametric richness estimator), nonparametic (np) Shannon diversity index, Simpson Evenness and Equitability. Structure of the pyrosequenced microbial communities was compared with two different methods. Principal coordinate analysis (PCoA) maps the samples on a set of orthogonal axes in order to explain the maximum amount of variation by the first coordinate and the second largest amount of variation by the second coordinate based on weighted UniFrac values (Lozupone et al., 2010). The UPGMA (unweighted pair group method with arithmetic mean) distance tree is based on a hierarchical clustering in which topological relationships are identified in order of similarity. The robustness of the UPGMA clusters was tested with jackknife analysis, based on 100 randomized subsamples.

BCO-DMO Processing Notes:
- added conventional header with dataset name, PI name, version date
- modified parameter names to conform with BCO-DMO naming conventions
- changed n/a (not available) to nd (no data)
- converted latitude and longitude to decimal degrees
- reformatted date from dd-Mon-yy to yyyy-mm-dd
- created links to NCBI accession pages