16S rDNA sequencing of fecal samples.

The fecal microbiome was sequenced using Roche 454 pyrosequencing in (n = 33) healthy subjects. DNA was extracted with a commercially available kit, FastDNA Spin Kit for Soil (MP Biomedicals, Solon, OH) using the manufacturer’s recommended protocol. The adequacy of the amount of extracted DNA from samples was verified with fluorometric quantitation (Qubit, Life Technologies, Grand Island, NY), and samples with inadequate amounts of template DNA were not sequenced. Forward primer 28F: 5′ GAGTTTGATCNTGGCTCAG 3′ and reverse primer 519R: 5′ GTNTTACNGCGGCKGCTG 3′ were used to pyrosequence the 16S rDNA on a 454 GS FLX platform, with barcoding, using titanium kits at Research and Testing Laboratories (Lubbock, TX) for the analysis of the bacterial 16S rRNA phylotypes (28). Python scripts in the Quantitative Insights Into Microbial Ecology (QIIME) software pipeline (VirtualBox versions 1.5, 1.6, and 1.7) and other custom scripts were used to process the sequencing files at Research and Testing Laboratories and Rush University (3, 12, 27, 36). The sequence outputs were filtered for low-quality sequences (defined as any sequences that are <200 bps or >1,000 bps; sequences with any nucleotide mismatches to either the barcode or primers; sequences with homopolymer runs >6; sequences with an average quality score of <25; sequences with ambiguous bases >6; and sequences truncated at the reverse primer). Sequences were denoised with USEARCH (6), and chimera were checked with UCHIME (7) and Chimera Slayer (10). Operational taxonomic units (OTUs) were picked with uclust (6) at a 97% similarity threshold, and representative sequences were generated. Sequences were aligned with PyNAST (2), and taxonomy assignment was performed in QIIME 1.6VB against the QIIME 1.6 version of Greengenes database (16, 35) using the RDP classifier (35) at a 80% bootstrap value threshold. An approximately-maximum-likelihood phylogenetic tree was created using FastTree v2.1.3 (18). Multivariate reduction analyses using principal coordinates with a Unifrac metric was done to determine the global microbiome composition based on OTUs using QIIME VB 1.6 distances. Analysis of similarities (ANOSIM) implementation in QIIME VB 1.7 was used to perform a randomization test of significance of pseudo F values, with 999 randomizations for each model, on rarified sequence data and was used to statistically assess differences in beta diversity. Linear discriminant analysis effect size (LEFSE) was calculated (26), and graphs of the data were generated using its open source Galaxy implementation at the Huttenhower laboratory website (https://huttenhower.sph.harvard.edu/galaxy/). Bacterial taxa that were not present in at least 10 of the samples (i.e., taxa that were present in only a handful of subjects) were not reported as differentially abundant, to prevent erroneous results stemming from absence of taxa rather than their presence. SPSS (v.19.0.0; Chicago, IL) was used to analyze clinical metadata. In SPSS, t-tests or ANOVA was used to analyze differences for parametric data satisfying test assumptions; Kruskal-Wallis, Mann-Whitney, or median tests were used to analyze nonparametric data; χ² or Fisher exact tests were used to detect differences in proportions between groups, as appropriate. Microsoft Excel and PowerPoint were also used to generate plots of bacterial taxa.