Subject recruitment and urine collection

The Value of Urodynamic Evaluation study was an institutional review board–approved, multicenter prospective randomized trial comparing surgical outcomes using 2 strategies for presurgical testing: multichannel urodynamic testing vs standardized basic office evaluation. Briefly, adult women were eligible if they had reported symptoms of SUI ≥3 months; a postvoid residual <150 mL; a negative urinalysis/standard urine culture; clinical assessment of urethral mobility; desire for SUI surgery; a positive provocative stress urinary test; and a qualifying Medical, Epidemiologic, and Social Aspects of Aging (MESA) questionnaire subscale score (stress > urge). Demographic and clinical characteristics were obtained by self-report including hormonal status, which was categorized by the study team into the hormone group that most appropriately described the patient’s hormone use: premenopausal, postmenopausal (with or without self-reported, current exogenous hormone use), or uncertain about status.

Participants in the main study provided written consent to contribute a single baseline urine specimen to the biorepository. Urine specimens were collected prior to surgery by a standard protocol and tested by dipstick to exclude UTI at study entry. Specimens were centrifuged at 500–1500 g for 10 minutes and the supernatant dispensed into ten 2-mL microfuge tubes, which were frozen at –80°C until shipped on dry ice to the National Institute of Diabetes and Digestive and Kidney diseases biorepository. Available baseline urine specimens with sufficient volume for the planned studies were shipped to Loyola University Chicago, (Maywood, IL) on dry ice, and stored at –80°C until processed for sequence analysis. Samples from 197 of the 630 (31%) Value of Urodynamic Evaluation study participants were used in this analysis; most (174) samples had been obtained by clean catch, with the remaining 23 by catheterization. Analyses for this report were approved by the Loyola University Chicago Institutional Review Board.

16S rRNA gene sequencing

Microbial composition was determined by sequencing the variable 4 (V4) region of the bacterial 16S rRNA gene, as described. The V4 region is ∼250 base pair (bp), ideal for MiSeq sequence technology (Illumina, San Diego, CA), and sufficient to classify most bacteria to the family or genus level. DNA isolation was performed in a laminar flow hood to avoid contamination. Genomic DNA was extracted from 1 mL of urine, using validated protocols. The V4 region was amplified by 2-step polymerase chain reaction (PCR), using modified universal primers 515F and 806R, as described. Extraction negative controls (no urine) and PCR negative controls (no template) were included to assess contribution of extraneous DNA from reagents. Final PCR products were purified from unincorporated nucleotides and primers using Qiaquick PCR purification kit (Qiagen, Valencia, CA) and Agencourt AMPure XP-PCR magnetic beads (Beckman Coulter, Brea, CA). Purified samples were normalized to equal DNA concentration, as determined by Nanodrop spectroscopy (Thermo Scientific, Waltham, MA). The sample library and PhiX sequencing control library (Illumina) were denatured and added to the 2 × 250 bp sequencing reagent cartridge, according to manufacturer’s instructions (Illumina).

Sequence processing

Each specimen was sequenced in duplicate and classified by phylogenetic diversity as measured by Bray-Curtis dissimilarity. A phylogenetic tree was generated and compared to percent total classified reads (relative abundance) at each taxonomic level (phyla, class, order, family, genus). For a genus level example, see Figure 1 .

Using phylogenetic similarity to determine similar profiles (aka urotypes)

For each taxonomic level (phylum, class, order, family, genus), all samples were compared to each other using Bray-Curtis similarity, which produces a phylogenetic tree, or dendrogram, in which shorter branches link similar samples, and longer branches link more dissimilar samples. Therefore, each tree can be divided into groups or clades. When aligned to relative sequencing abundance, clades of each tree separate by identity of predominant organism. This is one example, genus classification from the first replica. Urotype indicates clades that fall into the same urotype. Each urotype is named for the predominant genus. If no one genus is predominant, then the urotype is considered nonpredominant. All corresponding graphs for each replica and each taxonomic level can be found in the Supplementary Figure .

Thomas-White et al. Urinary microbiota of uncomplicated stress urinary incontinence. Am J Obstet Gynecol 2017 .

Each major branch or clade (termed urotype) in the phylogenetic tree was named for the predominant classified taxon (eg, Lactobacillus ). When there was no predominant taxon, we used the term “nonpredominant” to describe the urotype. MiSeq sequence reads were processed following mothur MiSeq standard operating procedure at http://www.mothur.org/wiki/MiSeq_SOP , with minor modifications. Mothur software (version 1.34.4) was used to process raw reads and, using default mothur parameters, to remove low-quality and chimeric sequences. Taxonomic classification from phylum to genus level of sequence reads was performed by the RDP classifier (version 2.5) using the default 0.8 confidence threshold. The sampling depth for this analytic set was set at 2000 reads; the Kolmogorov-Smirnov test confirmed that when subsampling depth exceeded 2000 reads, the distribution of subsampled and original reads distribution had >95.9% similarity among all samples.

Most (171/197) samples had detectable DNA with 338,000 and 340,000 reads subsampled for replicates 1 and 2, respectively. The 26 samples without detectable DNA following PCR amplification were classified as below the detection threshold. Due to read depths <2000, 2 samples from replica 1 and 1 sample from replica 2 were also classified as below the detection threshold, for a total of 28 in replica 1 and 27 in replica 2. Using mothur’s built-in average-linkage clustering algorithm, the cleaned high-quality sequences were clustered into species-level operational taxonomic units (OTUs) based on the commonly used 97% similarity cutoff, resulting in 2579 and 3082 OTUs for replicas 1 and 2, respectively. We used the resultant OTU count table and the R package vegan to determine the Chao1 richness estimate, Pielou evenness index, and Shannon diversity index, which accounts for both richness and evenness, 2 measures of microbial diversity. Richness is a measure of the total number of unique taxa within a given individual, but does not take into account the distribution of those taxa. In contrast, evenness is a measure of distribution, or equality of representation, of taxa within an environment. Samples below the detection threshold lack diversity measurements; these were excluded from subsequent diversity comparisons.

Statistical analysis

Generalized estimating equations (GEE), extensions of generalized linear models that account for correlation between replicas, were used to describe associations between demographic and clinical factors with diversity measurements after adjusting for genus urotype. A gamma distribution with a log link was assumed for Shannon, Chao, and Pielou diversity measurements, due to their skewed nature. To be inclusive, we did not make adjustments for type I error in the GEE analyses when determining potential clinical and demographic associations with microbiota characteristics. Only results from the lowest detected resolution level (ie, genus) are reported. There was insufficient sample size and power to compare urotypes between catheterized and voided samples. All statistical analyses were conducted using software (SAS, v9.4; SAS Software Cary, NC) and statistical significance was assessed at the α=0.05 level.

Subject recruitment and urine collection

The Value of Urodynamic Evaluation study was an institutional review board–approved, multicenter prospective randomized trial comparing surgical outcomes using 2 strategies for presurgical testing: multichannel urodynamic testing vs standardized basic office evaluation. Briefly, adult women were eligible if they had reported symptoms of SUI ≥3 months; a postvoid residual <150 mL; a negative urinalysis/standard urine culture; clinical assessment of urethral mobility; desire for SUI surgery; a positive provocative stress urinary test; and a qualifying Medical, Epidemiologic, and Social Aspects of Aging (MESA) questionnaire subscale score (stress > urge). Demographic and clinical characteristics were obtained by self-report including hormonal status, which was categorized by the study team into the hormone group that most appropriately described the patient’s hormone use: premenopausal, postmenopausal (with or without self-reported, current exogenous hormone use), or uncertain about status.

Participants in the main study provided written consent to contribute a single baseline urine specimen to the biorepository. Urine specimens were collected prior to surgery by a standard protocol and tested by dipstick to exclude UTI at study entry. Specimens were centrifuged at 500–1500 g for 10 minutes and the supernatant dispensed into ten 2-mL microfuge tubes, which were frozen at –80°C until shipped on dry ice to the National Institute of Diabetes and Digestive and Kidney diseases biorepository. Available baseline urine specimens with sufficient volume for the planned studies were shipped to Loyola University Chicago, (Maywood, IL) on dry ice, and stored at –80°C until processed for sequence analysis. Samples from 197 of the 630 (31%) Value of Urodynamic Evaluation study participants were used in this analysis; most (174) samples had been obtained by clean catch, with the remaining 23 by catheterization. Analyses for this report were approved by the Loyola University Chicago Institutional Review Board.

16S rRNA gene sequencing

Microbial composition was determined by sequencing the variable 4 (V4) region of the bacterial 16S rRNA gene, as described. The V4 region is ∼250 base pair (bp), ideal for MiSeq sequence technology (Illumina, San Diego, CA), and sufficient to classify most bacteria to the family or genus level. DNA isolation was performed in a laminar flow hood to avoid contamination. Genomic DNA was extracted from 1 mL of urine, using validated protocols. The V4 region was amplified by 2-step polymerase chain reaction (PCR), using modified universal primers 515F and 806R, as described. Extraction negative controls (no urine) and PCR negative controls (no template) were included to assess contribution of extraneous DNA from reagents. Final PCR products were purified from unincorporated nucleotides and primers using Qiaquick PCR purification kit (Qiagen, Valencia, CA) and Agencourt AMPure XP-PCR magnetic beads (Beckman Coulter, Brea, CA). Purified samples were normalized to equal DNA concentration, as determined by Nanodrop spectroscopy (Thermo Scientific, Waltham, MA). The sample library and PhiX sequencing control library (Illumina) were denatured and added to the 2 × 250 bp sequencing reagent cartridge, according to manufacturer’s instructions (Illumina).

Sequence processing

Each specimen was sequenced in duplicate and classified by phylogenetic diversity as measured by Bray-Curtis dissimilarity. A phylogenetic tree was generated and compared to percent total classified reads (relative abundance) at each taxonomic level (phyla, class, order, family, genus). For a genus level example, see Figure 1 .

Using phylogenetic similarity to determine similar profiles (aka urotypes)

For each taxonomic level (phylum, class, order, family, genus), all samples were compared to each other using Bray-Curtis similarity, which produces a phylogenetic tree, or dendrogram, in which shorter branches link similar samples, and longer branches link more dissimilar samples. Therefore, each tree can be divided into groups or clades. When aligned to relative sequencing abundance, clades of each tree separate by identity of predominant organism. This is one example, genus classification from the first replica. Urotype indicates clades that fall into the same urotype. Each urotype is named for the predominant genus. If no one genus is predominant, then the urotype is considered nonpredominant. All corresponding graphs for each replica and each taxonomic level can be found in the Supplementary Figure .

Thomas-White et al. Urinary microbiota of uncomplicated stress urinary incontinence. Am J Obstet Gynecol 2017 .

Each major branch or clade (termed urotype) in the phylogenetic tree was named for the predominant classified taxon (eg, Lactobacillus ). When there was no predominant taxon, we used the term “nonpredominant” to describe the urotype. MiSeq sequence reads were processed following mothur MiSeq standard operating procedure at http://www.mothur.org/wiki/MiSeq_SOP , with minor modifications. Mothur software (version 1.34.4) was used to process raw reads and, using default mothur parameters, to remove low-quality and chimeric sequences. Taxonomic classification from phylum to genus level of sequence reads was performed by the RDP classifier (version 2.5) using the default 0.8 confidence threshold. The sampling depth for this analytic set was set at 2000 reads; the Kolmogorov-Smirnov test confirmed that when subsampling depth exceeded 2000 reads, the distribution of subsampled and original reads distribution had >95.9% similarity among all samples.

Most (171/197) samples had detectable DNA with 338,000 and 340,000 reads subsampled for replicates 1 and 2, respectively. The 26 samples without detectable DNA following PCR amplification were classified as below the detection threshold. Due to read depths <2000, 2 samples from replica 1 and 1 sample from replica 2 were also classified as below the detection threshold, for a total of 28 in replica 1 and 27 in replica 2. Using mothur’s built-in average-linkage clustering algorithm, the cleaned high-quality sequences were clustered into species-level operational taxonomic units (OTUs) based on the commonly used 97% similarity cutoff, resulting in 2579 and 3082 OTUs for replicas 1 and 2, respectively. We used the resultant OTU count table and the R package vegan to determine the Chao1 richness estimate, Pielou evenness index, and Shannon diversity index, which accounts for both richness and evenness, 2 measures of microbial diversity. Richness is a measure of the total number of unique taxa within a given individual, but does not take into account the distribution of those taxa. In contrast, evenness is a measure of distribution, or equality of representation, of taxa within an environment. Samples below the detection threshold lack diversity measurements; these were excluded from subsequent diversity comparisons.

Statistical analysis

Generalized estimating equations (GEE), extensions of generalized linear models that account for correlation between replicas, were used to describe associations between demographic and clinical factors with diversity measurements after adjusting for genus urotype. A gamma distribution with a log link was assumed for Shannon, Chao, and Pielou diversity measurements, due to their skewed nature. To be inclusive, we did not make adjustments for type I error in the GEE analyses when determining potential clinical and demographic associations with microbiota characteristics. Only results from the lowest detected resolution level (ie, genus) are reported. There was insufficient sample size and power to compare urotypes between catheterized and voided samples. All statistical analyses were conducted using software (SAS, v9.4; SAS Software Cary, NC) and statistical significance was assessed at the α=0.05 level.