Assessing the Concordance Between Urogenital and Vaginal Microbiota: Can Urine Specimens Be Used as a Proxy for Vaginal Samples?

Abstract

Background: The vaginal microbiota play a key role in defense against reproductive tract infections; however, many population-based women's health studies do not collect vaginal samples. Molecular examinations of urine samples have revealed common vaginal bacteria. We sought to assess the extent that community state type assignments of archived random-catch and clean-catch urine samples agreed with the paired vaginal samples in both reproductive-age and peri/post-menopausal women.

Results: Using archived samples, we evaluated the microbiota concordance among women in three studies: two with paired mid-vaginal/random-catch urine (N=91 reproductive-age participants and N=13 peri/post-menopausal participants), and one with paired mid-vaginal/clean-catch urine (N=99 reproductive-age participants). Microbiota composition was characterized by sequencing amplicons of the 16S rRNA gene V3-V4 regions and assigned to community state types. Similarity of paired samples was gauged using agreement of community state types and Yue-Clayton θ indices. Analysis of Composition of Microbiomes II indicated which taxa were differently relatively abundant in paired vaginal and urine samples. In reproductive-age women, random-catch and clean-catch urines were 89.0% and 86.9% concordant on five community state types with paired mid-vaginal swabs, and Kappa statistics indicated almost perfect agreement (κ_random-catch=.85, κ_clean-catch=.81, p<0.0001). A small number of pairs of samples were discordant (23/190, 12%), and discordant pairs tended to be between samples classified to L. iners-dominated and/or low-Lactobacillus states. Concordance and agreement remained similar when dichotomizing the microbiota to Lactobacillus-dominated versus low-Lactobacillus microbiota, as well as when evaluating separately the three subtypes of the low-Lactobacillus community state type IV. Median similarity of paired urine/vaginal samples was high (θ_random-catch=.85, θ_clean-catch=.88), and a comparison of the random-catch and clean-catch similarity scores showed no significant difference (p=.80). Concordance and similarity were lower for peri/post-menopausal women, but agreement remained substantial (76.9% concordant, κ_random-catch= 0.64, θ_random-catch=.62). Taxonomic-level analysis confirmed these findings.

Conclusions: Random-catch and clean-catch urine samples showed substantial agreement on bacterial composition to paired mid-vaginal samples, indicating that the genitourinary microbiota may be a reliable proxy for assessing the overall composition of the vaginal microbiota via community state types. This data suggests that urine samples can, with proper interpretation, be utilized as a surrogate for developing preliminary data and hypothesis-generating studies.

Keywords: 16S rRNA gene amplicon sequencing; clean-catch urine; community state type; random-catch urine; vaginal microbiota.

Figure 1

Heatmap displaying the relative abundance of the 25 most abundant taxa in random-catch (A) or clean-catch (B) urine samples and paired vaginal swabs. Community state types (CSTs) are indicated in the second row from the top.

Figure 2

Example taxonomic composition for paired urine and vaginal samples that were concordant on CST in (A) reproductive-age women (paired clean-catch urine IDs: 1003, 1036. Paired random-catch urine IDs: 14, 15, 51) and (B) peri/post-menopausal women (all random-catch urine). Assigned CSTs are indicated above each sample pair.