O-279 The lower reproductive tract microbiota composition alters upon ovarian stimulation and is associated with, but not accurately predictive of reproductive outcomes following ART

Abstract

Abstract Study question Are ovarian stimulation (OS) and ART success associated with the microbiota of the lower reproductive tract (LRT), i.e. vagina and cervix? Summary answer Estrogen rise during OS has a pivotal role in the LRT microbiota variation. The LRT microbiota pre-OS, although associated, may not accurately predict ART success. What is known already OS is a key component to enhance success rates in ART. OS triggers a significant rise in serum estrogen levels, which are known to influence the vaginal microbiota. The role of the female LRT microbiota in ART success is increasingly recognized, yet the impact of OS, particularly the estrogen rise, on these microbial communities and how the change may influence ART success remains largely unknown. Additionally, finding putative biomarkers within LRT microbiota could aid in predicting ART success. Understanding this relationship could refine and personalize fertility treatments, but robust evidence is needed to guide microbiota-based clinical decisions in ART. Study design, size, duration Vaginal and cervical swabs were collected longitudinally from patients undergoing ART and this at baseline (pre-OS) and at oocyte retrieval (post-OS). We gathered a final dataset comprising 275 vaginal samples at baseline and 240 samples at oocyte retrieval. Additionally, we defined a prospective cohort of n = 228x2 samples corresponding to patients for which both pre- and post-OS samples were available. The main results as detected within the vaginal samples were confirmed in the cervical samples. Participants/materials, setting, methods Bacterial DNA from −80 °C preserved swabs was extracted in a Tecan Nucleic Acid Extraction Platform, using the QIAamp DNA-Stool-Mini-Kit. Samples for paired-end Illumina MiSeq were constructed using a two-step PCR amplicon approach targeting the V4 region of the 16S rRNA. Then, LotuS, DADA2, the RDP classifier trainset 16 and variance stabilisation (poscounts, DESeq2) were used for read de-multiplexing, pre-processing, taxonomy assignment and transformation, respectively, as recommended for zero-inflated compositional data. Main results and the role of chance First, a confounder analysis revealed that both the menstrual phase and intercourse correlated with the LRT’s baseline microbiota composition (multivariate distance-based redundancy analysis [dbRDA], AdjP<0.05, N = 275). Next, in the paired cohort, we observed that the OS-associated rise in estrogen was correlated with confounder-independent compositional (confounder-adjusted dbRDA, AdjP=0.002, R2=1.14%, N = 228x2) and microbial-feature changes (Wilcoxon test, AdjP<0.05, Ndelta=228) in the LRT. Interestingly, post-OS we noted a decrease in Lactobacillus proportions compared to baseline (in participants sampled at either the follicular or luteal phase of the menstrual cycle). Additionally, post-OS microbiota composition was not associated with the live birth rate (LBR, following the first embryo transfer of the cycle) or the cumulative LBR (CLBR, following all embryo transfers of the cycle), nor with the total number of high-quality (HQ) embryos available (dbRDA, AdjP>0.05, N = 240). Notably, pre-OS microbiota communities and LRT Anaerococcus and Peptoniphilus proportions were associated with (C)LBR (confounder-adjusted dbRDA, AdjP<0.05; Logistic regression, odds-ratio<1, AdjP<0.05, N = 275). Machine learning models determined that vaginal Peptoniphilus and Anaerococcusproportions alone did not have good accuracy/ROC for predicting (C)LBR (test-set;<70%). Notably, the number of HQ embryos showed good accuracy/ROC for CLBR (test-set; 80%/84.5%). Including vaginal Anaerococcus proportions only slightly improved the model’s accuracy/ROC (test-set; 82.5%/85.6%). Limitations, reasons for caution The data’s proportional nature restricted assessing whether the estrogen rise decreased Lactobacillus proportions or increased other bacteria. Luteal phase sampling was avoided given the intravaginal progesterone support. Larger datasets may further improve predictive models. Wider implications of the findings Our study highlights the confounding impact of baseline variables and the influence of the OS-associated estrogen rise on LRT microbiota. It questions the clinical relevance of the LRT microbiota’s impact on reproductive outcomes following ART, warning for caution in microbiota-based decisions in ART. Trial registration number NCT03105453