Patient cohort, characteristics and outcomes

A total of 452 patients with infertility undergoing IVF were assessed for eligibility in 13 reproductive clinics in Europe, America and Asia between August 2017 and February 2019 (Fig. 1). Forty-four patients were excluded from the study because they did not meet the inclusion criteria (n = 42) or declined to participate (n = 2). The remaining 408 women were recruited and their EB and EF microbiota composition was assessed by 16S rRNA sequencing. However, 66 patients were lost to follow-up. Of the 342 remaining patients, 198 (57.9%) became pregnant [141 (41.2%) had a LB, 27 (7.9%) had a BP and 28 (8.2%) a CM], whilst 144 (42.1%) did not become pregnant (Supplementary Table 1). Moreover, 2 patients experienced an ectopic pregnancy, but their results were not considered for further comparisons due to the small sample size (Fig. 1). Analysed patients had a mean age of 36 years (range 21–49), and a mean body mass index (BMI) of 23.3 (range 18.5–30.0). The ethnic distribution was Caucasian (57.3%), East Asian (14.0%), Hispanic (11.4%) and others (17.3%). The indications for IVF were advanced maternal age, male factor infertility, unexplained infertility and ovarian pathology. The assessed clinical and embryological variables displayed homogeneity and no bias towards the clinical outcome categories was observed (Supplementary Table 2).

Fig. 1
figure1

Flowchart of the study population. Of all patients assessed for eligibility (n = 452), 44 were excluded from the analysis and 66 were lost to follow-up. Thus, 342 patients were ultimately included in our assessment of the impact of the endometrial microbiome on pregnancy outcomes. Abbreviations: BMI, body mass index; ERA, endometrial receptivity analysis; ET, embryo transfer; HRT, hormonal replacement therapy

Endometrial microbiota composition in EF and EB

The EM was profiled in EF from 336 patients and EB for 296 patients, with paired EF–EB results in 290 (84.8%) participants. The mean total sequencing reads per sample was 302,299 (range, 110,050–394,659) in EF samples, and 335,659 (range, 237,889–430,675) in EB samples, with an average of 89,883 (range, 27,960–137,956) and 103,539 filtered reads (range, 61,650–162,653), respectively (Supplementary Table 3).

Because the endometrium presents a low-abundance microbiota, stringent analysis was performed to ensure that contaminating reads did not interfere with downstream analysis. Samples were classified as detectable and not-detectable by comparing them to blank samples included in each run and assessing certain quality parameters (see criteria in Materials and Methods). Despite starting with equivalent amounts of extracted DNA, detectable samples showed a different clustering behaviour as compared with not-detectable/low-biomass samples (with a higher 16S amplicon concentration), which clustered together with blank controls (Fig. 2). After applying these criteria, 208 EF samples and 190 EB samples were classified as detectable and included in the analysis (Fig. 2).

Fig. 2
figure2

Distribution of sequencing data. PCA showing the clustering of the (A) endometrial fluid samples (n = 336) and (B) endometrial biopsy samples (n = 296) and their corresponding blank controls, based on quality parameters such as percentage of empty reads, dispersion index for each sample and the ratio between the filtered and mapped reads. Samples are coloured using a filtered/mapped reads threshold of 0.65 for EF and 0.7 for EB. Abbreviations: BP, biochemical pregnancy; EP, ectopic pregnancy; LB, live birth; CM, clinical miscarriage; NP, no pregnancy

Lactobacillus was the major genus in both EF and EB samples. Bacterial genera such as Anaerococcus, Atopobium, Bifidobacterium, Corynebacterium, Gardnerella, Haemophilus, Microbacterium, Prevotella, Propionibacterium, Staphylococcus and Streptococcus were also commonly identified in both sample types (Supplementary Fig. 1). Streptomyces, Clostridium and Chryseobacterium were detected in EF but not EB, whereas Cupriavidus, Escherichia, Klebsiella, Bacillus, Finegoldia, Micrococcus and Tepidimonas were detected in EB but not EF (Supplementary Fig. 1).

The co-occurring EM bacterial networks showed several differences between the sample types: (i) the EF microbiota had two linked communities, whilst the EB microbiota had four linked communities and two isolated nodes; (ii) the EF microbiota network was more strongly connected than the EB community and (iii) Lactobacillus had positive and negative connected neighbours in the EF microbiota, but only negative relations in the EB microbiota (Fig. 3). In the EB microbiota, Lactobacillus was negatively correlated with pathogenic bacteria Gardnerella, Bifidobacterium and Atopobium, whereas in EF, Lactobacillus was negatively correlated with Gardnerella, Bifidobacterium, Atopobium, Staphylococcus, Streptococcus and Chryseobacterium, and positively correlated to commensal bacteria (Clostridium and Streptomyces).

Fig. 3
figure3

Co-occurrence bacterial networks in endometrial fluid (A) and endometrial biopsy (B) samples. Each network was created by computing the co-occurring bacteria with significant Pearson correlation coefficients. Samples from all reproductive outcomes are represented. Node properties: (i) circle size, proportional to the normalised and standardised bacterial relative abundances; (ii) colour, communities as retrieved by the Louvain algorithm. Edge properties: (i) thickness, proportional to p value of Pearson correlation coefficient, from the most significant (thicker) to the less significant (thinner); (ii) colour, red for negative and grey for positive Pearson correlation coefficients

Endometrial microbiota composition and reproductive outcome

To analyse the association between EM composition in EF and EB and reproductive outcome, we built microbiota networks for each reproductive outcome. We found that the LB category was denser and had a higher node degree distribution than co-occurrence networks of unsuccessful reproductive outcomes. Additionally, we found potential interactions that only occurred in patients with LB, reflecting the relevance of these relationships to successful pregnancy and how their disruption may lead to ecosystem instability. We also noted that in the EF microbiota of patients who had a LB, Lactobacillus was negatively related to dysbiotic bacteria such as Chryseobacterium, Staphylococcus and Haemophilus, and positively correlated to Streptomyces, which in turn was part of a dense community mainly composed of commensal bacteria such as Corynebacterium, Microbacterium, Propionibacterium and Clostridium. In patients with NP, we identified a similar behaviour, with Lactobacillus negatively correlated with Gardnerella, Bifidobacterium, Atopobium, Staphylococcus, Streptococcus and Chryseobacterium, and positively related to Streptomyces. Interestingly, in the group of patients with BP and CM, these interactions disappeared, and the resulting networks were disconnected and formed sparse communities (Fig. 4A). Finally, the EB microbiota networks were more dispersed than the EF ones, with fewer interactions between Lactobacillus and other taxa. Thus, in this case, the eventual beneficial/deleterious connections amongst taxa were less evident (Fig. 4B).

Fig. 4
figure4

Co-occurrence bacterial networks associated with reproductive outcomes. Co-occurrence bacterial networks in A endometrial fluid samples and B endometrial biopsy samples for each ART outcome. Each network was created by computing the co-occurring bacterial communities with significant Pearson correlation coefficients. Node properties: (i) circle size, proportional to the normalised and standardised bacterial relative abundances; (ii) colour, communities as retrieved by the Louvain algorithm. Edge properties: (i) thickness, proportional to p value of the Pearson correlation coefficient, from the most significant (thicker) to the less significant (thinner); (ii) colour, red for negative and grey for positive Pearson correlation coefficients. For association graphs, the same criteria were applied, with the thickness of the circle and colour intensity being proportional to the corresponding Pearson correlation coefficients. Pairs of bacteria without a circle have no significant Pearson correlation coefficient. BP, biochemical pregnancy; CM, clinical miscarriage; LB, live birth; NP, no pregnancy

To avoid potential bias when comparing samples analysed in different runs, bacterial profiles were transformed into centred log ratio (clr) data, and the bacterial communities were analysed according to the difference between Lactobacillus and other reproductive tract taxa using z-score-normalised values. Using these conditions, higher abundance of Lactobacillus was observed in both EF and EB in patients with LB compared to patients with negative reproductive outcomes (Fig. 5A). Taxa with a higher average abundance in unsuccessful outcomes than in LB included Streptococcus, Chryseobacterium, Corynebacterium, Haemophilus, Bifidobacterium, Staphylococcus, Atopobium, Gardnerella and Propionibacterium in the EF microbiota and Gardnerella, Klebsiella, Atopobiumi Finegoldia, Escherichia, Propionibacterium, Haemophilus, Anaerococcus and Bacillus in the EB microbiota (Supplementary Fig. 2). Predictive probability analysis using a Bayesian inference model showed a different highest posterior density (HPD) interval of the difference ‘Lactobacillus – other taxa’ for each ART outcome: LB (−0.12–1.51), NP (−1.05–0.65), BP (−3.45–0.81) and CM (−2.81–0.55). Patients with a LB were more likely to have a higher abundance of Lactobacillus (Fig. 5B). This increased probability of higher Lactobacillus abundance in LB was especially distinct in EF samples, where the HPD for LB samples showed less overlap with unsuccessful outcome intervals.

Fig. 5
figure5

Lactobacillus is more abundant than other taxa in reproductive success vs failure. A Difference between Lactobacillus and other reproductive tract taxa using z-score-normalised values in endometrial fluid (left panel) and endometrial biopsy (right panel) samples. B Predictive model showing the probability of each reproductive outcome based on the EM profile. Posterior predictive distribution density plot of z-score differences between Lactobacillus and other reproductive tract taxa by reproductive outcome. BP, biochemical pregnancy; CM, clinical miscarriage; LB, live birth; NP, no pregnancy

Finally, we compared the EM bacterial profiles in patients who achieved a successful pregnancy with LB versus those with unsuccessful outcomes (BP, CM and NP). Our hypothesis was that the microbiota composition in patients with LB is the physiological scenario and does not interfere with functional reproductive potential. Therefore, we evaluated the distance between the abundance of each bacterial taxon and reproductive outcome in the upper and lower 95% confidence intervals established for patients with LB (Supplementary Fig. 3). In patients with NP, the EF taxa with significantly higher abundance exceeding the established upper confidence interval were Atopobium, Bifidobacterium, Chryseobacterium, Gardnerella and Streptococcus, and in those with CM, Haemophilus and Staphylococcus exceeded the physiological levels. By contrast, taxa with a significantly higher distance to the lower established confidence interval were Lactobacillus and Microbacterium in NP patients, and Lactobacillus in patients with CM (Fig. 6A). In the EB microbiota of NP patients, Bifidobacterium, Gardnerella and Klebsiella were significantly more abundant, and the abundance of Cupriavidus, Finegoldia, Lactobacillus and Tepidomonas was significantly below the established confidence interval (Fig. 6B). The remaining comparisons did not reach statistical significance, possibly due to the small number of patients with BP and CM.

Fig. 6
figure6

Pathogenic bacterial profiles significantly associated with reproductive outcome. Box plots showing taxa with significant differential abundance in no pregnancy (NP), biochemical pregnancy (BP) and clinical miscarriage (CM) compared to live birth (LB). Differential abundance was calculated using the distance of each value to the upper (U) or lower (L) bounds for the 95% CI in LB (Supplementary Fig. 3) in A endometrial fluid and B endometrial biopsy samples. Only taxa with significant differential abundance, calculated with a two-sided Mann-Whitney U test, are represented in the graphs. BP, biochemical pregnancy; LB, live birth; CM, clinical miscarriage; NP, no pregnancy

Endometrial microbiota composition in chronic endometritis and reproductive outcome

We also evaluated the abundance of the main pathogenic bacteria reported to cause chronic endometritis (CE): Enterococcus, Escherichia, Klebsiella, Streptococcus, Staphylococcus, Gardnerella, Mycoplasma, Ureaplasma, Chlamydia and Neisseria. These bacteria are considered to be a potential cause of infertility as well as obstetric and neonatal complications [28, 30]. We compared the abundance of these bacteria with the confidence interval generated for infertile patients that achieved a LB.

Of the CE pathogens, Gardnerella, Klebsiella and Streptococcus were significantly increased in the EF microbiota of NP patients, whereas Enterococcus was increased in patients that experienced BPs, and Klebsiella and Staphylococcus were increased in CM (Fig. 7A). In the EB microbiota, Gardnerella, Neisseria and Klebsiella were significantly enriched in women with NP compared to those that achieved LB, whilst Enterococcus abundance was below the confidence interval (Fig. 7B). In the remaining unsuccessful reproductive categories (BP and CM), no significant taxa were detected. Interestingly, Gardnerella and Klebsiella were the only common pathogens significantly enriched in both EF and EB from patients with NP.

Fig. 7
figure7

Chronic endometritis profile associated with reproductive outcomes. Box plots showing differential abundance in chronic endometritis-causing bacteria in no pregnancy (NP), biochemical pregnancy (BP) and clinical miscarriage (CM) compared to live birth (LB). Differential abundance was calculated using the distance of each value to the upper (U) or lower (L) bounds for the 95% CI in LB (Supplementary Fig. 3) in A endometrial fluid and B endometrial biopsy samples. Only taxa with significant differential abundance, calculated with a two-sided Mann-Whitney U test, are represented in the graphs. BP, biochemical pregnancy; LB, live birth; CM, clinical miscarriage; NP, no pregnancy

Endometrial microbiota composition fingerprinting is associated with reproductive outcome

In summary, the pathogenic profile associated with reproductive failure in our cohort of infertile patients consisted of Atopobium, Bifidobacterium, Chryseobacterium, Gardnerella, Haemophilus, Klebsiella, Neisseria, Staphylococcus and Streptococcus. In contrast, Lactobacillus was consistently enriched in the EM (both in EF and EB) from patients that achieved LB. Therefore, patients with a higher abundance of lactobacilli are more likely to achieve reproductive success. Also, some commensal bacteria such as Cupriavidus, Finegoldia, Microbacterium and Tepidimonas were positively associated with LB.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Disclaimer:

This article is autogenerated using RSS feeds and has not been created or edited by OA JF.

Click here for Source link (https://www.biomedcentral.com/)

By admin