Description Of The Studied Variables
The study included 48 participants, who provided 273 samples, of which 264 were analyzed. Figure 1 details the number of patients and samples initially included in the study, losses and net inclusion data and analysis. Sociodemographic characteristics and clinical outcomes of the patients included in the study were recorded for the whole sample and according to achievement of an evolutionary pregnancy (Table 1), as well as patients with and without repeated implantation failures (RIF and NO RIF). At the level of the graph analysis, the “NO RIF” group is labeled as a control group for methodological issues, though it was not treated as such. Participants’ mean age was 39.44 years, and their mean weight was 63.41 kilos. Smokers made up 14.29% of the sample, while 0.67% had had previous pregnancies—58.33% of which ended in miscarriage. In addition, 43.75% had carried out the treatment with donated oocytes. All the patients underwent endometrial preparation under hormonal replacement therapy. In terms of clinical data, 54.2% of the patients had a positive pregnancy test. The clinical gestation rate was 43.8%, with an evolutionary pregnancy rate of 37.5%. The rate of biochemical miscarriage was 10.4%, and of clinical miscarriage, 14.28%. Of the patients who achieved pregnancy, 38.9% had a history of miscarriages, compared to 70.0% of those who did not conceive (p = 0.034).
Table 1
Patients socio-demographic characteristics and clinical outcomes, by pregnancy outcome
Variables | Total sample N = 48 | Pregnancy N = 26 | No pregnancy N = 22 | P value |
Age, years, mean ± SD | 39.44 ± 3.82 | 38.28 ± 3.39 | 40.13 ± 3.95 | 0.09235 |
Weight, kilos, mean ± SD | 63.41 ± 9.79 | 59.57 ± 7.98 | 64.69 ± 10.16 | 0.1944 |
Height, cm, mean ± SD | 162.33 ± 6.89 | 160.25 ± 10.41 | 163.09 ± 5.22 | 0.4806 |
Tobacco user (%) | 14.29 | 16.7 | 11.1 | 0.5975 |
N previous pregnancies, mean ± SD | 0.67 ± 0.60 | 0.56 ± 0.51 | 0.73 ± 0.64 | 0.2955 |
Previous miscarriages (%) | 58.33 | 38.90 | 70.0 | 0.0343* |
N of previous miscarriages, mean ± SD | 1.44 ± 1.80 | 1.11 ± 1.79 | 1.63 ± 1.81 | 0.3343 |
Previous treatments (%) | 77.08 | 72.20 | 80.0 | 0.5348 |
Semen analysis, normozoospermia (%) | 77.08 | 83.30 | 73.30 | 0.6164 |
Donated semen (%) | 20.83 | 22.20 | 20.0 | 0.8544 |
Endometrial thickness, mm, mean ± SD | 8.40 ± 2.04 | 8.54 ± 2.01 | 8.32 ± 2.09 | 0.7254 |
B-HCG + (%) | 54.2 | — | — | — |
Clinical gestation (%) | 43.8 | — | — | — |
Evolutionary pregnancy (%) | 37.5 | — | — | — |
Biochemical miscarriage (%) | 10.4 | — | — | — |
Clinical miscarriage (%) | 14.28 | — | — | — |
Generalized Mixed Model (logistic Function)
A generalized mixed model was constructed to determine the evolution of the vaginal microbiome pattern and its association with the gestation rate. The variation in alpha diversity (Shannon index) at different times of the cycle was assessed, and there were no statistically significant differences in the evolution of the vaginal microbiome patterns between visits, according to the treatment, or based on the achievement of clinical gestation (p = 0.412). Analyzing this evolution by RIF versus NO RIF groups (p = 0.019), we observed a variation in the vaginal microbiome pattern over time in the NO RIF group. Specifically, these patients showed a decrease in alpha diversity from the follicular to the luteal phase. In contrast, the RIF group showed a stable microbiome pattern across different timepoints. This lack of dynamism in the pattern of the vaginal microbiome in RIF patients could entail a lack of adaptation to endometrial physiology and preparation, and therefore a worse prognosis for embryo implantation (Fig. 2).
Vaginal And Endometrial Microbial Patterns At Visit 1
We observed statistically significant differences alpha diversity between endometrial and vaginal samples (p = 0.0139 for Shannon index and p = 0.046 for Simpson index), with higher values in endometrial samples (Fig. 3a).
Using PERMANOVA, the matrices with beta diversity measures showed statistical differences in composition according to the type of sample (p = 0.001). The unweighted UniFrac PCoA revealed a clear pattern of separation between vaginal and endometrial samples (Fig. 3b). The endometrial samples are grouped at the extreme right of the graph. The percentage of the variance explained by each component is shown on the axes (Principal component (PC), PC1: 34.16%, PC2: 20.31%, PC3: 8.58%). The first and second components would together explain more than 50% of the variability between our samples.
Regarding the taxonomic characterization in all the samples of the study, there was a clear dominance of the genus Lactobacillus in both the vaginal and the endometrial microbiome. The bar chart of the relative frequency of the most abundant genera grouped by sample type (vaginal/endometrial) is shown in Fig. 3c.
Microbiome profiles showed relative differences in genera and species present in the vaginal and endometrial samples. The univariate analysis reached statistical significance for Lactobacillus spp., Streptococcus spp., Ureaplasma spp., Delftia spp., Anaerobacillus spp. and L. Helveticus spp. Several genera were more abundant in the vagina than in the endometrium: Lactobacillus, 84.82% versus 83.17% (p < 0.0001); Streptococcus, 7.74% versus 1.59% (p = 0.014), and Ureaplasma, 0.89% versus 0% (p = 0.0062). The other genera showing significant differences were more abundant in the endometirum: Delftia 0.95% versus 0% in the vagina (p < 0.0003); Anaerobacillus, 1.59% versus 0% (p -value = 0.0004), and Ralstonia 3.17% versus 0% (p = 0.0006) (Table 2).
Table 2
Differences in the genera present in microbiome profiles in vaginal and endometrial samples
Genus | Endometrium | Vagina | p-value |
Lactobacillus spp. | 83.17% | 84.82% | < 0.001 |
Delftia spp. | 0.95% | 0.00% | 0.0003 |
Anaerobacillus spp. | 1.59% | 0.00% | 0.0004 |
Ralstonia spp. | 3.17% | 0.00% | 0.0005 |
Ureaplasma spp. | 0.00% | 0.89% | 0.0062 |
Streptococcus spp. | 1.59% | 7.74% | 0.0187 |
Figure 3d shows the relative frequency of the most abundant species for each sample type. Lactobacillus iners presents a higher relative abundance in endometrial samples (64% versus 40% in vaginal samples; p > 0.05). There was a significant difference in the abundance of L. helveticus spp.: 28% in the endometrium versus 47% in the vagina (p = 0.0001).
Microbiome patterns by diagnosis of implantation failure
Vaginal microbiome pattern
Regarding the vaginal microbiome pattern in the samples obtained at visit 1, we found no differences in alpha diversity between the RIF and NO RIF groups according to either the alpha Shannon or Simpson indices (Fig. 4a). A more detailed analysis showed that the results were statistically significant only for the Faith index. The box diagram for the alpha Faith phylogenic diversity index (phylogenetic analogue of taxon richness expressed as the number of tree units which are found in a sample) yielded a p value of 0.027, representing a significantly lower Faith alpha diversity index in the RIF group compared to the NO RIF group.
In relation to beta diversity, no statistically significant differences were observed between the two groups (Fig. 4b). Likewise, the univariate analysis showed no statistically significant results. In relation to the taxonomic allocation, the RIF group had a lower relative abundance of the genus Streptococcus, and a higher abundance of Prevotella spp., Ureaplasma spp., and Dialister spp. The NO RIF group presented a higher relative abundance of Streptococcus spp., Veionella spp., and Aerocuoccus spp. As for the genus Lactobacillus, no differences were observed between groups (Fig. 4c). At the species level, we observed a higher relative abundance of L. helveticus in the RIF patients, and of L. iners, L. jensenii, L. gasseri and L. agalactiae in the NO RIF group patients (Fig. 4d).
Endometrial microbiome pattern
The alpha diversity of the endometrial microbiome at visit 1 was significantly higher in the NO RIF group (Fig. 4e; p = 0.0206 for both Shannon and Simpson indices). There were also statistically signficant differences in beta diversity, as seen in the PCoA graph (Fig. 4f). There is a clear pattern of separation between the RIF group and the NO RIF group: the RIF samples fall in the top center of the graph and those collected in the NO RIF group are clustered in the centre. A larger sample size would help us to corroborate this difference. On the axes, there is the percentage of the variance, explained by each component (PC1: 31%, PC2: 14.1%, PC3: 8.9%). The results for the first and second component explain more than 45% of the variability between our samples.
The taxonomic assignment in the frequency table represents the relative abundance of the different taxa present in the samples for the RIF and NO RIF groups. A greater abundance of the genus Prevotella is observed in the RIF group (Fig. 4g). In the univariate analysis we found statistically significant differences for the genus Ralstonia, observing a much higher relative abundance in the NO RIF group compared to the RIF group (0.73% versus 0.09%; p = 0.0012).
Figure 4h shows the differences in relative abundance at the species level. L. iners and L. jensenii were more abundant in the NO RIF group, while L. helveticus and Sneathia amnii had a larger presence in the RIF group.
Evolution Of The Vaginal Microbiome (whole Sample)
Diversity analysis
There are no statistically significant differences in alpha or beta diversity between the samples over the different visits.
Taxonomic characterization
We did not observe any statistically significant difference between the visits for either the composition of genera or species (Fig. 5). There were some apparent changes in abundance of the genera Lactobacillus, Streptococcus and Prevotella: both Lactobacillus and Streptococcus were more abundant on visits 1 and 2, showing a decrease on visit 3. Prevotella shows a higher abundance on visit 1 and 3, especially on the latter timepoint. In the univariate analysis there were no statistically significant differences.
At the species level, the bar chart shows some differences in relative abundance for the following species: L. helveticus, L. iners, L. gasseri and L. jensenii (Fig. 5). L. helveticus was most abundant on visit 2; L.iners, on visit 1; and L. gasseri, on visit 3. At that timepoint, results showed a smaller proportion of L. jensenii. However, these differences were not statistically significant.
Association of the vaginal sample taken at different visits with the gestation rate
Diversity analysis
Analyzing diversity as a function of the gestation rate, we observed a greater alpha diversity in patients who do not achieve pregnancy, obtaining a trend without reaching statistically significant values (Shannon p = 0.0748 and Simpson p = 0.0856). Regarding the beta diversity, no statistically significant differences were found at visit 1 according to gestation rate. For the samples collected at visit 2, the differences in alpha diversity were not statistically significant; however, there is a trend suggestive of a negative correlation between the gestation rate and alpha diversity (p = 0.1518). For beta diversity, we found no difference in relation to visit 2 and the pregnancy rate. The samples taken at visit 3 show no difference in alpha or beta diversity.
Taxonomic characterization
At visit 1, participants who achieved pregnancy presented a significantly greater abundance of Lactobacillus spp. than those who did not, while Streptococcus spp. and Prevotella spp. were more abundant in the latter group (Fig. 6a). Streptococcus and Prevotella may thus be associated with a poor prognosis with regard to gestation. On the other hand, an abundance of Lactobacillus spp. could be indicative of more favorable conditions. The differences were observed at the genus level for Lactobacillus spp. (91% with no gestation vs 99% with gestation; p = 0.0445) and at the species level for L. reuteri (0.39% vs 0.17%; p = 0.0397; Fig. 6b).
Similar results were obtained at visit 2. Those who achieved pregnancy presented a greater relative abundance of Lactobacillus spp.than those who did not (97.69% versus 94.63%; p = 0.0268; Fig. 6c-6d). The opposite was true for the case of Streptococcus spp. (Fig. 6c).
Findings at visit 3 were similar (Fig. 6e). The univariate analysis showed statistically significant differences (p = 0.0492) for the genus Lactobacillus spp. (99.74% with gestation versus 97.73% without) and the species L. reuteri (0.30% versus 0.15%, respectively; p = 0.0591; Fig. 6f).