1. Introduction
In the last decades, increasing evidence has highlighted the presence of a unique microbiome not only in the lower, but also in the upper female reproductive system. Indeed, several studies have refuted the, hitherto assumed, sterile environment of the uterus and described distinct endometrial microbiota profiles [
1]. The accumulating evidence suggests that the female reproductive tract is an open system with a continuum of microbiota gradually changing from the outer to the inner organs, with an increasing pH and a decreasing bacterial abundance from the vagina to the endometrium [
2]. Yet, the number of well-powered clinical studies is still limited. As such, in contrast to the vaginal niche, a healthy or eubiosis-related endometrial microbiome is still not defined. The microbiota-immunity crosstalk seems of major importance in the upper female reproductive tract. Current theories suggest that a dysbiosis-related state triggers an inflammatory response in the endometrium that affects the embryonic implantation [
3]. As such, a close relationship is suspected between the endometrial microbiome and conception, (sub)fertility, and pregnancy outcome [
4,
5,
6].
A recent review article summarizes the findings of the currently available data with regard to female reproductive tract microbiota. The authors highlight the potential implication of these data in reproduction medicine and suggest further research and protocol standardization [
7]. To date, all studies investigating the endometrial microbiome, apply 16S rRNA or metagenomics-based methodologies. Both approaches include whole genome sequencing (WGS) to analyze the genetic material present in the microbiome [
8]. WGS enables a culture-free analysis of the microbiome, within a relatively short time and with limited effort. Despite the many advantages of this technique, there are also disadvantages to take into consideration.
Caution is specifically needed when sequence-based techniques are applied to study low biomass biological samples. DNA/RNA contamination in laboratory reagents and extraction kits can significantly bias the results of microbiota studies. This contamination is a concern for both rRNA gene sequencing requiring polymerase chain reaction (PCR) amplification, but also for shotgun metagenomics that do not require PCR. Many recent sequence-based studies describing microbial communities of low-biomass environments do not report the DNA/RNA quantification on the initial samples and do not include negative controls [
9]. Another important drawback is the lack of standardization in the extraction and sequencing protocols [
10]. Finally, a sequencing analysis cannot differentiate between living organisms and genetic fragments [
11,
12].
Because of the above-mentioned limitations of a sequencing-based microbiome analysis and the recent appreciation of the endometrial microbiome in female health and fertility, the application of alternative methodologies could be of relevance for upcoming studies and clinical practice. High-throughput culturing, or culturomics, combines a plethora of different enrichment broths, agar plates, and (an)aerobic incubation conditions to cultivate virtually all viable microbiota [
13]. The cultured species are identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) or 16S rRNA sequencing. Although culturing and MALDI-TOF MS are routinely used in a clinical laboratory setting and predate sequencing techniques, the high workload and poor traceability limit their application in microbiome research. Yet, culturomics-obtained microbiome data can be complementary to the metagenomics approach [
14]. With the use of several culturing media and the inclusion of adequate negative controls, the depth and DNA-extraction bias of a metagenomics-based microbiome analysis are bypassed in culturomics. A major advantage of the culturing approach is that viable bacteria and fungi are identified on a species level. In addition, microorganisms present at a very low biomass level can be detected, allowing for the detection of minority populations that potentially have a substantial effect on the ecology of microbiota [
12]. Even though, culturomics provides additional information, this information is only available in qualitative terms, where sequencing approaches also give a representation of the quantitative composition of microbiota.
Previous studies highlight certain challenges in a successful culturomics set-up. Fastidious anaerobic species, such as
Fusobacterium nucleatum and
Peptostreptococcus anaerobius, often grow very slowly, require special nutrients, are rapidly grown over by other species, and die in the presence of oxygen [
15]. As a result, most of these species are not suitable for standard cultivation on agar plates. A primary incubation in an anaerobic liquid enrichment medium provides a suitable environment for the growth of these species. The subsequent agar inoculation from these broths enables the detection of these slow-growing anaerobes [
14]. Compared with other ecosystems, culturomics has mostly been applied to the human gut microbiota. However, this technique also showed its utility on vaginal microbiota. Since the application of culturomics at least 15 new species from the vaginal tract haven been reported [
13,
15].
In the current proof-of-concept study, we applied a culturomics-based analysis of endometrial samples. The enrichment and incubation conditions were based on previously described protocols [
14]. To lower the workload and standardize the culturomics approach, we applied a semi-automated inoculation and incubation system: WASPLab
® (Copan Diagnostics, Brescia, Italy). As such, we are the first to describe a culturomics-based endometrial microbiome profile.
2. Results
Ten endometrial biopsies were included in the current study. The characteristics of the women are depicted in
Table 1. A total of 2933 colonies were identified using the MALDI Biotyper
® system. Eighty-five different microorganisms are listed in
Figure 1 and
Figure 2 and
Table 2, including two fungi (
Candida glabrata and
Candida parapsilosis). Twenty-six of these bacterial isolates (31%) were Gram-negative and 33 (39%) were obligate anaerobes. These 85 species belonged to 40 different genera and 28 different families. Four species
(Lactobacillus coleohominis, Prevotella colorans, Porphyromonas bennois and
Paenibacillus xylanilyticus) were not identifiable with MALDI-TOF MS and were identified using 16S rRNA gene sequencing. One bacterial isolate could only be identified on the genus level (
Demequina species).
Corynebacterium, Lactobacillus, Prevotella and
Staphylococcus are the best-represented genera among the ten samples in terms of the diversity in the species, as we identified seven species within each of them.
Stenotrophomonas maltophilia was found in one of our negative control samples. Therefore, the observation of this species was excluded from the analysis. The mass spectrometry used in this study cannot differentiate between Peptoniphilus harei and Peptoniphilus indolicus. Therefore, in subsequent analyses, they have been treated as one species as this was considered to cause less of an error than leaving them out completely.
In an average endometrial biopsy, 18.6 different species were cultured (8.6 with direct inoculation and 12.9 with pre-incubation). The minimum amount of species found in a sample was seven, the maximum was 41. A very similar number of species were found without the pre-incubation, with the aerobic pre-incubation or with the anaerobic pre-incubation. This holds true both for the absolute number of species, as well as the unique number of species not found in either of the other two incubation options, illustrated in
Figure 1,
Figure 2, and
Table 3. The different incubation-techniques all contribute substantially to the identification of the most clinically relevant species, such as the various
Lactobacillus and bacterial vaginosis and aerobic vaginitis-related species (
Supplementary Table S1). Yeasts, however, were exclusively found after pre-incubation.
As can be expected, there is quite some overlap in the species found within each sample, for each technique. Nevertheless, they each contribute enough additional information to justify performing each, in addition to the other two techniques (
Figure 3 and
Figure 4). Removing any of the three incubation options in this study would have led to an average loss of 20% to 25% of the species found per sample. Surprisingly, the number of obligate anaerobic species found in the anaerobic pre-incubation was slightly lower than in the aerobic pre-incubation (
Figure 5). For example,
Anaerococcus murdochii, Peptoniphilus duerdenii, Peptostreptococcus anaerobius and
Prevotella buccalis are obligate anaerobes and were only identified using aerobic pre-incubation.
It appears that a non-negligible part of the endometrial microbiome consists of slow growing species, as almost one sixth of the species found only showed up after 30 days of pre-incubation (
Figure 5).
Actinomyces radingae, Bacteroides coagulans, Demequina species,
Dermacoccus nishinomiyaensis, Paenibacillus xylanilyticus, Porphyromonas bennonis, Prevotella colorans and
Trueperella bernardiae were only identified after 30 days of pre-incubation.
Aside from the position of sample one, clustering was fairly consistent across single, average, and complete linkages. There appears to be some correlation between patients’ age and the clusters they are in, although this is most likely a result of random chance (
Figure 6).
3. Discussion
Culturomics has proven to be a very valuable tool in exploring human microbiota and has vastly enhanced our understanding of it [
13]. This is the first time the approach has been applied to endometrial biopsies, despite the very high need in expanding the knowledge about it. In this study, we conducted culturomics with the help of total lab automation (TLA) on 10 endometrial biopsies. The data show that, in strong contrast to what has been reported in other studies [
17], many microorganisms out of the endometrium can be cultured. Furthermore, we demonstrated that the TLA systems, as for example the WASPLab
® used in our study, could be valuable tools in tackling the extremely high workload of culturomics and provide standardization and a perfect traceability.
Based on the review article of
Punzón et al. [
7], we can report a good concordance with the previous described metagenomics-based endometrial microbiota in subfertile and infertile women. Seventy-seven (91%) of the identified species belonged to previously described genera, present in the endometrial microbiome. We described 53 species and seven genera for the first time in the endometrial microbiome. However, we did not find species belonging to
Acinetobacter or
Pseudomonas, two of the most frequently described genera in the endometrial microbiome, based on sequencing. This may be explained by the fact that the species within the genera
Acinetobacter and
Pseudomonas are known to be contaminants of DNA-extraction kits, DNA-free water, PCR primers and other laboratory reagents. This contamination is a real challenge to study low-biomass microbiota such as, the endometrial microbiome, which provide very little template DNA to compete with that in the reagents for sequencing. The importance of this issue when analyzing low biomass samples, despite the multiple reports of reagent contamination, remains underappreciated in the microbiota research community [
9]. A culture-based approach such as the one we used, could overcome these biases.
Interestingly, we found more obligate anaerobic species using the aerobic pre-incubation bottle than with the anaerobic pre-incubation bottle. We cannot give an unequivocal explanation for this, but it was notable that the
Lactobacillus species often grew much more abundantly in the anaerobic bottle than in the aerobic bottle. This lowers the pH, which inhibits the growth of some anaerobic bacteria [
18]. Finally, the aerobic atmosphere in the aerobic bottle is only temporary, since all of the oxygen present in the bottle is used after a few days of incubation. Plate streaking is a major part of the culturomics workload. The inoculation procedures critically influence the specimen identification, mainly by the ability to generate single colonies. When performing plate streaking manually, reproducibility is poor. When using WASPLab
® for plate streaking, a perfect reproducibility is almost guaranteed. The automated plate inoculation will also lead to a higher frequency of single-colony recovery, resulting in an increased detection of species. On top of that, inoculation with WASP
® is time-saving: Quiblier et al. showed 25 min saved “hands-on time” for every 100 plates streaked [
19]. Following the inoculation, there is a swift and automatic transfer into the incubators with a more stable temperature control than classic incubators. A perfect traceability is guaranteed due the unique label each streaked plate is given by the system.
Culture-based studies on endometrial samples are rare. Only very few studies describe the use of culture techniques on endometrial samples, almost all of them looking for very specific and possible pathogenic species, such as
Escherichia coli and
Gardnerella vaginalis. To our knowledge, only Smolnikova et al. tried to culture endometrial samples with extended nutrient non-selective media [
16]. In their study, they enrolled 80 patients with fertility problems. Following a routine embryo transfer, the most distal five mm portion of the embryo transfer catheter was used for the culture with extended nutrient media. However, these nutrient media were not defined in their article. Disseminated over these 80 samples, they managed to identify 33 different bacterial species. In contrast with our culturomics approach, they established a low bacterial diversity: 28% of the endometrial samples showed a monoculture and 12% of the samples showed no bacterial growth at al. Of the 33 species they described, we found 21 (64%) back in our study. They described that the
Lactobacillus species were the most frequently isolated, with a frequency of 80%. We confirmed this observation as we isolated the
Lactobacillus species in all 10 samples, with
Lactobacillus jensenii as most commonly isolated species. Studies based on metagenomics also confirm the significant part of
Lactobacilli in the endometrial microbiome, suggesting
Lactobacillus as the most represented genus in a endometrial microbiome [
7]. However, there is still no consensus. Additional findings of our study highlight the importance of aerobic and anaerobic incubation, with and without pre-incubation, as dysbiosis-related strains seem to dominate without pre-incubation enrichment (
Supplementary Table S1). As such, the inclusion of different types of sample incubation conditions are very important to achieve a non-biased culture-based approach on the microbiota composition of endometrial samples.
The first drawback of this study is the relatively small amount of used culture conditions that were selected, based on previous studies performed on stool samples [
14]. Despite the high bacterial diversity resulting from our bacterial culture media, a more extensive bacterial cultivation would probably expand the amount of identified species. Although being very challenging, it could be interesting for subsequent studies to take into account the physicochemical properties of the endometrium when selecting additional culture conditions. For example, one could experiment with pH, temperature and hormones: three important and dynamic factors in the endometrium [
20]. These factors are variable across the follicular, ovulatory and luteal phases, resulting in dynamic uterine microbiota during the menstrual cycle [
21]. Although the selection of additional culture media has to be considered very carefully and is not directly proportional to the discovery of additional species, such as was highlighted by Diakite et al. One of the most important factors is probably the use of rumen fluid in the blood culture bottles, which showed to have a great impact in the isolation of strict anaerobic bacteria [
14]. A second and insurmountable drawback of this study is the possible sample contamination with microorganisms originating from the lower reproductive tract. The impact of various contaminating interventions makes it hard to only detect the low-abundance microorganisms originating from the sampling site [
22]. Indeed, due to the extremely high biomass in the lower reproductive tract, compared to the uterine microbiome and the transcervical approach that cannot be avoided, the misrepresentation of the species distribution cannot be excluded. We have to keep in mind that surgical manipulations and instruments disrupting the cervical barrier, such as, for example a hysteroscopy, could carry over traces from the lower reproductive tract to the uterus. However, the exact impact of these manipulations is unclear and needs to be further investigated. For example, by comparing cultures of endometrial biopsies before and after a hysteroscopy. The influence of the blood microbiome should also be taken into account in subsequent studies, as it seems likely that some types of microorganisms could move to the endometrium via the blood [
23]. Then, we have to keep in mind that only a relatively small number of patients was included. Because 38 of the discovered species were found in only one of the 10 samples, the number of species would probably increase strongly when expanding the study-population. As such, the study-population is too small to consider results from clustering in any way representative of the general population. While no conclusions should be drawn from any specific clusters in this data set, it is clear that the samples are quite informative, as there is sufficient variance and range in the distances. This is partly due to the fact that the distances can be calculated down to the species level. Therefore, it clearly illustrates the potential of culturomics for differentiating between microbiota relative to certain clinical parameters.
Finally, no additional amplicon/metagenome analysis was performed on our own samples. Due to the very small volume of the biopsy, we chose to use the whole sample for culture. We tried to compensate for this missing part by comparing our results with the literature (similar tissues, different techniques).