Metabolomics as a tool to identify biomarkers to predict and improve outcomes in reproductive medicine: a systematic review

Abstract

Introduction: Infertility is a complex disorder with significant medical, psychological and financial sequelae. With live birth rates per cycle below 30% and a drive from the HFEA to encourage single embryo transfer, there is significant research currently in the field of reproductive medicine to improve success rates. Understanding the causes of infertility at a molecular level could lead to better success rates in fertility treatments, and metabolomics techniques provide an ideal platform for studying relevant biofluids at a molecular level.

Aim: The aim of this systematic review is to examine the recent findings for the potential application of metabolomics to female reproduction, specifically to the metabolomics of follicular fluid (FF), embryo culture medium and endometrial receptivity. To our knowledge no other systematic review has investigated this topic.

Search methods: A systematic search of English peer-reviewed journals of four computerized databases was undertaken, with no time restriction set for publications.

Results and discussion: There were 19 studies which met the inclusion criteria and were included in the systematic review. Metabolomic studies have been employed for the compositional analysis of various biofluids in the female reproductive tract, including follicular fluid, embryo culture medium and endometrial fluid. Some studies found 1H NMR spectroscopy of embryo culture media was able to predict viability of individual embryos and implantation rate better than standard embryo morphology. However, this data was not supported by larger multi-centre randomised controlled trials. This systematic review provides guidance for future metabolomics studies on biofluids of the female reproductive tract, with a summary on the current findings, promise and pitfalls in metabolomics techniques. The approaches discussed can be adapted to other metabolomics studies.

Conclusion: Sophisticated modern metabolomics techniques have the potential to provide a more comprehensive understanding of the reproductive tract environment and nutritional status of the developing oocyte and embryo, with resulting implications for diagnostics and therapeutics in the field of fertility. This review has revealed the sparsity of metabolomics studies in the field of fertility, as well as a lack of research examining the effects of various gynaecological diseases. With the continued development of metabolomic technology, the possibility of identifying biomarkers of IVF outcome could become an exciting reality.

Introduction

Infertility is a complex disorder with significant medical, psychological and financial sequelae. The Human Fertilisation and Embryology Authority (HFEA) report one in six to seven couples in the UK to be affected [1]. While many fertility conditions will be treated with drug regimens, a significant number of patients will need to undergo in vitro fertilization (IVF). In the UK in 2011, 2% of all babies born were conceived through IVF [2]. An improved understanding of fertility and embryology has allowed the procedures and techniques to improve accordingly, resulting in a steady improvement in live birth rates from assisted reproductive technologies (ART) since the early 90s [3]. In the UK in 2014 overall pregnancy and live birth rates per embryo transfer were 36.3% and 26.5% respectively [3]. While improved, these success rates are still unsatisfactorily low.

The completion of the Human Genome Project in 2000 [4] resulted in a significant development of understanding in several fields of molecular biology, which when grouped together are termed ‘omics’ studies. With 35,000 genes and hundreds of thousands of human proteins to identify, it is now accepted that studies investigating single genes or protiens were outdated. Instead ‘omics’ studies are considered the future in molecular biology [5]. ‘Omics’ studies can investigate differences in DNA sequencing (genomics); gene expression profiling and differences in messenger RNA levels (transcriptomics); the differing composition and concentration of proteins (proteomics); and the variability and concentration of metabolites (metabolomics). Transcriptomics and proteomics have provided information in the observation of the effect of drugs on gene expression; however they provide little data regarding the overall response to disease or drug effects since they don’t reflect the dynamic metabolic status of the organism [6].

With live birth rates per cycle below 30% and a drive from the HFEA to encourage single embryo transfer, there is significant research currently in the field of reproductive medicine to improve success rates. Major objectives include finding other methods for identifying the best embryo for transfer, as well as enhancing the time during the stimulated cycle that the endometrium is most receptive to the blastocyst, known as the ‘window of implantation’ (WOI). Therefore a significant amount of research has attempted to identify biomarkers for embryo and oocyte quality. Most of these studies were targeted analyses, where one class of biological compound is examined for its connection with embryo or oocyte quality [7]. Despite this research, currently no definitive biomarkers to predict IVF outcome have been identified. Studies investigating multiple classes of compounds for their predictive ability of reproductive outcome simultaneously are therefore essential, because it is likely that research has failed to identify single biomarkers that effect fertility outcome, as only combinations of markers are diagnostic. Altered concentrations of several metabolites together would also make the chances of the finding being incidental less likely than with a single metabolite. Therefore, in recent years research into metabolism and reproductive medicine has moved towards global metabolite profiling, or metabolomics.

Metabolomics

Metabolomics is defined as the non-targeted identification and quantification of all low molecular weight end products of metabolism (metabolites) [6]. It reflects events well downstream of gene expression and gives valuable information about the metabolism within those cells, that other ‘omics’ technologies cannot [8]. The human genome consists of over 250,000 genes that encode for ~ 200,000 transcipts and over 1 million proteins, whilst in contrast the human metabolome consists of only ~ 3000 metabolites [9]. This lower number of metabolites means metabolomic analyses can be performed faster in comparison to genomic and proteomic analyses [10].

The response of an organism’s cells to a diseased state is to modify the concentrations of numerous metabolites, and thus maintain homeostasis [11]. The metabolites inside the cells of an organism are in dynamic balance with the metabolites inside the biofluids that perfuse the cells [6]. Therefore, the effect of an insult should be reflected in an organism’s biofluid composition. Eventually it should be achievable to distinguish between biofluids from different physiological states, such as diseased and non-diseased, based purely on their metabolic fingerprint.

Immediately after collecting these biofluids, metabolic activity needs to be stopped until the time of analysis, which is achieved by snap freezing in liquid nitrogen and storing the sample at -80°C. By far the two most commonly used analytical platforms for metabolomics are nuclear magnetic resonance (NMR) spectroscopy and mass spectroscopy (MS). The first of these is the more commonly used, and calculates the presence of certain nuclei in a sample, most commonly the hydrogen nuclei (1H), but 13C, 15N and 31P are also investigated. A spectrum is produced with the metabolites identified by their position on the x-axis, measured in chemical shift, in units of parts per million (ppm). The intensity of the NMR signal is on the y-axis and is directly related to the concentration of the nucleus giving rise to the signal. Metabolites are small molecules with a small number of given nuclei, and therefore each line along the x-axis represents a particular molecule in the sample. NMR is an ideal platform to analyse biofluids because it is the only method that requires little or no sample preparation, is non-destructive to the sample and requires small sample volumes. This is significant, since these biofluids are often not available in large volumes. The other platform is MS, which is significantly more sensitive than NMR spectroscopy, and generally requires less material than that utilized in NMR experiments [12]. However, NMR has numerous advantages over MS, such as minimal sample preparation, quicker analysis time, non-destructive to the tissue sample and most significantly greater reproducibility. MS suffers from variable detection responses because different molecules have altered ionization efficiencies, and as a result this method is less reproducible [13].

Metabolomics can be applied to reproductive medicine as the causes of subfertility may result in an imbalance of normal metabolism, and it is therefore postulated that a better understanding of the metabolic effects of the various aetiologies of infertility, may help to improve reproductive outcomes [14]. In addition, it is anticipated that an improved knowledge in this area would lead to the identification of non-invasive biomarkers for diagnostic and prognostic purposes [14]. Biofluid compositions of interest include those to assess the quality of the oocytes (follicular fluid), the quality of the embryos (embryo culture medium), assess the implantation capacity of the endometrium (endometrial fluid), barrier for transport of spermatozoa (cervical mucus) and the overall metabolic status of the patient (blood plasma).

There has been a progression from targeted single metabolite analysis to metabolomics in recent years, and the aim of this systematic review is to examine the recent findings for the potential application of metabolomics to female reproduction, specifically to the metabolomics of follicular fluid (FF), embryo culture medium and endometrial receptivity. To our knowledge no other systematic review has investigated this topic.

Methods

Search strategy

The systematic search followed PRISMA guidelines [15]. A bibliographic search of English language publications in four computerized databases (PubMed, Googlescholar, Science Direct and SciFinder) was conducted. The search terms are listed in Table 1 and were used in all possible combinations. The search was augmented by identifying additional studies from references cited in primary sources and review manuscripts.

Study selection

Given that metabolomics is a relatively recent practice there were no restrictions placed on publication date and inclusion. Only global metabolic studies analyzing reproductive tract biofluid composition were considered for this review. All single or targeted metabolic studies were excluded from this review, as were those using NMR spectroscopy for structural determination of molecular components from the female reproductive tract. Magnetic resonance imaging and NMR microimaging studies were also excluded. This review aimed to synthesize all available data on the topic, so no studies were excluded based on design. All English language pee-reviewed animal and human studies were included. However studies that focused on male subfertility were excluded, as were animal studies on any species other than mammals.

Study screening

All manuscripts following the first general search (n=633) were independently reviewed by the first author (T.B-M) based on the inclusion and exclusion criteria. Following an initial screen, 166 studies were excluded due to the title alone. After applying the exclusion criteria to these abstracts, 38 studies were evaluated for inclusion (Figure 1). An additional 2 studies were included from snowballing the references of studies identified. A total of 19 studies were included in this review. The screening process was cross-checked by the second author (S.S).

Data extraction

A data extraction spreadsheet was developed and agreed between the authors. The selected studies were comprehensively examined and relevant data was extracted for each paper and inputted to the spreadsheet by the first author (T.B-M) and cross-checked by the second author (S.S). Information selected included author details, year of publication and country of the study, study aim, sample size, methodology, sample characteristics, outcome measures and summary of findings. Disagreements regarding extracted data were resolved by discussion and deliberated on by the most senior authors (M-Y.T and M.J).

Results and discussion

The study characteristics, sample size, methods and aims can be found in Table 2. Individual study results are discussed in detail in this section and are therefore not included in the table. The sample sizes varied from 10 [16] to 485 patients [17]. Of the 19 studies, the majority were conducted in the USA (5), followed by Spain (3), UK (2), Holland (2), France (1), Ireland (1), Belgium (1), China (1), Sweden (1), Italy (1) and Denmark (1). Of the metabolomics studies identified, eleven were performed on embryo culture media, seven on follicular fluid and one on endometrial fluid.

Metabolomic studies of follicular fluid

The follicular fluid (FF) supplies the in vivo microenvironment for an oocyte during its development, and therefore must contain all the metabolites that are essential for a follicle’s growth and oocyte maturation, along with the metabolites excreted by the oocytes. It is therefore hypothesized that the metabolic make up of FF could indicate the quality of an oocyte. FF is easily obtained in IVF patients, since it is aspirated at the time of egg collection and then disposed of. Numerous studies over recent decades have performed targeted analyses of low molecular weight molecules in the FF. However, no single metabolite has been identified as a clinically useful biomarker for oocyte quality or pregnancy outcome [10]. Overall 7 studies were identified which investigated global metabolic profiles of follicular fluid, with regard to oocyte quality and follicular development.

With regard to animal studies, Gerard et al (2002) was the first study to investigate global metabolic changes in composition of equine FF during various physiological stages of follicular development [18]. A transvaginal ovarian puncture of the dominant follicle was performed and a small volume of follicular fluid was aspirated at the early dominant, late dominant and pre-ovulatory phases [18]. The growth of the follicles was not affected by this puncture, with follicular growth characeterised by an increase in oestradiol and progesterone concentrations, while maturation post gonadotrophin injection resulted in a decrease in oestradiol and a further increase in progesterone [18]. The authors demonstrated a decrease in lipoproteins and alanine during follicular growth and an increase in lipoprotiens during pre-ovulatory follicular maturation [18]. Gerard et al (2002) postulated the change in lipoprotein concentration was related to the steroidogenic activity of follicular cells, since the concentration of progesterone within the follicles followed the same profile [18]. The authors also demonstrated a significant decrease in acetate and trimethylamine concentrations from the late dominant to pre-ovulatory stages, and they suggested this could be linked to polyamine metabolism, since they are both degredation products of cellular metabolism [18]. This was the first study to demonstrate these metabolic changes, and to compare FF content to changes in steroidal concentrations. However, the study did not correlate these metabolic changes to the equine fertilization outcomes. Following ovulation induction 3 out of 10 of equines had a positive pregnancy test, but the FF content was not linked to these pregnancy outcomes.

Bender et al (2010) investigated the metabolic differences between FF from the dominant follicle of lactating cows and heifers using gas chromatography mass spectrometry (GC-MS) based metabolomics [19]. FF and serum were aspirated over three stages of follicular development: early dominant follicles, pre-ovulatory pre-LH surge follicles and pre-ovulatory post LH surge follicles [19]. The authors reported significant differences in 24 fatty acids and 9 water-soluble metabolites between the cows and heifers. Of note were the higher levels of the saturated fatty acids, palmitic acid and stearic acid, which meant the cows had higher FF concentrations of detrimental saturated fatty acids overall [19]. Bender et al (2010) postulated that this different follicular microenvironment in cows would negatively impact the maturation of their oocytes and effect early embryo development, which may contribute to their lower fertility in comparison to heifers [19].

Piñero-Sagredo et al (2010) studied the metabolic profile of human FF using 1H-NMR spectroscopy from 30 the FF of oocyte donors under 35 years of age [20]. They assigned a total of 131 chemical shifts by using different monodimensional (1D) and bidemensional (2D) NMR experiements, and 42 metabolites were identified [20]. They identified a 2:1 ratio of glucose and lactate, supporting the presence of anaerobic metabolism in hyperstimulated follicles [20]. The authors reported a high correlation between glucose, lactate and pyruvate concentrations, and they concluded this supports the theory FF might have a role in supporting oocyte maturation by supplying lactate and pyruvate as an energy source [20]. Piñero-Sagredo et al (2010) also reported a strong connection between the glycolytic pathway and fatty acid synthesis [20], and they found this association was stronger in younger donors, and in whom fertilization rates were improved [20]. The authors concluded the potential for metabolic profiling of FF in the search of biomarkers to predict oocyte quality, and also implied these findings had the potential to influence culture conditions of oocytes and embryos in fertility treatments in the future [20]. This is the first study to profile human FF using 1H-NMR spectroscopy.

Wallace et al (2012) performed a study investigating the relationship between the metabolomics profile of human FF, oocyte developmental potential and implantation outcome [21]. 1H-NMR spectroscopy was used and found that oocytes which failed to cleave as an embryo had higher FF levels of glucose and high density lipoproteins (HDL), and lower levels of lactate, choline and phosphocholine [21]. Patients who had a positive pregnancy test were found to have higher FF levels of proline, lactate, leucine and isoleucine, and lower levels of glucose [21]. This was the first study to show the potential of human FF metabolomic analysis and IVF outcome.

McRae et al (2012) studied the follicular fluid and blood plasma from natural cycle IVF patients, who were receiving their fertility treatment due to unexplained or male factor, to identify changes in the follicular and periovulatory phases of their menstrual cycles [16]. By obtaining samples from the same patient twice, and by using natural cycle IVF patients, it was possible to identify metabolic variations during the menstrual cycle without the potential effects of gonadotrophins on FF metabolic contents [16]. Periovulatory FF exhibited high levels of lactate, pyruvate and lower levels of glucose [16]. Periovulatory plasma contained higher levels of glucose and acetate and lower levels of glycoprotein [16]. This showed the potential of metabolomics for identifying FF metabolite changes throughout the menstrual cycle and in studying any impact of exogenous HCG administration. However, because only ten patients were studied, the representation of FF biomarkers and resulting fertility treatment, could not draw conclusions.

Petra et al (2012) investigated whether the presence of endocrine-disrupting chemicals, such as polychlorinated biphenyls, polybrominated diphenyl ethers and organochlorine pesticides, in human follicular fluid negatively impacts oocyte development in vivo [22]. Follicular fluid (n=40) and blood plasma (n=20) samples were analysed by gas chromatography and mass spectrometry, and the group reported an overall higher EDC contamination in the follicular micr-environment was a ssociated with statistically significantly poorer fertilization rates and also poorer chance for the oocyte to develop into a high quality embryo [22].

More recently Xia et al (2014) used GC-MS based metabolomics to prospectively compare the metabolites in FF in repeated IVF failure patients to standard IVF patients [23]. They found that FF between the two groups differed with regard to 20 metabolites, with FF from the recurrent failed IVF group showing raised levels of the amino acids valine, threonine, isoleucine, cysteine, serine, proline, alanine, phenylalanine, lysine, methionine and ornithine; and reduced levels of dicarboxylic acid and cholesterol. The authors postulated the possibility of unique metabolomic profiles in the FF of recurrent failed IVF patients, which raises the exciting possibility that metabolomic profiling of FF could be a realistic method to improve the selection of oocytes and embryos in the future.

It has been suggested that oxidative stress could play a pivotal role in female infertility [24], and indeed numerous studies have used targeted analyses to investigate this [25-30]. The results of these studies were conflicting, and it is not known whether oxidative stress alone is deleterious to oocyte development, or whether to an extent it is essential for oocyte viability, but when a threshold level is crossed the result is detrimental. However, no global metabolomics studies investigating markers for oxidative stress in the female reproductive tract were identified. Oxidative stress could be investigated by metabolomics studies, since many metabolites involved in the oxidative stress process could be identified and quantified.

The use of FF as a biomarker for potential outcome in assisted reproduction has some significant limitations in its current state. First, and perhaps the biggest limitation is that for patients who have two or three embryos transferred with a FF sample linked to one oocyte means that FF samples cannot be connected to a singleton pregnancy, since it cannot be known which embryo prevailed. Therefore FF samples need to be obtained from women undergoing single embryo transfer, which decreases the potential recruitment of patients significantly. Second, some patients seeking fertility treatment may have unique metabolic profiles that could affect the discrimination between profiles from good and bad quality oocytes. Third, a study has demonstrated that the process of ovarian stimulation can alter the composition of FF [31], and therefore women should preferrably all be on the same IVF drug regimen or ideally from natural cycle IVF. Fourth, there is the potential for FF contamination with the flushing medium used by some centres during the egg collection process. Flushing medium contain numerous metabolites, such as glucose, meaning a FF sample contaminated by a flushing medium would not represent the true metabolic make up of the FF. FF from the previously aspirated follicle also has the potential to contaminate the aspirate of the next follicle collected. This highlights that research surrounding metabolomics analysis of FF is in its infancy, and the need for more studies of this kind. Such studies should take into account the above challenges, and ideally study the FF from natural IVF patients with no drug regimens and who undergo single embryo transfers, with no flushing medium used. Natural cycles in particular are rare, and these limitations mean it would take a long time to obtain the required number of samples to generate statistically significant findings. This highlights that research surrounding metabolomic analysis of FF is in its early stages, and that larger prospective studies taking the above factors into account are required.

Metabolomics studies of embryo culture medium

In clinical practice today embryos are selected as suitable for transfer based on morphological assessment and cleavage rates [32]. These assessments are quick and economical, but have limited predictive value, significant inter-observer variability and fail to detect genetic abnormalities [33]. In response to this a morphokinetic model was developed, allowing automated assessment of various parameters over time, and was found to predict embryos most likely to develop to blastocysts [34, 35]. Biomarkers for embryo quality could overcome some of the weaknesses of morphological assessment, and although unlikely to replace them, could certainly complement them.

After targeted and single metabolic profiling studies, over the last decade studies of embryo metabolism have also moved towards metabolomics. Seli et al (2007) performed the first metabolomics study to determine whether metabolomic profiling of embryo culture media correlated with the reproductive potential of individual embryos [36]. Sixty nine embryo culture media samples from 30 patients after day 3 embryo transfer were collected and analysed using Raman and/or near-infrared (NIR) spectroscopy [36]. The spectra obtained from each instrument were analysed separately using a wavelength selective genetic algorithm to determine regions that predict implantation outcome, and viability indices reflective of reproductive potential were calculated for each sample [36]. Higher viability scores were associated with live birth [36]. Scott et al (2008) performed a blinded trial at a different fertility centre to assess the validity of the findings by Seli et al (2007) [36, 37]. The group concurred that viability scores were significantly higher in embryos implanting successfully compared to those that did not, and the overall accuracy of the technique for predicting outcome was 82% [37]. The group concluded that embryos with greater reproductive potential impact their local environment differently to embryos with lesser potential [37].

Seli et al (2008) analysed embryo culture media from samples from 17 patients with day 3 embryo transfer resulting in pregnancy, and 17 samples from patients whose embryos failed to implant using 1H NMR spectroscopy [38]. Using a multivariate analysis, a model was developed that calculates a viability index for each specrum and the weighted coefficients of glutamate and alanine/ lactate ratio quantities were higher for embryos that implanted and resulted in successful pregnancy [38]. Overall, 1H NMR spectroscopy predicted viability of individual embryos with a sensitivity of 88.2% [38]. Building on this work in 2010, Seli et al (2010) collected embryo culture media from a larger numbers of IVF patients after single embryo transfer was performed on day 2 and 3 embryos (n=485) using NIR spectroscopy [17]. Mean viability scores of successful implantation were significantly higher compared with embryos that failed to implant for both day 2 and day 3 embryos [17]. The authors demonstrated that metabolomics profiling of human embryo culture media is independent of morphology and correlates with the reproductive potential of embryos [17]. The same group of authors then attempted to use receiver operating characristic (ROC) analysis to compare their mean viability score to morphological assessment to predict pregnancy outcome in women receiving a day 5 single embryo transfer [39]. 198 embryo culture media samples were analyzed by NIR, and an algorithm was developed to produce a viability score [39]. The authors concluded that the viability score alone or together with morphologic grading has the ability to be a better predictor for IVF outcome compared to morphology alone [39].

Vergouw et al (2008) explored whether metabolomic biomarker profiling of embryo cluture medium by NIR spectroscopy had an association with pregnancy in single embryo transfer, with 133 patients scheduled for IVF treatment. The authors found that NIR spectral analysis demonstrated that embryos with reproductive potential produced unique metabolomic profiles, and resulting viability scores correlation with pregnancy outcome was statistically significant [40]. The authors concluded that a low relative viability score was highly predictive of a poor pregnancy outcome, which implies that metabolomics profiling can be used in conjunction with morphology to select the best embryos [40]. This viability score was produced using fresh embryos, but the same group showed it to be predictive of the viability of frozen embryos that were thawed and transferred as day 5 embryos [41]. The authors concluded this might be a useful tool in selecting the best embryo to transfer when embryos have similar morphological assessment [41].

Hardarson et al (2012) conducted a single centre prospective randomized controlled trial to assess whether NIR spectroscopy on spent embryo culture media could positively effect the pregnancy rate after day 2 and 5 single embryo transfers [42]. The NIR group (n=164) was compared with a control group (n=163), and there was no significant difference found in the ongoing pregnancy rate [42]

Marhuenda-Egea et al (2010) used high performance liquid chromatography (HPLC)-MS to identify metabolomic differences in embryo culture media from embryos that implanted successfully (n=10), and in those that failed to implant (n=15) [43]. They used a classification chemometric tool, soft independent modeling of class analogy (SIMCA), to classify samples as ‘pregnancy’ or ‘non-pregnancy’ based on amino acid concentrations [43]. The authors demonstrated differences in relative amino acid concentrations in embryo culture medium between embryos with differing implantation potential and therefore hypothesized that amino acids play a crucial role in the metanolism of embryos [43].

D’Alessandro et al (2012) performed a fast HPLC-MS metabolomics approach to examine blastocele fluid [44]. Metabolomic assessment of embryo culture medium allows indirect assessment of embryos, but by withdrawing fluid from a blastocyst cavity prior to cryostorage, they were able to perform a direct metabolic assessment of the embryo [44]. The authors concluded that it was plausible to identify the majority of metabolites of biological interest for embryo quality from blastocoele fluid [44], however they did not assess whether there were any differences in metabolite concentrations among blastocele fluids obtained from blastocysts that implanted or failed to.

More recently, Kirkegaard et al (2014) studied 161 fertility patients undergoing assisted conception to investigate whether the metabolomic profile, obtained with 1H NMR spectroscopy of spent culture media from human embryos would correlate with reproductive potential [45]. They found no correlation between the meatbolomic profiles and pregnancy outsome and therefore could not develop a model for prediction of pregnancy based on metabolomics [45].

The studies detailed above have indicated that significant metabolic differences between embryos may be indicative of their potential to result in implantation, however its application to the clinical setting has numerous hurdles [46]. First, the studies described above used embryo culture media that were frozen at the point of collection, and then transported to a central laboratory. Metabolomic profiling has not been shown to be a reproducible, robust test when performed at local sites [42]. Second, fertility clinics use different types and volumes of embryo culture media, and have differing practices in terms of timing of embryo transfer. Whether the validity of metabolomics profiling is influenced by these factors needs further study. Finally, aneuploidy is an important factor in failed implantation and miscarriage, and whether aneuploidy will have a unique and detectable metabolomic profile requires further investigation [9]. Metabolomics has the potential to identify numerous biomarkers concurrently and its eventual aim, with regard to embryo assessment, is to produce a model that generates a scoring system, so that ultimately expert knowledge may not be needed, only the means of obtaining compositional data. Additionally in the future, if only a small number of biomarkers are identified, it would be conceivable to develop simple, rapid and cheap tests that could be performed in the individual fertility centre’s embryology laboratory.

Metabolomics studies of the endometrium

The human endometrium is a dynamic tissue that responds to ovarian steroid hormones, cytokines and chemokines to undergo growth, differentiation and regression throughout the mentrual cycle [47, 48]. The principle role of the endometrium is to provide support to allow the embryo to implant and to allow for further fetal growth. The human endometrium is receptive for implantation for a short period of time during the natural menstrual cycle, known as the ‘window of implantation’ (WOI) [49]. This window varies according to different menstrual cycle lengths, but occurs in the mid-secretory phase, coinciding with days 19-21 of a normal length cycle [50]. Investigating new molecules that are differentially expressed throughout the WOI will allow us to identify putative biomarkers and also aid clinicians to obtain receptive endometrium [51]. Studies have investigated cytokine profiles at different points in the menstrual cycle [52-54], as well as proteomic profiles of the endometrial fluid [55], to attempt to detect functional changes in the endometrium during the menstrual cycle by non-invasive methods. One of the main problems with these techniques is that a biopsy is essential for analysis, necessitating waiting until the next cycle to transfer the embryo. It has been successfully demonstrated that aspirating endometrial fluid does not impact implantation or pregnancy rates in the same cycle [56]. There are significant clinical implications for metabolomics to identify biomarkers associated with a receptive endometrium that favors implantation and allows dialogue between the developing embryo and the endometrium.

Only one study has been published investigating the endometrium at the window of implantation [57]. Vilella et al (2013) investigated the lipodomic profile of endometrial fluids from fertile patients at different stages of their menstrual cycle, focusing on the receptivity phase using liquid chromatography (LC)/MS techniques [57]. The lipodome is characterized by global changes in lipid metabolites, forming part of the metabolome, and lipids have different chemical properties compared to most water soluble metabolites, meaning separate analysis is warranted [58]. The authors demonstrated a statistically significant increase in the concentration of PGE2 and PGF2a during the WOI.

The development of metabolomics to use biofluids opens a new avenue to possible non-invasive diagnosis, allowing analysis of the endometrium during the same cycle as embryo transfer. Metabolomics certainly lags behind other ‘omics’ technologies and is in its infancy in assessing endometrial receptivity. Identifying biomarkers associated to endometrial receptivity is clearly desireable and could significantly impact pregnancy rates in patients seeking reproductive treatments. Further research in this area could add potentially valuable information into this understudied field.

Metabolomics of cervical mucous, ovarian tissue and blood plasma

Cervical mucous functions as a transport medium for spermatozoa as well as a barrier to infections [59, 60]. Sahrbacher et al (2002) used 1H NMR spectroscopy to quantitavely and qualitatively analyse the molecular composition of human cervical mucous in the aqueous phase [60]. The group attempted to standardize the cervical mucous properties by pre-treating the individuals with ethinylestradiol for 7 days, but still found the concentrations of molecular components of cervical mucous varied significantly [60]. Interestingly, these concentrations were not related to concentrations of molecular components found in the blood plasma, which implies that cervical mucous is not just a transudate [60]. The authors postulated that the strict individual regulation of cervical mucous constituents is connected to its function in fertility with regard to its interaction with sperm, but this was not studied further [60]. No global metabolomic studies were identified concerning changes to the composition of cervical mucous with the menstrual cycle, or connected to human reproductive physiology or disease states. Cervical mucous provides a transport medium for spermatozoa and as a defensive barrier for infection, so learning more about the mechanisms that underly these actions is potentially very useful in understanding infertility, and might give more insight into patients in whom their infertility is unexplained. Metabolomics permits the comparison of many metabolites taken from different individuals in varied physiological states, and could be a powerful technique in unraveling reasons for infertility within cervical mucous.

Metabolomic studies of ovarian tissue are scarce and concentrate on ovarian cancer, with research aiming to identify reliable biomarkers for epithelial ovarian cancer, which often presents clinically at a late stage. Studies have found metabolomics differences using 1H NMR spectroscopy between epithelial cancer cells and healthy controls [61, 62]. It is feasible that the same metabolomics technologies could be used to identify conditions within the ovary rendering the individual subfertile, such as poor ovarian reserve or premature ovarian failure

Blood plasma could influence oocyte quality, since FF is derived from blood plasma and metabolites are exchanged via the blood-follicle barrier. However, no global metabolomics studies of blood plasma with regard to oocyte quality or IVF outcome were identified. This would be intriguing and would give the capability of measuring a great number of metabolites, many of which have not been investigated in fertility patients. Metabolomic studies have been performed investigating the blood plasma of polycystic ovarian syndrome (PCOS) patients [63, 64] and found numerous metabolic differences, such as lower levels of citrulline, arginine, lysine, ornithine, proline, glutamate, acetone, citrate and histidine in PCOS compared to controls [63]. These studies reveal the future potential of metabolomics to not only aid the non-invasive diagnosis of PCOS, but also improve our grasp of the underlying metabolic pathways and pathogenesis behind PCOS, and thus potentially offering better treatments in the future. No metabolomics studies have been undertaken investigating endometriosis or pelvic inflammatory disease (PID), and since these disorders can be difficult to diagnose by methods other than invasive surgery, a non-invasive method to identify biomarkers associated with these conditions at an early stage would be highly desireable. Blood plasma has less potential than follicular fluid or embryo culture medium for predicting oocyte and embryo quality, since it doesn’t directly interact with the oocyte or blastocyst. In addition, blood plasma is more susceptible to variation from other things that can affect metabolism, such as diet, exercise and disease. However, blood plasma gives valuable information on the general health of the individual, and therefore metabolomic assessment of blood plasma could be very useful as a supplementary test in fertility patients, but is unlikely to ever be the only investigation required for global metabolic fertility assessment.

Conclusion and future perspectives

A significant amount of knowledge has been gained in recent decades from targeted metabolic analysis of embryo culture medium, FF and blood plasma, however no single biomarker has been identified, presumably because the metabolic picture is far more complex than to be explained by a single metabolite imbalance. Therefore in recent years there has been a move from targeted analyses to global metabolomics. Infertility is a significant problem with multiple aetiologies. Sophisticated modern metabolomics techniques have the potential to provide a more comprehensive understanding of the reproductive tract environment and nutritional status of the developing oocyte and embryo, with resulting implications for diagnostics and therapeutics in the field of fertility. This systematic review has revealed the sparsity of metabolomics studies in the field of fertility, as well as a lack of research examining the effects of various gynaecological diseases, such as PCOS, PID and endometriosis.

This technology has the potential to discover previously unknown mechanisms of infertility. This research remains in its infancy, and we have discussed various hurdles that will need to be overcome before useful groups of biomarkers are in routine clinical practice. Metabolism is a dynamic process that alters with time and therefore studies collecting samples over a period of time would be useful to investigate this further and confirm these changes in metabolic equilibrium. It also remains to be seen whether metabolomic studies can detect epigenetic factors, which may not significantly affect embryo implantation and early pregnancy development, but which could negatively impact fetal development and the resulting wellbeing of the offspring. Metabolomic studies have currently focused on oocyte and embryo quality by studying FF and embryo culture medium respectively. However, other factors of significance in female fertility, such as endometrial receptivity and cervical mucous composition would be of great interest. In the future large scale multi-centre randomized controlled trials are required to investigate the metabolomic profile of FF and embryo culture media further, and to look into the above mentioned biofluids of interest in the female reproductive tract. Metabolomic studies have also drastically under investigated and compared the metabolic profiles of the different fertility patient groups, such as unexplained infertility patients, tubal infertility patients, recurrent failed IVF patients and recurrent spontaneous miscarriage patients. With the continued development of metabolomic technology, the possibility of identifying biomarkers of IVF outcome could become an exciting reality.

Authors’ roles

T.B-M was responsible for the original manuscript design, drafting and revision for important intellectual content. S.S and E.H were responsible for providing important intellectual input into the work and preparation, drafting and final approval of the manuscript. M.J and M-Y.T are the guarantor for this paper and accepts full responsibility for the work and/or conduct of the study.

Funding

No external funding was used for this study.

Conflict of interests

The authors have no financial, personal, or professional competing interests to declare.

References

1.(HFEA), H.F.a.E.A., Fertility treatment in 2013: trends and figures. 2013.

2.(HFEA), H.F.a.E.A., Fertility Treatment in 2012: trends and figures. 2012.

3.(HFEA), H.F.a.E.A., Fertility treatment 2014 trends and figures. 2014.

4.Venter, J.C., et al., The sequence of the human genome. Science, 2001. 291(5507): p. 1304-51.

5.Weinstein, J.N., Searching for pharmacogenomic markers: the synergy between omic and hypothesis-driven research. Dis Markers, 2001. 17(2): p. 77-88.

6.Nicholson, J.K., J.C. Lindon, and E. Holmes, 'Metabonomics': understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica, 1999. 29(11): p. 1181-9.

7.McRae, C., V. Sharma, and J. Fisher, Metabolite Profiling in the Pursuit of Biomarkers for IVF Outcome: The Case for Metabolomics Studies. Int J Reprod Med, 2013. 2013: p. 603167.

8.Allen, J., et al., High-throughput classification of yeast mutants for functional genomics using metabolic footprinting. Nat Biotechnol, 2003. 21(6): p. 692-6.

9.Seli, E., C. Robert, and M.A. Sirard, OMICS in assisted reproduction: possibilities and pitfalls. Mol Hum Reprod, 2010. 16(8): p. 513-30.

10.Revelli, A., et al., Follicular fluid content and oocyte quality: from single biochemical markers to metabolomics. Reprod Biol Endocrinol, 2009. 7: p. 40.

11.Lindon, J.C., E. Holmes, and J.K. Nicholson, So what's the deal with metabonomics? Anal Chem, 2003. 75(17): p. 384a-391a.

12.Lei, Z., D.V. Huhman, and L.W. Sumner, Mass spectrometry strategies in metabolomics. J Biol Chem, 2011. 286(29): p. 25435-42.

13.Goldsmith, P., et al., Metabonomics: a useful tool for the future surgeon. J Surg Res, 2010. 160(1): p. 122-32.

14.Baskind, N.E., et al., Understanding subfertility at a molecular level in the female through the application of nuclear magnetic resonance (NMR) spectroscopy. Hum Reprod Update, 2011. 17(2): p. 228-41.

15.Moher, D., et al., Reprint--preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Phys Ther, 2009. 89(9): p. 873-80.

16.McRae, C., et al., Metabolic profiling of follicular fluid and plasma from natural cycle in vitro fertilization patients--a pilot study. Fertil Steril, 2012. 98(6): p. 1449-57.e6.

17.Seli, E., et al., Noninvasive metabolomic profiling as an adjunct to morphology for noninvasive embryo assessment in women undergoing single embryo transfer. Fertil Steril, 2010. 94(2): p. 535-42.

18.Gerard, N., et al., Analysis of the variations of follicular fluid composition during follicular growth and maturation in the mare using proton nuclear magnetic resonance (1H NMR). Reproduction, 2002. 124(2): p. 241-8.

19.Bender, K., et al., Metabolite concentrations in follicular fluid may explain differences in fertility between heifers and lactating cows. Reproduction, 2010. 139(6): p. 1047-55.

20.Pinero-Sagredo, E., et al., NMR metabolic profile of human follicular fluid. NMR Biomed, 2010. 23(5): p. 485-95.

21.Wallace, M., et al., An investigation into the relationship between the metabolic profile of follicular fluid, oocyte developmental potential, and implantation outcome. Fertil Steril, 2012. 97(5): p. 1078-84.e1-8.

22.Petro, E.M., et al., Endocrine-disrupting chemicals in human follicular fluid impair in vitro oocyte developmental competence. Hum Reprod, 2012. 27(4): p. 1025-33.

23.Xia, L., et al., Metabolomic profiling of human follicular fluid from patients with repeated failure of in vitro fertilization using gas chromatography/mass spectrometry. Int J Clin Exp Pathol, 2014. 7(10): p. 7220-9.

24.Agarwal, A., S. Gupta, and R.K. Sharma, Role of oxidative stress in female reproduction. Reprod Biol Endocrinol, 2005. 3: p. 28.

25.Attaran, M., et al., The effect of follicular fluid reactive oxygen species on the outcome of in vitro fertilization. Int J Fertil Womens Med, 2000. 45(5): p. 314-20.

26.Oyawoye, O., et al., Antioxidants and reactive oxygen species in follicular fluid of women undergoing IVF: relationship to outcome. Hum Reprod, 2003. 18(11): p. 2270-4.

27.Pasqualotto, E.B., et al., Effect of oxidative stress in follicular fluid on the outcome of assisted reproductive procedures. Fertil Steril, 2004. 81(4): p. 973-6.

28.Pasqualotto, E.B., et al., The role of enzymatic antioxidants detected in the follicular fluid and semen of infertile couples undergoing assisted reproduction. Hum Fertil (Camb), 2009. 12(3): p. 166-71.

29.Das, S., et al., Reactive oxygen species level in follicular fluid--embryo quality marker in IVF? Hum Reprod, 2006. 21(9): p. 2403-7.

30.Basini, G., et al., Reactive oxygen species and anti-oxidant defences in swine follicular fluids. Reprod Fertil Dev, 2008. 20(2): p. 269-74.

31.de los Santos, M.J., et al., Hormonal and molecular characterization of follicular fluid, cumulus cells and oocytes from pre-ovulatory follicles in stimulated and unstimulated cycles. Hum Reprod, 2012. 27(6): p. 1596-605.

32.Sakkas, D. and D.K. Gardner, Noninvasive methods to assess embryo quality. Curr Opin Obstet Gynecol, 2005. 17(3): p. 283-8.

33.Singh, R. and K.D. Sinclair, Metabolomics: approaches to assessing oocyte and embryo quality. Theriogenology, 2007. 68 Suppl 1: p. S56-62.

34.Cruz, M., et al., Timing of cell division in human cleavage-stage embryos is linked with blastocyst formation and quality. Reprod Biomed Online, 2012. 25(4): p. 371-81.

35.Meseguer, M., et al., The use of morphokinetics as a predictor of embryo implantation. Hum Reprod, 2011. 26(10): p. 2658-71.

36.Seli, E., et al., Noninvasive metabolomic profiling of embryo culture media using Raman and near-infrared spectroscopy correlates with reproductive potential of embryos in women undergoing in vitro fertilization. Fertil Steril, 2007. 88(5): p. 1350-7.

37.Scott, R., et al., Noninvasive metabolomic profiling of human embryo culture media using Raman spectroscopy predicts embryonic reproductive potential: a prospective blinded pilot study. Fertil Steril, 2008. 90(1): p. 77-83.

38.Seli, E., et al., Noninvasive metabolomic profiling of embryo culture media using proton nuclear magnetic resonance correlates with reproductive potential of embryos in women undergoing in vitro fertilization. Fertil Steril, 2008. 90(6): p. 2183-9.

39.Seli, E., et al., Receiver operating characteristic (ROC) analysis of day 5 morphology grading and metabolomic Viability Score on predicting implantation outcome. J Assist Reprod Genet, 2011. 28(2): p. 137-44.

40.Vergouw, C.G., et al., Metabolomic profiling by near-infrared spectroscopy as a tool to assess embryo viability: a novel, non-invasive method for embryo selection. Hum Reprod, 2008. 23(7): p. 1499-504.

41.Vergouw, C.G., et al., Non-invasive viability assessment of day-4 frozen-thawed human embryos using near infrared spectroscopy. Reprod Biomed Online, 2011. 23(6): p. 769-76.

42.Hardarson, T., et al., Non-invasive metabolomic profiling of Day 2 and 5 embryo culture medium: a prospective randomized trial. Hum Reprod, 2012. 27(1): p. 89-96.

43.Marhuenda-Egea, F.C., et al., A crucial step in assisted reproduction technology: human embryo selection using metabolomic evaluation. Fertil Steril, 2010. 94(2): p. 772-4.

44.D'Alessandro, A., et al., A mass spectrometry-based targeted metabolomics strategy of human blastocoele fluid: a promising tool in fertility research. Mol Biosyst, 2012. 8(4): p. 953-8.

45.Kirkegaard, K., et al., Nuclear magnetic resonance metabolomic profiling of Day 3 and 5 embryo culture medium does not predict pregnancy outcome in good prognosis patients: a prospective cohort study on single transferred embryos. Hum Reprod, 2014. 29(11): p. 2413-20.

46.Botros, L., D. Sakkas, and E. Seli, Metabolomics and its application for non-invasive embryo assessment in IVF. Mol Hum Reprod, 2008. 14(12): p. 679-90.

47.Wilcox, A.J., D.D. Baird, and C.R. Weinberg, Time of implantation of the conceptus and loss of pregnancy. N Engl J Med, 1999. 340(23): p. 1796-9.

48.Salamonsen, L.A., et al., Society for Reproductive Biology Founders' Lecture 2009. Preparing fertile soil: the importance of endometrial receptivity. Reprod Fertil Dev, 2009. 21(7): p. 923-34.

49.Harper, M.J., The implantation window. Baillieres Clin Obstet Gynaecol, 1992. 6(2): p. 351-71.

50.Bergh, P.A. and D. Navot, The impact of embryonic development and endometrial maturity on the timing of implantation. Fertil Steril, 1992. 58(3): p. 537-42.

51.Berlanga, O., et al., How endometrial secretomics can help in predicting implantation. Placenta, 2011. 32 Suppl 3: p. S271-5.

52.Boomsma, C.M., et al., Cytokine profiling in endometrial secretions: a non-invasive window on endometrial receptivity. Reprod Biomed Online, 2009. 18(1): p. 85-94.

53.Dominguez, F., et al., CXCL10 and IL-6 induce chemotaxis in human trophoblast cell lines. Mol Hum Reprod, 2008. 14(7): p. 423-30.

54.Simon, C., et al., Localization of interleukin-1 type I receptor and interleukin-1 beta in human endometrium throughout the menstrual cycle. J Clin Endocrinol Metab, 1993. 77(2): p. 549-55.

55.Casado-Vela, J., et al., Comprehensive proteomic analysis of human endometrial fluid aspirate. J Proteome Res, 2009. 8(10): p. 4622-32.

56.van der Gaast, M.H., et al., Endometrial secretion aspiration prior to embryo transfer does not reduce implantation rates. Reprod Biomed Online, 2003. 7(1): p. 105-9.

57.Vilella, F., L.B. Ramirez, and C. Simon, Lipidomics as an emerging tool to predict endometrial receptivity. Fertil Steril, 2013. 99(4): p. 1100-6.

58.Griffiths, W.J., et al., On the future of "omics": lipidomics. J Inherit Metab Dis, 2011. 34(3): p. 583-92.

59.Brunelli, R., et al., Globular structure of human ovulatory cervical mucus. Faseb j, 2007. 21(14): p. 3872-6.

60.Sahrbacher, U., et al., High resolution proton magnetic resonance spectroscopy of human cervical mucus. J Pharm Biomed Anal, 2002. 28(5): p. 827-40.

61.Iorio, E., et al., Alterations of choline phospholipid metabolism in ovarian tumor progression. Cancer Res, 2005. 65(20): p. 9369-76.

62.Odunsi, K., et al., Detection of epithelial ovarian cancer using 1H-NMR-based metabonomics. Int J Cancer, 2005. 113(5): p. 782-8.

63.Atiomo, W. and C.A. Daykin, Metabolomic biomarkers in women with polycystic ovary syndrome: a pilot study. Mol Hum Reprod, 2012. 18(11): p. 546-53.

64.Sun, L., et al., Metabonomics reveals plasma metabolic changes and inflammatory marker in polycystic ovary syndrome patients. J Proteome Res, 2012. 11(5): p. 2937-46.