Background
There is growing evidence that an increase in body mass index (BMI) is associated an increased risk of miscarriage [
1‐
3]. It is however unknown if this association is due to an adverse effect on the embryo, endometrium or both.
The possibility of an endometrial defect was first postulated in a study of assisted conception cycles using the oocyte donation model [
4]. In an earlier study, we attempted to identify the nature of any possible endometrial defect by examining several markers of endometrial function, including endometrial morphology, leukocyte populations, leukaemia inhibitory factor and steroid receptors using immunocytochemical staining, but were unable to identify any clear effect on any of these factors [
5].
However there are a vast number of endometrial factors involved in the implantation process and it is possible that a specific endometrial defect was missed. To prospectively examine all possible endometrial factors using immunocytochemical staining or other molecular techniques could be a time consuming and possibly unrevealing process, so we postulated that by mapping the endometrial protein profile around the time of implantation, potential target proteins could be identified for further investigation.
The aim of this study was to determine firstly if overweight and obesity are associated with a change in the endometrial protein profile and secondly if such a change could reflect an endometrial cause for the increased risk of miscarriage in overweight and obese women.
Discussion
Differential expression of the endometrial protein profile in the different phases of the menstrual cycle has been described by Rai et al. [
10] and since then the use of endometrial proteomics has continued albeit mainly concerning the study of endometriosis [
11‐
13] and endometrial cancer [
14,
15]. To the best of our knowledge, this pilot study is the first to use proteomic analysis to study the endometrial protein expression around the implantation period in women with recurrent miscarriage and to use this technique to investigate the possible effect of increased BMI on the endometrium.
When examining the whole population (obese and lean), the first finding of this study was that PCA had a poor ability to discriminate the samples. Although this may be due to the absence of a distinctive difference in endometrial protein expression between miscarriage and control samples, it may also be due the presence of confounding factors that led to heterogeneity of the samples.
We hypothesised that the presence of an increased BMI may be one such confounding factor and repeated the analysis after exclusion of samples from patients with an increased BMI. The analysis was repeated (Figure
2) and PCA then showed good discrimination of recurrent miscarriage and control samples indicating that an increased BMI may indeed have acted as a confounding factor in the initial analysis. Similarly, obese and lean control samples could also be discriminated using PCA, further providing evidence for a difference in endometrial protein expression (Figure
3).
Haptoglobin alpha chain (spot 1163) appears to contribute to the discrimination observed in all the comparisons of interest. The endometrial concentrations of this protein are lower in the miscarriage groups compared to control groups, but higher in the obese miscarriage group compared to the lean miscarriage group. Haptoglobin therefore seems to be affected both by the presence of recurrent miscarriage as well as by increased BMI, as both conditions seem to have an opposite effect on haptoglobin concentrations. Other proteins that were significantly increased in the obese miscarriage group were transthyretin (pre-albumin), and beta globin.
Haptoglobin is a glycoprotein synthesised in the liver. Its primary function is to bind excess haemoglobin thus protecting the kidneys in cases of intravascular haemolysis [
16]. However haptoglobin has several other functions that may be relevant to both implantation and obesity.
Firstly haptoglobin is an important component of the body’s response to inflammatory conditions [
17‐
19]. One such inflammatory condition is obesity, where an increase in the central fat compartment leads to a state of relative hypoxia in the adipocytes and a release of a number of inflammatory markers including haptoglobin. Indeed haptoglobin concentrations have been shown to positively increase in proportion to the severity of obesity [
17,
18]. It is therefore possible that haptoglobin is a marker of an ongoing inflammatory reaction in the endometrial lining of obese women with recurrent miscarriage, which may explain their adverse reproductive outcomes.
Secondly haptoglobin is produced by the endometrium [
20] and has been shown in several animal studies to be expressed in increasing amounts in the peri implantation period and is an important component of the extra embryonic matrix [
16,
19,
21]. This increase in endometrial haptoglobin around the time of implantation may play a role in modulating the maternal reaction to the implanting blastocyst [
16,
19].
The haptoglobin molecule is also known to display genetic polymorphism where some genotypes have been shown to be associated with better reproductive outcomes compared to others [
22]. So although obese miscarriage women showed an increased expression of endometrial haptoglobin, it is possible that haptoglobin in this case is of a different genotype that results in a less favourable pregnancy outcome. Future studies should analyse the specific genotype of the haptoglobin molecule using techniques such as Polymerase chain reaction and gel electrophoresis [
23].
Obese recurrent miscarriage samples were also associated with a significantly increased expression of transthyretin and beta globin, both of which are common intravascular products normally found in relative abundance in tissue. However, their increased expression in the obese miscarriage cohort may indicate some form of vascular or endothelial dysfunction in these women. Indeed vascular dysfunction is a characteristic of chronic inflammatory conditions such as obesity [
24,
25].
There are several mechanisms that could support the hypothesis of a local vascular dysfunction in obese women. Firstly hyperleptinaemia has been linked to the occurrence of vascular dysfunction possibly through dysregulation of endothelial nitric oxide [
24,
25]. Similarly the production of cytokines has also been linked to vascular dysfunction [
26]. Finally haptoglobin has been found to have an important role in angiogenesis and vascular dysfunction in chronic inflammatory conditions and this may be more common with certain haptoglobin genotypes [
27,
28].
It would be interesting to know whether obesity produces any structural changes in fertile women as well as woman with miscarriage and therefore a comparison between obese and non-obese controls would have been ideal. However due to relatively small numbers of the control subgroups, this was not practically possible, but should be addressed in future studies.
A possible confounding factor in this study is the presence of four samples from women with antiphospholipid syndrome which may act as a potential source of bias. Therefore careful consideration was given as to whether these samples should be excluded from the analysis. However the potential effect of antiphospholipid antibodies on pro inflammatory mediators such as tumour necrosis factor α (TNFα) [
29] has only been demonstrated in animal studies and recent human studies have been unable to demonstrate similar findings [
30]. So even though we cannot rule out a possible confounding effect, it is highly unlikely particularly in view of the small number of women with this condition relative to the whole population.
It was not possible to employ a second analysis technique to confirm the results since all the tissue was consumed in the process of protein extraction and analysis and we were unable to conserve tissue for further analysis. Endometrial samples in this study were obtained in an out-patient setting which results in a relatively small amount of tissue being retrieved and repetition of sampling more than once to obtain more tissue would not have been acceptable due to the associated patient discomfort.
A general limitation of proteomic analysis is that only abundant proteins can be detected. It is possible that other proteins that are relevant to the process of implantation were missed in this analysis either because they were expressed in small amounts or had a molecular weight below the level of detection using our particular method of analysis.
This raises the question as to the best method of approaching the same research question in future studies. Validation of changes in protein expression levels observed with a 2D gel-based approach could be carried out using orthogonal techniques such as Western blotting [
31] or a targeted proteomic approach using stable isotope-labeled peptides and multiple reaction monitoring [
32]. While the former approach is feasible when appropriate antibodies are available, the latter would require a considerable amount of effort and expense. The problem of abundant proteins in proteomic studies, especially of serum, can be addressed by antibody-based depletion, for example with the Agilent MARS columns. However, a major drawback of using a depletion method is the potential loss of important markers through their non-specific binding to depleted proteins, especially albumin [
33]. Other possible approaches include micro arrays and genomic or metabolomic analysis. It is yet unclear whether these technologies would offer an advantage over proteomics when trying to answer the current question.
We would like to emphasise that the results of this study are mainly limited to women with mild obesity; further studies would be needed to examine the effect of more severe forms of obesity on the protein profile. However this may prove difficult since patients with severe degrees of obesity may present more with problems related to infertility rather than recurrent miscarriage. Nevertheless, it may be possible to achieve the necessary numbers in these subgroups in the context of a large multicentre study.
Finally, as there were no previous similar studies, we had no guidance by which to perform a sample size calculation when designing this study. Our findings however have now provided data regarding the necessary sample size for future studies as stated above. Larger studies are now needed to confirm our findings.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MM, main investigator, responsible for recruitment of participants, obtaining the samples and writing the manuscript. RP and JT processed the samples, performed the proteomic analysis and assisted with writing the results section. WL and TCL contributed to the study design, interpretation of results, and preparing of the manuscript. All authors read and approved the final manuscript.