Upregulation of immune genes in the proliferative phase endometrium enables classification into women with recurrent pregnancy loss versus controls

Abstract

STUDY QUESTION Does the transcriptome of preconceptional endometrium in the proliferative phase show a specific profile in women with recurrent pregnancy loss (RPL)? SUMMARY ANSWER A specific differential gene expression signature in endometrial samples for women experiencing RPL compared with IVF control women was identified including an RPL subgroup characterized by upregulation of immune-related genes and pathways. WHAT IS KNOWN ALREADY RPL affects 1–3% of couples trying to become parents with both short- and long-term health implications; furthermore, the underlying pathophysiology is complex and heterogeneous with no explanations found for more than half of the couples. Some studies indicate that immunological dysfunction plays a role even preconceptionally and during implantation; however, the few published studies of endometrial transcriptomes from women with RPL have had small sample sizes and focused on the secretory phase of the menstrual cycle. STUDY DESIGN, SIZE, DURATION The study was based on two cohorts of women: an RPL cohort (n = 108) and a control cohort (n = 27). Endometrial samples were collected at two university hospital clinics from March 2013 until February 2019. Dating of the endometrium was made by histological examination by experienced pathologists. PARTICIPANTS/MATERIALS, SETTING, METHODS All women were between 18 and 42 years at the time of collection of the biopsy. RPL was defined as three or more consecutive pregnancy losses or two second-trimester losses or stillbirths. The control group consisted of women referred to IVF/ICSI treatment with a presumed healthy endometrium. All biopsies, except one, were collected in a natural menstrual cycle. In total, 108 women with RPL were subjected to RNA-seq analysis. Seventy-six biopsies were in the proliferative phase, 29 were in the secretory phase, and three could not be classified. For the control women, in total, 27 were included in the RNA-seq analysis; 22 biopsies were in the proliferative phase, one was in the secretory phase, and four could not be classified. Total RNA was extracted from the endometrial biopsies, which had been stored in RNA stabilization solution at −80°C. RNA-seq reads were mapped and quantified using a reference transcriptome and analysed using principal component analysis (PCA), hierarchical clustering, and differential gene expression methods. MAIN RESULTS AND THE ROLE OF CHANCE PCA showed a clear separation of biopsies collected either in the proliferative or secretory phase. For the main analyses, we focused on the women with biopsies in the proliferative phase. PCA and differentially expressed genes (DEGs) revealed that RPL patients were characterized by upregulation of a limited number of immune-related genes and pathways. Further analyses revealed that subjects could describe a gene expression continuum, separable into four different subgroups by the gene expression data, where one subgroup consisted only of IVF controls, one was mixed, and two were composed of RPL patients only. The final analyses showed a distinct subgroup in the RPL cohort with stronger upregulation of immune-related genes, and deconvolution analyses of the bulk RNA-sequencing data together with immunohistochemistry analyses of the CD56 marker indicated an increased number of natural killer cells in this subgroup. A machine-learning model based on the Random Forest algorithm and gene expression data from the 157 DEGs (RPL vs controls) was trained on a subset of the cohort and validated using the remaining subjects, reaching on average 96.6% accuracy (95% CI: 93.3–96.7%), 95.7% sensitivity (95% CI: 95.7–95.7%), and 99.5% specificity (95% CI: 85.7–100.0%). The same analysis using only the most informative genes increased validation accuracies further. LIMITATIONS, REASONS FOR CAUTION The size of the IVF control group and difficulties in defining the most optimal type of control group is a recurrent limitation in this and other RPL studies. Some of the women from the control group might be subfertile in relation to endometrial factors. However, the current control group is a mix of women with different pregnancy and fertility records, which is a strength. Alternative control groups could be women with one to two healthy pregnancies or one to two pregnancy terminations. Furthermore, in only about 10% of the RPL cases information on foetal pathology or chromosomal aneuploidy was obtained. WIDER IMPLICATIONS OF THE FINDINGS The findings presented here indicate that a gene expression signature exists, which can be used to classify RPL patients versus control (non-RPL) women. An interesting aspect is that a pregnancy loss in some women thereby might result in a specific signature detectable as a specific endometrial gene expression profile possibly irrespectively of the cause of the pregnancy loss. Aside from contributing to a further understanding of the pathophysiology and development of new treatments, this finding could lead to a specific and cost-effective test that at an early stage could identify women with high risk of experiencing RPL. To this end, further perspectives include a prospective study, which would explore further utility of the predictive model by analysing endometrial samples collected in the proliferative phase from a cohort of women, who have experienced one pregnancy loss. STUDY FUNDING/COMPETING INTEREST(S) This work was supported by the Region Zealand Health Sciences Research Foundation, Zealand University Hospital through the ReproHealth Research Consortium ZUH, the Frimodt-Heineke Foundation, ‘Direktør Emil C. Hertz og Hustru Inger Hertz’ Fond’, the Torben and Alice Frimodt’s Foundation, the Novo Nordisk Foundation, and the Independent Research Fund Denmark. L.H.J.A., R.S.M., Y.D., K.J.H., H.S.N., A.S., and T.V.F.H. are inventors on a patent application (number EP24171923.6) submitted by Region Zealand, The Capital Region of Denmark, and the University of Copenhagen that covers a diagnostic test based on the results of the study. The authors declared no other competing interests. TRIAL REGISTRATION NUMBER N/A.