The predictive value of cytotoxicity-related molecules for pregnancy outcome in patients with recurrent reproductive failure

Abstract

Recurrent reproductive failure (RRF), including recurrent miscarriage (RM) and repeated implantation failure (RIF), represents reproductive disorders characterized by complex etiologies. It brings heavy financial burden and mental stress to the patients. Therefore, it is urgent to explore the potential biomarkers and evaluation strategy for RRF. Although natural killer (NK) cells and gamma delta-T (-T) cells are essential during physiologic pregnancy, the association between NK cells, -T cells and RRF remains uncertain. First, it was found that there was no correlation between NK cells cytotoxicity to K562 cells and its cytotoxicity to trophoblast cells, HTR-8 cells. It indicated that K562 cells could not replace HTR-8 cells to evaluate cytotoxic activity of NK cells to trophoblast cells during pregnancy. To further explore the potential biomarkers for RRF, the prognosis capacity of cytotoxic-related granules and receptors of peripheral blood NK cells and -T cells in predicting pregnancy outcomes of RRF was analyzed. We found that RM patients were more likely to have an immunosuppressive status with increased expression of inhibitory receptor CD158a. However, it was not associated with subsequent pregnancy outcomes of RM patients. In RIF patients, we found that the percentage of granzyme B and NKG2D of peripheral blood -T cells in lymphocytes were significantly increased in RIF patients with failed clinical pregnancy, when compared to that in patients with successful clinical pregnancy. In addition, the prognosis ability of uterine NK (uNK) cells in RRF patients is attractive, however, it remains controversial. In this study, it was found that the correlation between the number of uNK cells and subsequent pregnancy outcomes of RRF patients is not linear. A significantly increased risk of pregnancy failure was observed in RRF patients with either high, or low uNK cells. Furthermore, the uNK cell subsets, which are classified based on CD11b and CD27, were determined with multi-plex staining immunofluorescences in RRF patients. It was found that the percentage of uNK subset with cytotoxic activity, CD45+CD56+CD11b+CD27-, was significantly increased, and the percentage of uNK subsets with regulatory activity, CD45+CD56+CD11b-CD27+ and CD45+CD56+CD11b-CD27- cells, were significantly decreased in patients with RRF. The more indicators that are uncovered, the harder it is for clinicians to diagnose the cause of the disease. Without any automated tool, it will be tough to assess interaction and synergistic effect of the immune factors and other characteristics on the pregnancy outcomes of RRF patients. Here, we further propose to apply artificial intelligence (A.I.) to analyze the medical information of RRF patients. Through various deep learning technologies, the prediction models of pregnancy outcomes were established based on the following panels: basic characteristic, hormone level, autoantibodies, peripheral blood immune status, endometrial immune status as well as the embryo parameters were conducted. Finally, four models were established to predict pregnancy outcomes at different gestational nodes, including biochemical pregnancy, clinical pregnancy, ongoing pregnancy, and live birth, respectively. These results give us new insight on reproductive immunology and can assist clinicians to carry out more precise and personalized diagnosis and treatment for RRF patients.