Characterization of micro-RNA in women with different ovarian reserve

Women undergoing infertility treatment are routinely subjected to one or more tests of ovarian reserve. Therefore, an adequate assessment of the ovarian reserve is necessary for the treatment. In this study, we aimed to characterize the potential role of microRNAs (miRNAs) as biomarkers for women with different ovarian reserves. A total of 159 women were recruited in the study and classified according to their anti-Müllerian hormone (AMH) level into three groups: (1) low ovarian reserve (LAMH, n = 39), (2) normal ovarian reserve (NAMH, n = 80), and (3) high ovarian reserve (HAMH, n = 40). SurePrint Human miRNA array screening and reverse transcription-quantitative PCR (RT-qPCR) were respectively employed to screen and validate the miRNA abundance level in the three tested groups. Compared with NAMH, the abundance level of 34 and 98 miRNAs was found to be significantly altered in LAMH and HAMH, respectively. The abundance level of miRNAs was further validated by RT-qPCR in both, the screening samples as well as in an independent set of validation samples. The abundance levels of the validated miRNAs were significantly correlated with the AMH level. The best AUC value for the prediction of the increase and decrease in the AMH level was obtained for the miR-100-5p and miR-21-5p, respectively. The level of miRNAs abundance correlates with the level of AMH, which may serve as a tool for identifying women with a different ovarian reserve and may help to lay the ground for the development of novel diagnostic approaches.

Assessment of the ovarian reserve has become essential to determine the strategy for the treatment of female infertility. For this purpose, several non-invasive clinical, endocrinological, and ultrasonographic examinations, at the early follicular phase, have been applied. Yet, to our best knowledge, no ideal test for ovarian reserve assessment exists. Serum levels of anti-Müllerian hormone (AMH) and total antral follicle count (AFC) performed better than all other known markers, and are the most predictive, direct tests for evaluation of ovarian reserve 1 . Currently, no molecular biomarkers are used, in combination with these conventional tests, to brush up the predictive accuracy of different forms of ovarian reserves. MicroRNAs (miRNAs) have received increasing attention due to their likely role in the regulation of nearly every cellular process. Currently, there are 2300 "real" miRNAs in the miRNA database based on miRBase.org 2 . Changes in miRNA abundance level are associated with many reproductive pathologies in both male and female partners [3][4][5][6][7][8][9][10][11] . In the female partner, miRNAs were found to be involved in ovarian development and function 11 . Specifically, the literature supports a crucial role for miRNAs during various stages in either somatic cells and/or ovarian follicles. In turn, a functional and crucial role in the production of the mature, and viable oocytes that are capable of fertilization and subsequent embryo development and implantation of miRNA has been suggested [12][13][14] . Circulating miRNAs have been utilized as a potential independent predictive system for different diseases due to their abundant and unique merits in body fluid (i.e., stable, easy to be detected, and potentially disease-specific) 15,16 . As miRNAs seem to be an important regulator of gene expression during follicular development and maturation, we hypothesized that circulating miRNAs in the early follicular phase could serve as potentially useful biomarkers for predicting ovarian reserve. To address this hypothesis, we characterized the abundance level of circulating miRNAs in women undergoing assisted reproductive technology (ART) using a large panel of human miRNA arrays. Then, we validated the deregulated miRNAs in a large independent cohort of samples with individual quantitative reverse transcriptase-polymerase chain reaction (RT-qPCR) assays. Data were analyzed to determine whether circulating miRNA profiles correlate Screening of differentially abundant miRNAs using microarray. To identify miRNAs that are differentially abundant in the blood samples of included women, we analyzed the abundance level of 2549 human mature miRNAs of miRBase v21. The abundance levels of circulating miRNAs were screened in 38 women including LAMH (n = 12), NAMH (n = 13), and HAMH (n = 13) (phase I). By considering the miRNAs with a significant adjusted P value of < 0.05 with fold change ≥ 2 (lower and higher abundant level), only 34 and 98 miRNAs showed differential abundance levels in LAMH and HAMH groups, as compared to NAMH group, respectively (Fig. 1). As shown in Table 2, of the 34 differentially abundant miRNAs in LAMH versus NAMH, 18 miRNAs were significantly lower and 16 miRNAs were significantly higher in the abundance level, whereas 44 miRNAs were significantly lower, and 54 miRNAs were significantly higher in the HAMH versus NAMH group. The two abnormal groups i.e., LAMH versus HAMH were also compared with one another, however, no significantly differently abundant miRNA was identified (data not shown).
Correlation of the validated miRNAs with age and AMH. A woman's fertility gradually declines with age and this decline significantly correlates with the number and quality of her eggs. To exclude the age-related changes in miRNA abundance level, correlation analysis between the validated miRNA abundance levels (phase III, i.e., 29 miRNAs) (Fig. 1) and the age of the included women was performed. As shown in Fig. 2A, the miRNA abundance levels were shifted slightly towards older women with lower AMH levels (i.e., women with poor ovarian reserve) and were shifted slightly towards younger women with higher AMH levels. However, this slight shift in the abundance level of miRNAs was not significantly correlated to age, suggesting that the altered abundance level of miRNAs that we observed occurs regardless of age, as depicted in Fig. 2B. In contrast to age, the validated miRNA abundance levels (phase III, i.e., 29 miRNAs) in the LAMH, HAMH, and NAMH were mostly abundant in the range of low, high and normal AMH concentrations, respectively, as shown in Fig. 2C. To prove that the alteration in miRNA abundance level is due to the changes in AMH levels, a spearman's correlation was carried LAMH NAMH HAMH AMH (ng/ml)    2D). Of these correlated miRNAs, 7 miRNAs were correlated with AMH level in the LAMH versus NAMH group and 14 miRNAs in HAMH versus NAMH group. While 5 miRNAs were correlated with AMH in both tested groups i.e., LAMH versus NAMH and HAMH versus NAMH. These findings provide evidence that the miRNA abundance levels changed significantly in the groups, depending on the level of AMH. Additionally, miR-144-5p and miR-140-5p that were previously shared in both groups (i.e., LAMH and HAMH, Fig. 1) were found to be correlated with AMH level in only LAMH versus NAMH. Similarly, miR-7-1-3p and miR-27a-3p were previously shared in both (i.e., LAMH and HAMH, Fig. 1), were found to be correlated with AMH level in only HAMH versus NAMH. Interestingly, the abundance level of miRNAs that were positively correlated with AMH in the LAMH versus NAMH is negatively correlated with the AMH in the HAMH versus NAMH (Fig. 2D, P < 0.05).

Diagnostic accuracy of the validated miRNAs. The miRNAs that have been validated in phase III by
RT-qPCR and correlated with AMH level (26 miRNAs, Fig. 2D, Supplementary Fig. 2) were tested for their suitability as biomarkers for assessing the ovarian reserve of a woman presenting at an infertility clinic. The receiver operating curve (ROC) analysis was performed, in which the miRNAs were tested for their predictive ability to detect women with low, normal, and high-ovarian reserve. All identified miRNAs that correlated with AMH have an AUC > 0.5. The best AUC value was observed for the miR-100-5p for the prediction of an increased AMH level with an AUC = 0.756. The calculated AUC values for the validated miRNAs correlated with AMH are shown in Fig. 2D.

Discussion
In this study, the difference in miRNA abundance level was determined in women with normal, low, and high ovarian reserve by miRNA microarray and RT-qPCR analyses. Considering miRNAs with an adjusted P value < 0.05 exhibiting ≥ 2-fold change in abundance level, the abundance level of 34 and 98 miRNAs was significantly altered in the high (HAMH) and low LAMH groups, respectively, as compared to the normal NAMH group. The result was validated by RT-qPCR for 12 and 23 miRNAs in the LAMH and HAMH groups, respectively. Using an independent set of samples, 14 and 24 miRNAs were validated in LAMH and HAMH groups, respectively (Fig. 1). Despite that the mean age of the three groups differed significantly, statistical analysis indicated that no significant correlation exists between the abundance level of the validated miRNAs and the age of women in the three tested groups (Fig. 2B). On the other hand, the abundance level of the validated miRNAs was  www.nature.com/scientificreports/ significantly correlated with the AMH level in the tested groups (Fig. 2D). Accordingly, we may argue that the changes in miRNA abundance level have resulted from differences in the AMH level rather than differences in the women's age. A statistically significant difference was observed between the three groups in terms of the mean age, PRL, LH, FSH, Basal E2, Testosterone, Androstenedione, DHEA-S, and AFC. Low levels of FSH and Basal E2 were found to correlate with improved ovarian response 17 . Higher levels of LH and testosterone contribute to early ovarian reserve failure 18,19 , while PRL could directly act on the ovary to suppress follicular development 20 . Additionally, higher level of PRL have been correlated with menstrual disorders because of its restraining effect on pulsatile Gonadotropin-releasing hormone (GnRH) secretion as well as inhibition of FSH and LH release 21 .
Our results concord with previous reports showing that many of our dysregulated miRNAs play a role in female reproduction and/or infertility-associated diseases of women and/or her male partner [3][4][5][6][7][8][9][10][11] . In more details, MiR-330-3p, miR-144-5p, and miR-221-3p showed lower abundance levels in the cumulus cells of women with polycystic ovary syndrome (PCOS) 22 , and miR-144-5p and miR-133b expression in the granulosa cells and was linked to female infertility 23 . In particular, miR-144-5p inhibits prostaglandin E2 (PGE2) secretion, which is an important regulator of ovulation and can therefore lead to fertility problems through changes in its synthesis 24 . Xiao et al. found a higher abundance level of miR-133b in meiosis I oocytes when the insulin-like growth factor-I (IGF-1) was overexpressed 25 . IGF-1 is involved in the development of the primordial follicles and the growth of the oocytes 25,26 . The abundance level of miR-140-5p was significantly decreased in the follicular fluid of women with PCOS as compared to controls 27 . Regulation of miR-140-5p by the Estrogen receptor α (ERα) was detected in women with breast cancer and in women with PCOS 27,28 . Similarly, miR-126-3p was also associated with PCOS 29 . The hypermethylation in the promoter site of the miR-126-3p gene was found in granulosa cells of patients with PCOS, which resulted in a reduction of the abundance level of miR-126-3p 29 . Another two miR-NAs, namely miR-148a-3p and miR-28-5p were associated with endometriosis 30,31 . As for miR-148a-3p, He et al. identified it as a modulator for the Estrogen (E 2 )-induced epithelial-mesenchymal transition (EMT), which leads to endometriosis 31 . Besides, Liu et al. found that the reduced abundance level of miR-148a-3p leads to the higher abundance level of its target gene, the small nucleolar RNA host gene 4 (SNHG4), and thus has a direct influence on the ectopic growth of the endometrium outside the uterus 32 . As for miR-28-5p, Vanhie et al. concluded that miR-28-5p can be used as a non-invasive biomarker for endometriosis in infertile women 30 . Additionally, miR-28-5p was involved in the pathogenesis of PCOS via its target gene prokineticin 1 (PROK1) which plays a role in ovarian physiology, implantation of embryos in the endometrium, and success of pregnancies 33 .  www.nature.com/scientificreports/ MiR-21-5p was associated with the female reproductive system in many ways, besides its role in various types of cancer, such as breast 34 , ovarian 35 , and cervical cancer 36 . MiR-21-5p has been associated with diseases that affect female fertility including the development of endometriosis, PCOS, and primary ovarian insufficiency (POI). In-depth, miR-21-5p was found to promote angiogenesis associated with the development of endometriosis 37 . Park et al. observed an up-regulation of miR-21-5p in endometrial cells 38 , while Papari et al. found a reduced expression level of miR-21-5p in the plasma of women with endometriosis 39 . A physiological reduction in miR-21-5p was observed in human endometrial stromal cells (hESC) in preparation for pregnancy (decidualization) 40 . The abundance level of miR-21-5p was also investigated when the ovarian reserve was changed. Karakaya et al. found a higher abundance level of miR-21-5p in cumulus cells of women with low ovarian reserve (poor responder) presenting at an infertility clinic, while the abundance level of miR-21-3p was lower 41 , suggesting that elevated miR-21-5p abundance level in cumulus cells is not regulated at the pre-miR-21 level in women with low ovarian reserve. Very recently, a reduction in miR-21-5p abundance level was observed in the plasma of infertile women with abnormal AMH levels 42 . Overall, it became clear that the miR-21-5p is changed in many processes that are related to the ovarian reserve. It can be assumed that miR-21-5p is a key regulator of female fertility.
MiR-100-5p has recently been identified as a potential biomarker for various female reproduction disorders, in ectopic pregnancies 43 , endometriosis 44 , recurrent implantation failure 45 , and decreased ovarian reserve 46 . A lower expression level of miR-100-5p was found in women with ectopic pregnancies compared to that in normal pregnancies 43 . Similarly, a reduction in miR-100-5p expression level was observed in serum and plasma of women with unsuccessful embryo transfer compared to successful embryo transfer 45 , suggesting that miR-100-5p, may serve as a potential biomarker for recurrent miscarriage and/or recurrent implantation failure 38 . Woo et al. also observed a reduction in miR-100-5p expression level in granulosa cells of women with diminished ovarian response 46 and reported that miR-100-5p targets fibroblast growth factor receptor 3 (FGFR3), insulinlike growth factor 1 receptor (IGF1R), and cyclin E and cyclin-dependent kinase (CDK), which play a role in the proliferation and steroidogenesis of granulosa cells. Therefore, Woo et al. concluded that miR-100-5p limits the proliferation of granulosa cells by binding to these target genes. In agreement with our finding, miR-100-5p decreased in abundance in women with lower and higher ovarian reserve compared to normal suggesting that this miRNA can be used as a biomarker for ovarian reserve in women undergoing infertility treatment. Based on these studies, our finding of an altered abundance of miRNAs (especially MiR-21-5p and miR-100-5p) in women with reduced or increased ovarian reserve suggests that they may be involved in the pathogenesis of the condition and lay the ground for the development of novel diagnostic approaches for women manifesting diminished ovarian reserve and subsequent fertility complications.
Various parameters and biomedical markers have been proposed to detect the ovarian reserve including age, FSH, Estradiol, Inhibin, AMH, and AFC 1,47 . Of these markers, AFC and AMH levels have been considered good predictors of the ovarian reserve during ART compared with other traditional measures 1,47-50 . AMH level was observed to be highly correlated with the AFC, age, and basal FSH level 1,49 . Therefore, AMH has been labelled as an adequate predictor of the ovarian reserve before IVF/ICSI treatment, in both, high and poor ovarian responders 1,47-50 . Although the AMH level is highly correlated with the AFC, in clinical practice there is a discrepancy between the AMH level and the AFC in about 18-32% of women presenting at an infertility clinic 48,51 . In our study, due to the controversial discussion over the decades between the AMH and AFC, we opted to classify our included women based on the AMH level and because all included women in our study have been tested for AMH, but not for AFC. Furthermore, AMH was correlated with LH, FSH, testosterone, and androstenedione and this is likely to be attributed to the interaction of other hormones during the menstrual cycle. The age (negatively correlated) and the AFC (positively correlated) were more significantly correlated with the AMH than to other markers and this is in agreement with the existing literature 1,[47][48][49][50][51] . Age is the single biggest factor affecting a woman's chance to achieve pregnancy, and therefore age cannot be ignored in infertility research. It has been shown that a relevant portion of human miRNA changes depending on age and sex 52 . In our study, age was highly correlated with the AMH level, but not with the miRNA abundance level of the validated miRNAs, and thus provide evidence that the alteration in miRNA abundance level was not linked to age, and most probably linked to the changes in AMH level. The AMH level was positively correlated with the abundance level of miRNAs in LAMH versus NAMH group and negatively correlated with the abundance level of miRNAs in HAMH versus NAMH group, indicating that when the AMH level decreases, the abundance level of the identified miRNAs decreases and when the AMH level increases, the miRNA abundance level also decreases. Accordingly, an abnormal AMH level always leads to a reduction in the abundance level of miRNAs. This could explain the observed non-significant correlation and non-significant alteration in the miRNA abundance level in the screening and validation phases when the two abnormal groups (i.e., LAMH versus HAMH) were compared to one another. These findings suggest that miRNA abundance level is always reduced in case of an abnormal AMH level and probably in the cases with manifestations associated with diminished ovarian reserve. Based on these findings it is, however, premature to draw such a conclusion, as the number of included subjects in each group is too low.
One of the primary goals of this study was to find a diagnostic biomarker that can assess a woman's ovarian reserve and thus her fertility and to supplement or replace previously used biomarkers. The altered miRNAs that are correlated with AMH were therefore subjected to a ROC analysis, which allows a statement to be made about the quality of the miRNA as a biomarker. For this purpose, the AUC values were determined and found greater than 0.5 in all validated miRNAs correlated with AMH in both comparison groups LAMH versus NAMH and HAMH versus NAMH. This means that the miRNAs identified in the LAMH versus NAMH group can predict a low AMH level and thus a low ovarian reserve. MiRNAs identified as biomarkers in HAMH versus NAMH can predict a high AMH level and thus a high ovarian reserve. MiRNAs identified in both groups can predict abnormal AMH levels or abnormal ovarian reserve. The miRNAs, which can predict either low or high ovarian reserve, would tend to be more useful as biomarkers based on the information about the current state of the www.nature.com/scientificreports/ ovarian reserve. The best AUC value was determined for the miR-100-5p. It indicates with a probability of 76% that the ovarian reserve is higher than normal. In summary, the alteration in miRNA abundance level is in part associated with female reproduction and with diseases that affect female fertility status, and other miRNAs were newly identified in this area, suggesting that some miRNAs are involved in the maintenance of female fertility and changes of some other miRNAs might adversely and /or negatively affect female fertility. Altered abundance levels of miRNAs, particularly miR-100-5p, can provide new insights into the underlying mechanisms of female fertility and thus improve the diagnosis and treatment of infertility. The identified and validated miRNAs in our study should be further validated in a larger number of samples to confirm their predictive ability as biomarkers that could complement or possibly replace the previous markers as genetic biomarkers.

Methods
Study population and sample collection. Blood samples were collected from a total of 159 consecutive women undergoing infertility treatment with IVF-ICSI at Saarland University School of Medicine IVF Center (Homburg/Saar) between November 2016 and May 2020. The mean age was 32 ± 4.8 years (range 20-44 years). At enrollment, ultrasonography was conducted on the second day of the menstrual cycle to evaluate the anatomical characteristics of the female reproductive system and determine the antral follicular count (AFC). Peripheral blood samples were then collected from each woman into serum tubes and PAXgene blood tubes (Becton-Dickinson, Heidelberg, Germany) on the third day of the menstrual cycle. The serum was immediately prepared by centrifugation at 1800g for 15 min and used to determine the level of Free Thyroxine 4 (FT4), Thyroid-Stimulating Hormone, Prolactin (PRL), Luteinizing Hormone (LH), Follicle Stimulating Hormone (FSH), Estradiol (E2), Testosterone, Androstenedione, Dehydroepiandrosterone Sulfate (DHEA-S), and anti-Müllerian hormone (AMH). All PAXgene blood tubes were stored at room temperature for at least 24 h to ensure complete lysis of the blood cells, then stored at − 20 °C for several days and finally transferred to − 80 °C for long-term storage until RNA including miRNA isolation. This study was approved by the Saarland University Institutional Review Board committee (Ärztekammer des Saarlandes Nr. 160/15) and informed consent was obtained from each participant and the study complies with the Declaration of Helsinki.
The determination of miRNA abundance levels was performed in three successive phases as indicated in Fig. 1. In phase I (screening phase), samples were randomly grouped based on their AMH level into low AMH level (LAMH, n = 12), normal AMH level (NAMH, n = 13), and high AMH level (HAMH, n = 13).
These samples were used to identify the differential miRNA abundance level by applying a miRNA microarray. In phase II (validation of the screening phase), the initially identified miRNAs (phase I) were evaluated, by RT-qPCR assay. In phase III (Validation of the independent cohort), the selected miRNAs in phase II were additionally validated by RT-qPCR in an independent set of samples from 27, 67, and 27 women from the low, normal, and high level of AMH, respectively. Isolation of total RNA, including miRNAs. Total RNA including miRNAs was isolated from blood samples using PAXgene Blood miRNA Kit on the QIAcube robot (Qiagen, Hilden, Germany) following the manufacturer's recommendations. DNase I treatment (Qiagen, Hilden, Germany) was carried out during the isolation to eliminate any genomic DNA contamination as previously described 9 . The total RNA concentration was measured using the NanoDrop ND-2000 spectrophotometer (Thermo Fisher Scientific, Massachusetts, United States). RNA purity was assessed by determining the OD 260/280 and the OD 260/230 ratios. The quality of total RNA was assessed using the Agilent Bioanalyser 2100 Eukaryote Total RNA Nano Series II (Agilent Technologies, California, United States).
MiRNA microarray. MiRNA expression profiles in LAMH, NAMH, and HAMH were determined by hybridization to the Sureprint G3 Human v21 miRNA microarray chips, 8 × 60 K (release 21.0), each containing 2549 human miRNAs (Agilent Technologies). Hybridizations were carried out following the manufacturer's recommendations. In brief, 120 ng RNA from each sample was processed using the miRNA Complete Labeling and Hybridization Kit (Agilent Technologies) to generate fluorescence-labeled miRNA. The microarrays were loaded and incubated at 55 °C for 20 h with rotation. After several washing steps, microarrays were scanned with the Agilent Microarray Scanner at 3 microns in double path mode. Data was acquired using Agilent AGW Feature Extraction software version 10.10.11 (Agilent Technologies).
Reverse transcription and quantitative real-time PCR (RT-qPCR) of miRNA. The abundance level of circulating miRNAs was quantified by RT-qPCR using the Biomark HD System (Fluidigm Corporation, California, United States) and the TaqMan microRNA Assays (Thermo Fisher Scientific) according to the as previously described 53 . Briefly, complementary DNA (cDNA) was generated in 8 µL reactions by reverse transcription of 350 ng total RNA using the TaqMan MicroRNA Reverse Transcription Kit and RT Primers Pool (10×) (Thermo Fisher Scientific). Following reverse transcription, 2.5 μL of the generated cDNA was preamplified by mixing 12.5 µL of TaqMan PreAmp Master Mix (2×) and 3.75 μL of PreAmp Primers Pool (10×) (Thermo Fisher Scientific) in 25 µL reaction volume. Following the preamplification of the cDNA, RT-qPCR was carried out with 96.96 Dynamic Array IFC for Gene Expression arrays (Fluidigm Corporation) as indicated in Fluidigm's protocol (PN 68000130 E1). Briefly, every 10× Assays contained 3 µL TaqMan Primer Assay (20×) (a mixture of forward and reverse primers, and probe) (Thermo Fisher Scientific), and 3 µL Assay Loading Reagent (2×) (Fluidigm, PN 85000736) Statistical analysis. Microarray images were scanned using the Feature Extraction Software (Agilent Technologies) and the extraction of data was carried out using GeneSpring GX software (version 14.9.1, Agilent Technologies). Microarray measurements were normalized using quantile normalization and the differential abundance levels of miRNAs were identified for each sample. An unpaired two-sample t test with Benjamini-Hochberg correction for multiple testing was applied, and adjusted P value of < 0.05 was considered statistically significant. Fold change for the LAMH and HAMH groups was obtained with respect to the NAMH group. Consequently, the lower and higher abundant miRNAs with a fold change > 2 were considered for further studies. The diagnostic value for each validated miRNA was evaluated by receiver operating characteristic (ROC) curves analysis and subsequently the area under the ROC curve (AUC) was computed to assess the potential use of miRNA(s) as a biomarker. The relative quantitative method of 2 −ΔΔCt was used to measure the dynamic change of selected validated miRNAs using RNU6B small nuclear RNA (snRNA) as an endogenous reference miRNA as previously validated for this type of sample [53][54][55][56][57][58][59][60][61] . The correlation analysis was carried out by the Spearman correlation coefficient and the differences in clinical characteristics among the three tested groups i.e., LAMH, HAMH, and NAMH were analyzed by analysis of variance (ANOVA) and data was presented as mean ± standard deviation with a P value of < 0.05 was considered statistically significant.