Introduction
Polycystic ovary syndrome (PCOS) is a heterogeneous endocrine disorder with an overall prevalence ranging from 6 to 10% according to the diagnostic criteria used [
1,
2]. The impact of PCOS on patients is not limited to oligo- or anovulation, as a large number of patients experience poor reproductive outcomes [
3‐
5]. The current research on PCOS mainly focuses on improving ovulatory function, while the mechanisms associated with adverse fertility are rarely mentioned [
6,
7]. In other words, PCOS patients still face the risks and challenges of adverse fertility outcomes.
A plethora of data showed that poor reproductive outcomes are associated with endometrial dysfunction [
3]. Several studies have found that some clinical and biochemical factors can exert a deleterious effect on endometrium [
8‐
10]. This indicates that clinical and biochemical factors may affect reproductive outcomes.
Autophagy, the primary intracellular degradation system, plays a pivotal role in cellular renovation and homeostasis by recycling waste materials [
11]. Previous studies have elucidated the intricate interplay between autophagy, apoptosis, and necrosis. For instance, autophagy can trigger other forms of cell death through selective degradation [
12]. Recently, some studies found that autophagic degradation of ferritin leads to ferroptosis due to elevated levels of labile iron and ROS [
13,
14]. Some studies have proved that defects in autophagy can lead to follicular development disorders [
15,
16]. Furthermore, emerging research has unveiled a link between autophagy and pregnancy loss as it influences immune tolerance at the maternal–fetal interface [
17].
This study conducted a secondary analysis of PCOS proteomic and clinical data to investigate the association between autophagy and endometrium, as well as their impact on reproductive outcomes. We analyzed and screened autophagic proteins and biochemical indicators that have a critical impact on the pregnancy outcomes of PCOS. Subsequently, prognostic models were constructed based on characteristic proteins and clinical data, both of which demonstrated robust predictive power. This research significantly contributes to the existing knowledge regarding the relationship between autophagy and pregnancy outcomes in PCOS.
Materials and methods
Samples collection
This study is a secondary analysis based on the proteome dataset of endometrium samples obtained from PCOS patients and controls, aged from 21 to 40 years, collected from The Reproductive Center of the Second Hospital of Lanzhou University during the period from September 2019 to September 2020. This dataset included 33 PCOS patients and 7 normal control subjects. The patients who were recruited had to satisfy the Rotterdam criteria meet the following 2–3 items: (1) Oligo-and/or anovulation; (2) Clinical and/or biochemical signs of hyperandrogenism; (3) Polycystic ovaries. The exclusion criteria were as follows: (1) Subjects suffer from hypothyroidism, hyperprolactinemia, adrenal disease, hypertension, and diabetes; (2) hormone-medication and drugs affecting glucose metabolism within the last 3 months. The control group was non-PCOS with successful pregnancy and live birth. They had regular menstrual cycles and normal ovarian morphology via routine ultrasound scans. Informed consent was obtained from all participants before collecting samples. The study was authorized by the Ethics Committee of Lanzhou University Second Hospital (2017A-057).
The endometrial samples were the proliferative endometrium. The endometrial samples were obtained using a pipelle endometrial aspirator and stored at-80℃.
Clinical and prognosis data collection
Demographic characteristics, including age and BMI, were recorded from outpatient medical records. Serum samples collected during the 2–5 days of menstruation were utilized for the analysis of biochemical indicators, coagulation index, and sex hormones. The analyzed biochemical indicators encompassed serum lipid concentration, fasting plasma glucose levels(FPG), insulin levels, thyroid hormone levels, homocysteine levels, vitamin D3 levels, CA125 levels, and D-dimer. Sex hormones include basal testosterone (T), basal luteinizing hormone (LH), basal follicle-stimulating hormone (FSH), and the anti-mullerian hormone (AMH). The insulin resistance index (IR) is calculated by the HOMA-IR index, which was calculated as fasting plasma glucose (FPG) (mmol/l) × fasting insulin (lU/ml)/22.5, and a value of > 2.6 was considered IR [
18]. Endometrial thickness (ET) was examined by ultrasound scanning.
Reproductive outcomes and gestational duration were used as prognostic data, Reproductive outcomes include live birth and adverse fertility. Gestational duration includes the gestational time of live birth and adverse gestational time weeks. Gestational time was estimated in weeks.
Sample preparation and fractionation, data-dependent acquisition (DDA) mass spectrometry, mass spectrometry data analysis, and database search have been described in detail in previous articles [
4].
Obtain the DEPs and the autophagy related proteins
The differential expression protein analysis was based on R package (limma). The screening criteria were |Log
2fold change (Log
2Fc)|> 0.585 and adjusted
P < 0.05 [
4]. Autophagy-related proteins (ARPs) derived from the Autophagy Database (
http://www.autophagy.lu/).
The functional enrichment analysis of DEPs
Import DEPs into
http://metascape.org/gp/index.html for metascape analysis. Functional and pathway enrichment analysis by Gene Ontology (GO) enrichment analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis. Min overlap = 3 and Min Enrichment = 1.5 were the screening conditions. The
P-value < 0.01 was considered significant.
Identification of candidate autophagy proteins
We overlapped the ARPs and endometriosis-related proteins. Univariate Cox regression analysis was used to identify the proteins related to pregnancy outcomes. To further identify more reliable autophagy proteins, we conducted LASSO regression algorithm. The “glmnet” package was used to construct the LASSO model with penalty parameter tuning conducted by ten-fold cross-validation. The expressions of candidate autophagy proteins were used to establish a risk model.
Establishment and evaluation of model
Based on the expressions of candidate autophagy proteins, multivariate Cox regression analysis was used to establish AutoSig Risk Model, Forward and backward method were empolyed for filtering models. The risk score was evaluated by formula as follows: \(\mathrm{AutoSig }({\text{PCOS}})={\sum }_{i=1}^{n}coef({Autopro}_{i})*expr({Autopro}_{i})\). AutoSig (PCOS) represents a prognostic risk score, \(coef({Autopro}_{i})\) represents the risk coefficient of ith prognostic autophagy protein. \(expr({Autopro}_{i})\) is the expression level of the ith prognostic autophagy protein for the patient. The PCOS samples were separated into high-risk and low-risk by the risk score cutoff value (median risk score). Kaplan–Meier method was used to estimate the reproductive outcomes of different groups in R package (survival and survminer). At the same time, Logistic regression was performed for clinical data. The outcome variable was the presence or absence of a live birth. Similarly, forward and backward methods were used for filtering models. We obtained a CliSig Risk model formula as follows: \({\text{P}}=1/\left(1+\mathrm{exp }\left(- \left(\beta +{\beta }_{1}*{x}_{1}+{\beta }_{2}*{x}_{2 }+{\beta }_{3}*{x}_{3}+{\beta }_{4}*{x}_{4}\right)\right)\right)\)
The ROC curves were evaluated for the AutoSig Risk Model and CliSig Risk Model. The decision curve analysis (DCA) curves was performed to assess the net benefits with the Risk Model.
Statistical analysis
The StataSE 15.0 software was used to calculate clinical data. The proteomic data were analyzed by R software. The results were shown as mean ± standard deviation (SD) or median (interquartile range) according to the normal disttribution assumption. The binary logistic regression model was used to develop a CliSig Risk Model, the Cox regression model was used to develop an AutoSig Risk Model. All statistical tests were two-sided, and P values of < 0.05 were considered significant.
Discussion
PCOS is a common gynecological disease characterized by reproductive and metabolic disorders which are related to the occurrence and progression of diseases [
19,
20]. The comorbidities of PCOS, including (obesity, metabolic syndrome, hyperinsulinemia or hyperandrogenism), may contribute to pregnancy loss [
21]. Obesity and T2DM associated features such as dyslipidemia, oxidative stress, hyperglycemia, hyperinsulinemia could interrupt and compromise autophagy [
22]. While poor endometrial receptivity can lead to adverse reproductive outcomes [
23]. We could hypothesize that downregulation of autophagy in PCOS patients might lead to poor endometrial receptivity, thereby increasing the incidence of miscarriage. A study using an obese mouse model showed that autophagy was more up-regulated in decidualizing cells of control mice compared to high-fat/high-sugar diet mice [
24]. This study attempts to screen the factors most closely related to pregnancy loss in PCOS based on the analysis of PCOS proteomic data and clinical data. Providing a reference for the mechanism research and clinical decision-making.
Molecular functions and pathways could explain the reasons for poor endometrial receptivity in PCOS patients. In our study, the DEPs were shown to be involved in metabolism of RNA pathway. Different types of RNA and RNA-related complexes are recruited to and degraded by autophagy pathway [
25]. Lots of studies have demonstrated that inhibitors of autophagosome formation significantly block starvation-induced RNA degradation [
26,
27]. The autophagy pathway is damaged, while the metabolism of RNA pathway will be inhibited. This has been positively validated in our research.
In our research, we established two prognostic models based on proteomics and clinical data. After evaluating the models, we found that both models had good predictive performance. The autophagic protein model based on 3 proteins (ARSA, ITGB1, GABARAPL2). Interestingly, our new autophagy proteins model achieved an AUC of 0.922 with only 3 feature proteins, surpassing our previous model which used 5 feature proteins and had an AUC of 0.884. This demonstrates the superiority of our current model. ARSA, ITGB1, GABARAPL2 were rarely studied in PCOS in previous studies. ITGB1 is integrin, which can affect tumor process by regulating angiogenesis, apoptosis, and metastasis [
28,
29]. It is widely recognized that ITGB1 involved and promotes the adhesion ability of NCAM1
birgh NK cells at the maternal–fetal interface [
30‐
32]. This indicates that there is significant research value in exploring the relationship between ITGB1 and PCOS, and it can also serve as a predictor of pregnancy outcomes in individuals with PCOS. RASA is a lysosomal enzyme that catalyze degradation of sulfatides into galactosylceramides (GalC) [
33,
34]. Research has found that the lack or complete absence of ARSA presents metachromatic leukodystrophy which is characterized by the degradation of intellectual function and motor skills and often fatal in early childhood [
35‐
37]. This indicates ARSA may be an important factor in pregnancy loss in PCOS. GABARAPL2 (also called GATE-16) belongs to the GABARAP subfamily of Atg8 proteins [
38]. The Atg8 proteins play a key role in the sealing of the isolation membrane which is a vital role in autophagy [
39]. This further demonstrates the important significance of autophagy in PCOS.
The clinical data model is based on 4 variables (TSH, VD3, TPOAB, Insulin). The results show that the insulin level is a reliable predictor of pregnancy loss in PCOS. Hyperinsulinemia affects the immune response of the endometrium by decreasing the expression of glycodelin and IGF-binding protein-1 [
8], a large number of studies have found that TSH is associated with adverse pregnancy outcomes [
40,
41]. Multiple studies have demonstrated the impact of vitamin D on PCOS phenotype and pregnancy loss [
42‐
44]. Research has found a significant correlation between TPO-AB and infertility in patients with PCOS [
45].
Interestingly, we conducted a linear analysis of the variables in both models. ITGB1 plays an important role in beta cell development and function, while some studies have found a positive effect of EIF4G1 on insulin secretion [
46,
47]. Recent studies have shown that inactivation impairs insulin function [
48], which supports the reliability of our results. This may be the mechanism behind pregnancy loss in PCOS. In the present study, we observed a significant correlation between ARSA Insulin and TSH expression, however, there is limited research on the relationship between ARSA and insulin which deserves further study.
Our study still has some limitations that require further study. Firstly, it was a retrospective study, the sample size was relatively small and the public database PCOS proteomics data was few. These findings need to be verified in future intervention studies. Secondly, although we obtained only three proteins with good predictive performance, the use of machine learning algorithms may miss some useful predictive factors that can’t be ignored. Thirdly, Numerous experiments are needed to verify how these proteins and pathways affect the receptive mechanism of the endometrium in PCOS patients.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.