Preeclampsia (PE) is diagnosed by de novo onset of hypertension with blood pressure over 140/90 mmHg and substantial proteinuria of $\ge$300 mg in 24 hours at or after 20 weeks of gestation and affects 2% to 8% of pregnancies [1, 2, 3]. PE is associated with maternal and offspring morbidity, including fetal growth restriction, preterm birth, oligohydramnios, and maternal end-organ damage [4]. Besides, it is the second leading cause of maternal mortality just behind maternal hemorrhage. This pregnancy complication presents a major risk factor for other disorders, such as diabetes, cardiovascular disease, and renal complications [5, 6, 7]. Although the genuine etiology of PE still remains to be clarified, it is widely accepted that dysregulation of angiogenesis and trophoblast apoptosis is the major contribution to PE [8, 9]. After implantation, the extravillous trophoblast (EVT) cells migrate into the decidua and then remodel uterine arteries by replacing the vascular endothelial and muscle cells. The impaired invasion of EVT cells could affect angiogenesis and disturb the remodeling of the maternal myometrial spiral artery, then inducing excessive pregnancy-incompatible factors production, such as inflammatory cytokines and anti-angiogenic factors [10, 11]. The presence of inflammation can aggravate trophoblast apoptosis which can further disrupt cell migration and placental vascularization [12, 13].

Autophagy, a common phenomenon in eukaryotic cells, is an important metabolic process for cytoplasmic proteins and organelles degrading into fatty acids and amino acids [14]. Previous studies have shown that autophagy is relevant to various diseases, including respiratory diseases, malignancies, and hypertension in pregnancy [15, 16, 17]. Some studies suggested that autophagy protects the placenta against pathogens and stress, and has physiological functions for maintaining a normal pregnancy [18, 19]. Hyperactivated autophagy induced by oxidative stress may affect cell invasion and placental vascularization, and then promote the development of PE [20]. Other studies demonstrated that under oxidative stress conditions, autophagy could improve cell survival and reduce cell apoptosis [21]. Thus, exploring the roles of autophagy in PE may help clinicians better understand the etiology of PE and seek for early diagnostic and prognostic markers. At present, over 30 kinds of autophagy-related genes (ATGs) are demonstrated to be closely related to autophagy. For instance, it has been shown that trophoblast-specific conditional autophagy related (Atg)7 knockout mice exhibited a significant elevation in blood pressure, a smaller placenta and reduced expression of placental growth factor [22]. However, the role of ATGs in PE has not been fully understood.

Using published gene expression data from the Gene Expression Omnibus (GEO) database, differentially expressed ATGs were obtained in PE and healthy samples. The least absolute shrinkage and selection operator (LASSO) logistic regression and support vector machine-recursive feature elimination (SVM-FRE) algorithm were applied to select the feature genes among ATGs. Then, a predictive model was constructed based on feature genes, which could efficiently distinguish PE patients from the healthy pregnancies, and may be used for early diagnosis of hypertension in pregnancy.

Preeclampsia is characteristic of placental shallow implantation and insufficient spiral artery recasting, which is considered to be of placental origin [38]. It remains a significant contributor of short- and long-term maternal and fetal morbidity. It is generally accepted that impaired angiogenesis and trophoblast apoptosis play important roles in the occurrence of PE. In recent decades, an association between autophagy and pregnancy has been demonstrated. Autophagy is the most fundamental phenomenon in eukaryotes, through which senescent or damaged structures are degraded to maintain microenvironment stability. Studies showed that autophagy can influence the trophoblast cell homeostasis via maintaining the reactive oxygen species (ROS) balance so to preserve the angiogenic capacity [39]. It is also reported that autophagy could preserve the function of trophoblast cells by regulating vascular endothelial growth factor A (VEGFA) and fms-related tyrosine kinase 1 (FLT1) expression and protecting against cell apoptosis at the maternal-fetal interface [17]. However excessive or impaired autophagy can result in pregnancy related disorders. Li et al. [17] found that excessive autophagic activity in trophoblasts or endothelial cells could affects trophoblast invasion and the placenta vasculature, thus participating in the development of preeclampsia. This suggests that to some extent the development of preeclampsia is regulated by the autophagic mechanism. To better understand the role of autophagy in PE, we used published gene expression data from GEO database to construct bioinformatics analysis of ATGs in PE.

In this study, 20 differentially expressed ATGs were obtained in placentas from women diagnosed with PE. Most of these autophagy genes were highly-expressed, indicating that hyperactivated autophagy may be dominant in the development of PE. Functional analysis indicated these genes were related to behaviors like autophagy, apoptosis, angiogenesis, inflammation, and immune response. The inflammatory response is a well-recognized contributor of PE and imbalance in immune response may cause abnormalities in angiogenesis and placental structure [40]. Studies demonstrated that inflammation can further activate and promote trophoblast apoptosis and the activation of apoptosis can backfire on trophoblast through disrupting cell migration and placenta vasculature, indicating that autophagy damage may play a pivotal role in the progress of PE via inducing inflammation status and affecting angiogenesis and apoptosis. KEGG analysis showed that these genes may be involved in hypoxia-inducible factor 1 (HIF-1), forkhead box O (FoxO) and AMP-activated protein kinase (AMPK) signaling pathway. Severe or persistent hypoxia in PE placenta induces the overexpression of hypoxia-inducible factor 1-alpha (HIF-1ɑ), which further causes an increase in soluble fms-like tyrosine kinase 1 (sFlt-1) and ultimately lead to placental dysfunction [41]. Studies also illustrated that hypoxia induces autophagy in a HIF-dependent induction of BCL2/adenovirus e1B 19 kDa protein interacting protein 3 (BNIP3) and BCL2/adenovirus e1B 19 kDa protein interacting protein 3-like (BNIP3L) [42]. AMPK, a crucial kinase regulating energy homeostasis, plays a pivotal role in promoting autophagy via regulating phosphorylated autophagy-related proteins in mTORC1, ULK1 or indirect proteins to stimulate the expression of autophagy transcription factors such as FoxOs [43, 44]. Overall, these findings indicate that autophagy and ATGs may play a significant role in the develpoment of PE.

Since PE remains a serious threat to the health and life of the mother and the fetus, it is of vital importance for early diagnosis in order to expand the window of treatment and reduce complications. The predictive effect of well-known sreening indicators is not satisfactory, such as placental growth factor (PIGF) and soluble fms-like tyrosine kinase 1 (sFlt-1) [45]. Therefore, there is an urgent need for novel biomarkers with high specificity and sensitivity. To elucidate the sensitivity and specificity of the autophagy genes in predicting PE, 12 feature genes were screened out and presented a remarkable predictive efficiency of PE and normal pregnancies. Considering the difficulties in accessing placenta samples during pregnancy, we assessed the discrimination power of these selected feature genes in maternal blood samples. Results shown that feature ATGs also succeeded in separating patients with PE from the healthy subjects. Ultimately, four ATGs, LEP, ERO1L, MAPK8, and PIK3CB, were selected as hub diagnostic genes of ATG-based signature, which were both differentially expressed in placenta and blood. Upregulation of leptin (LEP) was detected in placenta samples from PE patients when compared with those from normal pregnancy and depletion of LEP could repress apoptosis and promote proliferation, migration, and invasive capacity [46]. The protein product leptin of LEP was also increased in the serum of PE patients in comparison to those with normotensive pregnancies [47]. MAPK8 is reported to dissociate the complex of BCL2L1 (BCL-like 1) and BECN1 (beclin 1) to trigger autophagy and TNFSF10 (tumor necrosis factor superfamily member 10)-induced MAPK8 activation and autophagy can be effectively suppressed by knockdown of TRAF2 or RIPK1 [48]. Several studies shown an association of decreased MAPK8 activity with pulmonary hypertension [49, 50]. Yang et al. [51] comfirmed a decreased expression level of MAPK8 in PE placentas. The role of MAPK8 in pre-capillary pulmonary hypertension gives us a hint that it might also exert function on placenta vasculature. The hypermethylation of ERO1L has been reported to be triggered by increased binding of DNA (cytosine-5)-methyltransferase 1 (DNMT1) to the ERO1L promoter which contributes to trophablast cell apoptosis in the placenta of PE rats [52]. However, the role of PIK3CB in PE has not been well studied. In this study, the expression patterns of ERO1L and PIK3CB in PE placental samples and blood samples were quite contrary. We assumed that this phenomena may be caused by different microenvironments. Gene expression is quite a complex process, which is related not only to the gene itself, but also to the upstream and downstrem regulators.

There are also some shortcomings in our study. Firstly, this study was based on bioinformatics analysis and the expression patterns of ATGs differed between blood and placental samples. Therefore, further validation from both in vivo and in vitro experiments is needed. Secondly, though the ATGs-based signature model presented a powerful predictive value in PE, the efficiency remains to be further validated in multicenter, large-scale prospective studies.

In this study, we identified 20 potential autophagy-related genes of preeclampsia via bioinformatics analysis. Moreover, four feature genes LEP, ERO1L, PIK3CB, and MAPK8 may be used as potential biomarkers for preeclampsia prediction. Identification of these ATGs might help to understand the molecular mechanism underlying the development of preeclampsia and might provide a new perspective on the management and treatment of preeclampsia.

The raw data of the present study are downloaded from the GEO data portal (https://www.ncbi.nlm.nih.gov/geo/; Accession number: GSE75010, GSE48424, GSE84260), which is a publicly available database.

LQW made substantial contributions to conception and design. JYS and XYT collected and analyzed the data, as well as draft the paper. JYZ and YLF made contributions to interpretation of data for the work and reviewing the manuscript for important intellectual content. LQW gave final approval for the version to be published. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

The raw data of the present study are downloaded from the GEO data portal, which is a publicly available database. Neither ethics approval nor consent is needed.

Not applicable.

This research received no external funding.

The authors declare no conflict of interest.