Circulating miRNAs Associated with Dysregulated Vascular and Trophoblast Function as Target-Based Diagnostic Biomarkers for Preeclampsia

Preeclampsia (PE) is a pregnancy-specific disorder associated with hypertension and proteinuria. Since there is no proven method to treat PE, early prediction and accurate diagnosis are essential for appropriate management of the disease. Thus, reliable biomarkers for diagnosing PE need to be identified and evaluated. We analyzed serum-soluble factors and miRNAs in 92 patients with PE and an equal number of healthy controls to identify new useful biomarkers for PE. Serum miR-31-5p, miR-155-5p, and miR-214-3p levels were significantly elevated in these patients and highly correlated with clinical symptoms of hypertension and proteinuria, whereas the miR-1290-3p level was decreased. The increased miRNAs were upregulated in an NF-κB-dependent manner and suppressed endothelial nitric oxide synthase (eNOS) and placental growth factor (PlGF) expression. The level of each miRNA had greater than 90% diagnostic accuracy, which was further increased by analyzing its ratio relative to that of miR-1290-3p. Taken together, the ratios of miR-31-5p, miR-155-5p, or miR-214-3p to miR-1290-3p may serve as reliable diagnostic or prognostic tools for PE, thereby providing evidence that these miRNAs are promising mechanism-based targets for designing therapeutic and preventive strategies for the clinical management of PE.


Introduction
Preeclampsia (PE), a pregnancy-specific disease associated with hypertension, proteinuria, thrombocytopenia, or renal insufficiency, is a common disorder affecting 5% to 7% of all pregnancies [1]. It is a major cause of maternal mortality (10-15%) worldwide and accounts for a substantial proportion of low-birth-weight neonates and iatrogenic preterm delivery [2,3]. enzyme levels, twin or multiple pregnancies, or any evidence of previous medical illness were excluded. PE was defined as systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg on at least two occasions 4 h apart but no more than 7 days apart, which developed after 20 weeks of gestation in previously normotensive women; proteinuria was defined as ≥300 mg protein in a 24 h urine specimen, as previously described [20]. No women with pre-existing hypertension, with or without superimposed PE, were included in the present study. All experiments were carried out in accordance with the approved guidelines and regulations, in line with the tenets of the Declaration of Helsinki.

Sample Collection
The serum fractions were obtained according to standard protocols from all participants. Briefly, at the time of clinic attendance, maternal venous blood samples were collected into standard Vacutainer tubes (EDTA, Becton-Dickinson, Franklin Lakes, NJ, USA) and processed within 1 h by centrifugation at 2000× g for 10 min at 4 • C. The supernatants were quickly removed, aliquoted, and stored immediately at −80 • C. For analysis, serum samples were thawed on ice and centrifuged at 3000× g for 5 min, avoiding traces of red blood cells and other cellular debris that could affect miRNA analysis. Placentas were also obtained after full-term normal (spontaneous vaginal) deliveries, and maternal arteries were isolated from the placental bed from the patients as described previously [15].

miRNA Isolation and Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR)
RNA was isolated from serum (200 µL) and cultured cells using the miRNeasy Mini Kit (#217184; Qiagen, Hilden, Germany) according to the manufacturer's protocol and eluted with 14 µL of nuclease-free water. cDNA for analyzing miRNAs was prepared from 1 µg of total miRNA using a miScript II RT kit (#218161; Qiagen, Hilden, Germany). qRT-PCR was performed with a miScript SYBR Green PCR kit (#218073; Qiagen, Hilden, Germany) according to the manufacturer's instructions. The levels of miRNAs were analyzed by miScript Primer Assay with target miRNA-specific and universal primers. The relative levels of miRNAs were normalized to SNORD-95. In addition, total mRNA was isolated using TRIzol reagent (#15596018, Invitrogen, Carlsbad, CA, USA) from cultured cells and placental tissues. mRNA levels of eNOS in placental bed-derived arteries, inducible nitric oxide synthase (iNOS) in placental tissue, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were determined by HiPi Real-Time PCR 2X Master Mix (SYBR Green, Elpis, Daejeon, Korea) with a Rotor-Gene Q real-time PCR cycler (Qiagen) using target-specific primers as previously described [15].

Statistical Analysis
Statistical analyses were undertaken using GraphPad Prism version 6.0 (GraphPad Software, San Diego, CA, USA). All clinical samples, except those for Western blotting, were analyzed in triplicate, and data are expressed as mean ± SEM. Statistical significance was determined using Student's t-test between two groups, and the correlation between the two factors was determined by linear regression analysis. Receiver operating characteristic (ROC) curves were used to analyze the diagnostic utility of differentially expressed miRNAs or produced serum factors. The optimal cut-off point for the plasma miRNA expression level was determined by the Youden index (J = sensitivity + specificity − 1). Statistical significance was established at p < 0.05.

Levels of sFlt-1, sEng, and PlGF in Sera from Patients with PE
Clinical characteristics of the study participants are shown in Table 1. No significant difference in patient age was noted between healthy pregnant women and preeclamptic patients. However, preeclamptic participants had significantly increased blood pressure (systolic and diastolic, 115.20 ± 0.84 vs. 157.80 ± 1.90 mmHg; 74.89 ± 0.78 vs. 99.85 ± 1.20 mmHg, p < 0.001) and proteinuria (1.89 ± 0.19 g/24 h, p < 0.001). Moreover, these women delivered low-birth-weight fetuses (3.08 ± 0.06 vs. 2.39 ± 0.07 kg, p < 0.001) approximately 1.8 weeks earlier (p < 0.001) than the healthy controls. On the basis of the relationship between gestational age and birth weight, the percentage of small for gestational age (SAG) infants (or intrauterine growth retardation; IUGR), appropriate for gestational age (AGA) infants, and large for gestational age (LGA) infants were 29.3%, 68.5%, and 2.2%, respectively, in the women with PE. AGA was defined as a birth weight between 10% and 90% of the average body weight in infants at the same gestational age, while LGA and SGA or IUGR were defined as a birth weight of more than 90% and less than 10% of the average body weight in infants at the same gestational age, respectively [22,23]. Maternal serum concentrations of the antiangiogenic factors sFlt-1 and sEng were significantly higher (p < 0.001) in the preeclamptic patients than in the healthy subjects, whereas levels of the angiogenic factor PlGF were decreased in patient sera (p < 0.001; Figure 1A-C), consistent with previous reports [6,24]. As expected, the sFlt-1/PlGF ratio, known as a predictive marker of PE, was significantly higher in women with PE than in healthy subjects (p < 0.001) and showed a better predictive value than either biomarker alone ( Figure 1D). This suggests that the pathological symptoms and analytical characteristics observed in the patients belonged to the typical PE spectrum.

Comparative Levels of eNOS, iNOS, NO, cGMP, and TNF-α in PE Patients
The NO/cGMP axis plays a crucial role in vascular relaxation, and impairment of the axis causes endothelial and vascular dysfunction, which are associated with vascular constriction and glomerular endotheliosis and result in hypertension and proteinuria [25]. We determined the levels of NO and cGMP in the sera of PE patients and control subjects. Unexpectedly, the levels of both molecules were significantly higher in PE patients compared to healthy control subjects (Figure 2A,B). The NOx levels were positively correlated with the cGMP levels ( Figure 2C), indicating that the increased cGMP levels are derived by NO-dependent activation of sGC. Both NOx and cGMP levels also positively correlated with the serum sFlt-1/PlGF ratio, a suitable diagnostic marker of PE ( Figure 2D,E), indicating that the elevated NO and cGMP levels might be serological features of PE. To investigate which isotype of NOS was responsible for the elevated serum NOx and cGMP levels in the patients, we examined the expression levels of eNOS and iNOS in umbilical cords and placental tissues immediately after delivery. In contrast to serum NOx and cGMP levels, eNOS mRNA levels were significantly decreased in maternal arteries of the placental bed from the PE patients ( Figure 2F), whereas iNOS expression was significantly increased in their placental tissues ( Figure 2G). Furthermore, the elevated NOx levels in the patients were negatively correlated with eNOS mRNA levels and positively correlated with iNOS mRNA levels ( Figure 2H,I), suggesting that the increased NO/cGMP levels in the patients might be derived from increased iNOS expression. Since inflammatory cytokines, including TNF-α, stimulate the iNOS/NO/cGMP pathway [26] and inhibit the eNOS/NO/cGMP pathway [15][16][17][18][19], the relationship between serum levels of TNF-α and eNOS or iNOS expression was assessed. Our results showed that the serum TNF-α levels were significantly increased in PE patients ( Figure 2J) and positively correlated with the sFlt1/PlGF ratio ( Figure 2K). More notably, the serum TNF-α levels were inversely correlated with eNOS mRNA levels and positively correlated with iNOS mRNA levels ( Figure 2L,M), suggesting that TNF-α stimulated the iNOS/NO/cGMP pathway but inhibited the eNOS/NO/cGMP axis, probably by activating the transcription factor NF-κB [17,26]. Collectively, these data suggest that PE is likely to be associated with TNF-α-mediated downregulation of eNOS, although total serum NO/cGMP levels are elevated through TNF-α-induced induction of iNOS in preeclamptic patients.

Identification of miRNAs that are Positively and Negatively Regulated in PE Patients
Inflammation-dependent and NF-κB-responsive miR-31-5p and miR-155-5p have been shown to impair endothelial and vascular function via the downregulation of eNOS, sGC, and PKG in the vasculature [16][17][18][19]. Circulating levels of both miRNAs were significantly elevated in PE patients compared to healthy controls ( Figure 3A,B), consistent with previous reports [15,16]. Since miR-214 and miR-1290 have been shown to be positively and negatively regulated by hypoxia and estrogen, respectively, which are responsible for the pathogenesis of PE [4,27], we examined the alteration of their expression levels in the sera from patients with PE. The circulating miR-214-3p levels were also increased in the patients ( Figure 3C), whereas miR-1290-3p was downregulated ( Figure 3D). Notably, miR-31-5p, miR-155-5p, and miR-214-3p levels were highly correlated not only with TNF-α levels but also with the sFlt-1/PlGF ratio of these patients ( Figure 3E-G,I-K), whereas miR-1290-3p levels were negatively correlated with TNF-α levels and the sFlt-1/PlGF ratio in the patients ( Figure 3H,L). Collectively, these data suggest that increased serum levels of miR-31-5p, miR-155-5p, and miR-214-3p and decreased serum levels of miR-1290-3p are useful biomarkers for the pathogenesis and prognosis of PE.

Correlation between the miRNAs and the NOS/NO/cGMP Axis or PlGF Level in PE Patients
We next investigated which miRNA is possibly involved in the regulation of the eNOS-or iNOS-dependent NO/cGMP pathway in patients with PE. Circulating levels of miR-31-5p, miR-155-5p, and miR-214-3p were inversely correlated with eNOS mRNA levels ( Figure 4A-C) and positively associated with the levels of iNOS expression, NOx production, or cGMP synthesis ( Figure 4E-G, I-K, and M-O). By contrast, circulating miR-1290-3p levels were partially correlated with the eNOS mRNA level and negatively associated with the levels of iNOS mRNA, NOx, and cGMP ( Figure 4D, H, L, and P). Because miR-214-3p has been shown to target PlGF transcripts, we examined the correlation of miR-214-3p and PlGF levels in the sera of PE patients. Serum levels of miR-214-3p were inversely correlated with PlGF levels ( Figure 4Q). These data suggest that all four miRNA levels are positively or inversely correlated with the levels of eNOS, iNOS, NO/cGMP, and PlGF in PE patients.

Discussion
It is well-known that placental lesions and uteroplacental vascular insufficiency cause an imbalanced angiogenic state, such as high sFlt1 and sEng levels and low PlGF levels, which are associated with the pathogenesis of PE [6,8]. Accumulating evidence has demonstrated that the sFlt-1/PlGF ratio is significantly increased in maternal blood in the early stage of the clinical development of PE and is considered a diagnostic biomarker [9,10]. This suggests that circulating levels of sFlt-1 and PlGF can be used as pathogenic risk factors and diagnostic and prognostic biomarkers of PE. Unfortunately, pregnant women with a high sFlt-1/PlGF ratio do not always develop hypertension and proteinuria [11,12]. Therefore, new accurate biomarkers directly linked to the clinical symptoms of PE are needed to improve their predictive values.
PE is a pregnancy-specific vascular disease that is also considered an inflammatory disorder caused by the elevation of circulating proinflammatory cytokines in the patients [30][31][32]. We and others have previously demonstrated that serum levels of inflammatory cytokines, including TNF-α, which induces vasoconstriction and hypertension, are significantly elevated in the patients [15][16][17][18][19]33]. Infusion of TNF-α into pregnant rodents and nonhuman primates induces PE-like syndrome [24,34], implying the critical involvement of TNF-α in the pathogenesis of PE. In this study, we also found a potential role of TNF-α in the pathogenesis of PE by confirming a correlation between TNF-α levels and the sFlt-1/PlGF ratio or clinical symptoms. In addition, circulating miR-31-5p, miR-155-5p, and miR-214-3p levels were increased in PE patients and highly correlated with TNF-α levels, the sFlt-1/PlGF ratio, and clinical symptoms. In contrast, miR-1290-3p levels were decreased and inversely correlated with clinical symptoms, although miR-1290-3p expression was not altered by TNF-α and NF-κB activation. Collectively, our data suggest that these miRNAs are promising diagnostic or prognostic biomarkers for PE. However, to utilise the miRNAs as predictive factors of PE, further studies are needed for the prospective analysis of the predictive and correlative values between the circulating mRNA levels before the onset of symptoms and late-onset PE.
The eNOS/NO pathway plays a crucial role in vasorelaxation, angiogenesis, and vascular remodeling. Thus, inhibition or knockdown of eNOS elicits PE-like syndromes, such as hypertension and proteinuria, in rats and mice [25,35,36]. This suggests that impairment of the eNOS/NO/cGMP axis causes endothelial dysfunction, which is directly associated with the pathogenesis of PE. More recent studies showed that elevated TNF-α in PE is responsible for the NF-κB-mediated biogenesis of miR-31-5p and miR-155-5p and subsequent inhibition of the eNOS/NO axis in cultured human endothelial cells, resulting in the impairment of vascular relaxation and trophoblast invasion [15,16]. Similarly, our data demonstrated a strong correlation between TNF-α levels and miR-31-5p or miR-155-5p level and a concomitant decrease in eNOS expression in PE subjects. We also directly demonstrated that TNF-α inhibits eNOS expression in HUVECs by NF-κB-dependent biogenesis of miR-31-5p and miR-155-5p. These findings suggest that preeclamptic hypertension can be caused by the downregulation of the eNOS/NO axis via NF-κB-mediated biogenesis of miR-31-5p and miR-155-5p.
NF-κB-regulated miR-155-5p downregulates the NO-dependent sGC/cGMP/PKG axis in vascular smooth muscle cells (vSMCs) [18,19], leading to additional vasocontraction and hypertension. Unexpectedly, we found that serum NOx and cGMP levels were elevated in PE, although their levels have been shown to be equivocally altered, i.e., decreased, increased, or remained unchanged, in the blood of PE patients [37]. These phenomena might be associated with the differential expression of eNOS and iNOS via the activation of NF-κB in response to inflammatory cytokines. Indeed, NF-κB is significantly activated in the placenta in PE patients [38]. NF-κB has dual functions in pathophysiological NO production, through the transcriptional induction of iNOS in various types of cells, including vSMCs, hepatocytes, and macrophages [39], and the post-translational downregulation of eNOS in endothelial cells via the biogenesis of miR-31-5p and miR-155-5p [16,17]. In general, iNOS is able to synthesize high (micromolar) amounts of NO for many hours or even days, whereas eNOS produces lower (picomolar) and transient levels of NO. Therefore, increased NOx/cGMP levels in PE are thought to be due to the significant induction of iNOS, although eNOS was downregulated. Because NO synthesized by eNOS and iNOS has different biological functions, such as vascular functions and immune regulation, respectively [39], the decreased eNOS/NO pathway is responsible for hypertension and proteinuria in PE, while the elevated iNOS levels stimulate immune activation and tissue damage. Based on all these findings, increased NOx/cGMP levels are due to NF-κB-dependent iNOS induction and are unlikely to be linked to the regulation of vascular function. Thus, NOx and cGMP levels are not better diagnostic biomarkers for PE than miR-31-5p and miR-155-5p levels, which are associated with vascular dysfunction by downregulating eNOS, sGC, and PKG in the endothelial cells and vSMCs of PE patients.
Besides miR-31-5p and miR-155-5p, we also identified that miR-214-3p was significantly elevated in the sera of the patients and upregulated in trophoblasts stimulated with TNF-α. miR-214-3p is located on the opposite strand of an intron of dynamin 3 [40], and its expression is modulated by NF-κB ( Figure 5) and hypoxia-inducible factor (HIF)-1α [41]. Given the evidence that inflammation and hypoxia are crucial risk factors for PE [4,32], miR-214-3p is expected to be increased in women with PE. Although miR-214-3p has been shown to be equivocally linked to PE, with microarray data showing its downregulation and qRT-PCR data indicating its upregulation [41], we demonstrated that miR-214-3p was upregulated in an NF-κB-dependent manner in the human trophoblast cell line HTR-8/SVneo treated with TNF-α. Consistent with a previous study [29], our data showed that NF-κB-dependent miR-214-3p biogenesis is responsible for TNF-α-mediated suppression of PlGF production, suggesting that miR-214-3p expression precedes the downregulation of PlGF and is an earlier and better diagnostic biomarker of PE than PlGF. We also found that there is a strong inverse correlation between circulating levels of miR-214-3p and PlGF in PE patients. PlGF is highly expressed in the placenta in all gestational stages and promotes trophoblast invasion into the decidua, as well as spiral artery remodeling. Therefore, increased miR-214-3p levels can cause preeclamptic characteristics via PlGF downregulation and defects in spiral artery remodeling, thereby representing a reliable biomarker for the early diagnosis of PE. miR-1290-3p is upregulated in tumor stem cells [42] or downregulated during neuronal differentiation [43], suggesting that it somehow participates in cell division and cellular differentiation in the embryo. A recent study showed that miR-1290-3p was highly upregulated in the endometrium on Day 3 of the bovine estrous cycle, followed by rapid downregulation [44], suggesting that its level is regulated in an ovarian hormone-responsive manner and that it plays an important role in endometrial contribution to the implantation and survival of embryos. In fact, miR-1290-3p expression is downregulated in estrogen receptor α-positive breast cancer tumors [45], indicating that estrogens negatively regulate miR-1290-3p. Notably, estrogen concentrations were elevated in PE versus healthy controls [27,46,47]. This suggests that miR-1290-3p is downregulated in PE, consistent with our present findings. Although its expression regulation and target genes are largely unknown in PE, miR-1290-3p can be used as a downregulated biomarker of PE, similar to PlGF.
The pathogenesis of PE occurs as a consequence of interactions among multiple factors, including hypoxia, the immune system, the hormonal system, and oxidative stress [48]. Thus, a single factor is not sufficient to explain the pathogenic mechanism or determine the diagnosis of PE. The strengths of this study include the identification of multiple mechanism-based miRNAs, including inflammation-responsive miR-31-5p and miR-155-5p, and NF-κB/HIF-α-dependent miR-214-3p and miR-1290-3p ( Figure 9). Their serum levels were highly correlated with the sFlt-1/PlGF ratio and clinical symptoms and had high diagnostic accuracy as demonstrated by ROC value analysis. However, miR-155-5p has been shown to be elevated in several pathogenic conditions, such as cancer, inflammatory disorders, obesity, and diabetes [49,50], suggesting that miR-155-5p is useful, but not ideal, for the diagnosis for PE. The other miRNAs are considered more specific to PE than miR-155-5p because they are expressed to a limited extent in few other pathological conditions. More notably, all the miRNAs, except miR-1290-3p, the function of which is unknown, are considered risk factors for the development of PE and are thus useful as mechanism-based biomarkers for diagnosing PE. Because miR-1290-3p, considered an inflammation/NF-κB-insensitive and estrogen-responsive biomarker, was downregulated in PE patients, the ratios of miR-31-5p, miR-155-5p, or miR-214-3p to miR-1290-3p may be clinically useful as effective diagnostic determinants for PE, as shown by the diagnostic value of the sFlt1/PlGF ratio. Using the current analytical techniques with qPCR and silver nanocluster DNA probes [51], these miRNAs can be analyzed quickly in a small volume of blood, representing a test that could aid the risk stratification of women with suspected PE.
In conclusion, we identified that the serum levels of several biomarkers were significantly elevated in patients with PE. Among them, NF-κB-responsive miR-31-5p, miR-155-5p, and miR-214-3p could be used as novel mechanism-based diagnostic biomarkers and therapeutic targets for PE ( Figure 9). Furthermore, the ratios of those miRNAs to miR-1290-3p, which was reduced in PE, showed greater diagnostic power than that of their individual miRNA, providing evidence that these ratios might be of potential value in the diagnosis and prognosis of women with suspected PE, although the case numbers are still too low to make sufficient differentiation. Figure 9. Possible roles of the identified miRNAs in the pathogenesis of PE. Hypoxia, inflammation, and decreased estrogen levels caused by inappropriate implantation. Inflammation and hypoxia increase the biogenesis of miR-31-5p, miR-155-5p, and miR-214-3p in endothelial cells (ECs), vSMCs, and the trophoblast (TB) via activation of NF-κB and HIF-1α. They inhibit the expression of eNOS in ECs [16][17][18], sGC and PKG [19,20] in vSMCs, and PlGF in the TB ( Figure 5), resulting in the impairment of vasorelaxation and TB invasion, which cause hypertension, proteinuria, and spiral artery remodeling. Meanwhile, reduced estrogen levels downregulate miR-1290-3p in placental tissue (PT); the function of miR-1290-3p, however, has not been identified in the development of PE. Nevertheless, decreased miR-1290-3p level is considered a "rule-out" test factor, like the reduced PlGF level that is used as a critical factor to discriminate against women with suspected PE from healthy pregnant women.

Conflicts of Interest:
We declare no conflicting interests.