Maternal plasma concentrations of sST2 and angiogenic/anti-angiogenic factors in preeclampsia

Abstract

Objectives: Angiogenic/anti-angiogenic factors have emerged as one of the promising biomarkers for the prediction of preeclampsia. Since not all patients with preeclampsia can be identified by these analytes, the search for additional biomarkers continues. The soluble form of ST2 (sST2), a protein capable of binding to interleukin (IL)-33 and thus contributing to a Th1-biased immune response, has been reported to be elevated in maternal plasma of women with preeclampsia. The aims of this study were to examine: (1) differences in maternal plasma concentrations of sST2 and IL-33 between women diagnosed with preeclampsia and those having uncomplicated pregnancies; (2) the relationship between sST2, umbilical and uterine artery Doppler velocimetry, and the severity of preeclampsia; and (3) the performance of sST2 and angiogenic/anti-angiogenic factors in identifying patients with preeclampsia at the time of diagnosis.

Methods: This cross-sectional study included women with preeclampsia (n = 106) and women with an uncomplicated pregnancy (n = 131). Plasma concentrations of sST2, IL-33, soluble vascular endothelial growth factor receptor (sVEGFR)-1, soluble endoglin (sEng) and placental growth factor (PlGF) were determined by enzyme linked immune sorbent assay. Area under the receiver operating characteristic curve (AUC) for the identification of preeclampsia was examined for each analyte.

Results: (1) Patients with preeclampsia had a higher mean plasma concentrations of sST2 than those with an uncomplicated pregnancy (p < 0.0001), while no significant difference in the mean plasma concentration of IL-33 between the two groups was observed; (2) the magnitude of this difference was greater in early-onset, compared to late-onset disease, and in severe compared to mild preeclampsia; (3) sST2 plasma concentrations did not correlate with the results of uterine or umbilical artery Doppler velocimetry (p = 0.7 and p = 1, respectively) among women with preeclampsia; (4) sST2 correlated positively with plasma concentrations of sVEGFR1-1 and sEng (Spearman’s Rho = 0.72 and 0.63; each p < 0.0001), and negatively with PlGF (Spearman’s Rho = −0.56, p < 0.0001); and (5) while the AUC achieved by sST2 and angiogenic/anti-angiogenic factors in identifying women with preeclampsia at the time of diagnosis were non-significantly different prior to term (<37 weeks of gestation), thereafter the AUC achieved by sST2 was significantly less than that achieved by angiogenic/anti-angiogenic factors.

Conclusions: Preeclampsia is associated with increased maternal plasma concentrations of sST2. The findings that sST2 concentrations do not correlate with uterine or umbilical artery Doppler velocimetry in women with preeclampsia suggest that elevated maternal plasma sST2 concentrations in preeclampsia are not related to the increased impedance to flow in the utero-placental circulation. The performance of sST2 in identifying preeclampsia at the time of diagnosis prior to 37 weeks of gestation was comparable to that of angiogenic/anti-angiogenic factors. It remains to be elucidated if an elevation of maternal plasma sST2 concentrations in pregnancy is specific to preeclampsia.

• Biomarker
• interleukin-33
• pregnancy
• sflt-1
• Th2 immune response
• VEGFR-1

Introduction

Preeclampsia, one of the “great obstetrical syndromes” [1,2], remains a leading cause of maternal and neonatal morbidity [3–8]. A defect of deep placentation [9–13] is proposed to generate utero-placental ischemia [14–16], placental endoplasmic reticulum and oxidative stress [17–22], and subsequently systemic intravascular inflammation [23–25] and endothelial dysfunction [17,26–34]. The circulating concentrations of pro-inflammatory cytokines, such as interleukin (IL)-1-β and tumor necrosis factor (TNF)-α [35–39], as well as other inflammatory mediators (such as IL-6) [36,40–42] are typically elevated in patients with preeclampsia, although this is not a universal feature [43,44]. The placenta is essential for the development of preeclampsia, while the fetus is not [45–47]. Preeclampsia has been referred to as “toxemia of pregnancy” because it is believed that “toxins” released from a damaged placenta are responsible for the clinical manifestations of the disease [3,48–50].

It has become increasingly clear that the magnitude of derangements underlying preeclampsia is greater when it is diagnosed prior to 34 weeks of gestation [51]. The frequency of placental lesions [15,52,53], as is the frequency of multisystemic involvement [54–57], is higher in these patients than those who are diagnosed later. For example, hemolysis, elevated liver enzymes, low platelet (HELLP) syndrome is more common in patients with early-onset disease than in those diagnosed at term [58–60]. The increased perinatal morbidity associated with preeclampsia is also closely linked to early-onset disease because indicated preterm delivery is associated with an increased risk of neonatal complications [61]. Late-onset disease is, however, much more common than early-onset preeclampsia [56,57,62].

Considerable effort has been made to identify biomarkers that could predict women who will develop preeclampsia [63–78]. Although markers of systemic inflammation [79,80] and endothelial dysfunction [81–88] in maternal blood have been associated with preeclampsia, their prognostic performance for the identification of patients with preeclampsia prior to diagnosis has been disappointing [79–81,83,89–91]. This indicates that either inflammation occurs late (even during the subclinical phase of the disease), or that the current methods available to detect intravascular inflammation are not sensitive enough.

Preeclampsia [63–67,69,70,72,92–103], as well as other obstetrical syndromes [46,104–118], is characterized by an anti-angiogenic state. Plasma concentrations of the angiogenic factor, placental growth factor (PlGF), are decreased [69,70,92,93,96,100–102], and the anti-angiogenic factors (soluble vascular endothelial growth factor receptor-1 [sVEGFR-1] and soluble endoglin [sEng]) are elevated both prior to [67,72,119–125] and at the time of diagnosis [94,95,97–99]. Moreover, the concentrations of these factors can be used to predict the outcome of the disease in patients presented to the obstetrical triage area [73,75,126,127]. However, the prognostic performance of these biomarkers evaluated in the mid-trimester for late-onset preeclampsia has been suboptimal [105,128–131]. Recently, an elevated plasma concentration of sST2 was observed prior to and at the time of the diagnosis of preeclampsia in a case–control study [132]. Therefore, sST2 was proposed as a novel candidate biomarker for the prediction of preeclampsia [132].

ST2 is a member of the IL-1 receptor family gene, and the ST2 protein has four isoforms, of which the best characterized are a membrane-bound receptor form (ST2 receptor or ST2L) and a soluble form (sST2). ST2L expresses on mast cells, macrophages and exclusively on Type 2 T-helper cells [133–140]. Upon binding to ST2L, IL-33 is capable of stimulating the Th2 immune response and cytokine production [141–144]. In contrast, sST2 acts as a decoy receptor for IL-33 and is thought to inhibit IL-33 function, thus favoring a shift toward the Th1 immune response [145]. Moreover, sST2 can be secreted in response to inflammatory signals by endothelial cells [146,147]. The involvement of sST2 has been observed in a variety of pathological conditions characterized by: (1) intravascular inflammation [145,148–151] (i.e. myocardial infarction [152–154], atherosclerotic vascular disease [155], sepsis and trauma [149,150]); and (2) an imbalance of Th1/Th2 immune response [141,156,157] (i.e. asthma and other allergic conditions [145,148,158,159]), both of which are also observed in patients with preeclampsia [23,25,132,160,161].

The objectives of this study were to: (1) determine whether plasma concentrations of sST2 and IL-33 in patients with preeclampsia at the time of diagnosis differ from those in women with uncomplicated pregnancies; (2) examine the relationships between plasma concentrations of sST2 and Doppler velocimetry in the uterine and umbilical arteries, gestational age at diagnosis and severity of preeclampsia; and (3) compare the performance of sST2 and angiogenic/anti-angiogenic factors at different gestational ages in identifying preeclampsia at the time of diagnosis.

Material and methods

Study design and population

A retrospective cross-sectional study was conducted by searching the clinical database and bank of biologic samples of Wayne State University/Detroit Medical Center/Perinatology Research Branch. All patients were enrolled at Hutzel Women’s Hospital in Detroit, MI, USA. The following groups were included: (1) women with an uncomplicated pregnancy; and (2) patients with preeclampsia. Chronic hypertension, multiple pregnancies, or fetuses with chromosomal or structural anomalies were excluded. Women with an uncomplicated pregnancy between 20 and 42 weeks of gestation were enrolled from either the labor and delivery unit (in case of scheduled cesarean section) or from the antenatal clinic, and followed until delivery. Patients with preeclampsia had a venipuncture upon diagnosis. Doppler velocimetry examination of the uterine and umbilical arteries was performed in women with preeclampsia within 48 hours of venipuncture.

All women provided written informed consent prior to the collection of plasma samples. The collection and utilization of the samples for research purposes was approved by the Institutional Review Board of Wayne State University and the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD/NIH/DHHS). Many of these samples were used previously in studies of intravascular inflammation, angiogenesis and cytokine biology in normal and complicated pregnancies.

Clinical definition

Patients were considered to have an uncomplicated pregnancy outcome if they did not experience any obstetrical, medical, or surgical complications of the pregnancy, and delivered a normal term (≥37 weeks) neonate whose birth weight was between the 10th and 90th percentile for gestational age without complications [162]. Preeclampsia was defined as hypertension (systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg on at least two occasions, 4 hours to 1 week apart) and proteinuria (≥300 mg in a 24-hours urine collection or one dipstick measurement ≥2+) [163]. Early-onset and late-onset preeclampsia were defined as cases diagnosed before and after 34 weeks of gestation, respectively [51]. Severe preeclampsia was defined as previously described [164].

Doppler velocimetry

Pulse-wave and color Doppler ultrasound examinations of the uterine and umbilical arteries were performed in a subset of patients with preeclampsia as previously described (Acuson, Sequoia, Mountain View, CA) [104,164]. Uterine artery Doppler velocimetry was defined as abnormal if the mean (average of right and left) resistance index (RI) was above the 95th percentile for gestational age (using the reference range proposed by Kurmanavicius et al.) [165]. Umbilical artery Doppler velocimetry was defined as abnormal if either the pulsatility index (PI) was above the 95th percentile for gestational age (using the reference range proposed by Arduini and Rizzo) [166] or in the presence of abnormal waveforms (absent or reversed end-diastolic velocities) [167]. The patients were classified into the following groups: (1) normal Doppler velocimetry in the uterine and umbilical arteries; (2) Doppler abnormalities in the uterine arteries alone; (3) Doppler abnormalities in umbilical arteries alone; or (4) Doppler abnormalities in both vessels.

Sample collection and human sST2 and IL-33 immunoassay

The blood collected by venipuncture was stored in tubes containing EDTA. Following centrifugation, the samples were stored at −70 °C. Specific enzyme-linked immunoassays were used for the determination of maternal plasma concentrations of sST2 (R&D Systems, Minneapolis, MN) and IL-33 (BioLegend, San Diego, CA). The quantitative sandwich enzyme immunoassay technique was used. The inter- and intra-assay coefficients of variation (CVs) were 4.6% and 3.9% for sST2 and 14.9% and 12.2% for IL-33, respectively. The sensitivities of the assays for sST2 and IL-33 were 17.5 and 12 pg/mL, respectively.

Angiogenic and anti-angiogenic factor immunoassay

Maternal plasma concentrations of PlGF, sVEGFR-1, and sEng were determined by sensitive and specific immunoassays (R&D Systems, Minneapolis, MN). All three immunoassays utilized the quantitative sandwich enzyme immunoassay technique. The inter- and intra-assay CVs obtained were 6.02% and 4.8% for PlGF; 1.4% and 3.9% for sVEGFR-1, and 2% and 4% for sEng, respectively. The sensitivity of the assays were: PlGF: 9.52 pg/mL; sVEGFR-1: 16.97 pg/mL and sEng: 0.08 ng/mL, respectively. The results of sVEGFR-1 and sEng have been reported in previous publications to examine the relationships between these proteins and Doppler velocimetry in uterine and umbilical arteries [104,164].

Statistical analysis

A Kolmogorov–Smirnov or Shapiro–Wilk test and visual plot inspection were used to assess normality of arithmetic data distributions. For analyte concentrations below the limit of the detection of the assays, the concentration was replaced by 99% of the lowest detectable concentration. A Kruskal–Wallis with post hoc Mann–Whitney U test was used to compare variables that were not normally distributed among and between groups. Comparison of proportions was performed using Chi-square or Fisher’s exact test, as appropriate. Correlation between two continuous variables was determined using Spearman’s rank correlation.

Logistic regression models were used to determine the magnitude of association between biochemical markers and preeclampsia. They were also used to construct receiver-operating characteristic (ROC) curves to assess the performance of plasma concentrations of sST2, PlGF, sVEGFR-1 and sEng in identifying preeclampsia at the time of diagnosis. Covariables included in adjusted models were selected based on clinical knowledge. Model reduction was performed additionally based on the plausibility of regression coefficients, association with independent/dependent variables and the magnitude of change in the main effect parameter estimates. Models constructed to represent clinical factors, selected biomarkers, Doppler velocimetry in the uterine and umbilical arteries, both independently and in combination, were evaluated overall and additionally stratified by gestational age at the diagnosis of preeclampsia (≤34 weeks; >34 to <37 weeks; ≥37 weeks). Paired sample non-parametric statistical techniques were used to compare area under the ROC curves (AUC) of models constructed using logistic regression for the identification of preeclampsia at the time of diagnosis.

A general linear model was fit to determine whether the predicted geometric mean analyte concentration varied by more than 5% when adjusted for one or more of the following: gestational age at venipuncture, maternal age, race, nulliparity and smoking. A p value <0.05 was considered statistically significant. Analyses were performed with SPSS, version 19 (IBM Corp, Armonk, NY) and SAS version 9.3 (Cary, NC).

Results

Demographic and clinical characteristics of the study population

Demographic and obstetric characteristics of women with uncomplicated pregnancies (n = 131) and patients with preeclampsia (n = 106) are summarized in Table 1. The proportion of nulliparous women was greater among patients diagnosed with preeclampsia than in those with uncomplicated pregnancies [61% (65/106) versus 28% (36/131); p < 0.0001]. As expected, the median gestational age at delivery and the median birth weight were significantly lower in women with preeclampsia than in those without preeclampsia. No differences were observed by age, race, pre-pregnancy weight, smoking habit or gestational age at venipuncture. Among patients with preeclampsia, 39% (41/106) had early-onset disease and 80% (85/106) were diagnosed with severe preeclampsia (Table 2).

Table 1. Clinical characteristics of the study population.

	Uncomplicated pregnancy (n = 131)	Preeclampsia (n = 106)	p
Age (years)	25 (21–27.7)	21 (18–28.7)	0.12
Race			0.68
African American	103 (78.6%)	88 (83%)
Caucasian	15 (11.5%)	11 (10.4%)
Hispanic	7 (5.3%)	5 (4.7%)
Other	6 (4.6%)	2 (1.9%)
Nulliparous	36 (27.5%)	65 (61.3%)	<0.0001*
Pre-pregnancy  weight (pounds)	145.0 (125–184)	154.5 (130–180)	0.54
Smoker	23 (17.7%)	14 (13.2%)	0.35
GA at venipuncture  (weeks)	38.0 (31.5–39.3)	36.0 (30.6–38.7)	0.18
GA at delivery  (weeks)	39.5 (38.6–40.1)	36.0 (31.9–38.7)	<0.0001*
Birth weight  (grams)	3250 (3065–3610)	2280 (1480–2812)	<0.0001*

Values are expressed as median (interquartile range) or number (percent). GA: gestational age.

*p < 0.05.

Table 2. Clinical characteristics of patients with preeclampsia (n = 106).

Parameter
Blood pressure (mmHg)
Systolic	171 (160–180)
Diastolic	104 (98–110)
Mean arterial pressure	127 (119–133)
Aspartate aminotransferase (SGOT; U/L)	28 (21–43)
Platelet counts (x10^3; cells/µL)	175 (133–223)
Birth weight <10th percentile	48 (45.3%)
Severe preeclampsia	85 (80%)
Early-onset preeclampsia	41 (38.7%)

Values are expressed as median (interquartile range) or number (percent).

Plasma concentrations of sST2 in preeclampsia, early- and late-form of preeclampsia and clinical severity

Patients with preeclampsia had a higher median plasma concentration of sST2 than women with uncomplicated pregnancies (median 76.1 ng/mL, interquartile range (IQR) 48.4–130.3 versus median 30.8 ng/mL, IQR 14–52.3; p < 0.0001). The magnitude of the difference in sST2 plasma concentrations between women with preeclampsia and women with uncomplicated pregnancies was greater among patients diagnosed at ≤34 weeks, >34 but <37 weeks compared to those diagnosed at term (≥37 weeks; Figure 1). There was no significant difference in the median plasma sST2 concentrations among patients with preeclampsia diagnosed at different gestational age categories (p = 0.87; Figure 1). While plasma sST2 concentrations in uncomplicated pregnancies increased as a function of gestational age (Spearman’s Rho = 0.77, p < 0.0001), there was no significant relationship between plasma sST2 concentration and gestational age at the diagnosis of preeclampsia (Spearman’s Rho = −0.09; p = 0.38; Figure 2).

Figure 1. Plasma concentrations of sST2 sub-classified according to gestational age at the diagnosis of preeclampsia. The median plasma concentration of sST2 in women with preeclampsia was significantly higher than in those with uncomplicated pregnancies (≤34 weeks, median 93 ng/mL, interquartile range (IQR) 48.4–159.3 ng/mL versus median 12.7 ng/mL, IQR10.6–20; p < 0.0001; between >34 and <37 weeks, median 62.4 ng/mL, IQR 39.8–101.3 ng/mL versus median 22.5 ng/mL, IQR 12.2–32.5 ng/mL; p < 0.0001; and ≥37 weeks, median 73.1 ng/mL, IQR 51–101.7 ng/mL versus median 49 ng/mL, IQR 35.1–73.4 ng/mL; p ≤ 0.027). The magnitude of the difference was more pronounced in earlier gestation than at term. However, no differences were observed in plasma concentrations of sST2 in women with preeclampsia stratified by gestational age at diagnosis (p = 0.87). The y axis is presented in log 2 scale.

Figure 2. Relationship between gestational age at venipuncture (weeks) and plasma concentrations of sST2 (ng/mL) in women with uncomplicated pregnancies (ˆ; Spearman’s Rho = 0.77, p < 0.0001) and those with preeclampsia (•; Spearman’s Rho = −0.09, p = 0.38). The y axis is presented in log 2 scale.

Patients with severe preeclampsia had a significantly higher median plasma concentration of sST2 than those with mild preeclampsia (median 88.9 ng/mL, IQR 52–145 versus median 52.4 ng/mL, IQR 42–85, p = 0.01). Among women diagnosed with preeclampsia, plasma concentrations of sST2 were significantly correlated with the amount of urine protein in 24 h (Spearman’s Rho = 0.4, p = 0.02), gestational age at delivery (Spearman’s Rho = 0.17, p < 0.0001) and neonatal birth weight (Spearman’s Rho = 0.13, p = 0.001). In contrast, sST2 was not associated with mean arterial pressure (Spearman’s Rho = −0.2, p = 0.4) or platelet count (Spearman’s Rho = −0.2, p = 0.4).

Use of general linear models to evaluate the need for multivariable adjustment (gestational age at venipuncture, maternal race, maternal age, nulliparity, and smoking) revealed little change (<5%) in the predicted geometric mean of sST2 concentrations compared to unadjusted estimates, meaning the results shown above are robust and reliable in that they persist after multivariable adjustment.

Plasma concentrations of IL-33 in women with uncomplicated pregnancies and those with preeclampsia

Plasma concentration of IL-33 was below the detection limit in 8% (10/131) of women with uncomplicated pregnancies and in 5% (5/106) of women with preeclampsia. There was no significant difference in the median plasma concentration of IL-33 between patients with preeclampsia (92.5 pg/mL, IQR 34.9–258.2) and women with uncomplicated pregnancies (90.9 pg/mL, IQR 42.6–179.4) overall, or in any of the gestational age intervals (p = 1.0 for all comparisons).

Plasma concentrations of sST2 and IL-33 do not correlate with the result of Doppler velocimetry of the uterine and umbilical arteries among women with preeclampsia

Information about Doppler velocimetry was available in 61% (65/106) of patients with preeclampsia. The median uterine artery RI was 0.68 (IQR 0.58–0.75) and the median umbilical artery PI was 1.2 (IQR 1.0–1.4). Abnormal uterine artery Doppler velocimetry was observed in 82% (53/65) of patients, while abnormal umbilical artery Doppler velocimetry was present in 18% (12/65) of patients diagnosed with preeclampsia.

Plasma concentrations of sST2 were not correlated with uterine artery RI (Spearman’s Rho = 0.05, p = 0.7), or umbilical artery PI (Spearman’s Rho = −0.001, p = 1). Similarly, plasma concentrations of IL-33 were not correlated with uterine artery RI (Spearman’s Rho = 0.21, p = 0.12), or umbilical artery PI (Spearman’s Rho = −0.03, p = 0.85). There were no significant differences in the median plasma concentration of sST2 among patients with preeclampsia with and without abnormalities in the uterine and/or umbilical artery Doppler velocimetry (Figure 3). Similar results were observed after taking into account the effect modification term including gestational age at venipuncture and abnormalities in Doppler velocimetry.

Figure 3. Plasma concentrations of sST2 in uncomplicated pregnancies and patients with preeclampsia sub-classified according to the results of uterine (Ut) and umbilical (Umb) artery Doppler velocimetry. When compared to women with uncomplicated pregnancies, each sub-group of preeclampsia had a significantly higher median plasma concentration of sST2 (Kruskal–Wallis test p < 0.0001). There was no significant difference in the median plasma concentration of sST2 among each subgroup of preeclampsia (Kruskal–Wallis test p = 0.93). The y axis is presented in log 2 scale.

The performance of sST2 and angiogenic/anti-angiogenic factors in identifying patients with preeclampsia at the time of diagnosis

Nineteen percent (20/106) of patients with preeclampsia had plasma concentrations of PlGF below the detection limit. The median plasma concentrations of PlGF, sEng, and sVEGFR1 and the ratio between PlGF/sVEGFR1 and PlGF/sEng in women with uncomplicated pregnancies and those with preeclampsia are shown in Table 3. There was a positive correlation between plasma concentrations of sST2 and anti-angiogenic factors (for sEng: Spearman’s Rho = 0.63; for sVEGFR-1: Spearman’s Rho = 0.72; each p < 0.0001) and a negative correlation between sST2 and PlGF (Spearman’s Rho = −0.56), PlGF/sVEGFR1 ratio (Spearman’s Rho = −0.69), and PlGF/sEng ratio (Spearman’s Rho = −0.63; each p < 0.0001; Table 4). No correlation was observed either between IL-33 and sST2, or between IL-33 and angiogenic/anti-angiogenic factors (Table 4).

Table 3. Plasma concentrations of angiogenic and anti-angiogenic factors in women with uncomplicated pregnancies and patients with preeclampsia.

	Women with uncomplicated pregnancies n = 131	Patients with preeclampsia n = 106	p
PlGF (pg/mL)	502.4 (250–896)	86.7 (55.6–138)	<0.0001*
sEng (ng/mL)	7.2 (5.1–10.5)	41.5 (18.8–73.5)	<0.0001*
sVEGFR1 (ng/mL)	1.3 (0.74–2.0)	5.4 (3.6–9.6)	<0.0001*
PlGF/sVEGFR1	0.41 (0.18–0.89)	0.02 (0.005–0.03)	<0.0001*
PlGF/sEng	73 (28–146)	2 (0.5-6.6)	<0.0001*

Values are expressed as median (interquartile range).

*p < 0.05.

Table 4. Correlation between sST2, IL-33 and angiogenic/anti-angiogenic factors.

Correlation coefficient	sST2	p	IL-33	p
PlGF	−0.56	<0.0001*	−0.10	0.1
sEng	0.63	<0.0001*	0.03	0.7
sVEGFR1	0.72	<0.0001*	0.06	0.3
PlGF/sVEGFR1	−0.69	<0.0001*	−0.09	0.2
PlGF/sEng	−0.63	<0.0001*	−0.07	0.3
IL-33	0.07	0.3	–

Spearman’s correlation coefficient was applied for the evaluation of correlation.

*p < 0.05.

Table 5. Area under the ROC curves (AUC) achieved by sST2 and angiogenic/anti-angiogenic factors (adjusted for clinical factors) in identifying women with preeclampsia at the time of diagnosis.

	AUC (95% CI)	p
sST2	0.86 (0.81–0.91)	Reference
PlGF	0.96 (0.93–0.98)	0.0001*
sEng	0.94 (0.91–0.97)	0.002*
sVEGFR1	0.97 (0.95–0.99)	<0.0001*
PlGF/sVEGFR1	0.98 (0.96–0.997)	<0.0001*
PlGF/sEng	0.97 (0.95–0.99)	<0.0001*

p Value refers to the comparison of AUC between each angiogenic/anti-angiogenic factor and sST2.

CI, confidence interval.

*p < 0.05.

The AUC for models constructed to identify women with preeclampsia at the time of diagnosis using sST2 alone significantly exceeded that achieved by clinical factors (age, smoking, nulliparity and race) [0.85 (95% CI 0.80–0.90) versus 0.70 (95% CI 0.63–0.78), respectively (p = 0.0007)]. An addition of clinical factors to sST2 did not significantly improve the performance of sST2 in identifying women with preeclampsia at the time of diagnosis as reflected by change in AUC (0.86 vs. 0.85; p = 0.5). The AUCs for PlGF, sEng, sVEGFR1, PlGF/sVEGFR1 and PlGF/sEng ratio (adjusted for clinical factors) are shown in Figure 4; each performed better than sST2 in identifying women with preeclampsia at the time of diagnosis (Table 5). The combination of sST2 with the PlGF/sVEGFR1 ratio or the PlGF/sEng ratio did not result in significant improvement (AUC for PlGF/sVEGFR1 alone: 0.977 versus AUC for PlGF/sVEGFR1 plus sST2: 0.978; p = 0.7; and AUC for PlGF/sEng alone: 0.971 versus for sST2 plus PlGF/sEng 0.974; p = 0.6).

Figure 4. Receiver operating characteristic (ROC) curves of plasma sST2 concentrations, each angiogenic/anti-angiogenic factor and their ratios (adjusted for clinical factors) for the identification of women with preeclampsia at the time of diagnosis. Area under the ROC curve for each analyte was also presented (sST2 = 0.86, PlGF = 0.96, sEng = 0.94, sVEGFR1 = 0.97, PlGF/sVEGFR1 = 0.98, and PlGF/sEng = 0.97).

Results stratified according to gestational age at the diagnosis of preeclampsia showed that in women diagnosed before 37 gestational weeks, sST2 performed similarly to angiogenic/anti-angiogenic factors in the identification of patients with preeclampsia at the time of diagnosis (Table 6). In contrast, in women evaluated at or beyond 37 gestational weeks, the AUC achieved by sST2 in identifying women with preeclampsia at the time of diagnosis (AUC 0.73) was significantly lower than that achieved by angiogenic/anti-angiogenic factors (AUCs ranging from 0.85 for sEng to 0.97 for PlGF/sVEGFR-1 ratio).

Table 6. Area under the ROC curves (AUC) for sST2 and angiogenic/anti-angiogenic factors (without adjustment for clinical factors) sub-classified according to gestational age at the diagnosis of preeclampsia.

	AUC (95% CI) ≤34 weeks	p	AUC (95% CI) >34 – <37 weeks	p	AUC (95% CI) ≥37 weeks	p
sST2	0.989 (0.974–1)	Ref	0.945 (0.872–1)	Ref	0.73 (0.63–0.83)	Ref
PlGF	0.989 (0.971–1)	0.96	0.959 (0.892–1)	0.7	0.88 (0.82–0.95)	0.02*
sEng	0.981 (0.948–1)	0.6	0.941 (0.841–1)	0.9	0.85 (0.77–0.93)	0.055
sVEGFR1	0.993 (0.982–1)	0.5	0.950 (0.867–1)	0.9	0.95 (0.90–0.99)	0.0002*
PlGF/sVEGFR1	0.999 (0.998–1)	0.1	0.964 (0.898–1)	0.5	0.97 (0.94–0.998)	<0.0001*
PlGF/sEng	0.995 (0.985–1)	0.3	0.977 (0.93–1)	0.2	0.93 (0.88–0.98)	0.0005*

CI: confidence interval; ref: reference.

p value refers to the comparison of AUC between each angiogenic/anti-angiogenic factor and sST2.

*p < 0.05.

Discussion

Principal findings of this study

(1) Preeclampsia is associated with a higher plasma sST2 concentration than in uncomplicated pregnancies; the magnitude of this difference is greater in early- compared to late-onset disease, and in severe than in mild preeclampsia; (2) among patients with preeclampsia, there was no association between plasma sST2 concentrations and abnormalities of Doppler velocimetry of the uterine and umbilical arteries; (3) the performance of sST2 in identifying patients with preeclampsia diagnosed at <37 weeks of gestation was similar to that of angiogenic (PlGF), anti-angiogenic (sVEGFR-1, sEng) factors and their ratios (PlGF/sVEGFR-1, PlGF/sEng). However, the performance of angiogenic/anti-angiogenic factors and their ratios was better than that of sST2 plasma concentrations in patients who were diagnosed at term; and (4) no significant differences in plasma concentrations of IL-33 were observed between women with and without preeclampsia.

Plasma concentrations of soluble ST2 is elevated in preeclampsia

The finding that the median plasma concentration of sST2 in preeclampsia was higher than in uncomplicated pregnancies are consistent with a previous study by Granne et al. [132]. Moreover, the current study extends these observations by demonstrating that the magnitude of the difference between preeclampsia and uncomplicated pregnancies is markedly higher in early- than in late-onset disease, (Figure 1), and in severe than in mild preeclampsia.

Two hypotheses can be invoked to explain these findings. First, a higher plasma sST2 concentration may be a consequence of an intravascular inflammatory response which is exacerbated in preeclampsia when compared to an uncomplicated pregnancy [23,25,168–170]. Indeed, it has been demonstrated that pro-inflammatory cytokines are increased in the plasma or serum of women with early-onset preeclampsia [35,41,42] and severe preeclampsia [36,40,41] when compared to women with uncomplicated pregnancies. Moreover, IL-1β or TNF-α (pro-inflammatory cytokines which are increased in preeclampsia) are able to stimulate vascular endothelial cells to secrete sST2 into supernatant [146,147]. This hypothesis is also consistent with the view that preeclampsia is characterized by a Th1/Th2 imbalance towards a Th1 type immune response [160,161,171], since sST2 can inhibit IL-33 function, thus favoring a shift toward a Th1 response.

An alternative view is that an elevation of sST2 in preeclampsia could be a response of the myocardium and vascular endothelium to pressure and/or volume (hemodynamic) overload. Indeed, patients with preeclampsia have increased plasma concentrations of natriuretic peptides that reflect changes in ventricular size and volume load [172–175]. In chronic heart disease, serum sST2 concentration increases as natriuretic peptides and indexes of hemodynamic load increase [146,153]. In this case, the source of sST2 could be cardiac myocytes, as shown by the up-regulated ST2 gene expression in neonatal rat myocytes exposed to mechanical strain [176]. However, Bartunek et al. provided evidence that adult human myocardium is not a source of an increased sST2 concentration in patients with cardiac hypertrophy induced by pressure overload or those with congestive cardiomyopathy [146]. These investigators have postulated that elevated plasma sST2 concentrations in patients with diastolic overload were more likely to originate from endothelial cells rather than cardiac myocytes [146]. Collectively, these observations support the hypothesis that the increased plasma sST2 concentration in preeclampsia may derive from myocardial and endothelial cells stimulated by hemodynamic changes in pregnancy, which becomes exacerbated in women with preeclampsia.

Relationship between plasma sST2 concentrations and Doppler velocimetry, gestational age at diagnosis, and severity of preeclampsia

The involvement of the placenta in early-onset preeclampsia is more extensive than in the late-onset disease [15,52,177]. The finding that the magnitude of the difference in plasma sST2 concentrations between preeclampsia and uncomplicated pregnancies was greater in early- than in late-onset disease may indicate a greater contribution of sST2 from an injured placenta in preeclampsia. Indeed, the ST2 gene [178] and protein [132] expressions were observed in the placenta, and sST2 can be released into the maternal circulation in a placental perfusion model [132]. Although sST2 secretion from the placenta is increased by hypoxia/reperfusion or treatment with pro-inflammatory cytokines (TNF-α or IL-1β), there was no significant difference in ST2 and IL-33 protein expression in placentas obtained from patients with preeclampsia when compared to placentas from uncomplicated pregnancies [132]. Moreover, the findings that there was no significant relationship between plasma concentrations of sST2 and abnormalities in Doppler velocimetry in the uterine and umbilical arteries do not support the view that the increased impedance to blood flow in the placenta is responsible for the increased plasma sST2 concentration in preeclampsia. Consistent with this hypothesis, plasma concentrations of sST2 in preeclampsia is only mildly correlated with neonatal birth weight (Spearman’s Rho = 0.13, p = 0.001). Therefore, the finding that the plasma concentrations of sST2 were higher in severe than in mild preeclampsia may reflect a greater degree of endothelial dysfunction and intravascular inflammation, which characterizes severe preeclampsia [168].

Relationship between sST2 and angiogenic/anti-angiogenic factors

A novel finding of the current study is that plasma concentrations of sST2 in preeclampsia had a strong correlation with plasma concentrations of anti-angiogenic factors (sVEGFR-1 and sEng), and a moderate inverse relationship with the angiogenic factor (PlGF). Although individual angiogenic/anti-angiogenic factors reported herein can identify patients with preeclampsia at the time of diagnosis, the ratio between angiogenic and anti-angiogenic factor concentrations seems to provide a slightly larger AUC than the individual marker alone. The performance of sST2 is better in preterm than in term preeclampsia, with the AUC comparable to that of angiogenic/anti-angiogenic factors for preeclampsia diagnosed in preterm gestations. sST2 did not perform as well when compared to angiogenic/anti-angiogenic factors in identifying preeclampsia diagnosed at term. The integration of sST2 to models using PlGF/sVEGFR1 or PlGF/sEng to identify preeclampsia at the time of diagnosis did not improve the performance compared to the latter biomarker ratios alone. This indicates that the combination of these markers does not result in a clinical advantage over using each marker alone, most likely due to their moderate to strong correlation.

Strengths and limitations:

This study is the first to examine the association between plasma concentrations of sST2 and abnormalities of Doppler velocimetry in uterine and/or umbilical arteries, gestational age at diagnosis, and severity of preeclampsia. The cross-sectional design, however, prevents inferences about temporal changes. Also, like all observational studies, we are unable to separate causation from association in this study.

Conclusion

Preeclampsia is associated with higher plasma sST2 concentrations than in uncomplicated pregnancies. The magnitude of this difference is markedly increased in early- than in late-onset disease, and in severe than mild preeclampsia. Among women with preeclampsia, the plasma sST2 concentration does not correlate with uterine or umbilical artery Doppler velocimetry. This indicates that elevated maternal plasma sST2 concentrations in preeclampsia are not related to the increased impedance to flow in the utero-placental circulation. The performance of sST2 in identifying preeclampsia at the time of diagnosis is similar to that of angiogenic/anti-angiogenic factors among women diagnosed prior to 37 weeks of gestation, but worse in identifying women diagnosed with preeclampsia thereafter. It remains to be elucidated if an elevation of maternal plasma sST2 concentrations in pregnancy is specific to preeclampsia.

Declaration of interest

This research was supported, in part, by the Perinatology Research Branch, Division of Intramural Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services (NICHD/NIH); and, in part, with Federal funds from NICHD, NIH under Contract No. HSN275201300006C.

Notes

*Presented at the 33rd Annual Meeting of the Society for Maternal Fetal-Medicine on February 11–16, 2013, in San Francisco, CA, USA.