Circulating pregnancy hormone relaxin as a first trimester biomarker for preeclampsia

Objective: Preeclampsia, a multi-system hypertensive disorder, is associated with perturbations in the maternal cardiovascular system during early pregnancy. The corpus luteal hormone relaxin, a potent vasodilator, may contribute to physiological circulatory changes especially in early gestation when circulating levels are highest. This study investigated whether first trimester circulating relaxin may be a suitable biomarker for the early prediction of preeclampsia. Methods: Relaxin was initially measured in first-trimester samples of women who developed late-onset preeclamptic (LO-PE; delivery ≥ 34 weeks; n = 33) and uncomplicated pregnancies (n = 25) in Pittsburgh, USA. Subsequently, to expand the group numbers, relaxin was measured in women who developed LO-PE (n = 95), early-onset preeclamptic (EO-PE; delivery < 34 weeks; n = 57), and uncomplicated pregnancies (n = 469) in Utrecht, the Netherlands. Results: In the Pittsburgh subjects, low relaxin levels (lowest centile: < p10) showed an adjusted odds ratio (OR) of 5.29 (95%CI 1.10–25.5) for LO-PE. In the Utrecht population, low relaxin levels (< p10) demonstrated adjusted ORs of 1.45 (95%CI 0.54–3.90) and 2.03 (95%CI 1.06–3.88) for EO-PE and LO-PE respectively, the latter increasing to an adjusted OR of 3.18 (95%CI 1.41–7.20) when newborn weight was < 10%. Serum relaxin concentrations slightly improved the detection rate of a previously derived prediction model for LO-PE from 42.5% to 45.1% at a fixed 10% false-positive rate. Conclusion: Relaxin shows little improvement in the performance of first trimester prediction models, which does not support its clinical implementation as a biomarker. Although this study was only correlational, the results point to a possible pathophysiologic role for low relaxin levels in pregnancies that later develop LO-PE.

of cardiovascular disease later in life for the mother [13,14], and her offspring [15]. Therefore, the early identification of pregnancies at high risk for preeclampsia is still one of the major challenges of modern obstetrics [16,17].
Relaxin is a peptide hormone secreted by the corpus luteum into the maternal circulation during the late secretory phase of the menstrual cycle and in pregnancy [18,19]. Relaxin is emerging as an important regulator of the maternal vascular adaptations to pregnancy [20,21]. Moreover, relaxin may hold promise as a novel therapeutic intervention for hypertensive pathologies, such as reduction of vascular wall stiffness as demonstrated in preclinical models [22][23][24]. A low first trimester serum relaxin concentration is associated with increased blood pressure in the third trimester of pregnancy [25]. Moreover, pregnancies achieved by oocyte donation that lack a corpus luteum, have no detectable circulating relaxin and are at increased risk for hypertensive disorders of pregnancy including preeclampsia [26,27]. Locally at the maternal-fetal interface relaxin may be important in the vascular preparation of the endometrium for placentation. In a non-human primate model of early pregnancy, relaxin was shown to stimulate vascularization as demonstrated by a significantly higher number of arterioles in the endometrium [28][29][30]. In vitro studies corroborate these findings as relaxin was shown to be a potent inducer of human endometrial maturation (decidualization) and increases expression of angiogenic factors such as vascular endothelial growth factor in endometrial cells [31][32][33]. Therefore, women with low concentrations of circulating relaxin could experience defects in the establishment of a functional maternal-fetal interface or fail to adequately vasodilate in early pregnancy, predisposing them to develop preeclampsia. The aim of this work was to establish whether relaxin may be a suitable first trimester biomarker in the early prediction of preeclampsia. We also evaluated pregnancy-associated plasma protein A (PAPP-A), free beta human chorionic gonadotrophin (fb-hCG), A Disintegrin And Metalloprotease 12 (ADAM-12) and placental growth factor (PlGF), previously established as biomarkers of preeclampsia, in the context of relaxin.

Study populations and ethics statement
We first performed a prospective pilot study at Magee-Womens Research Institute at the Department of Obstetrics, Gynecology and Reproductive Sciences, University of Pittsburgh, Pittsburgh, PA, USA ('Pittsburgh group'). To corroborate the findings of this initial study we used a large Dutch nested case-control group collected at the Center for Infectious Disease Research, Diagnostics and Screening at the National Institute for Public Health and the Environment, Bilthoven, The Netherlands ('Dutch group'). The University of Pittsburgh Institutional Review Board (protocol 0404159) and the Medical Ethical Committee of the University Medical Center Utrecht (protocol 11-002) approved the research protocols for the Pittsburgh and Dutch studies respectively.
The Pittsburgh population was a nested case-control study with samples derived from a longitudinal, prospective group of pregnant women studied for the development of preeclampsia and adverse pregnancy outcomes at Magee-Womens Hospital, Pittsburg, PA. For a detailed description we refer to a previous publication of this group [34]. The overall prevalence of preeclampsia in this group was 3.8% and gestational hypertension 6.4%. The study consisted of 33 women who developed LO-PE and 25 normotensive women with uncomplicated pregnancies matched for gestational age of sample collection, parity, and race. Blood samples were collected at 5 +0 -12 +1 gestational weeks and plasma was aliquoted and stored at −80°C until analysis. Maternal characteristics were obtained through interviews and detailed medical record abstraction and review. Pregnancy outcomes, including the diagnosis of preeclampsia, were determined retrospectively based on medical chart review by a jury of research and clinical investigators.
The Dutch population was a nested case-control study with serum samples derived from a large national group of women participating in the routine first trimester Down syndrome screening and has been used for previous studies by our group. For a detailed description we refer to two previous publications of this group [35,36]. In short, blood samples were collected at 9 +1 -13 +6 gestational weeks and serum was aliquoted and stored at −80°C until analysis. Pregnancy outcomes including chromosomal disorders, date of birth, birth weight and hypertensive disorders (pregnancy induced hypertension, preeclampsia, HELLP syndrome) were collected through self-reporting of participating women and confirmed by consultation with the participating clinics, where also maternal characteristics were obtained (i.e. medical history, parity, weight, height, first trimester blood pressure, smoking status) that were recorded by a midwife or gynecologist. The LO-PE group was selected to have a small for gestational age (SGA) infant in half of the cases.

Outcome measures
Preeclampsia was defined as the new onset of hypertension (≥140/ 90 mmHg) after 20 weeks of gestation measured on at least two separate occasions at least four hours apart, combined with the presence of proteinuria (a 24-hour collection of urine with ≥ 300 mg/24 h or at least 2 + by dipstick on a spot urinalysis), according to the criteria of the International Society for the Study of Hypertension in Pregnancy [37]. (The definition of preeclampsia has since changed, insofar as proteinuria is no longer a requirement for the diagnosis [38]. But when the current study was conducted, the older definition was in place.) EO-PE was defined as preeclampsia necessitating delivery < 34 weeks gestational age, and LO-PE as preeclampsia in pregnancies delivering ≥ 34 weeks. To calculate birth weight z-score, the 2008 growth charts from The Netherlands Perinatal Registry (PRN, now Perined) were used [39]. SGA was defined as a birth weight under the 10th centile.
MAP was calculated by adding 1/3 of the pulse pressure (difference between diastolic and systolic blood pressure) to the diastolic pressure: MAP = 1/3 (SBP-DBP) + DBP.

Sample analysis
Sample analysis for the Pittsburgh group was performed at R&D Systems, Minneapolis, MN, USA. The Dutch group was analyzed at the Laboratory for Prenatal Screening, National Institute for Public Health and the Environment, The Netherlands. In both groups samples were analyzed blinded for outcome in duplicate using the Human relaxin-2 Quantikine enzyme-linked immuno-sorbent assay (ELISA), according to the manufacturer's instructions (DRL200, R&D Systems, Abingdon, United Kingdom). All measurements were above the detection limit of the ELISA kit (assay range 7.81-500 pg/mL). The ELISA kit has an intraand inter-assay coefficient of variance of 3.2% and 7.3% respectively. Values with an intra-assay coefficient of variance above 15% were excluded from analysis (Pittsburgh group n = 0; Dutch group n = 2).

Statistical analysis
Study population characteristics were expressed as numbers and percentages for categorical variables and median and interquartile ranges (IQR) for continuous variables (as data were not normally distributed), and were compared between preeclampsia cases and controls using Fisher's Exact and Mann-Whitney U tests, respectively. Bonferroni corrections for multiple testing were applied when both EO-PE and LO-PE were compared to controls. Relaxin concentrations were also expressed as median and IQR. Based on the centiles of the relaxin concentrations in the control population, two cut-off points were determined at the 25th (p25) and the 10th (p10) centile, respectively. Subsequently, to study the association between relaxin concentrations and preeclampsia, unadjusted and adjusted odds ratios (OR) were calculated using multivariate logistic regression analysis. In the Dutch group, the LO-PE group was selected to have a SGA infant in half of the cases. Again, multivariate logistic regression analysis was used to study associations between relaxin and preeclampsia among these subgroups.
Based on the predefined cut-off point of relaxin < p10, detection rates (sensitivity) were calculated for relaxin as a single biomarker, as well as in combination with maternal characteristics (prior risk based on: age, BMI, nulliparity) and MAP.

Results
Demographic and clinical characteristics of the study populations are presented in Table 1. Data from a total of 25 control pregnant and 33 LO-PE women were analyzed in the Pittsburgh group. The control group consisted of 16 (64%) Caucasian and 9 (36%) African-American; the LO-PE group 23 (70%) Caucasian and 10 (30%) African-American. The Dutch group consisted of 469 control pregnant, 57 EO-PE and 95 LO-PE women. Data on ethnicity were not collected in the Dutch group, but the make-up of the Dutch pregnant population is reported to be largely Caucasian (84%) [40]. Women who developed preeclampsia had a significantly higher MAP in both groups. In the Pittsburgh group, other characteristics did not reach statistical significance. In the Dutch group there were several statistically significant differences between the study groups. Preeclamptic women had a higher BMI and were more often nulliparous. Furthermore, preeclamptic women had more often a history of hypertensive pregnancy disorders. There was no significant difference in smoking status between study groups. Preeclamptic women delivered earlier, their infants had a lower birth weight, and were smaller for gestational age.

Pilot study in 'Pittsburgh group'
In the Pittsburgh group relaxin concentrations did not prove to be significantly different in LO-PE from controls (median with IQR for controls 802 [571-966] pg/mL and LO-PE 783 [452-1081] pg/mL; p = 0.561). Moreover, relaxin concentrations showed no correlation with gestational age in the control group (Fig. 1). However,   Table 2). When the OR was adjusted for gestational age at sampling, BMI and smoking it increased to 5.29 (1.10-25.5). A cut-off value at the p25 of the control group (< 570.8 pg/mL) showed a similar trend, but did not reach statistical significance (OR unadjusted 1.81 and adjusted 2.12, respectively).

Validation in 'Dutch group'
The same analysis strategy was applied for the independent data of the Dutch group. Again there was no change in relaxin concentration with gestational age (Fig. 2). Relaxin concentrations were higher in the Dutch group compared to the Pittsburgh group; the latter was sampled on average 4-5 weeks earlier in pregnancy. As in the Pittsburgh group, a relaxin concentration below p10 (< 1054.4 pg/mL) was associated with LO-PE: OR 2.1 (p = 0.015) and adjusted OR 2.03 (p = 0.032) ( Table 2). A cut-off value at p25 (< 1342 pg/mL) resulted in an OR of 1.60 (p = 0.05), which was not significant after adjustment (OR 1.50; p = 0.124). There was no significant association with EO-PE.
In the LO-PE group we performed a sub-analysis to assess the influence of an SGA infant (Table 3). In LO-PE women who delivered an SGA infant (LO-PE with SGA), the correlation with a relaxin concentration below p25 and p10 was stronger, with adjusted ORs of 2.17 (p = 0.034) and 3.18 (p = 0.005) respectively. In LO-PE women without a SGA infant (LO-PE without SGA), the correlations with a relaxin concentration below p25 and p10 were lost (adjusted OR's of 1.06, p = 0.871; and 1.30, p = 0.582 respectively). We did not include Odds ratios were adjusted for gestational age at sample collection, Body Mass Index (BMI) and smoking. < p25: below 25th percentile; < p10: below 10th percentile; OR: Odds ratio; CI: confidence interval; EO-PE: early-onset preeclampsia; LO-PE: late-onset preeclampsia. a SGA sub-analysis for EO-PE because of low case numbers (12 out of 57 EO-PE cases delivered a SGA infant), and lack of correlation between low relaxin concentrations (< p10) with EO-PE in the initial analysis.

Screening performance in 'Dutch group'
In the Dutch group, prediction models for the complete LO-PE group (n = 95) and for the subgroup of LO-PE with SGA (n = 47) were fitted based on maternal characteristics (prior risk), MAP and relaxin < p10 ( Table 4). Comparison of the performance of these models showed that the highest detection rate (DR) for the complete LO-PE group was obtained for the model combining maternal characteristics, MAP and relaxin < p10, with a DR of 45.1% at a false positive rate (FPR) of 10% (16.6% gain in DR compared to the model using prior risk only; 2.6% gain in DR compared to the model using prior risk + MAP). For the prediction of LO-PE with SGA the best performing model combined maternal characteristics with relaxin < p10 only, with a DR of 37.1% at a FPR of 10% (2.7% gain in DR compared to the model using prior risk only).

Correlation of relaxin with other circulating biomarkers
We previously reported pregnancy-associated plasma protein A (PAPP-A), free beta human chorionic gonadotrophin (fβ-hCG), A Disintegrin And Metalloprotease 12 (ADAM-12) and placental growth factor (PlGF) concentrations for the study subjects of the Dutch group [18], enabling us to correlate relaxin concentrations with these other circulating biomarkers. There was a statistically significant positive, but low correlation of relaxin with PAPP-A (Spearman's rho 0.098; p = 0.014) and fβ-hCG (Spearman's rho 0.137; p = 0.001). There was no correlation between relaxin and ADAM-12 (Spearman's rho 0.072; p = 0.074) and between relaxin and PlGF (Spearman's rho 0.065; p = 0.111).

Discussion
To the best of our knowledge, this two-stage study, in which patients were well-characterized, is the first to investigate relaxin as a predictor for women destined to develop preeclampsia. We observed a clear association of low (lowest centile: < p10) first trimester relaxin concentrations with LO-PE (adjusted OR 2.0-5.3) but not with EO-PE. However, in a multivariate model, a relaxin concentration below the 10th centile only showed a~2.5% gain in detection rate of prediction models for LO-PE, which combined maternal characteristics (prior risk based on: age, BMI, nulliparity) with or without MAP.
Preeclampsia in combination with a SGA infant is suggestive of a distinct and more severe (placental) pathophysiology [41,42], which generally results in different performance of biomarkers for this subgroup [35,43]. A low relaxin concentration (< p10) indeed showed a stronger association with LO-PE in combination with SGA, compared to LO-PE without SGA. This, as well as the association between low relaxin concentration with LO-PE but not with EO-PE, underscores the heterogenic nature of the preeclampsia syndrome and is in line with accumulating evidence for differences in the underlying etiology. EO-PE is thought to result from poor placental development (placental preeclampsia) [10], whereas LO-PE is more likely to result from poor maternal systemic cardiovascular adaptations to pregnancy (maternal preeclampsia) [44,45]. However, logic dictates that this pathophysiologic division is likely to be graded and not absolute, with no definite gestational age break point. Nevertheless, the association of low relaxin with LO-PE but not EO-PE may indicate that adequate circulating relaxin levels are more important for the maternal systemic vascular adaptations to pregnancy, but less so for processes at the maternal-fetal interface that guarantee the establishment of a functional placenta.
The finding of a stronger association between low circulating relaxin and LO-PE in combination with SGA may at first glance seem paradoxical. However, given the higher occurrence of LO-PE compared to EO-PE [8], SGA is more commonly associated with LO-PE. On a  pathophysiological basis, low circulating relaxin in early human pregnancy may impair maternal vascular adaptation, thereby compromising uterine perfusion and fetal growth [46]. In support of this hypothesis, relaxin was shown to increase uterine blood flow velocity and to increase uterine arterial compliance in conscious rats [47]. There was a low but significant correlation between relaxin and placenta-derived biomarkers measured in the 469 control samples of the Dutch cohort: Pregnancy-associated plasma protein A (PAPP-A) and free beta human chorionic gonadotrophin (fb-hCG). Adding relaxin to a prediction model of PAPP-A and/or fb-hCG (including MAP and prior risk) did not significantly improve the performance of these prediction models to predict EO-PE or LO-PE (data not shown). No correlation was found between concentrations of relaxin with A Disintegrin And Metalloprotease 12 (ADAM-12) and placental growth factor (PlGF) [35]. Relaxin concentrations showed a slight inverse correlation with BMI and MAP (data not shown).
Overall strengths of the present study are the two well-characterized study populations with similar assays run in blinded fashion, and the consistency in resulting findings of both groups adding to the generalizability. The strengths of the Pittsburgh group are the very well characterized population with a jury of investigators reviewing and confirming the diagnosis of preeclampsia. Limitations of the Pittsburgh group include the relatively small sample size, the selection of participants in a tertiary clinical setting and the lack of EO-PE cases. Furthermore, preeclampsia cases in this group had a relatively high BMI adding to the stronger association with low relaxin concentrations than found in the Dutch group. The Dutch group was much larger, includes both EO-PE and LO-PE cases, and comprises an unselected general population, as samples derived from surplus material of the national Dutch first trimester down syndrome screening program. Consequently this population has a relatively high maternal age (~33 years) compared to the Pittsburgh group which may affect study comparison, although both studies were internally controlled. Moreover, the gestational age at sampling was on average 4-5 weeks later in pregnancy in the Dutch group than the Pittsburgh group. This may partly explain differences in absolute relaxin concentrations between these studies, as circulating relaxin concentrations are known to rise after ovulation until the end of the first trimester [18,19]. Furthermore, other differences in study population characteristics of the Pittsburgh compared to the Dutch group (predominately Caucasian) may also have contributed.
In conclusion, relaxin only provides little improvement in performance of prediction models for LO-PE and no improvement for EO-PE, which does not support clinical implementation. The association between low relaxin concentrations (< p10) with LO-PE, particularly when complicated by a SGA infant may indicate a potential pathophysiologic role for deficient circulating relaxin in one subtype of preeclampsia.