Circulating SPINT1 is a biomarker of pregnancies with poor placental function and fetal growth restriction

Placental insufficiency can cause fetal growth restriction and stillbirth. There are no reliable screening tests for placental insufficiency, especially near-term gestation when the risk of stillbirth rises. Here we show a strong association between low circulating plasma serine peptidase inhibitor Kunitz type-1 (SPINT1) concentrations at 36 weeks’ gestation and low birthweight, an indicator of placental insufficiency. We generate a 4-tier risk model based on SPINT1 concentrations, where the highest risk tier has approximately a 2-5 fold risk of birthing neonates with birthweights under the 3rd, 5th, 10th and 20th centiles, whereas the lowest risk tier has a 0-0.3 fold risk. Low SPINT1 is associated with antenatal ultrasound and neonatal anthropomorphic indicators of placental insufficiency. We validate the association between low circulating SPINT1 and placental insufficiency in two other cohorts. Low circulating SPINT1 is a marker of placental insufficiency and may identify pregnancies with an elevated risk of stillbirth.


Supplementary Figure 2:
Circulating analytes screened at 36 weeks that were detected in plasma but not changed among women destined to deliver a small-for-gestational-age (SGA) infant. These are 13 of 22 proteins screened in maternal plasma in a case control cohort at 36 weeks' gestation. There were 105 cases of SGA (birthweight <10 th centile) and 210 matched controls. Graphed here are proteins we screened that were detectable in the maternal plasma, but were not different in the two groups: COBLL1 (a), CRH (b), CSH1 (c), ENDOU (d), MAO (e), MICALL1 (f), PAPPA (g), S100P (h), SERPINB2 (i), FIBULIN-1 (j), hCG (k), SIGLEC 6 (l) and TFPI (m). Individual symbols represent individual patient analyte levels. The full names of the proteins are listed in supplementary table S3. Source data included as a source data file.

Supplementary Figure 4: SPINT1 levels are not associated with umbilical artery resistance, infant body fat percentage, or fat mass.
A subgroup from the FLAG cohort had antenatal ultrasound measurements, and/or infant body composition assessed after birth using the PEAPOD air displacement plethysmography device. There was no relationship between circulating SPINT1 concentrations at 36 weeks and umbilical artery pulsatility index (a), neonatal body fat percentage (b) or neonatal fat mass (c). We also examined whether SPINT1 concentrations were similar in plasma and serum (d). We measured SPINT1 in women who delivered a <10 th centile baby at <34wks relative to levels in healthy controls -where each woman, we obtained both a plasma and serum sample at the same blood draw. While SPINT1 was significantly reduced in the serum of women with SGA, the degree of change was far less than that observed in plasma. Individual symbols represent an individual patient. Panel a; n=327, b,c; n=281, d, data expressed as mean± s.e.m ****p<0.0001, *p<0.05 using two tailed Mann-Whitney U tests vs control. Source data included as a source data file. Figure 5: Comparing the association between circulating SPINT1 and PlGF at 36 weeks with clinical with markers of placental insufficiency. These were a subset of participants from the FLAG cohort. Compared to PIGF, circulating SPINT1 concentrations appeared to have a stronger association with birth weight centile (a vs f), placental weight (b vs g), neonatal lean mass (c vs h) and uterine artery resistance (d vs i). Neither were significantly correlated with umbilical artery resistance (e and j). Each datapoint represents an individual patient. a; n=999, f; n=1002, b,g; n=96, c,h; n=136, d, e, i, j; n=63. Source data included as a source data file. Figure 6: Murine fetal, but not placental weights are reduced by maternal exposure to hypoxia. Following exposure to maternal hypoxia, fetal weight (a; n=9 normoxic, n=9 hypoxic from separate litters. c; n=13 normoxic, n=11 hypoxic from separate litters) but not placental weight (b; n=9 normoxic, n=9 hypoxic from separate litters. d; n=13 normoxic, n=11 hypoxic from separate litters) was significantly reduced. Fetal and placental weights shown in a,b relate to the placentas where SPINT1 mRNA expression was measured (Figure 4e, main manuscript), whilst fetal and placental weights shown in c,d relate to the placentas where SPINT1 protein expression was assessed ( Figure  4f, main manuscript). Each individual symbol represents an individual fetus or placenta. Data expressed as mean± s.e.m ****p<0.0001 using two-tailed Mann-Whitney U tests. Source data included as a source data file.

Discovering circulating biomarkers of placental insufficiency
To identify new biomarkers of placental insufficiency we performed the Fetal Longitudinal Assessment of Growth (FLAG) study. This included prospective collection of blood samples from pregnant women at 28 (27 +0 -29 +0 ) and 36 (35 +0 -37 +0 days) weeks' gestation from 2015 participants. The FLAG study was undertaken at Mercy Hospital for Women, a tertiary referral hospital in Melbourne Australia. This study was approved by the Mercy Health Research Ethics Committee (Ethics Approval Number R14/12) and written informed consent was obtained from all participants. Identification of pregnancies that ended in the birth of a neonate that was small for gestational age (birthweight centile <10 th ) was pre-specified as the primary outcome in the study protocol.
We also recruited a sub-cohort consisting of 347 nulliparous participants chosen from those enrolled in the FLAG study to undergo more intensive studies. For this sub-cohort, we performed ultrasound assessments at 28 and 36 weeks' gestation, where several parameters were measured including Doppler assessment of blood flow resistance in the uterine, umbilical and the fetal middle cerebral arteries. Post birth, where possible, we measured neonatal body composition (lean body mass and fat mass) within 4 days of birth by performing air displacement plethysmography studies using a PEAPOD device (COSMED, Concord, CA, USA). Further methods for recruitment for the FLAG study are described below.
To identify new biomarkers of placental insufficiency near term gestation we focused on the 36 week samples, where 1996 plasma samples were available. We divided the cohort approximately in half to discover, and then subsequently validate markers. Samples from the first 997 consecutively recruited participants constituted Cohort 1 and those from the second 998 consecutively recruited participants constituted Cohort 2.
We initially screened 22 circulating proteins in a 1:2 nested case (105 neonates born small-forgestational-age, SGA, birthweight <10 th centile) control (210 neonates born with a birthweight ≥10 th centile) set selected from Cohort 1. The controls were group-matched to cases for maternal age, booking body mass index, smoking status, gestational diabetes mellitus, and parity.
We selected proteins that are highly expressed in placenta relative to other tissues by referencing two bioinformatic databases. We selected proteins that were both 1) highly expressed at the mRNA level in placenta relative to all other human tissues using BioGPS (Biogps.org); and 2) abundantly expressed in its protein form on the membrane surface of the placenta (using Protein Atlas [www.proteinatlas.org]). The function of many of these proteins remain poorly understood. However, we postulated that because these are highly expressed in the placenta, most will play important biological roles and the expression of some may be perturbed in the presence of placental insufficiency.
The proteins screened where circulating concentrations were different among cases of SGA compared to controls were re-assayed in a new batch assay run on all samples from Cohort 1. Those that remained significantly associated with SGA were then assayed in Cohort 2. We also measured placental growth factor (PIGF) and soluble fms-like tyrosine kinase-1 in Cohort 2. Further methodology detailing how the proteins were measured is included below.

Developing diagnostic tests for placental insufficiency
The intent of the FLAG study was to examine whether a combination of these markers was better at identifying fetal growth restriction, compared to one in isolation. We examined the diagnostic performance of potential markers, either alone or in combination, to predict neonates born at birthweight centiles; <20 th , <10 th , <5 th and <3 rd ; and those with birthweight <5 th centile who required nursery admission (the latter is a small cohort likely to have suffered significant placental insufficiency).
We generated potential diagnostic tests by examining whether we could combine the different proteins in results obtained in Cohort 2. We tried to develop modelling by setting the specificity at around 90%, which would equate to a 10% screen positive rate. We found SPINT1 performed the best and none of the other markers added to its performance. Therefore, we validated the diagnostic performance of SPINT1 in Cohort 1. To adjust for the fact that the research ELISA we used exhibited variability in reporting the absolute SPINT1 concentrations between the batches run for Cohorts 1 and 2, we expressed SPINT1 results as multiples of the median (MoMs).
We also developed (Cohort 2) a 4-tier model of risk for these different low birthweight ranges based on different SPINT1 MoM cut off levels. These cut-off levels were arbitrarily chosen when developing the test in Cohort 2.
Further methods on describing the statistical analyses are detailed below.

SPINT1 and other parameters of placental insufficiency
We further investigated SPINT1 by correlating circulating levels with several other clinical indicators of placental insufficiency using data obtained from our sub-cohort study. We also measured circulating SPINT1 concentrations and placental expression in a separate cohort with fetal growth restriction delivered at <34 weeks gestation, where we performed further laboratory investigations in vitro, and in a mouse model of fetal growth restriction. Detailed methods on the laboratory studies are described below.

Recruitment of samples for the FLAG cohort
Women were screened for eligibility and were invited to participate at their oral glucose tolerance test, universally offered around 28 weeks' gestation to test for gestational diabetes mellitus. Inclusion Criteria -English-speaking women aged over 18 years, with a singleton pregnancy and normal midtrimester fetal morphology examination were eligible to participate. Samples from women where a SGA fetus was suspected at the time of blood sampling were not excluded. Exclusion criteria -multigestation pregnancies or pregnancies identified as having fetal anomalies.

Maternal blood sample collection and preparation for the FLAG cohort
Participants donated blood samples at between 27 +0 to 29 +0 weeks' and/or 35 +0 to 37 +0 weeks' gestation inclusive. Whole blood was collected in a 10ml ethylenediaminetetraacetic acid tube. Plasma was stored at -80°C until the time of sample analysis.

Outcomes and definitions of cases for the FLAG cohort
Maternal characteristics and pregnancy outcomes were obtained from review of each participant's medical record, investigation results and hospital database entry, by a single clinician, blinded to any protein levels.

Birthweight centile calculations
Infant birthweights were assigned a customised centile using the GROW software 1 (www.gestation.net), which generates a 'term optimal weight' based on an optimised fetal weight standard. We adjusted for the following non-pathological factors: maternal height, weight and parity; infant sex; and exact gestational age. Coefficients for the Australian dataset of GROW were informed by a local dataset; the multiple regression model has a constant to which weight is added or subtracted for each of the adjusted variables. SGA was defined as customised birthweight <10 th centile. We compared the circulating protein levels among SGA cases, to those of the controls. We determined a priori plans to further investigate our most promising biomarkers for their association with several clinical parameters associated with placental insufficiency and fetal growth restriction by: (i) validating our findings in the entire first 1000 sample cohort; (ii) validating our findings in the second 1000 sample cohort; (iii) investigating the predictive potential of the biomarker in the 28 week blood samples; (iv) correlating candidate biomarkers with 36 week Doppler ultrasound parameters that are antenatal indicators of uteroplacental function; and (v) correlating candidate biomarkers with neonatal body composition measures -indicators of in utero nutrient supply.

Doppler ultrasound parameters at 36 weeks
Some nulliparous participants were also involved in the ultrasound-based arm of the FLAG study. For this, 347 women underwent a 36 (35 +0 -37 +0 ) week ultrasound assessment where transabdominal colour and pulsed-wave Doppler were used to measure the mean maternal uterine artery pulsatility index (PI) and the umbilical artery PI. Measurements were taken during periods of fetal apnoea and inactivity with the angle of insonation close to zero. The umbilical artery PI was measured in a free loop of umbilical cord away from cord insertion sites. For the maternal uterine artery the probe was placed in each of the iliac fossae, and the waveform recorded within 1cm of the uterine artery crossing the external iliac artery 2 . PI values were measured in triplicate and the mean calculated. Average uterine artery PI values were obtained for both the right and left vessels, and these averaged to provide the overall mean PI. For each of the PI values, the gestation-dependent centile (if normally distributed), or the multiples of the median (MoM) were determined. Treating clinicians were blinded to the uterine artery PI results.

Inclusion and exclusion criteria for samples from the MAViS clinic
To validate the observation that SPINT1 is associated with placental insufficiency we measured SPINT1 in a high-risk cohort from the United Kingdom -the Manchester Antenatal Vascular Service (MAViS clinic). Women gave written informed consent to donate samples for future research studies. The study was approved by the NRES Committee North West 11/NW/0426.
The inclusion criteria for women in the MAViS study were: 1. chronic hypertension BP ≥140/90 at ≤20 weeks; 2. chronic hypertension requiring antihypertensive treatment ≤ 20 weeks; 3. pre gestational diabetes with evidence of vascular complications (hypertension, nephropathy); 4. history of ischeamic heart disease and 5. previous early onset preeclampsia.
Prespecified outcomes in the protocol for the MAViS biobank collection: primary outcome was the development of pregnancy complication requiring preterm birth (<37 weeks), secondary outcomes were the development of preeclampsia, birth of a small for gestational age neonate and early preterm birth (<34 weeks' gestation).
For our validation study cases and controls were selected from the MAViS biobank to specifically address whether SPINT1 was differentially expressed in pregnancies destined to deliver a neonate <10 th centile birthweight (see supplementary table 10). Hence, for the purposes of our present study, the delivery of a small for gestational age was our primary prespecified outcome.

Recruitment and collection of samples from women with preterm fetal growth restriction
Blood samples or placental samples were obtained from women with a diagnosis of preterm fetal growth restriction and were delivered at <34 weeks' gestation, or gestation matched controls. Sample collection as part of the Mercy Tissue Bank was approved by the Mercy Health Research Ethics Committee (Ethics Approval Number R11/34) and written informed consent was obtained from all participants. Use of samples for this study was approved by the Mercy Health Research Ethics Committee (Ethics approval number R18/55). In this cohort fetal growth restriction <34 weeks' gestation was pre-specified as the primary outcome.
For the fetal growth restriction cohort, women were invited to participate when it was diagnosed on ultrasound. We also only included their samples if they were delivered <34 weeks' gestation and growth restriction was confirmed following birth (by the fact that the neonatal weight was <10 th centile).
Controls were women who consented to blood collection at matched gestations but went on to deliver a normal sized infant (>10 th centile) at term gestation.
For the control group for the preterm placental samples we obtained placentas from women who were delivered for reasons other than fetal growth restriction or hypertensive diseases (the mother remained normotensive), such as vasa praevia, maternal medical conditions, spontaneous preterm labour or antepartum haemorrhage.
For the placental studies all women were delivered by caesarean section for both the fetal growth restriction and control cohorts.
Whole blood was collected in a 10ml ethylenediaminetetraacetic acid tube. Plasma was stored at -80°C until the time of sample analysis. Placental tissue was obtained immediately following delivery. Maternal and fetal surfaces were removed, and a sample was then washed briefly in sterile phosphatebuffered saline (PBS). Samples for RNA or protein extraction were fixed in RNALater for 48 hours and then stored at -80°C.

Mouse model of maternal hypoxia
To assess the effect of maternal hypoxia on SPINT1, placentas were obtained from mothers exposed to hypoxia (10% inspired O2) or normoxia (21% inspired O2) from gestational day 14.5-19.5 as previously described 5 . Briefly, virgin C57BL/6 J female mice, aged 6-8 weeks, were housed in groups of two to five per cage under a 12:12 h light:dark cycle at 22°C and were mated overnight with C57BL/6 J males. The presence of a copulatory plug was designated as day (D)1 of pregnancy (term ∼D20.5). All animals had ad libitum access to water and food [RM3, energy from fat 11%, protein 26%, carbohydrate 62% (simple sugar 7%), 15.3 MJ kg −1 , diet code 801066; Special Diet Services, Witham, Essex, UK]. Mated females were weighed daily and the daily consumption of food and water was measured per cage to calculate intake per mouse per day. Pregnant mice were exposed to chronic, normobaric hypoxia for 5 day periods by placing their cages into an isolated PVC chamber (PFI Plastics Ltd, Milton Keynes, UK) in which the oxygen content was reduced to 10% by displacing oxygen with nitrogen using a nitrogen generator (N2MID60; Domnick Hunter Ltd, Warwick, UK). Control mice were maintained at room oxygen (21%). All procedures described were approved by the Ethical Review Committee of the University of Cambridge (Cambridge, UK) and were carried out in accordance with UK Animals (Scientific Procedures) Act 1986 as previously reported 5 .
On day 19 of gestation, dams were anesthetised before death (by cervical dislocation) with an intraperitoneal injection of fentanyl-fluanisone and midazolam in sterile water (1:1L2, 10 ug/mL; Janssen animal health). The uterus was removed, and all placentas were dissected free of fetus and membranes, weighed and immediately snap frozen whole in liquid nitrogen and stored at -80C for molecular analyses. For RNA and protein studies, 1 placenta per litter was included. Proteins were extracted from the second lightest placenta per litter (where possible) in lysis buffer containing 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM Na2EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM b-glycerolphosphate, 1 mM Na3VO4, and complete mini proteases inhibitor cocktail (Roche Diagnostics, East Sussex, UK). Lysates protein concentration was determined using a Bicinchoninic acid assay (Sigma-Aldrich). RNA was extracted from the smallest placenta per litter (where possible) using the RNeasy Plus Mini Kit (Qiagen, Machester, UK) and the quantity of RNA determined using a NanoDrop spectrophotometer (NanoDrop Technologies, Inc.). From each sample, 2.5ug of total RNA was reverse transcribed to cDNA using High Capacity cDNA Reverse Transcription Kit with random primers (Applied Biosystems, Paisley,UK).

ELISAs to measure circulating levels of proteins
Commercially available protein-specific ELISAs were obtained and used to measure the concentration of each protein of interest in pg/ml (see table S3). Maternal plasma levels of sFlt-1 and PlGF were each measured with a commercial electrochemiluminescence immunoassay platform (Roche Diagnostics). These assays have received Conformité Européenne marking for use as in vitro medical devices.

RT-PCR on human and mouse placenta
To measure SPINT1 mRNA expression in human and mouse placenta, mRNA was extracted from 20-30 mg of RNA later preserved frozen human placental samples by homogenization or from cytotrophoblast using an RNeasy mini-kit (Qiagen Western Blot on human and mouse placenta To examine SPINT1 protein expression, western blot analysis was undertaken on human and mouse placenta. 20μg of placental lysates (n=23 preterm and n=13 FGR) or 15 μg of trophoblast (n=11 separate isolations) were separated on 10% SDS-polyacrylamide gels with wet transfer to PVDF membranes (Millipore, Billerica, MA). Membranes were blotted overnight with an antibody targeting SPINT1 (Rabbit anti-human SPINT1, Sigma, 1:250), an anti-GAPDH antibody for human placental samples (1:5000, Cell Signaling Technology, Danvers, MA, USA) or an anti-b-actin antibody for trophoblast samples (1:10,000, Sigma) and visualized using the Amersham ECLTM Prime Western blotting detection reagent (VWR International) and ChemiDoc XRS (BioRad, Hercules, CA, USA). Relative densitometry was determined in all samples using Image Lab (BioRad). For murine placental western blots, 20μg of placental lysates were separated on a 12% SDSpolyacrylamide gel with wet transfer to PVDF membranes (Millipore, Billerica, MA). Membranes were blotted overnight with an antibody targeting murine SPINT1 (R&D systems, 1:2000 dilution) and b-actin followed by an anti-mouse HRP. Protein bands were visulaised using the Amersham ECLTM Prime Western blotting detection reagent (VWR International) and ChemiDoc XRS (BioRad, Hercules, CA, USA). Relative densitometry was determined in all samples using Image Lab (BioRad).