Determinants of “placental” versus “maternal” preeclampsia

Abstract

Background:

The placenta is known to be critical in the etiology of preeclampsia (PE). However, there is a subset of PE cases without identifiable placental pathology. We evaluated which clinical PE classification system best distinguishes PE with placental pathology from PE without placental pathology.

Methods:

We evaluated five placental pathologic features in 197 placentas from PE patients grouped by three clinical PE subclasses. 1. PE with calculated infant birthweight <10th percentile (small) for gestational age (SGA PE) versus PE with birthweight ≥10th percentile for gestational age (NSGA PE). 2. PE with delivery before 34 weeks of gestation (EDPE) versus PE with delivery at or after 34 weeks of gestation (LDPE) and 3. PE with severe features versus PE without severe features. Clinical, histologic and molecular findings in patients with PE were compared to normotensive patients, with and without SGA infants (N = 1,078 total).

Results:

The SGA versus NSGA PE classification system performed best (likelihood ratios (95% CI) for ≥ 3 of 5 placenta pathologic findings: 15.7 (6.5, 38.1) in SGA PE vs. NSGA PE; 6.8 (4.3, 10.8) in EDPE vs. LDPE and 5.2 (1.95, 14.1) in PE with SF vs. PE without SF; all p <.0001). SGA PE and SGA normotensive placentas were abnormal and shared alterations in hypoxia, TNF alpha, glycolysis, unfolded protein response, estrogen response, UV response, p53, TGF beta and mTORC1 signaling pathways.

Conclusions:

Classifying PE based on birthweight percentile for gestational age is the most useful system for consistently identifying PE associated with placenta pathology.

Graphical Abstract

Introduction

Preeclampsia (PE) is a hypertensive disorder of pregnancy that the American College of Gynecology (ACOG) defines as new-onset hypertension (systolic blood pressure of 140 mm Hg or more and/or diastolic blood pressure of 90 mmHg or more) and proteinuria or new-onset hypertension with thrombocytopenia, renal insufficiency, impaired liver function, pulmonary edema, headache unresponsive to medication or visual changes at >20 weeks gestation [1]. PE is a major cause of maternal morbidity and mortality world-wide [2].

ACOG lists history of PE in a past pregnancy, multi fetal gestation, chronic high blood pressure, kidney disease, diabetes mellitus and autoimmune conditions as high risk factors for preeclampsia. Moderate PE risk factors include nulliparity, obesity (BMI >30), lower socioeconomic status, first-degree family history of PE, greater than 10 years between pregnancies, age 35 years or older, previous pregnancy complications, in vitro fertilization and Black race [1]. Both gestational and preexisting diabetes mellitus (DM) have been associated with increased risk of preeclampsia [3-4].

Multiple lines of evidence show that the placenta is critical to the etiology of preeclampsia: PE only occurs in the setting of pregnancy, PE can occur without a fetus, such as after molar gestations [5], PE can occur after extra-uterine pregnancies [6], PE is associated with placental abnormalities [7-9], PE often resolves following delivery of the placenta [6, 10] and PE-like symptoms can be induced in animals injected with human placenta extracts [11]. However, there is a subset of PE cases without identifiable macroscopic or histologic placental findings, where extra-placental factors appear to be sufficient to cause PE. This PE subset lacking identifiable placenta pathologic findings has been termed maternal PE by some authors [12-13]. In a study by Leavey et al., maternal PE accounted for 17.5% (14 of 80) of PE cases [13].

Subtyping PE as placental versus maternal is a way of attempting to address the underlying pathophysiology of this multisystemic disease. Other ways of subtyping PE include subtyping based on clinical evidence of disease onset, with early onset defined as PE diagnosed before 34 weeks of gestation and late-onset as PE diagnosed at or after 34 weeks of gestation [14]. PE has also been subtyped based on severity of maternal hypertension or organ dysfunction, with PE with severe features (PE with SF) defined as PE with systolic blood pressure ≥160 mm Hg or diastolic blood pressure ≥ 110 mm Hg or PE with any of the following: thrombocytopenia, impaired liver function, renal insufficiency, pulmonary edema, new-onset headache unresponsive to medication or visual disturbances; and PE without severe features (PE without SF) as PE without the aforementioned signs [1, 14].

Some of the short-comings of subtyping PE based on disease onset include the difficulty of definitely pinpointing disease onset. Likewise, subtyping PE based on severity of maternal hypertension is limited by blood pressure fluctuations. Furthermore, neither categorization clearly identifies underlying pathophysiology.

In our previous study addressing PE categorization based on placental histopathology, we found that placentas with pathologic lesions characterized as maternal vascular malperfusion (MVM) were associated with preterm PE with severe features and small for gestational age neonates [15].

To our knowledge, no study has compared the various PE categorization systems to one another. In their review article addressing clinically meaningful PE subtypes, Roberts et al. [16] suggest that a clinically useful PE subtype should: 1. Identify specific pathophysiologic pathway. 2. Indicate maternal or fetal outcome, 3. Be recognizable in a clinically useful time frame and 4. Be reproducible and generalizable, including in low resource settings.

The current study was undertaken to determine which clinical PE classification system most reliably identifies a specific pathophysiologic pathway (i.e. distinguishes PE associated with placental pathology (“placental” PE) from PE without associated placental pathology (“maternal” PE)), while also indicating maternal and fetal disease severity and being generalizable. The null hypothesis being that there are no differences in placental findings among the various clinical PE classification systems.

Methods

In order to minimize the possibility of unintentionally sharing information that can be used to re-identify private information, a subset of the data generated for this study are available at [Gene Expression Omnibus (GEO) and can be accessed at [https://www.ncbi.nlm.nih.gov/geo/] and at Sequence Read Archive (SRA) and accessed at https://www.ncbi.nlm.nih.gov/sra]

Study cohort

This was a single institution retrospective cross-sectional case-control study. Clinical and pathologic data were obtained from patients with singleton parturitions at UCSD-affiliated hospitals between January 2010 and November 2020 after having given written informed consent for accessing their electronic medical records and collection of placental tissue. Demographic, clinical and placenta histology data were collected in a REDCap-based obstetric registry. For each pregnancy, the clinical diagnoses of preeclampsia with or without severe features were adjudicated by two board-certified obstetricians, per ACOG criteria [1]. We did not have definite record of PE onset and thus used time of delivery instead; PE associated with delivery before 34 weeks of gestation (<34 weeks GA) was categorized as early delivery PE (EDPE) while PE with delivery at or after 34 weeks of gestation (≥ 34 weeks GA) was categorized as late delivery PE (LDPE). Birthweight percentiles for gestational age were calculated using Hadlock growth curves [17-18]. Small for gestational age (SGA) was categorized as birthweight less than 10th percentile for gestational age. Not small for gestational age (NSGA) was categorized as birthweight percentile at 10th percentile or greater. Demographic, clinical and placental pathologic information were obtained for 1,078 patients, including 197 with PE (19 African ancestry, 34 Asian ancestry and 144 European ancestry; including 37 of Hispanic ethnicity), 143 SGA normotensive (18 African ancestry, 34 Asian ancestry and 91 European ancestry; including 14 of Hispanic ethnicity) and 738 NSGA normotensive (40 African ancestry, 114 Asian ancestry and 584 European ancestry; including 105 of Hispanic ethnicity). Of these, RNA sequencing was performed on 203 placentas (46 SGA PE, 40 SGA normotensive and 117 NSGA normotensive). Study inclusion and exclusion criteria are presented in Figure S1.

The study was approved by University of California, San Diego (UCSD) institutional review board, approval #090652 (from Jan 2009 – Jan 2019) and #181917 (from Jan 2019 to Jan 2027).

Placental pathologic examination

Placenta macroscopic and histologic examinations were performed according to recommendations in the Amsterdam placental group consensus statement [19]. Placental discs, trimmed of cord and membranes were weighed. Placenta size percentile for gestational age was determined based on trimmed placenta weights compared to the collaborative perinatal study reference weights [20]. Small placentas for gestational age were categorized as placentas that weighed less than 10th percentile. A minimum of four tissue cassettes – one containing umbilical cord and membranes and three full thickness placenta disc sections – were formalin-fixed, paraffin-embedded, stained with hematoxylin and eosin, and microscopically examined; additional sections were submitted if placental lesions were grossly identified. Five placenta lesions were studied: 1. Small placenta for gestational age. 2. Villous hypermaturity a.k.a. accelerated villous maturation. 3. Decidual vasculopathy. 4. Infarct. 5. Fetal vascular malperfusion (FVM). Lesions 1-4 fall under the umbrella of maternal vascular malperfusion (MVM). FVM was diagnosed based on findings of fetal vascular thrombi, segmental avascular villi and/or villous stromal-vascular karyorrhexis. Both MVM and FVM have been associated with PE in multiple prior studies, including our own [8-9; 15, 21].

Clinical and pathologic analyses

Maternal age, parity, body mass index (BMI) at delivery, history of diabetes mellitus, maternal disease severity (adjudicated by two board-certified obstetricians), neonatal disease severity (as determined by NICU admissions noted from maternal charts at delivery) and placenta macroscopic and histologic findings were evaluated. Statistical analyses (including likelihood ratios) were performed to determine which clinical PE categorizations – SGA PE vs. NSGA PE, EDPE vs. LDPE and PE with versus without SF – best distinguished PE associated with placental abnormalities from PE without associated placenta abnormalities, and which were associated with maternal and infant disease severity. Maternal age, parity, BMI and history of diabetes mellitus were chosen as the studied maternal risk factors because these were consistently recorded in the course of obstetric visits.

GraphPad Prism (version 9.4.1) was used for statistical analysis; p values were calculated using t-test (with Welch’s correction) for normally distributed parametric data, Mann Whitney U test for nonparametric data and Fisher’s exact test for categorical variables. P-values were analyzed and adjusted for false discovery rate (FDR) using Benjamini, Krieger and Yekutieli two-stage linear step-up procedure with Q: 5% to yield adjusted p-values (q-values). Wilson-Brown was used for computing confidence intervals (CI), sensitivity (Sens), specificity (Spec), positive predictive value (PPV), negative predictive value (NPV), relative risk (RR) and likelihood ratio (LR). Logistic regression analyses were performed adjusting for maternal age, BMI, parity and diabetes mellitus (when comparing PE categorization systems: design = ≥3 of 5 placental pathologic findings ~ Intercept + SGA PE (or EDPE or PE with severe features) + P0 parity + DM + Maternal age + BMI. When comparing PE to normotensive: design = SGA PE (or NSGA PE) ~ Intercept + P0 parity + DM + ≥3 of 5 placental pathologic findings + Maternal age + BMI + BMI≥30 + 3 of 3 maternal risk factors). All statistical tests were 2-sided, with values <.05 considered statistically significant.

RNA isolation from placental tissue, bulk RNA sequencing and analysis

We performed RNA isolation from placental tissue and bulk RNA sequencing using methods we have previously described [15, 21-22]. Briefly, for each placenta, fresh placenta disc villous tissue: two 2x2x2 cm chunks from two different regions, each about half way between cord insertion site and margins were collected, dissected and stored in RNAlater (ThermoFisher). Total RNA was isolated using mirVana RNA isolation kit (ThermoFisher). RNA integrity was checked using RNA 6000 Nano chip, read by a 2100 bioanalyzer (Agilent). RNA samples in this dataset had RIN values ranging from 5.5 to 9.2, with a median of approximately 7.8, indicating high RNA integrity across the data set. RNA-seq libraries were prepared and sequencing performed at the UCSD IGM Genomics Center. Data were accounted for batch effect, gestational age and gender (i.e. design formula as follows; design = ~ batch effect + gestational age + gender + condition), and adjusted p-value <.05 was considered as differentially expressed. Gene set enrichment analysis (GSEA) was performed using the R (version 4.4.3) package Fast Gene Set Enrichment Analysis (version 1.32.2) using MsigDB Hallmark gene set pathway analysis [23]. Gene ontology analysis was performed using Enrichr [24]. Clustered bar graphs were constructed using Microsoft excel. Venn diagrams were constructed using BioVenn [25].

Results

Evaluation of clinical PE categorization systems

Placental abnormalities

With all three categorization systems, there were significantly more placental abnormalities in the more clinically severe subset, with at least 1 of the 5 evaluated placenta abnormalities present in 100%, 100% and 83.5% of SGA PE, EDPE and PE with SF categories respectively (vs. 64.2%, 71.1% and 61.4% of the NSGA PE, LDPE and PE without SF categories) (Table 1; Table S1). However, when multiple placenta abnormalities were considered, the SGA PE vs. NSGA PE categorization system outperformed the other systems in distinguishing PE associated with multiple placental abnormalities from PE with few to no placental abnormalities (LR (95% CI) for ≥ 3 of 5 placenta pathologic findings: 15.7 (6.5, 38.1) in SGA PE vs. NSGA PE; 6.8 (4.3, 10.8) in EDPE vs. LDPE and 5.2 (1.95, 14.1) in PE with SF versus PE without SF; AUC (95% CI) for ≥3 placenta pathologic findings as the dependent variable: 0.91 (0.86, 0.96) in SGA PE vs. NSGA PE with 52.5 (18.1, 190.6) SGA PE odds ratio; 0.83 (0.76, 0.90) in EDPE vs. LDPE with 30.9 (11.4, 97.2) EDPE odds ratio and 0.75 (0.68, 0.83) in PE with SF vs. PE without SF with 7.5 (2.7, 26.8) PE with SF odds ratio; all p <.0001) (Table 1; Figure 1A and 1B; Figures S2-4).

Table 1:

Placental pathologic findings evaluated according to preeclampsia (PE) subcategorizations. There were 197 cases of PE. The cases were compared based on three PE categorizations: 1. PE associated with calculated birthweight <10th percentile (small) gestational age (SGA PE) versus PE associated with calculated birthweight >10th percentile for gestational age i.e. not small for gestational age (NSGA PE). 2. PE associated with early delivery (<34 weeks gestation; EDPE) versus PE associated with late delivery (≥ 34 weeks gestation; LDPE). 3. PE with severe features (PE with SF) versus PE without severe features (PE without SF). Statistical analyses were performed using GraphPad Prism; p values were calculated using t-test (with Welch’s correction) for normally distributed parametric data, Mann Whitney U test for nonparametric data and Fisher’s exact test for categorical variables, with p values <.05 considered as statistically significant.

	SGA PE (n=63)	NSGA PE (n=134)	EDPE (n=31)	LDPE (n=166)	PE with SF (n=127)	PE without SF (n=70)	p-value SGA PE vs. NSGA PE	p-value EDPE vs. LDPE	p-value PE with SF vs. PE without SF
SPGA	48 (76.2%)	33 (24.6%)	22 (71%)	59 (35.5%)	60 (47.2%)	21 (30%)	<.0001****	.0003***	.023*
Plac. HM	39 (61.9%)	22 (16.4%)	31 (100%)	30 (18.1%)	52 (40.9%)	9 (12.9%)	<.0001****	<.0001****	<.0001****
Infarct	38 (60.3%)	29 (21.6%)	18 (58.1%)	49 (29.5%)	49 (38.6%)	18 (25.7%)	<.0001****	.003**	.084
DV	31 (49.2%)	26 (19.4%)	21 (67.7%)	36 (21.7%)	44 (34.7%)	13 (18.6%)	<.0001****	<.0001****	.021*
FVM	24 (38.1%)	18 (13.4%)	12 (38.7%)	30 (18.1%)	32 (25.2%)	10 (14.3%)	.0002***	.016*	.101
≥ 1 of 5 Placental abnl.	63 (100%)	86 (64.2%)	31 (100%)	118 (71.1%)	106 (83.5%)	43 (61.4%)	<.0001**** Sens: 100% (94.3-100%) Spec: 35.8% (28.2-44.2%) PPV: 42.3% (34.6-50.3%) NPV: 100% (92.6-100%) LR: 1.56 (1.37, 1.77)	.0001*** Sens: 100% (89-100%) Spec: 28.9% (22.6-36.2%) PPV: 20.8% (15.1-28%) NPV: 100% (92.6-100%) LR: 1.41 (1.28, 1.55)	.0009*** Sens: 83.5% (75.8-89.5%) Spec: 38.6% (27.2-51%) PPV: 71.1% (66.8-75.1%) NPV: 56.3% (44.1-67.7%) LR: 1.36 (1.11, 1.66)
≥ 3 of 5 Placental abnl.	37 (58.7%)	5 (3.7%)	24 (77.4%)	19 (11.4%)	38 (29.9%)	4 (5.7%)	<.0001**** Sens: 58.7% (46.4-70%) Spec: 96.3% (91.6-98.4%) PPV: 88.1% (75-94.8%) NPV: 83.2% (76.6-88.3%) LR: 15.7 (6.5, 38.1)	<.0001**** Sens: 77.4% (60.2-88.6%) Spec: 88.6% (82.8-92.6%) PPV: 55.8% (41.1-69.6%) NPV: 95.5% (90.9-97.8%) LR: 6.8 (4.3, 10.8)	<.0001**** Sens: 29.9% (22.1-38.7%) Spec: 94.3% (86-98.4%) PPV: 90.5% (78-96.2%) NPV: 42.6% (39.5-45.7%) LR: 5.2 (1.95, 14.1)

Figure 1:

Comparisons of preeclampsia (PE) subclasses. 197 cases of PE were compared based on three PE categorizations: 1. PE associated with calculated birthweight <10th percentile (small) for gestational age (SGA PE) versus PE associated calculated birthweight ≥10th percentile (not small) for gestational age (NSGA PE). 2. PE associated with delivery <34 weeks gestation (EDPE) versus PE associated with delivery ≥ 34 weeks gestation (LDPE). 3. PE with severe features (PE with SF) versus PE without severe features (PE without SF). A. Forest plot of the likelihood ratio of finding 3 or more of 5 evaluated placental abnormalities in the PE subclasses. B. ROC curves of logistic regression analyses performed adjusting for maternal age, BMI, parity and diabetes mellitus (design = ≥3 of 5 placental pathologic findings ~ Intercept + SGA PE (or EDPE or PE with severe features) + P0 parity + DM + Maternal age + BMI. C. Clustered bar graph comparing disease severity and placental abnormality percentages in the PE subclasses.

Disease severity

The percentage of patients who delivered before 34 weeks of gestation was high in the EDPE category (100%), by definition, but only 38.1% of SGA PE and 29% of PE with SF patients delivered before 34 weeks of gestation. Less than 5% of patients in the NSGA PE, LDPE and PE without SF categories delivered before 34 weeks gestation. The percentage of patients with SGA infants was high in the SGA PE category by definition (100%) as well as in the EDPE category (80.6%). SGA infants were present in 44.9%, 22.9%, 8.6% and 0% of PE with SF, LDPE, PE without SF and NSGA PE categories respectively. The SGA PE, EDPE and PE with SF categories contained high percentages of patients with severe maternal disease (i.e. PE with severe features; 90.5% in SGA PE, 93.5% in EDPE and 100% in PE with SF) and NICU admissions (84.1% in SGA PE, 100% in EDPE and 66.1% in PE with SF). The percentages of patients with severe maternal disease in the NSGA PE, LDPE and PE without SF categories ranged from 0% (PE without SF) to 59% (LDPE). NICU admissions occurred in 45.5%, 50% and 44.3% of NSGA PE, LDPE and PE without SF parturitions respectively. When both severe maternal disease and NICU admissions were considered, the SGA/NSGA PE categorization was similar to the EDPE/LDPE categorization (LR: 3.1 vs. 2.9; both p<.0001). The PE with/without SF categorization outperformed both, by definition (Table S2; Figure 1B). There were more female than male neonates in the SGA PE and EDPE categories, but this did not reach statistical significance (Table S2).

Validation of the SGA/NSGA PE categorization system by comparing to SGA and NSGA normotensive parturitions

Given that the SGA vs. NSGA PE categorization system outperformed the others in distinguishing PE associated with placenta abnormalities from PE with few to no placental abnormalities, and was also a relatively good indicator of maternal and neonatal disease severity, we further evaluated this PE categorization system in comparison to SGA and NSGA normotensive parturitions.

Placental abnormalities

Placental abnormalities were significantly more common in all the parturitions associated with clinical disease compared to NSGA normotensive parturitions. There was at least 1 of the 5 evaluated placental pathologies in 100%, 64.2% and 90.2% of SGA PE, NSGA PE and SGA normotensive parturitions respectively compared to 54.6% of NSGA normotensive parturitions. In other words, all of the SGA PE cases had MVM and/or FVM, and 35.8%, 9.8% and 45.4% of NSGA PE, SGA normotensive and NSGA normotensive respectively had no MVM or FVM abnormalities (Table 2). In comparison to placentas from NSGA normotensive patients, placentas from both SGA PE and SGA normotensive patients were significantly associated with small placentas from gestational age (p <.0001), but SGA PE placentas had significantly more placental hypermaturity, infarcts and decidual vasculopathy (each p <.0001) than SGA normotensive placentas (Table 2; Table S3). The presence of 3 or more of the 5 evaluated placental abnormalities performed well in distinguishing SGA PE from SGA normotensive (p<.0001; LR 7.0), SGA PE from NSGA normotensive (p<.0001; LR 24.1) and SGA normotensive from NSGA normotensive (p=.001; LR 3.4), but not NSGA PE from NSGA normotensive (p=.380; q=.099) (Table 2; Table S3; Figure 2A).

Table 2:

Evaluation of placental findings in preeclampsia associated with calculated birthweight <10th percentile (small) gestational age (SGA PE) and PE associated with calculated weight ≥10th percentile for gestational age (NSGA PE) compared to SGA and NSGA normotensive parturitions.

	SGA PE (n=63)	NSGA PE (n=134)	SGA Norm (n=143)	NSGA Norm (n=738)	p-value SGA PE vs. SGA Norm	p-value SGA PE vs. NSGA Norm	p-value NSGA PE vs. NSGA Norm	p-value SGA Norm vs. NSGA Norm
SPGA	48 (76.2%)	33 (24.6%)	105 (73.4%)	232 (31.4%)	.732	<.0001****	.126	<.0001****
Plac. HM	39 (61.9%)	22 (16.4%)	15 (10.5%)	50 (6.8%)	<.0001****	<.0001****	.0005***	.119
Infarct	38 (60.3%)	29 (21.6%)	16 (11.2%)	60 (8.1%)	<.0001****	<.0001****	<.0001****	.254
DV	31 (49.2%)	26 (19.4%)	15 (10.5%)	59 (8%)	<.0001****	<.0001****	.0002***	.324
FVM	24 (38.1%)	18 (13.4%)	34 (23.8%)	131 (17.8%)	.044*	.0003***	.262	.101
≥ 1 of 5 Plac.	63 (100%)	86 (64.2%)	129 (90.2%)	403 (54.6%)	.006**	<.0001****	.047*	<.0001****
≥ 3 of 5 Plac.	37 (58.7%)	5 (3.7%)	12 (8.4%)	18 (2.4%)	<.0001**** Sens: 58.7% (45.6-71%) Spec: 91.6% (85.8-95.6%) PPV: 75.5% (63.3-84.6%) NPV: 83.4% (78.9-87.2%) LR: 7.0 (3.9, 12.5)	<.0001**** Sens: 58.7% (45.6-71%) Spec: 97.6% (96.2-98.6%) PPV: 67.3% (55.5-77.2%) NPV: 96.5% (95.4-97.4%) LR: 24.1 (14.6, 39.7)	.380	.001** Sens: 8.4% (4.4-14.2%) Spec: 97.6% (96.2-98.6%) PPV: 40% (24.7-57.5%) NPV: 84.6% (83.9-85.3%) LR: 3.4 (1.7, 7.0)

Figure 2.

Comparisons of small for gestational age (SGA) and not small for gestational age (NSGA) preeclampsia (PE) to SGA and NSGA normotensive (Norm) parturitions. 197 cases of PE (63 SGA PE and 134 NSGA PE) were compared to 881 normotensive (143 SGA Norm and 738 NSGA Norm) parturitions. A. Forest plot of the likelihood ratio associated with 3 or more of the 5 evaluated placenta abnormalities in SGA PE and NSGA PE compared to SGA Normotensive and NSGA Normotensive placentas B. Forest plot of the relative risk associated with 3 of 3 evaluated maternal risk factors (BMI ≥ 30, P0 parity and diabetes mellitus) in SGA PE and NSGA PE compared to SGA Normotensive and NSGA Normotensive patients. C. ROC curves of SGA PE compared to NSGA normotensive: design = SGA PE ~ Intercept + P0 parity + DM + ≥3 of 5 placental pathologic findings + Maternal age + BMI + BMI≥30 + 3 of 3 maternal risk factors. D. ROC curves of NSGA PE compared to NSGA normotensive: design = NSGA PE ~ Intercept + P0 parity + DM + ≥3 of 5 placental pathologic findings + Maternal age + BMI + BMI≥30 + 3 of 3 maternal risk factors.

Disease severity

Gestational age at delivery was significantly lower in SGA PE, NSGA PE and SGA normotensive gestations compared to NSGA normotensive gestations. Average gestational age at delivery was 34.3 weeks, 37.9 weeks and 37.6 weeks in SGA PE, NSGA PE and SGA normotensive respectively compared to 38.4 weeks for NSGA normotensive. However, delivery at less than 34 weeks of gestation was significantly more common only in SGA PE compared to both SGA and NSGA normotensive gestations. NICU admissions were also significantly more common in SGA PE, NSGA PE and SGA normotensive gestations in comparison to NSGA normotensive gestations (Table S4). In comparison to NSGA normotensive gestations, infants of SGA normotensive patients were more frequently female (p=.008), but fetal/infant sex did not significantly differ between SGA PE and SGA normotensive, SGA PE and NSGA normotensive, and NSGA PE and NSGA normotensive gestations (Table S4).

Evaluation of maternal risk factors in SGA and NSGA PE compared to SGA and NSGA normotensive parturitions

We sought to determine which maternal risk factors predisposed to SGA versus NSGA PE in comparison to SGA and NSGA normotensive parturitions. We compared maternal age, parity, BMI at delivery and history of diabetes mellitus. These risk factors were consistently recorded in the course of obstetric visits, with information thus available for all included study patients.

Elevated maternal BMI and maternal obesity were significantly more common in SGA PE and NSGA PE compared to both SGA normotensive and NSGA normotensive patients (relative risk (RR) 2.06-2.49; p<.01) (Table 3). Interestingly, average maternal BMI was lower in SGA normotensive compared to NSGA normotensive patients (28.1 vs. 30.0; p=.028; q=.029) (Table 3; Table S5). P0 parity (primiparity) was significantly more common in SGA PE and SGA normotensive compared to NSGA normotensive gestations (66.7% and 55.2% in SGA PE and SGA normotensive compared to 39.3% in NSGA normotensive; p<.0001 and p=.0006; Table 3). All three evaluated maternal risk factors were significantly more prevalent in NSGA PE compared to NSGA normotensive (RR 2.49, 2.22 and 1.56 for BMI >30, DM and P0 parity respectively) (Table 3). The combination of all 3 maternal risk factors was significantly more prevalent only in NSGA PE compared to NSGA normotensive (RR 3.07 (2.02, 4.35); p <.0001; q=.0003;) (Table 3; Table S5; Figure 2B).

Table 3:

Evaluation of maternal risk factors in preeclampsia associated with calculated birthweight <10th percentile (small) gestational age (SGA PE) and PE associated with calculated birthweight ≥10th percentile for gestational age (NSGA PE) compared to SGA and NSGA normotensive parturitions.

	SGA PE (n=63)	NSGA PE (n=134)	SGA Norm (n=143)	NSGA Norm (n=738)	p-value SGA PE vs. SGA Norm	p-value SGA PE vs. NSGA Norm	p-value NSGA PE vs. NSGA Norm	p-value SGA Norm vs. NSGA Norm
Mat. age	31.1 ± 6.1	31.8 ± 5.2	31.7 ± 5.2	32.4 ± 5.0	.453	.097	.159	.162
Mat. BMI	31.7 ± 6.4	34.0 ± 7.8	28.1 ± 4.3	30.0 ± 6.8	<.0001****	.005**	<.0001****	.028*
BMI ≥ 30	36 (57.4%)	86 (64.2%)	44 (30.8%)	279 (37.8%)	.0006*** RR: 2.10 (1.39, 3.17)	.003** RR: 2.06 (1.28, 3.30)	<.0001**** RR: 2.49 (1.80, 3.45)	.129
Mat. DM	15 (23.8%)	56 (41.8%)	20 (14%)	157 (21.3%)	.107	.633	<.0001**** RR: 2.22 (1.63, 3.01)	.052
Parity P0	42 (66.7%)	70 (52.2%)	79 (55.2%)	290 (39.3%)	.167	<.0001**** RR: 2.83 (1.76, 4.66)	.006** RR: 1.56 (1.14, 2.12)	.0006*** RR: 1.71 (1.27, 2.31)
3 of 3 Mat.	3 (4.8%)	18 (13.4%)	5 (3.5%)	24 (3.3%)	.702	.464	<.0001**** RR: 3.07 (2.02, 4.35)	.801

Evaluation of both placental abnormalities and maternal risk factors in SGA and NSGA PE compared to SGA and NSGA normotensive parturitions

Logistic regression analysis designed to evaluate both placental abnormalities and maternal risk factors revealed ≥3 placental findings strongly discriminated SGA PE from NSGA normotensive (AUC (95% CI) and odds ratio of ≥3 placental findings (95% CI): 0.88 (0.82, 0.93) and 73.7 (34.4, 169.8) with all variables combined, AUC: 0.71 (0.64, 0.77) with just maternal risk factors, and AUC: 0.78 (0.71, 0.86), OR: 56.9 (29.2, 115.6) with just placental findings) (Figure 2C; Figure S5). Three or more placental findings also strongly discriminated SGA PE from SGA normotensive (AUC: 0.85 (0.78, 0.90), OR: 14.2 (6.37, 34.0); Figure S6A). Small placentas for gestational age discriminated SGA normotensive from NSGA normotensive (AUC: 0.75 (0.70, 0.79), OR: 5.6 (3.78, 8.57)) (Figure S6B), but the AUC and OR fell with ≥3 placental findings (Figure S6C). Maternal risk factors drove the discrimination of NSGA PE from NSGA normotensive; the presence of ≥3 placental findings was not useful (AUC (95% CI) and odds ratio of ≥3 placental findings (95% CI): 0.71 (0.66, 0.76) and 2.06 (0.64, 5.59) with all variables combined, AUC: 0.71 (0.66, 0.76) with just maternal risk factors and AUC: 0.51 (0.43, 0.56), OR: 1.55 (0.50, 3.97) with just placental findings) (Figure 2D; Figure S7).

Evaluation of the generalizability of the SGA PE vs. NSGA PE categorization system according to ancestry

A clinically useful categorization system should be generalizable across human ancestral groups. We thus separated patients according to geographic ancestral origins and re-evaluated the performance of the SGA vs. NSGA PE categorization system.

African ancestry

The presence of 3 or more of the 5 evaluated placenta abnormalities performed well in distinguishing SGA PE from NSGA PE (p=.003; LR >100), SGA PE from SGA normotensive (p=.0003; LR 13.1) and SGA PE from NSGA normotensive (p<.0001; LR >100), but not NSGA PE from NSGA normotensive (p >.999) (Table S6). Small placentas for gestational age were significantly more common in SGA normotensive compared to NSGA normotensive (p=.0003) (Table S6).

Asian ancestry

The presence of 3 or more of the 5 evaluated placenta abnormalities performed well in distinguishing SGA PE from NSGA PE (p=.013; LR 5.65), SGA PE from SGA normotensive (p=.005; LR 4.58), SGA PE from NSGA normotensive (p<.0001; LR 30.7) and SGA normotensive from NSGA normotensive (p=.025; LR 6.7), but not NSGA PE from NSGA normotensive (p =.114) (Table S7). Small placentas for gestational age were significantly more common in SGA normotensive compared to NSGA normotensive (p=.007) (Table S7).

European ancestry

The presence of 3 or more of the 5 evaluated placenta abnormalities performed well in distinguishing SGA PE from NSGA PE (p<.0001; LR 19.7), SGA PE from SGA normotensive (p<.0001; LR 7.33), SGA PE from NSGA normotensive (p<.0001; LR 20.6) and moderately in distinguishing SGA normotensive from NSGA normotensive (p=.025; LR 2.81), but not NSGA PE from NSGA normotensive (p>.999) (Table S8). Small placentas for gestational age were significantly more common in SGA normotensive compared to NSGA normotensive (p>.0001) (Table S8).

Molecular analysis

There were high rates of the evaluated placental abnormalities in both SGA PE and SGA normotensive gestations. To determine which signaling pathways are altered in placentas from patients with pregnancies complicated by PE and SGA infants, we performed RNA sequencing and Gene Set Enrichment Analysis using the Hallmark gene sets from human MSigDB in placentas from SGA PE and SGA normotensive patients compared to NSGA normotensive patients (Table S9). Hypoxia, TNF alpha, unfolded protein response, TGF beta, protein secretion, mTORC1, estrogen response, glycolysis, allograft rejection, UV response up, p53, inflammatory response, MYC targets, interferon gamma and androgen response pathways were significantly upregulated in SGA normotensive compared to NSGA normotensive (padj <.05), while UV response down and myogenesis pathways were down-regulated (Figure 3A; Tables S10-11). Hypoxia, unfolded protein response, glycolysis, heme metabolism, TNF alpha, mTORC1, TGF beta, UV response up, estrogen response, DNA repair, p53 and adipogenesis pathways were significantly upregulated in SGA PE compared to NSGA normotensive while UV response down, E2F targets, KRAS signaling, peroxisome, angiogenesis, bile acid metabolism and epithelial mesenchymal transition pathways were down-regulated (padj <.05) (Figure 3B; Tables S12-13). Ten pathways were altered in common in SGA PE and SGA normotensive placentas compared to NSGA normotensive placentas. Of the ten, nine pathways – hypoxia, TNF alpha, glycolysis, unfolded protein response, p53, estrogen response, UV response up, TGF beta and mTORC1 – were upregulated in common (Figure 3C).

Figure 3:

Molecular analysis and Model. RNA sequencing was performed on 203 placentas. A. Pathways significantly differentially expressed in placentas from normotensive patients with calculated birthweights <10th percentile (small) for gestational age (SGA Norm; N = 40) compared to placentas from normotensive patients with calculated birthweights ≥10th percentile (not small) for gestational age (NSGA Norm; N = 117). B. Pathways significantly differentially expressed in placentas from patients with preeclampsia and small for gestational age fetus/neonate (SGA PE; N = 46) compared to placentas from NSGA Norm. C. Shared pathways altered in SGA PE and SGA Norm in comparison to NSGA Norm. D. Model for the interplay between maternal, fetal and placenta factors in the pathophysiology of PE and SGA normotensive gestations.

Discussion

Some authors have given PE the moniker, “disease of theories” due to the enigma surrounding the etiology of PE. Studies have shown associations of PE with maternal congenital heart disease [26], maternal obesity [27-28], maternal diabetes mellitus [3-4], maternal autoimmune disease [29-30], vitamin D deficiency [31], fetal trisomy 13 [32], and maternal-fetal HLA/Kir incompatibility [reviewed in 33], among others. Regardless of the possible triggers, it is now widely accepted that a dysfunctional/malperfused/hypoxic placenta plays a central role in the disease [5-9]. However, not all cases of PE have macroscopic or histologic placental abnormalities, suggesting that extra-placental factors are sufficient to cause PE in a subset of patients.

In the current study, we found that classifying PE based on birthweight percentile for gestational age performed best in distinguishing PE associated with multiple MVM and/or FVM placenta abnormalities (“placental” PE) from PE associated with few to no MVM/FVM placenta abnormalities (“maternal” PE). This classification system performed well when all our cases were combined, as well as when our patients were divided into their geographic ancestral origins.

SGA PE is placental; all SGA PE cases in our study were associated with at least one MVM/FVM placental abnormality. Notably, placental abnormalities were also frequent in SGA normotensive placentas with at least one placental abnormality being present in 90.2% of the placentas. Both SGA PE and SGA normotensive placentas had a high frequency of small placentas for gestational age while SGA PE placentas additionally had increased placental hypermaturity, infarcts and decidual vasculopathy in comparison to SGA normotensive placentas. Maternal obesity was more prevalent in SGA PE compared to SGA normotensive patients. These findings all together suggest that placental abnormalities are important contributors to SGA fetuses/infants and maternal obesity may additionally contribute to the development of PE in a subset of patients with abnormal placentas or, more likely, that lower maternal BMI may protect from PE in a subset of patients with abnormal placentas.

The identification of small for gestational age fetus or neonate may thus be a harbinger or indicator of preeclampsia associated with placental abnormalities. Per ACOG recommendations, screening for growth restriction should begin at 24 weeks gestation [34].

In addition to high rates of placental abnormalities, both SGA PE and SGA normotensive placentas shared alterations in hypoxia, TNF alpha, glycolysis, unfolded protein response, p53, estrogen response, UV response, TGF beta and mTORC1 signaling pathways when compared to placentas from NSGA normotensive patients. Of the shared altered pathways, hypoxia and TNF alpha have been studied in animal models and shown to be maladaptive. Tong et al. found that chronic hypoxia in pregnant sheep led to increased placental hypoxia-mediated oxidative damage and fetal growth restriction [35]. TNF alpha infusion in pregnant rats led to placental mitochondrial dysfunction, increased blood pressures and fetal demise [36]. Studies of therapeutic interventions targeting these shared altered pathways are therefore warranted as they could potentially support fetal growth and prevent or alleviate PE.

Female sex was more commonly associated with SGA, particularly SGA normotensive gestations in comparison to NSGA normotensive gestations. In a study of 29,530 deliveries, growth restriction was higher in female newborns. However, male sex was more often associated with unfavorable outcomes like non-reassuring fetal heart rate and acidemia [37].

EDPE is also placental, with placental abnormalities detected in all of the EDPE cases in our study. However, the main advantage of the SGA/NSGA PE system over the EDPE/LDPE system lies in the ability of the SGA/NSGA PE system to identify cases with multiple placental abnormalities and increased risk for maternal and neonatal disease severity, beyond 34 weeks of gestation.

The limitation of the SGA/NSGA PE classification system lies in the heterogeneity of NSGA PE. Though 35.8% of NSGA PE placentas had no MVM or FVM abnormalities, at least one of the 5 evaluated MVM/FVM placental abnormalities was present in 64.2% of NSGA PE cases. This suggests that NSGA PE is multifactorial with MVM/FVM placental abnormalities contributing to a subset of cases and maternal (and/or fetal) factors being sufficient to cause PE in another subset. A combination of 3 maternal risk factors (BMI ≥ 30, diabetes mellitus and P0 parity) was significantly more common in NSGA PE patients (p <.0001; RR 3.07). Evaluation of maternal blood matched to placenta tissue in NSGA PE patients compared to normotensive controls may be informative in identifying additional maternal factors contributing to NSGA PE.

Overall, 63 of 197 (32%) of our PE cases were SGA PE (“placental” PE), 48 of 197 (24%) were NSGA PE without MVM or FVM abnormalities (“maternal” PE) and 86 of 197 (44%) were NSGA PE with one or more placenta MVM or FVM abnormalities (multifactorial). Our findings support interplay between maternal, fetal and placental factors in the etiology of PE (modeled in Figure 3D).

In their review article addressing clinically meaningful PE subtypes, Roberts et al. [16] suggest that a clinically useful PE subtype should: 1. Identify a specific pathophysiologic pathway. 2. Indicate maternal or fetal outcome, 3. Be recognizable in a clinically useful time frame and 4. Be reproducible and generalizable, including in low resource settings. The SGA vs. NSGA PE categorization mostly meets these criteria. 1. The SGA vs. NSGA PE categorization consistently identifies PE associated with macroscopic (small placenta for gestational age), microscopic (hypermaturity, decidual vasculopathy, infarcts and fetal vascular malperfusion) and molecular (hypoxia, TNF alpha, glycolysis, p53, etc.) placental abnormalities i.e. identifies PE caused by placental dysfunction (SGA PE). 2. SGA PE is associated with maternal and fetal/infant disease severity, regardless of gestational age at delivery. 3. Detection of poor growth in utero or SGA at birth may potentially be amenable to interventions, targeting pathways altered in abnormal placentas, to support placental health and fetal growth, or prevent recurrent SGA PE. 4. The SGA vs. NGSA PE categorization is generalizable and still performs well when patients are evaluated according to their geographic ancestral origins.

Study limitations

We had a limited number of NSGA PE placentas without MVM or FVM abnormalities in our RNA-seq data set. Thus, our study is limited by our inability to compare the molecular profile of “maternal” PE placentas to placentas from normotensive patients. Leavey et al [12-13] have however performed a similar analysis using unsupervised microarray analysis on placentas from patients with PE and found that PE placentas without placental abnormalities clustered with normotensive placentas.

Another limitation of our study is our use of delivery time at 34 weeks gestational age versus the more conventional use of disease onset at 34 weeks for the classification of early-onset (EOPE; <34 weeks) versus late-onset (LOPE; ≥34 weeks) PE. The exact time of PE onset was not consistently recorded in our database and thus we used time of delivery instead. With the EDPE/LDPE system, EDPE is EOPE as delivery before 34 weeks is an indicator of PE onset before 34 weeks of gestation. However, LDPE may have included some cases of EOPE that were managed medically till delivery on or after 34 weeks gestation.

The SGA/NSGPA PE categorization system likely outperformed the EDPE/LDPE and PE with/without severe features categorizations due to the categorization being based on an objective measure (birthweight) rather than more subjective assessments (such as determinations of when to induce delivery or reliance on self-reported symptoms).

Our study is also limited by our use of maternal BMI at delivery for evaluation of PE risk, rather than the more traditional use of pre-pregnancy BMI. Many patients in our study population were transferred to our tertiary center for higher level care and thus, pre-pregnancy BMI information was unavailable.

Other limitations of our study include the small sample size in the African ancestry subgroup analysis which limits statistical reliability and our ability to evaluate the influence of socioeconomic and/or environmental factors on our study findings. Validation in cohorts containing larger numbers of African ancestry participants is needed. Given the fact that our study was performed in a single institution, external validation of our RNA-seq findings is also needed. Of note, a previously published subset of our RNA-seq dataset [21] has been externally validated [38].

Finally, though we performed a rigorous case-control study validation, including comparing PE to normotensive patients with and without SGA, as a tertiary care center, our samples may over-represent complex cases. Studies in community settings are thus also needed.

Perspective

We evaluated clinical PE classification systems and found that classifying PE based on birthweight percentile for gestational age performed best in distinguishing PE associated with multiple MVM and/or FVM placenta abnormalities (SGA PE) from PE associated with few to no MVM/FVM placenta abnormalities (NSGA PE). We found high rates of placental abnormalities in SGA PE and SGA normotensive placentas with shared molecular alterations in ten signaling pathways. The implications of our study findings are two-fold: 1. The detection of poor fetal growth in utero or SGA at birth is highly suggestive of placental abnormalities, and suggests further studies into therapeutic interventions targeting the shared pathways altered in abnormal placentas are warranted. Low-dose aspirin for instance has been shown to inhibit hypoxia-induced sFLT-1 expression in placental trophoblasts and endothelial cells [39-40]. Low-dose aspirin may thus potentially be useful upon in-utero identification of SGA, or as a means of preventing recurrent SGA/SGA PE in subsequent pregnancies. 2. There is a high relative risk of NSGA PE in patients with high BMI, DM and/or P0 parity. Patients with NSGA PE, particularly without identified MVM or FVM placental abnormalities, may thus benefit from weight loss and/or DM control to prevent recurrent PE in subsequent pregnancies. Patients with PE should thus have their placentas examined after delivery.

Supplementary Material

Tables S1-S13

Figures S1-S7

Novelty and Relevance.

What is new?

Birthweight percentile for gestational age can be used as a surrogate for identifying patients with multiple placental abnormalities.

What is relevant?

Hypoxia, TNF alpha, glycolysis, unfolded protein response, estrogen response, UV response, p53, TGF beta and mTORC1 signaling pathways are altered in abnormal placentas.

Clinical/Pathophysiological implications?

The identification of small for gestational age fetus or neonate may be a harbinger or indicator of preeclampsia associated with placental abnormalities. Therapies targeting hypoxia, TNF alpha and other signaling pathways altered in abnormal placentas are worth further study as potential interventions to support fetal growth and possibly prevent or alleviate preeclampsia.

Nonstandard Abbreviations and Acronyms

Abnl

Abnormality

DM

Diabetes mellitus

DV

Decidual vasculopathy

Deliv.

Delivery

EDPE

Early delivery preeclampsia – preeclampsia associated with delivery before 34 weeks gestational age

FVM

Fetal vascular malperfusion

GA

Gestational age

HM

Hypermaturity

LDPE

Late delivery preeclampsia – preeclampsia associated with delivery at or after 34 weeks gestational age

Mat.

Maternal

MVM

Maternal vascular malperfusion

NICU

Neonatal intensive care unit

NSGA

Not small for gestational age

PE

Preeclampsia

Plac.

Placenta

Sens

Sensitivity

SF

severe features

SGA

Small for gestational age

Spec

Specificity

SPGA

Small placenta for gestational age

wk.

Week